Sciencemadness Discussion Board

Power Supplies

 Pages:  1  

Xenos - 4-1-2003 at 22:07

I am trying to preform electrolysis of various compounds and need a sufficient power supply. From my understanding, high amps and low volts is the best. I tried rewiring a microwave transformer, but the results were not pleasing. I need a reletivly cheap high amp power supply. Anyone have any ideas? I did find this on ebay. Is that any good?

rikkitikkitavi - 5-1-2003 at 00:59

It is a very good powersupply for that price.
The main disadvantage is that it is difficult to pull 48 A @ 5V unless you use thick wires and a several parallell connected eletrolysis cells. So maybe it is a bit oversized, but that increases the possibility to run it continously for 24/7 duty.

A cheap alterative is old PC-AT power supplies. You can usually pull about 20 A from the 5V.

trinitrotoluene - 5-1-2003 at 23:23

Maybe a car battery recharger will do the trick. Its 12 volts and 7 amps DC power.

rikkitikkitavi - 6-1-2003 at 02:03

that depends on the size of the charger. I have seen everthing from 12V/4A to 24V/50A for forklift batteries (24V, 400 Ah) . The later cost considerably more than 75 US$.

I m using a 18V-22V.75A transformer with 4 100A recitifying diodes+ a Variac .
And big cooling on the diodes..



Organikum - 20-1-2003 at 08:36

without a bench of capacitors it is no DC source at all. More some pulsed power supply.

Many reactions ask for a stable V or I as I remember and without the possibility for fixed I it is of reduced usability anyhow.
A solid state relay might be cheaper and much easier to control.

Just my 2 cents

lucifer - 20-1-2003 at 15:34

Use a big capacitor a coil and again a big capacitor and you will have a good dc current.

A computer supplies will still be the cheapest solution I think and if you need more than 5 V you can put two of them in series.

PC power supplies

Polverone - 20-1-2003 at 16:52

If you go for PC power supplies, make sure you get an older AT one and not one of the newer ATX models. I tried to modify an ATX supply for electrolysis with my meager knowledge of electronics a few weeks ago, and it was not successful. One problem is that you need to connect two pins to actually turn on the supply; just plugging it in or flipping on the power switch isn't sufficient. A more serious problem is that, according to a number of web pages that I consulted, the power supplies can easily be damaged or destroyed if they are run without a proper load attached to them.

I read one person's account who had gone through 6 PC power supplies in just a few months with electrolysis experiments. That is far too high a failure rate. I would use a battery charger or even invest in a nice professional, programmable DC power supply if you are going to do a lot of experimentation with it.

Maybe somebody who is handier with electronics than me could make the PC power supplies work well, but I would just scout garage sales and auctions until you find a power supply that can do the job.

Organikum - 22-1-2003 at 00:54

Putting two power supplies in series does not work. Pfft makes the fuse. If you are lucky. If you are unlucky you smell it.

Polverone: You have to shorten the green next to the blacks in the middle for a moment and you have to have a load on the main voltages aka 3,3V, 5V and 12V. A lightbulb is sufficient. Damage can occure when you have full load at 3,3V and nuthing else. Lightbulbs on 12V and 5V solve this problem.
But a battery loader with the capacitor - coil - capacitor attached as luzifer proposed will do better.
As parts are free or cheap just try a few ways. And tell.
Thats the real fun.


lucifer - 22-1-2003 at 14:28

Putting two power supplies in series will work, but only if the common (-) is not connected to the ground, or else one output will be shorted, but I guess you already new that.
If the ground is removed from the supplie that is in series and the casings of the supplies don’t touch each other it should also work.

you are right lucifer

Organikum - 8-2-2003 at 11:30

But it depends on some more factors. Having two identical power supplies and following the ways you described ok. If you don´t have two identical ones I wouldn´t advice to do so except you know enough on electrics. Somebody having this knowledge wouldn´t have to ask so......
And there is still the possibility of getting some serious oscillations, but it maybe worth a try as the parts are free.

The main trick is a large bench of big capacitors - this should enable you to (ab)use nearly everything. A truck or diesel automobil battery might be the the best choice for most applications if not of semi-industrial dimensions.

raistlin - 14-2-2003 at 16:18

The main problem with using a large number of capacitors in series is that they have a tendency to explode after a short period of time. I found this out in electronics after trying to connect 4 high uF capacitors in series. Ten minutes later there was a loud explosion, and part of the tabletop was on the floor....:D

Did I say series?

Organikum - 19-2-2003 at 01:42

No, puhhh
But I was sloppy in my last post, sorry. The big capacitor bench was thought for use with any transforming device which gives AC on the outlet,
the automobil battery was a own concept for a power source for electrochemistry - delivers best DC with lots of ampere. Two batteries with a loader should satisfy also higher demands.
Advantage is that often batteries and loader are available and costs zero.

Exploding capacitors are very valuable. I learned a lot on electrics after the first big one went the way of overload. The permanent high tone in my right ear for several weeks was nasty so.....


Dc power supply

Taz - 13-3-2005 at 07:51

I was looking at this on ebay

it is 0-100v but the problem in it is only
.75amp I am worried that while it may supply a steady volts it won't push enough amps
This would be my first attempt at electrolysis
One other thing I wanted to ask is if I understand this right does it take 10 ma to make 1 amp so 100ma would = 10amp???????

[Edited on 13-3-2005 by Taz]

BromicAcid - 13-3-2005 at 09:24

.75 A ... ?

I would not buy a power supply like that to do electrolysis as the actual productivity of the procedure is directly proportional to the amps. In most cases for me 15 A is not enough. Hence my recent aquisition of a PC power supply with 2 fans and a capacity of 28A @ 5 V.


Taz - 13-3-2005 at 09:30


Al Koholic - 13-3-2005 at 11:21

You can solve the load problem with PSU's with either a light bulb or, if you are like me and are into computer parts, one of your old 2 gig HD's sitting around collecting dust.

I've found that a suitable PSU for electrolysis is easily realized with a variac, transformer, full wave rectifier, and a nice 30,000uF cap. Depending on the resistance of my cell it usually is capable of 3-4 amps at low voltages...

hodges - 13-3-2005 at 14:05

Don't know if you still can or not but you used to be able to get a 12 volt 3 amp regulated and fused DC power supply from Radio Shack for around $50.

Well its a little costly But what about this for a power supply

Taz - 14-3-2005 at 18:42

rift valley - 14-3-2005 at 19:50

That looks like a nice power supply, a little too much for me though. I am also interested in starting experiments involving electroloysis of H2O . From one of my other projects I have a 120 volt power supply that is 5-15 amp variable, is this power supply too big to do small (or any) experiments with electroloysis of H2O? A while back I bought a car battery charger from walmart for 25 bucks to do electroloysis of NaOH When I tried to use it as a power source the "Battery is dead" light lit up and it would not send any voltage, so I returned it. Is there something I was missing?


P.S. Would labx be a good place to find a cheap power supply? I was browsing their inventory and it seems like it is difficult to locate a product unless you have a specific model number in mind, it seems like a very confusing website to me.

P.P.S. Radioshack has a battery charger that does 13.8 volts at up to 25 amps of current for 100 bucks I'll probably buy this right after i fiish my electroloysis cell (I orded strips of stainless steel today for it) anyway heres the like for the charger I was talking about
[Edited on 15-3-2005 by rift valley]

[Edited on 16-3-2005 by rift valley]

The_Davster - 11-7-2005 at 13:16

Dumb question:
How do you know if your PC power supply is AT or ATX?...I just removed one from an old computer of mine. I have read every word on the case and no AT or ATX anywhere to be found.

Craig - 11-7-2005 at 13:58

How do you know if your PC power supply is AT or ATX?...I just removed one from an old computer of mine. I have read every word on the case and no AT or ATX anywhere to be found.

ATX power supplies have an ATX connector. This site has the pinout of an ATX connector:

I assume AT power supplies don't have an ATX connector.

IrC - 11-7-2005 at 14:15

Look at the plug that powers the motherboard. If it is two inline plugs that both go to a single socket (twice the length of a single plug) it is AT. If it is a dual row rectangular single plug, it is ATX. Just load the 5 volt side with an amp or so and the supply will run fine. This is how many of the newer Galaxy base radios use a computer supply to replace the old transformer supply, they use a 10 watt 5 ohm resistor that gives the 5 volt side a load of one amp. You can then use either the 12 or 5 volt output for your experiments, staying within the ratings. If you load it correctly the power good line can be ignored, but I have found a few supplies that will not start without the power good signal (they also failed after a while). Some supplies have both connectors. It really does not matter as either supply can be made to work within their ratings.

12AX7 - 11-7-2005 at 17:15

Last one I was using, I took all the red wires and all the black wires on the mobo connector and joined them together for each electrode respectively. Better than just using just one wire at 30A! :o


The_Davster - 11-7-2005 at 19:26

Well my power supply has a dual row single plug going into the motherboard(indicates ATX by IrC's definition), however each pin only has one wire going to it(indicates AT by Craig's definition). which is it?

IrC - 11-7-2005 at 21:00

Only one wire to each pin. Not sure where that applies anywhere, but there is no reason only one wire would have to go to any single pin connector within a socket. However, none of this makes any difference to the question, what you have is an ATX supply, which is also the socket in craig's link, a correct view of an ATX plug. I would post a pic of an AT connector but I still can't get the ftp into the site to work for me, and I don't know of a link to a pic of an AT plug out there. They are getting so outdated nobody talks about them anymore I guess. But they are a cheap source of supplies, seeing as how so many were made and now are unused.

As to the almost $190 power supply on ebay that is 20 amps, not a bad supply but if you ever did any real plating it really is not big enough, especially for the bucks they want. H&Y Electronics sells a 52 amp power supply with voltage and current meters, all you would have to do is go in and slightly redesign it to allow full adjustability rather than a small range around 13.8 volts. For 52 amps under 200 dollars this would be a very worthwhile project, and finding a schematic to the supply would not be hard. It may even come with one. If not and you choose to go that way I can dig one up and even show what changes to make to convert it to a fully adjustable zero to 25 volt source of 52 amps.

Found a picture of an AT plug(s):

[Edited on 12-7-2005 by IrC]

The_Davster - 11-7-2005 at 22:04

Eh...I have no idea what I am talking about...In Craig's link:

I assumed the double circles in pin 11 meant 2 wires, and was confused when mine had only a single wire...oops

[Edited on 12-7-2005 by rogue chemist]

IrC - 11-7-2005 at 23:01

I see what you mean. Wouldn't worry about it though, as long as you have the rectangular connector like the pic you have an ATX supply. I don't see why you couldn't use one for an experimental supply though, I do it all the time, and like 12AX7 mentioned I parallel all the reds (5 volt) together, and the yellows (12 volt), and blacks (ground), to have the maximum output from both the 12 and 5 volt sources. Of course this is for limited voltage selections. I still firmly believe in a brute force transformer supply when I want lots of amps with fully adjustable voltage. If they were not so expensive and hard to find two variacs (one for current, one for voltage control), a big transformer, and some big bridges combined with about a one farad capacitor (pay attention to maximum possible voltage output to pick the voltage rating needed, or put two in series for a half a farad) make a very good all around supply.

A MOT transformer rewound for low voltage high current (with about a 30 uF 300 volt AC capacitor across the primary) makes a very good transformer if you wind it correctly. Sola power supplies also work well if you need good voltage regulation, if you can find one surplus.

Pinlayouts and schems for connectors

Lambda - 12-7-2005 at 12:22

For pinlayouts and schems for connectors just go here, and you will be able to sleep at night:

Download the offline-manual here:

ATX power supply pinlayout and schems

Lambda - 12-7-2005 at 15:01

And here you will find your ATX connector's pinlayout and schems (on this same website):

Archimede - 18-7-2005 at 14:42

If I remember this right, the power supply used in computers are called 'switching'.
If you look at that with an oscilloscope you will see a serie of constant pulses like a square wave. Thats how they can get soo much current output with the components used. Its probably ok with most of the work you are trying to do , as many people use it.
The car battery its interesting eccept that the voltage is probably too high and you cannot limit the current.
Of course more expensive (this one is selling now for around $100) seems to me the best equipment to use.

It let you set the Volt output (0-40V) and the max current output (up to 30A) so you can limit it if you need to.
Something like this would also let you recharge your car battery (set Volts at 13.8 and Current to around 5A) or any rechargable battery pack just setting the right Voltage and Current.

[Edited on 18-7-2005 by Archimede]

Twospoons - 18-7-2005 at 23:21

Oh dear...

A 'switching' supply is one that uses internal switching to regulate the output. It does not mean the output is a square wave! The output is flat DC.

Switching supply topologies are used because they are (or can be :D ) more efficient than linear supplies. So they don't get so hot.

12AX7 - 19-7-2005 at 08:21

Originally posted by Twospoons
Switching supply topologies are used because they are (or can be :D ) more efficient than linear supplies. So they don't get so hot.

They are also much lighter, smaller, and cheaper once it's been designed and a few ten thousand units have been produced. :)


Quibbler - 19-7-2005 at 08:29

Why do you guys need so much current. What you will get is all kind of reactions taking place as one kind of ion is discharged near the electrode another will take over. It will depend on how much convection and external mixing is taking place. Or maybe you're trying to boil the electrolyte.
I use a 317 to keep the current constant, and I would never use more than 100 mA if I wanted to get a clean reaction.

12AX7 - 19-7-2005 at 10:27

I guess you've never wanted to oxidize more than ten grams of a salt per week, or had to plate something larger than a ring.

100 amperes on a good two square feet of graphite, lead dioxide or platinum will go through a lot of chloride > chlorate nice and fast.

Oh, and it's not the amps that melt the cell, it's the volts.


IrC - 19-7-2005 at 10:56

"Oh, and it's not the amps that melt the cell, it's the volts."

Come on guys, you are bumming me out here, as I know you are smarter than that. It is the power, a product of volts times amps called watts, that produces cell melting heat. Not just volts, or amps alone. Heat is molecular or atomic vibration, needing energy in ergs per second (or joules/sec or watts/sec if you prefer) per unit area.

So it is the volts times amps per unit area that melts the cell.

12AX7 - 19-7-2005 at 11:26

I mean okay you can physically boil a cell with just a few watts, if you wrap it in thermal blankets. But a cell with the proper electrode surface area and voltage for the reaction taking place inside is going to be a certain size and that size is going to dissipate most of that power quite nicely. Yes, some reactions will need extra cooling or heating. In general, excess voltage (beyond what's needed for the reaction) still causes ohmic heating and unwanted side reactions, not to mention is a waste of power.

So yes, quite the wild generalization, but there are more (and more common) cell reactions in use that prove rather than disprove it, so it works!

(Salt is a good exception, typically being ran in the 10-20V range, well above the 4.1V required to split the ions. This is simply for convienience, using the excess 6-16V to heat the electrolyte and keep it molten, heat that could otherwise be supplied by fire or whatever.)


Quibbler - 19-7-2005 at 12:13

Ok fair enough if you're into industrial production. I'm normally doing something more subtle ever tried Hittoff's method if you put in more than 10mA you're wasting your time.

The_Davster - 28-7-2005 at 09:32

Done building the powersupply.:D

powersupply.JPG - 199kB

jimwig - 29-7-2005 at 08:25

where would be without hot glue?????

if the outputs (secondaries) of the power supplies are floating (not grounded) you can indeed series or parallel them.

Use MOTs all the time in this configuration to gain either voltage and or current.

also if you get a defunct large transformer like a old welder or large battery xformer.

use it on a large variac (variable transformer) through a full wave bridge with or without filtration.

got a welder that delivers up to 225 amps at what ever voltage i desire by varitying the input accordingly.

duty cycle is dependent on percentage usage of therating of the unit

10amps @ 12vdc goes much better than

100amps @ 30vdc.

[Edited on 29-7-2005 by jimwig]

[Edited on 29-7-2005 by jimwig]

Archimede - 31-7-2005 at 08:14

Originally posted by 12AX7
Originally posted by Twospoons
Switching supply topologies are used because they are (or can be :D ) more efficient than linear supplies. So they don't get so hot.

They are also much lighter, smaller, and cheaper once it's been designed and a few ten thousand units have been produced. :)


Yea, but they are also designed to work between a range of loads. I don't know about the ATX power supply , but some old ones used on computers wouldnt even turn on without a certain load connected on it.

hinz - 31-7-2005 at 14:19

Has anyone thought about using an old arc welder as high power supply. They normally give 100A or much more at something around 28V. They are really hard to kill, and they won't overheat.
The ionly problem is that they give 28V.
Mayby it's possible to connect their primary side (120V/240V) in serie and connect their secondary side parallel, so it should be possible to get 14V with two transformers, 9V with 3 transformers and so on.

12AX7 - 31-7-2005 at 21:35

MIG welder maybe, but stick or TIG has a constant-current output so the voltage varies from say 70V open (zero amperes) to 20V arc (at say 100A). One solution may be to max out the current control, which would work reasonably on my stick welder, which uses a variable magnetic shunt and tops out at 225A. Using it at something less, say 50-100A should have acceptable regulation.

Reducing the primary voltage would be an ideal solution, but then you need a massive buck transformer or variac. Droooool, 240V variac. :D


Twospoons - 31-7-2005 at 22:17

Originally posted by jimwig
where would be without hot glue?????

A lot better off! The problem with hot glue is
1/ if it gets hot - it lets go.
2/ usually it lets go anyway because most hot glue is useless for anything other than cardboard.

Power supply and electronics in general

Lambda - 13-8-2005 at 18:13

13.8 V / 15 A from a PC power supply:

13.8V 20A power supply:

200W ATX PC power supply (please check out the links on this page for power supply modification):

Power supplies and control schematics:

And for more schematics on the same website in various fields of electronics:

Alarms and security related schematics
Audio power amplifier schematics
Audio preamp circuits
Automative, car and motorcycle schematics
Data acquisition and data logging schematics
Filter schematics
Games and fun stuff (schematics)
Infrared based schematics
Laser related power supplies and data transmission
LED related schematics
Lighting schematics
Medical and health related schematics
Microcontroller based schematics
Misc audio (also see Music, Amplifiers, Preamp)
Miscellaneous schematics
Model and remote control schematics
Motor and general control schematics
Music related schematics (also see Audio)
PC related schematics
PDA interfaces and related schematics
Power supplies and control schematics
Radio-frequency schematics (also see Transmitters)
Solar-powered schematics
Telephone and intercom related schematics
Test equipment schematics
Timing, oscillators etc. (schematics)
Transmitter schematics (also see RF)

neutrino - 22-11-2005 at 17:25

Just to make sure I've got this right: to get 5V, connect all the +5V wires for the positive terminal and all the ground wires for the negative one, correct?

For an AT, this do I use red or orange for +5V? I'm guessing red because there are a lot of red wires and only one orange.

Is there any use for the -5V wires? I notice they have a very low amperage rating on my ancient AT, so I'm guessing that the answer is no.

[Edited on 23-11-2005 by neutrino]

12AX7 - 22-11-2005 at 22:06

Leave the other wires alone, +5 and +12 are the only ones with enough snuff behind them to be of use. -5 and -12 are almost always rated under an ampere, so unless you have an experiment you *know* will be safe and use less than rated current, leave 'em alone. ;)

If in doubt, you can open the case (mind the high voltage parts, they might carry a charge) and trace what wires go where. Usually, the red wires meet in a tight grove for +5V (something printed on the circuit board, or nearby such as a capacitor, should say something in that range), and likewise for the yellow +12V lines and a whole mess of black ground leads. Those are the most powerful so need the most wires.

Newer supplies may have a few 3.3V lines BTW. I forget what color that is.


neutrino - 23-11-2005 at 03:07

So red and black, then.

I have an older supply, the better part of a decade probably. None of that ATX junk. :D

One more little thing. When I got the unit, there was no power button to turn it on, just a bunch of wires where a switch once was. To get the thing working, I connected the wires labled 'L' with each other where the switch would have been, likewise with the 'N'. Sound good?

12AX7 - 27-11-2005 at 13:04

Careful with that. It'll probably work, but I'd check the circuit board or wiring to make sure you aren't either swapping wires (not too big a problem, but could cause trouble to a grounded power cord), or worse, shorting both sides! :o


neutrino - 27-11-2005 at 13:52

It seems to work fine the way I have it now. Don't worry, I was careful not to cross - and +.

The_Davster - 23-1-2006 at 15:38

Can two PC powersupplies be hooked up in parallel to get around 40A output on the 5V line? Or would bad things happen? And is there a limit for how many could be put into parallel?(I just got access to a large number of old computer parts :cool: ). I imagine at some point they would have to be plugged into separate wall outlets though...

[Edited on 23-1-2006 by rogue chemist]

IrC - 23-1-2006 at 16:12

It should work fine, especially if they are all coming on at the same time. Since you have many risk a few and try it. As long as your AC input is fused reasonably whats the worry?

bio2 - 23-1-2006 at 22:04

When paralleling transformers the impedances should closely match and preferably they should be identical make and model for the best results.

This mainly applies to the AC secondaries in parallel and there are specific techniques for doing this after the rectifiers to prevent circulating currents.

Try to use those computer power supplies that are matched
(model, brand) and you will have much better results without losing a lot of power due to mismatched rectifiers etc. One technique is to place a power resistor after the rectifier on each unit say.22ohms or so. This provides some
compensation and is very simple.

The_Davster - 23-1-2006 at 22:12

Why is it important that they both come on at the same time? Could current flowing into the PSU through the output cause 'bad things'(I seem to be overusing that phrase recently:P) The supplies will be on the main lines so the breaker for that room will be the regulator I guess(beats fuses, I went through way too many with my furnace). I have noticed with computer PSUs that inside they do have a fuse, although the filament looks pretty beefy, I think that only prevents major shorts as the PSU turns off when more amps are pulled from a line than what the line is rated for.

As I am using this for electrolysis I will not be doing this but for my own curiosity, can the PSUs outputs be hooked up in series(to get a higher voltage) or would that cause a short? I am wary of connecting outputs in series after an incident I had when I was little in which I thought if I conected the 'positive' and 'negative' parts of 2 separate wall outlets I could get 240V(I wanted to zap a pickle it higher voltage:D)...need less to say the wire I used(it was stranded Cu-wire exploded in a nice shower of sparks at the end when I stuck it in the socket...:P:o

Hmm...seeing as this is turning into some sort of super power supply I might as well add in an option to impress some AC on the DC, for those sensitive electrochemical oxidations:D

EDIT: Thanks Bio2, but I am afraid my source is pretty much the odds and ends, so matching brands, let alone models is not likely. I also understood very little of the terms in your post so I should likely do some electronics reading before I do this...
But a quick stupid question...what does a rectifier of the type found in the supply look like? I did a google image search and I do not see many similarities between the guts of my supply and the image result pictures.

[Edited on 24-1-2006 by rogue chemist]

woelen - 23-1-2006 at 23:44

Putting voltage sources in parallel is not a really wise thing to do. If the power supplies have a small difference in voltage (e.g 4.8 volts and the other one 5.0 volts), then a current flows, which can become really high (the resistance probably only is in the order of 0.05 Ohm or even less).

Another effect is that when one of the supplies powers on, and the other doesn't, then huge currents may flow into the not-yet powered device. It depends on the electronics what happens precisely, but I bet that these things usually are not fun :D.

Making a series connection of power supplies is safe, provided that the power outputs are "floating", relative to earth. IIRC, a PC power supply indeed has floating output, its GND (0 Volt) is not connected to earth.

I made myself a fool-proof powersupply from a PC power supply. I use the concept of current control for electrolysis experiments. That gives much more consitent results (I tried with electrolysis of NaBr, making KBrO3, by precipitating the resulting bromate with KCl). In order to keep things chaep, I use fixed resistor networks, but it works like a charm. I also put all resistors in small glass vials, reducing the chance of accidental short circuits. A complete description is on my website:

It is easy to build. Only the 12V output is needed, because I use current control. The PSU allows at least 10 A of current, but the resistor network limits the available current. Of course, with other resistors, you can obtain higher currents.

Anyway, at my situation, not the PSU or resistors are limiting the current, but my anodes. If I use more than 1 A of current, then my anodes pulverize very fast.

bio2 - 24-1-2006 at 01:43

A rectifier is a diode and looks like the package it's in.
Usually a bridge with 4diodes in one plastic DIP. It's the
first thing after the transformer secondary and may be multiple barrel shaped type or TO220 in a switcher. Just look for the "D" on the board and then measure the forward voltage and try to match if you have others.

Very nice article and site Woelen.

You should consider using an LM317K ($3)as a constant current regulator. Requires one 25 ohm pot and a 8ohm fixed for max current so this gives 50mA to 1500mA.

To produce higher currents the 317 can drive a large power transistor. There are many designs in the data sheet at National Semiconductor and the LM317 is very easy to find and apply.

Current limiting resistors on the output terminals

Lambda - 24-1-2006 at 09:29

Rogue chemist, by putting a ~0.1 Ohm resistor of ~20 watts (or a piece of maybe ~2 mm diameter coiled up Nichrome or Steel wire) in series with all PSU 5 Volt outputs (per PSU), you should be able to even out the differences sufficiently before connecting them parallel. I don't think the one PSU will short circuit into the others before they are turned on, for the rectifier is on the output stage with only a small capacitor, due to the high switching frequencies used. How the loop feedback sensing electronics will respond to the already turned on output, I don't know.

Just make shore that these units stay closed, so that any exploding capacitor, diode etc. may not hit you in the eye. This shit flies as fast as bullets when high currents are involved.

[Edited on 24-1-2006 by Lambda]

tumadre - 24-1-2006 at 11:40

the following is how I connect PSUs in parallel.

Get a diode and a resistor, sized for the full current and voltage drop needed to balence each psu's output voltage at full current, respectivly.
Connect the anode to the 5 volt output, the cathode to the resistor and connect all the resistors together, and the grounds are connected.
the result is that the PSUs can't "see" each other through the diodes
you can find six 30 ampere diodes in a 60 ampere alternator, find one at a junk yard for a few dollars.

Originally posted by Lambda

Just make shore that these units stay closed, so that any exploding capacitor, diode etc. may not hit you in the eye. This shit flies as fast as bullets when high currents are involved.

[Edited on 24-1-2006 by Lambda]

can't agree more, a friend of mine had a tantalem capacitor shoot though his hand

woelen - 24-1-2006 at 12:17

This trick may work, but the power delivered by all PSU's will be very unevenly spread, and you loose approximately 0.6 volts plus the drop accross the resistor. So, your voltage at the output will be 4.2 .. 4.3 volts, not more.

What this setup indeed does, however, is making power up safe. The PSU, which powers up first does not feel a big power surge from the other PSU's, the diodes of the other PSU's simply block.

But then I have a question, why do you need such a high current?

The_Davster - 24-1-2006 at 16:07

Partly for the "I have a 50A power supply" bragging rights:P, and partly because in my antimony separation from solder experiments I have pulled more amps than the supply could handle on the 12V line(although I have changed designs since then). I guess its mainly a 'I have the resources so why not' type thing. And I perhaps wanted to draw an arc to make a bit of carbide, but after reading the above posts the number of supplies in series and parallel needed for that would be downright mad!:P So I do not think I will be doing that right away...

Tumadre, you mentioned the diodes but did not say where they should go, but I am guessing in series with the output?.

So after reading the above I will start with only 2 in parallel, use resistors of the appropriate value to bring all outputs to the same voltage, the diodes once I find out where they go, and put a switch on one of the output lines of one PSU so I have the option to only have one supply worth of current. Sounds good? I will vary the amps with number of cells, resistors, electrode distance and such.

Luckily I have already built one powersupply from a single computer power supply in the past, so I am not completly clueless here:). It has served me well in the past.

Now any problems with impressing some AC in parallel on the DC output of a single power supply? Of course at the same voltage. And Does "impress" really mean to just have AC and DC going through the same wires at the same time?

[Edited on 25-1-2006 by rogue chemist]

IrC - 24-1-2006 at 17:37

"Why is it important that they both come on at the same time?"

Because the schematic of the supplies is unknown so who knows if one slightly higher source will sink into another source. But with a few tenths of a volt difference who cares, why waste all the hassle worrying about isolation? You have a bunch for free see if they go or blow with simple parallel connections. No loss, and likely no flow between units assuming all the output sees is the internal diodes and filter capacitors anyway. A small few tenths of a volt difference between them isn't dick if there is no way for power to flow backwards into one, all you will see is one or more source more current than others. So what. Part of mad science is the theory of go or blow isn't it? What have you got to lose and look how many parts and wiring you will save.

woelen - 25-1-2006 at 00:05

I do not agree with you IrC. What is the cost of a diode (< $0.50), and what is the cost of a complete PSU? I would not try it, unless you really don't care about the devices. Also, what would the result be on the long run? It may turn out OK one time, two times, etc. and then suddenly the whole thing blows up at the X-th time you switch on the stuff. I hate unreliable things, especially if there also is a risk of dangerous situations.

The diodes indeed must be in series with the output.

For each PSU, connect the anode of the diode to the +5 volt. Connect the resistor to the cathode of the diode. So, for N PSU's you need N diodes and N resistors.

Once, you have all PSU's with added diode and resistor, then connect all open ends of the resistors to each other, making a single large output.

[Edited on 25-1-06 by woelen]

IrC - 25-1-2006 at 09:32

I was looking at the fact that there already were diodes in the outputs and adding more merely reduces the output voltage available. If the voltage is more than high enough anyway then your added circuitry would failsafe the circuit. However so far these two I am trying are not showing any unusual warmth sitting here with no load, indicating indeed there is no appreciable current flow between them. If the possibility did exist then you are right the added circuitry would insure a problem free source. As I said before he got them for free so not much is lost trying a couple if his goal was to get as close to the 5 volts as possible. If I only needed 4 volts then probably I would add diodes for reliability also, but I am not sure the resistors are needed since the added diodes each have their own internal impedance in the "on" state. To each their own way.

12AX7 - 25-1-2006 at 10:30

If you have switchmode power supplies, they both actively regulate the 5V output (generally) so you have two feedback loops fighting. If you set them equal (there should be a potentiometer on the board inside with a dab of silicone on it), the current should share equally, to whatever extent the output impedance allows (a 0.2V droop across 30A range is all of 6.7 miliohm; variation in this figure will alter current distribution according to the Thevenin source impedances).

It is true that both supplies have diodes internally, so backflow isn't an issue. However, if a supply's output voltage is raised, the control loop chokes off the power flow, putting the entire load on the higher power supply. This is the concern.

Same applies to linear power supplies.

Paralleling anything constant-voltage (or equivalently, connecting constant currents in series) is a bad idea, and generally you need some external stimulus to keep things in order: master-slave configuration, master control loop, etc.


The_Davster - 25-1-2006 at 20:27

So I was at the place I get the PSU's, and I ended up with a 350W one, 35A on the +5V line:D. Looks like I may not have to risk the capacitors blasting through my hand:(. I think I'll just stick with a single PSU as I think 35A will work for most things. I really can't imagine ever needing more. I did learn some cool stuff from the replies above.

[Edited on 26-1-2006 by rogue chemist]

Power "good" sensor and load

Lambda - 25-1-2006 at 21:50

Rogue Chemist, just one final remark about these PSU's. If there is a sensing wire on the output connector and this is not hooked up, then the PSU will not stay on, for it will then encounter an error. This is a feedback sensor, and is used to check if the computer were they are to be used in has powered up properly.

If present, this wire must be connected in order for the PSU to stay on (power good sensor). Also, some PSU will not run without a load.

Please check this link out for more details:

How to Convert a Computer ATX Power Supply to a Lab Power Supply:

The_Davster - 25-1-2006 at 21:57

Thanks Lambda:)

That link is actually quite similar to the page I followed when making my first power supply from a computer power supply:
The only difference between the two pages seems to be the currents involved, and as such, the use of a power resistor instead of a sandbar type resistor. My current power supply has just a sandbar resistor pulling an amp, and it gets hot despite being next to fan. For a future power supply I will be sure to use a power resistor as those seem to be better intended for high current uses.

woelen - 26-1-2006 at 01:41

Indeed, I also had to connect the sense line (brown) to the orange CPU voltage line. Not before the voltage across the orange line is within tolerance, the total PSU powers up. In my page I described this, as connecting the brown wire to an orange wire.

Besides that, you also need to ground the green wire. That is the soft on-switch. Failing to ground the green wire also prevents the PSU from powering up properly.

My PSU only needs a few mA of load, but I have heard of other people, that their PSU requires much more load. But these PSU's were older. The modern ones apparently do not need much load.

IrC - 27-1-2006 at 12:35

My solution to my mad science power supply for electrochem experiments was to take a Pyramid Gold series (13.8V/30A) power supply and gut it for everything except the transformer and bridges. Then I designed my own supply as I wanted zero to 45 volts at a maximum of 30 amps. To eliminate dissipation problems I altered the transformer-bridge configuration to allow only 25 volts on low and 50 volts on high. This way in the low range the pass transistors only have 25 volts on them and on high 50. Of course you have to remember to switch since on high the supply will still adjust from zero to 50 volts. I also mounted a thermistor on the bridge heat sink, transformer, and one on each pass transistor sink assembly (4 total). This signal feeds a circuit which will remove all power if overheat is detected, as well as an overvoltage circuit intended to kill the juice and sound a buzzer and light in the case of a shorted pass transistor. The design is my own and works so well it is used for many things around my lab.

I of course used the existing voltage and current meters and altered the volt meter circuit to be 0-25 or X2 which would be of course 0-50. The overvoltage is set to trip at 47 volts, and 0-45 is all I adjust for meaning no matter what in the event of pass transistor failure the supply is guaranteed to stop doing it's thing if it sees the full 50 volts no matter where it is being run at the time. Later I am going to add a fully adjustable constant current output independant of the normal output but the case is really crammed full now so I am having a hard time figuring out where to mount another circuit board in there. The 4 pass transistors are 16 amp 180 volt 2N3773's and the predrivers are the high voltage 2N3055HV's (16A/150V).

I forgot to mention but the low value resistors in the ground end of the supply provide a kill voltage to the power supply in the case of extreme high currents, a simple short circuit protection circuit. I had 5 to begin with but removed one to allow a lower current before tripping. The bleeder on the 50 volt rail is overkill at 60 watts but they were in the junk box so why not. In fact the whole supply was made from my junk box.

[Edited on 27-1-2006 by IrC]

Lambda - 27-1-2006 at 13:28

IrC, you can also sense the output voltage with an OpAmp, which in turn drives a power relay via a transistor so that the Transformer secondary can be switched. A hysteresis has to be implemented into the OpAmp design to kill threshold relay vibrating. A mistake I often see in cheap OTC junk designs. More elegant, and a design I am working on now, is to regulate the transformer primary via Triac's or Thyristor's, to chop the input mains supply voltage sinus, which is regulated via the difference between the Transformer output voltage and regulated output voltage.

Just one tip, although depending on the manufacturer of the 2N3055's, the power dissipation may vary quite a bit. Generally speaking, the 2N3055 should not dissipate more than about 50 Watts, due to the ease of cooling you will then encounter. Cooling elements are often much to small or not sufficiently ventilated. In you tight-fit-enclosure, forced cooling will be very beneficial and save you a lot of hassle afterwards. Personally, I think there is nothing worse than the shutdown protection kicking in during my experiments because not that the current or power output was to high, but because the power supply just overheated. Imagine sitting there with molten Potassium or Sodium Chloride or Hydroxide during the synthesis of Potassium or Sodium and the juice dropping away !.

IrC - 27-1-2006 at 14:05

"IrC, you can also sense the output voltage with an OpAmp, which in turn drives a power relay via a transistor so that the Transformer secondary can be switched. A hysteresis has to be implemented into the OpAmp design to kill threshold relay vibrating. A mistake I often see in cheap OTC junk designs."

This is exactly what the 741 IS doing for the overvolt trip, monitoring the rail. However if you study the circuit I implemented a latching circuit so hysteresis does not apply since the latching takes less than a few milliseconds. In this way the problem can be looked into before the reset button is pushed to unlatch the protection circuit. I designed this thing while watching the movie Serenity late at night with only one eye open. What amazes me is it works so well. I would like to see some of your AC control circuitry sometime as I am always into learning something new having at one time been a design engineer in a R&D lab involving electronics and chemistry. Another story for another day. My main goal here was to build a brick shithouse of a supply that would go up to 45 volts and down all the way to zero (not a few tenths above as sometimes I may be playing around with some experiment where I only want a few tenths of a volt), provide generous amps and regulate within a tenth of a volt.

Also, I wanted it done that weekend as I never have time for projects like this anymore. I will say I loaded it to 300 watts at 45 volts and switching the load in and out the output rail only varied by 0.05 volt, not even a tenth of a volt drop!

The 2N3055HV's are heavy duty versions of the 2N3055 and merely drive the pass transistors, which are 3773's. I like the primary control idea but I wanted this up and running soon and built soley from what was in my junkbox without waiting to order anything, and to implement your ideas would have taken me somewhat more time to design. Maybe I will sooner or later play around with your ideas but I need to add the proper thyristors or similar components to my stores, believe it or not I just ordered some last night for another project, and I always order extra to play so I think I will try your ideas but it would be nice if you had a schematic you could post to give us a starting point on the general idea. The fans (2) are powered through the 100 ohm 10 watt resistor right from the 24 volt supply, they are 12 volt 120 mm computer types.

Here is a pic of it, I added red and green high output LED's to the meters so it would look cool when runnning.

[Edited on 27-1-2006 by IrC]

Lambda - 27-1-2006 at 18:18

IrC, please let me explain what I meant by implementing hysteresis into the OpAmp output voltage sensing design in order to drive a power Relay to minimize power consumption and thus heat dissipation of the Power Supply:

I will start at basics, although you already know this, it may be of interest for those who are not so familiar with these "old fashioned" analog Power Supply designs (I find them more reliable and robust than switching Power Supplies, "Murphy" may also have a word to say about this).

Lets start by assuming you want a Power Supply that can give 50 volts at 10 amps regulated output, and have a Transformer with two 25 volt secondary output leads that are more than sufficient to give the required output current. During a Metal Plating job to Chrome Plate a few nuts and bolts for use on your favorite Harley Davidson Motorcycle, only 5 volts is used with a current of 10 amps. What dose this then mean for the power consumption and heat dissipation of the Power Supply ?.

1 - The power (Watts = P) that the Chrome Plating bath requires will be P = I * V so that 10 amps * 5 volts = 50 Watts.
2 - The power that the Power Supply will consume during this operation will be roughly about P (more than) = I * V (more than, because of the Diode Bridge-Capacitor network) so that 10 amps * >50 volts = >500 Watts total power consumption to the output.

The efficiency will then be: 50 watts (output)/>500 watts (total) * 100% = <10% so that >500 watts (total) - 50 watts (output) = >450 Watts of power is wasted and has to be dissipated in the form of heat through the output Transistor regulator network. This means:

1 - A waste of precious energy.
2 - This heat will have to be dissipated, and the cooling plates will have to be made bigger (or made more efficient by means of forced cooling) and the use of more parallel output Transistors that can conduct this heat so that it may flow off. The design becomes more expensive, less efficient and a lot of heat dose not benefit long-life span of components.

How to overcome this problem:

I started of by mentioning that we have a Transformer with two secondary output leads of 25 volts, so if we only need 5 volts and 10 amps for the Chrome Plating procedure, then using only one 25 volt lead will mean: 10 amps * >25 volts = >250 watts power output. This puts the efficiency of the Power Supply at: 50 watts (output)/>250 watts (total) * 100% = <20% so that 250 watts (total) - 50 watts (output) = >200 watts of power is wasted. Less heat problems will be encountered here, the Power Supply efficiency is higher and thus less energy wasted. So by putting the secondary Transformer leads in series, one can chose between 25 volts and 50 volts if an output voltage of above 25 volts is required. This can be done in two ways: Manually or automatic.

The automatic way, the way of the lazy bum and the wise man:

The Potmeter that is used to regulate the output voltage has in this case two functions. First of all, the output of the moving slide is fead to the regulator electronics to function as reference voltage, the second function will be that to be used to switch the Transformer to the 50 volt conjunction. When this Potmeter has been turned to half of the slide, then an OpAmp senses this so that a Power Relay may switch the Transformer outputs to 50 volts. At a certain point the Relay will switch on or off. But just suppose you are using an output voltage at this switching point. This will mean that the Power Relay may turn on and off due to slight fluctuations (boarder line effect). By introducing a hysteresis in the sensing OpAmp, so that once it switches the Power Relay on (at about 25 volts output) then it will only turn the Power Relay off when an output of maybe 22 volts is required. This is what I meant by introducing a hysteresis in the sensing OpAmp. I have often seen OTC junk Power Supplies with a vibrating Relay at this critical point. The chances that a particular voltage is required that makes this possible may not be that big, but at the cost of a diode and a resistor of maybe $0.10 in total, what the hell !.

I have a 24 volt at 40 amps Transformer with no in between secondary junctions which I want to use in a Laboratory Power Supply. So I am aiming at primary regulation to minimize power consumption at high currents. I do however, have a Power Supply that I have built that makes use of this principle. I got the design from the Dutch electronics magazine Elektuur (written in Dutch). Elektuur is published in many countries, and the English version is Elektor. I will try to find this design for you IrC. The problem is however, that these magazines don't publish with symmetrical articles, that have only been translated. Maybe they do nowadays, for in the past I have found articles in the German version that I did not know of being in the Dutch version. When I have found the English version of this article, then it will be uploaded to the Forum.

As for my design, I have just started. Roque Chemist got me hooked onto this "drug" with his 40 amp dream Power Supply desire. After I had read 40 amps in his post, I wandered of into my storage room, picked up this massively heavy Philips transformer, and thought to myself, I also want a 40 amp lolly in my lab. When the design has been completed, and all the monkey quirks have been worked out of the trees, then I will make it available for all to use and abuse.

IrC - 27-1-2006 at 18:40

The dissipation problem is why I switch between 25 and 50 volts depending upon the desired range. I used to get Elector, they once paid me a for an article I wrote for them back in 1992. But I haven't written any for them since they vanished from US shelves. I really thought it was so much more advanced than the worthless build a led flasher promoting Popular and Radio Electronics magazines they puke out here. It used to be embarrasing to see that the US pubs were so far down the food chain in electronic sophistication as compared to the European magazines. Talking with the US representative of Elektor, I was told that Gersback pulled dirty tricks with the unions to keep Elector from being stocked on US shelves so soon it would only be available overseas. Paying for a yearly overseas subscription just got too expensive. Capitalist pigs at their best (Gernsback), and we here pay for it by not having good construction articles to read.

The article you mention may be Dutch but still you could post a schematic which is all I need to study anyway.

FPMAGEL - 28-1-2006 at 03:03

I'am more plamsa science, than elctro(the lines become blerd), but, two 12v batterys in seiers, == 24v
two 12v batterys(car batterys) in parrel == 120amps hours

"Indeed, I also had to connect the sense line (brown) to the orange CPU voltage line. Not before the voltage across the orange line is within tolerance"
just a question is that earth or negtive?

"This is exactly what the 741 IS doing for the overvolt trip, monitoring the rail. However if you study the circuit I implemented a latching circuit so hysteresis does not apply since the latching takes less than a few milliseconds"
Just so i'am on the right track, is ther a pice of horse shoe of steel in there.

[Edited on 28-1-2006 by FPMAGEL]

Lambda - 28-1-2006 at 05:53

IrC, apparently the article you requested is prior to 1995, because I could not find it on the "Elektor/Elektuur" web sites.

I do have the Photocopies, and maybe even the magazine in which this design has been described. They are all still in my storage boxes, the magazine will be the easiest to find if I have it, but to find it in my unsorted Photocopies (25.000 - 50.000 !!) may seem an impossible task at this moment.

Tomorrow (Sunday), I will go to the Library, and dig it up for you. When I find it, then I will download it from the Internet in the English (?) and Dutch language version.

Just one tip !!!, on eMule you can download this:

Everyday Practical electronics from ~ 1999 to 2006 (Complete) (<500 MB).
Elektor from ~ 1989 to 2006 (Complete) + Data Manuals & Compilations (designs, audio etc.) (<10 GB !!).

About the Power Supply, I may also go for PWM (Pulse Width Modulation). I am still in debate with myself about this one though.

IrC - 30-1-2006 at 18:01

Useful info:

Rosco Bodine - 2-2-2006 at 02:54

A question for the circuit gurus , please take a look
at the attached file and tell me if the modernized
and hexfet adapted " emitter follower " I have
drawn in the schematic will work as I am thinking it will .

On rare occasion I have been reading and have
encountered the term " electronic rheostat " used
in various ways for no specific circuit which of course
is never illustrated either . But what I have supposed
is that such a device would work similarly as do the
resistance wire rheostats , limiting current by their
IR heat loss , and voltage drop across the semiconductor
when it is operating in the active region .

Anyway , there are uses for these simpler circuits where the noisy waveforms and ripple generated by PWM and other switching type power supplies is unacceptable .

I have been thinking about possible ways of speed / power controlling permanent split capacitor asynchronous AC motors
such as 1/10 horsepower and less , using off the shelf standard motors , and not the more expensive inverter duty
motors as are required for PWM variable frequency drives .
The simple circuit I have been contemplating could work
well open loop and it could also very easily have the feedback loop closed with some modification and added components . Anyway it looks to me like the basic circuit should work , but I haven't tried it yet .

Take a look and tell me what you think .

Variable AC Voltage Regulator.jpg - 41kB

IrC - 2-2-2006 at 04:32

Likely the mosfet will blow when the gate goes negative with respect to the source.

Rosco Bodine - 2-2-2006 at 09:22

Originally posted by IrC
Likely the mosfet will blow when the gate goes negative with respect to the source.

I don't see how that would happen since the source has to follow the gate , the same as does the emitter have to follow the base in an NPN emitter follower . The source will
be at the same voltage as the gate which is Vref.
The idea is that a greatly current amplified Vref. appear on
the right hand side of the AC load , with the rest of the voltage above Vref . being dropped across the mosfet being
forced to act as a power resistor .

Hmmmm , please analyze and elaborate .

Please talk me through it by the half cycle ,
because I don't see it looking straight at it :D

Edit : see attachment for the mosfet modus operandi

[Edited on 2-2-2006 by Rosco Bodine]

Attachment: mosfet used as source follower.pdf (93kB)
This file has been downloaded 922 times

Rosco Bodine - 2-2-2006 at 16:40

Originally posted by IrC
Likely the mosfet will blow when the gate goes negative with respect to the source.

For a few hours I have been studying the proposed circuit
and I still don't see the problem . However should you prove to be correct because of something I'm not seeing in terms of direct or parasitic paths , I have been thinking about how to remedy the problem of the barbecued mosfet so that volunteer mosfet number two has a happier time of it .
What I am thinking is to self-bias the mosfet from a local
supply derived from its own DC rails across the power bridge rectifier , using an optocoupler to control the gate , and using
Vref. to drive the transmitter side of the optocoupler . The ultimate voltage would probably be 6 volts less across the AC load , but the mosfet would be bulletproofed . I am thinking a 400 volt rated mosfet minimum , 500 volts even better , for the same reason as the 370 volt capacitor rating if you have a PSC motor as the load , the peak to peak doubling of voltage that can happen because of the capacitor .

Edit : I have been retracing the paths using different colored highlighters for the two halves of the cycle , and now I see the problem . It wouldn't hurt the mosfet , but neither would the circuit provide fullwave regulation . When the bottom AC rail is negative going the mosfet can't function as a source follower but would be in saturation at any of the higher Vref. settings . The result would a baseline shift for the AC through the load with regulation only upwards from 50% and an asymmetrical output to the load . So ....back to the drawing board :D

I can do better and I will sketch this hopefully better idea
and post it here .

The alternative idea which probably has a better chance of working , involves a ganged pot directly between the AC rails with each separate Vref. from the wipers driving the gates of two separate N channel mosfets each having their drain connected to the respective AC rail and the source of each mosfet connected to opposite ends of the AC load . The mosfets operate sequentially as source followers during the half cycles when their drain is positive , with the return diode in the non-conducting mosfet completing the path through the load to the rail which is negative on that half cycle . The ganged pots are connected between the AC rails in the way that the wiper of each moves towards rail voltage of the respective mosfet whose gate is supplied increasing Vref. as the pot is turned clockwise . This should provide a balanced circuit with a symmetrical AC output and no baseline shift and using fewer parts than the other idea , as well as dividing the heat dissipation between two mosfets when the voltage is set for lower levels . I have never seen this circuit I have in mind , but it seems so simple
it would have to work .....I hope :D

[Edited on 3-2-2006 by Rosco Bodine]

12AX7 - 2-2-2006 at 18:38

At the very least, toss on a protection resistor and 12V zener diode, yeesh!


Rosco Bodine - 2-2-2006 at 19:23

See my edit above , I found the problem and
thought of a better circuit at about the same time :D

Twospoons - 2-2-2006 at 19:48

Having the gate more negative than the source is not a problem - its a trick often used to provide noise immunity to a mosfet thats supposed to be off.
The problem is the likelyhood of exceeding the gate/source breakdown voltage - which will fry your fet instantly.
The other issue is heat - you're running the fet linear, with potentially larfe current and large voltage across it. The poor thing could end up trying to dump 50W or more.

Rosco Bodine - 2-2-2006 at 20:56

Here's the simpler circuit that I am thinking
would be sure to work . Any criticisms or
comments or suggestions are welcome .

With regards to the breakdown voltage ,
I thought it was customary practice that
the manufacturers put zener protection
on the gates . I have seen the zener on
manufacturers data sheets , and sometimes
the return diode is a zener too , something
would have to be a higher rated zener than
the max allowable drain to source voltage .

One thing which would probably not hurt is
to parallel an external return diode across the
mosfet to split the heat dissipation from the
internal diode which occurs on each half cycle .
Yeah running any semiconductors in that
active region where they are in fact semiconducting
and acting as resistors is going to produce a whole
lot of heat , so big old heatsinks would be a good idea ,
even a fan cooled CPU heatsink would probably
be a good solution for this circuit when operating
at medium output voltages where the mosfets are
dumping as much heat as the motor .

[Edited on 3-2-2006 by Rosco Bodine]

Variable AC Voltage Control .jpg - 49kB

densest - 3-2-2006 at 15:53

I've got a couple of questions:
Wouldn't an autotransformer (Variac (tm)) work a lot better?
Is there a specific reason to take an analog-ish mosfet approach?

It will probably self-destruct releasing the magic smoke inside the transistors.
Sam Goldwasser and his friends have published the circuits inside of commercial
dimmers, motor controls, etc. See
Each application (motor control, etc.) mentioned in the thread probably
requires a different circuit design and approach if active devices are used.

What follows is a random rant about the subtleties of power circuits.
The FET manufacturers have extensive application notes and sample circuits to copy.
They're not simple - they can't be and still work reliably. Many of the
sample circuits include a sample PCB layout which can be copied as well.
This saves enormous grief.

If the intent is to make a low electrical noise attenuator for line power,
a transformer or autotransformer is the best power/cost for an amateur.
If you happen to have (for example) Crown (tm) power amplifier and a signal generator,
you can make clean sine waves. This costs more $$ unless you have the equipment already.
The third way (more exotic) is to get Texas Instruments' pulse density modulation
power amplifier modules and wire them up (carefully), using a correctly sized reconstruction
filter on the output.

Using today's power MOSFETs as analog devices is treacherous.
They oscillate, overheat, and generally are rebellious unless you've studied a lot of electronics.
They are not designed to be used "partly on" and they are not characterized for that.

In any case attaching the gate to a pot connected to the power line is
a recipe for much smoke. At least connect series resistors and parallel zeners
to make sure that the gate never exceeds + or - 15 volts with respect to the source.
Even instantaneous spikes will kill the FET. The built in zener diodes (if present) have
extremely limited power handling capability. Be sure that the zener diode power
limits are never exceeded.

There's a further problem here:
most power FETs can work well into high megahertz frequencies, and the parasitic
inductances and capacitances of the leads & nearby components frequently cause
oscillations which can destroy the FET quickly. The gate must be driven from a low
impedance (usually in the 1-100 ohm range) to prevent oscillation. This usually requires
an active circuit to drive the gate.

If anyone makes it this far and wants to know more, I'll deliver more prime bull.

Rosco Bodine - 3-2-2006 at 21:24


Look at the attached file showing the evolution of the
circuit design which I have been contemplating , and
consider why certain advantages are provided over
a variac , such as being able to set a minimum voltage
where the adjustment begins to effect adding power to the motor . The use of control circuitry also provides locations
where logic input is possible to be applied to make the circuit behave differently in response to sensed conditions .

For example when the circuit is first powered up , the existing
voltage setting could optionally be over-ridden by external signal specifying some selected number of complete cycles of full voltage to the motor to provide extra starting torque ,
and when the counted number of full voltage cycles has
completed , the circuit resumes operation at its setpoint voltage just after the motor breakaway into rotation .

What I am looking at is simple ways of improving the performance particularly of off the shelf 1/10 hp and less permanent split capacitor AC motors , using something
better for the task than triac based sine wave choppers .
The trapezoidal waveform from the circuit I have attached
is more nearly zero crossing with its waveform being continuous to less than 2 volts of zero on each half cycle
before dropout . This should be a very quiet circuit in
terms of any unwanted harmonic effects or spikes .

IGBT's could be used as easily as MOSFET's .

Edit : I may have the left side of the range adjustment pot
drawn wrong , and moving the wrong direction .
I just drew this and am looking at it . I think
its coil probably needs to be above ......yeah I'll have to redraw it and repost the corrected schematic . I have made
some changes and really should redraw the entire schematic
in a better layout . Long day .

I'm too tired to redraw the whole thing so forgive the smudges from my erasures and corrections . Here's the
corrected schematic for where I am now .

[Edited on 4-2-2006 by Rosco Bodine]

Variable AC Voltage Supply for AC Motor.jpg - 84kB

densest - 3-2-2006 at 22:09

@balanced_world: for temperature control (resistive heater) a triac driven by an asymmetrical diac (to prevent "snap on") is cheap, effective, and simple. If you need electrical quiet, use a cycle skipping zero crossing triac chopper which only delivers full cycles to the load. Thermal inertia takes care of temperature fluctuations.

@rosco: (please excuse typos - a cat is holding one hand hostage)
Sorry if I'm pedantic about this. I've been studying motor drives and power transmission recently. I'm assuming that your intended applications are not for large scale production, so scrounging good equipment on EBay, etc., can cut your costs a lot.

Your motor control application is reasonable. Is this intended for existing motors or new ones? What's the intended speed range (i.e. if 100 is "full speed", do you want 1-300 or 5-120?) Do you want constant torque, constant power, or ?? over the speed range? Remember that as speed decreases you must increase torque (magnet strength) to maintain power, and electromagnets fry if you saturate the steel.

For old motors or existing ones: remember stiction at startup or low RPM. At low voltages, the power and torque of AC motors is pretty low. Some of the chopper drives, ridiculously simple as they are, do provide more power as the load increases even at low speeds. If your goal is to control the rotor speed, you must have load feedback.

For a new design: consider a three phase motor. It inherently is easier to drive at full power or torque over a wide speed range. Rather than changing the voltage alone, changing the frequency as well gives you a more control. The capacitor phase shift in a 1PH motor doesn't like a wide frequency range. It's inherently simpler (no startup circuits, no cap to blow up...) Some new washing machines use IGBT choppers to drive a 3PH motor for all the washer functions with only a belt connecting the tub and the motor. No more transmissions to break.

If you derate them (like using a 1HP motor at 1/3 HP), "motor drive rating" doesn't matter very much. Saves a lot of money.

For a new design: consider a permanent magnet DC motor if you need high torque at low speed.

In any case, IGBTs and FETs are much happier used in a H-bridge configuration when driving an inductive load.

I'm looking forward to seeing your design.

Rosco Bodine - 3-2-2006 at 23:04


For reasons of economy and simplicity and practicality
for the application , the alternatives you suggest are
not valid . I want to control a single phase 120 Volt
60 cycle off the shelf permanent split capacitor motor
of 1/10 horsepower or less , using most desirably the
inherent abilty of a quadratic load to self regulate
the speed of the motor , without the use of closed loop
digital speed regulation . If the power and motor is
well matched to the load , particulaly a dual quadratic
load on a dual shaft motor where say 20% to 30% of
the motors torque is diverted into a parasitic quadratic
load like a squirrel cage cooling fan or a hysteresis disc ,
or even static field of DC injection into the stator itself ,
an asynchronous motor being driven under these conditions
will self speed regulate acceptably well at a fixed supply voltage . If the workload decreases on the motor , its
speed increase is limited by the parasitic load increase ,
and vice versa . And the response is instantaneous in
such a self-regulation scheme where feedback is an inherent
property of a quadratic load being exploited as a parasitic
load . I hope that made sense . This regulation effect
should occur best in the speed range from about 120 rpm to 1200 rpm for a four pole PSC motor where the torque speed
slope is linear . The shaft horsepower output should still be more than double at the same wattage for a PSC motor run
in such a scheme as is gotten from a shaded pole motor ,
and the speed regulation better , as well as the starting torque and entire low speed range torque .

I really can't understand the fascination of people with variable frequency drives for asynhronous motors which will never be run up to synchronous speed where the effciency
would be realized . And for a 75 watt motor the hardware investment is hardly worth consideration in the same sense as would be understandable for a 75Kw motor in a factory .

Zero crossing firing and cycle skipping

Lambda - 3-2-2006 at 23:54


I would like to know what you think about using a "cycle skipping zero crossing Triac chopper" in a Power Supply.

Transformer Specs.(Philips): 24 Volts and 40 Amps.

This is not the usual 1000 Watt Magnetron type Transformer kind of garbage. At least 4 Magnetron type Transformers would fit this Core, and the leads are about 8 mm diameter and made of solid Copper wire connected to M10 Brass studs. At 40 Amps, this "Monster" stays ice cold which makes me believe that 40 Amps is a rather conservative rating from the guy who sold it to me.

I would like to keep the "Bridge-Caps" voltage about 5 Volts (or even less) above the regulating Transistors output Voltage, to give them regulation space.

By using Triacs in the primary leads and firing them in the Sin Cycle, not only "bumps" the transformer, but also produces lots of Harmonics and heat in the Core. The Transformer efficiency drops drastically.

"Zero crossing firing and cycle skipping" seems to be a very elegant solution. I have 10, 25, 50 and 100 Amp zero crossing Solid-State relays from "Crouzet" and "Gunther" etc.. They are internally Optic-Coupled, and only need 3-32 Volts to trigger them and can switch 24-280 Volts AC. Even if they are fired during a Sin Cycle, they still take the first zero crossing.

[Edited on 4-2-2006 by Lambda]

Rosco Bodine - 4-2-2006 at 00:13

Lambda ,

Be careful with cycle skipping or people
will say the real truth is that you are
performing burst modulation :D

Page 10 on the attached file may interest you .

[Edited on 4-2-2006 by Rosco Bodine]

Attachment: Triac Inductive Load Control Circuits.pdf (92kB)
This file has been downloaded 750 times

bio2 - 4-2-2006 at 03:46

@ rosco

The circuit you have in mind controlling a series
connected AC load using the DC side of a bridge rectifier
does work but has many limitations such as non linear operation and not of much use except between 80-120 volt. The simplicity can't be beat except perhaps by a triac like the inductive rated "quadracs" made up to 40A
800V by Teccor.

This may well be adequate for your application and can be simply implemented using a duty cycle modulated transistor on the DC side. This circuit puts a lot of noise on the AC line if that matters.

Motorola had patented a circuit like this years ago but it never recieved much attention or use. The unit I built some years ago for a 3phase 20A 480V application used
a PWM duty cycle controlled bipolar transistor on each phase and worked well enough.

Rosco Bodine - 4-2-2006 at 10:43

The latest corrected revision of the mosfet circuit I am
considering is first attached at the end of my fourth
post above counting this post as one . Please take a
look . I think this particular circuit will work fine and
can be matched to the characteristic of a particular motor
in a particular application by using different value
power resistors for the fixed idle current value to the
motor , to supply just sufficient power that the motor
will barely turn , windmilling under light load . The value for these fixed resistors will have to be selected by first doing a rheostat test with the motor and 4 diodes ( two in series forward biased paralleled with two more in series reverse biased ) and measuring the resistance value needed for setting the idle current . These power resistors will divert much of the heat dissipation from the mosfets when the adjustable voltage is at minimum , reducing the heatsinking requirement . The redundant return diode I have added externally across the mosfet drain to source will also halve the return diode dissipation which would otherwise be there in the mosfet . The added zeners are likely to be 15 volt and are redundant gate protection . Candidate mosfets are Hitachi 2SK1837 , overkill for sure , but the overkill translates to ruggedness at little extra cost above using a minimal device for the mission critical component and having a failure
later from too little transient immunity .

I think I will reattach the schematic file here since I am further describing the circuit here in more detail and it
may be lost in the posts above departure to other related
circuits .

There is an idea which I have wondered if it has ever been
explored by the industry regarding the possible use of
many CdS photoresistor elements paralleled as regions
to form a " power photoresistor based optocoupler " ,
a " power vactrol " if you are familiar with the signal level
vactrol optocoupler by Perkin Elmer . Such devices are
high voltage AC photoresistors which might be arrayed
in sufficient number in parallel , deposited directly on a
heatsinkable substrate and driven by high output LEDS ,
to form a high current capable full wave line voltage AC controller , driven by TTL voltage levels , and passing
the AC power with the waveform completely unchanged
except in terms of controlling its voltage . Has anyone
heard of such a device ?

Anyway the revised mosfet circuit from the earlier discussion
is reattached here . I will likely prototype and test this
particular revision attached below since it seems workable in visualization of how it should work . Now if only Murphy's Law will not discover what I may have missed , it should be all right :D

I am going out on a limb here , but there is always the possibility when anyone is drawing a schematic for a
dedicated circuit for a niche application , that what they
design is entirely novel , since if they could find the circuit
already published in an application note as a commonly known circuit , then there would be no need to invent it
themselves in the first place . So it is just possible that
what I have done is propose an entirely new circuit in
one of those necessity is the mother of invention scenarios ,
and the fact it is a simple circuit in no way rules that out .
So if anybody should run across this circuit already published somewhere , please let me know since it also likely that often these sort of things turn out to be that in the lands where no wheel was to be found , it was reinvented by one genius after another who are all cousins of mine :D

[Edited on 4-2-2006 by Rosco Bodine]

Variable AC Voltage Supply for AC Motor.jpg - 84kB

12AX7 - 4-2-2006 at 11:20

Erm... *ANY*!!! *AC* motor will really highly NOT appeciate a "light dimmer" approach. This will ONLY work on universal type motors (the kind that you can use a dimmer switch on anyway).

Other than that, the circuit ought to work as shown, not counting the blown junctions of course (I hope you have a good way to dissipate all that wasted power).


[Edited on 2-4-2006 by 12AX7]

Rosco Bodine - 4-2-2006 at 11:47

Originally posted by 12AX7
Erm... *ANY*!!! *AC* motor will really highly NOT appeciate a "light dimmer" approach. This will ONLY work on universal type motors (the kind that you can use a dimmer switch on anyway).

Hmmm , there are dimmers of the sort we all know ,
and then there are other different circuits having different waveforms , like what I am illustrating , and indeed this should work for PSC and shaded pole motors better than
triac based wave choppers or series rheostat circuits which
are often used even though their behavior is counterintuitive
in that the inherent feedback to motor current increase under loading goes in the wrong direction , reenforcing the
speed decrease towards a stall . And then there's the buzzing created by switching harmonics of triac circuits which
cause stator hum and insulation breakdown over time .
What I propose should address those deficiencies and that
is the motivation .


Other than that, the circuit ought to work as shown, not counting the blown junctions of course (I hope you have a good way to dissipate all that wasted power).


Exactly where do you expect blown junctions and why ?
And who cares about dumping heat and " wasted power " from smoothly controlling a sub-fractional horsepower motor
when you don't get something for nothing in this case ,
and the wasted heat amounts to about what is dumped
from your computers CPU heatsink right now .....that's
wasted power too , but very necessary to get done
what business is at hand . All in all the waste is less than
what is dumped by the average desk lamp , and that waste
goes away completely as the voltage setting is raised to
maximum . The heat dumping only occurs at the lower
voltage settings .

When it comes to the matter of active closed loop speed regulation , there is a method which is fairly simple to implement to provide for fine control of the speed when
there is not a great variation of the load , perhaps in the range of plus or minus 10% . The method involves eddy
current braking being applied as a parasitic load to a motor
to increase its slip and reduce its stable rpms at a given voltage . If you have two iron core inductors with their
airgaps widened sufficiently that the opposite edges of an aluminum disk on the motor shaft is travelling through the airgaps during rotation , the magnetic field through the gap induces eddy currents in the disc and produces a torque loading on the motor which varies quadratically with rpms or with the strength of the field which may be varied by the current through the electromagnets . There is some inherent
speed regulation present with this sort of parasitic load ,
and the effect can be increased by active control of the current through the electromagnets , which can decrease or
increase the torque load on the motor which varies its
slip speed and rpms accordingly , even if the voltage to the motor is held constant . This is similar to the scenario where
a car with an automatic transmission is stopped on a steep uphill grade and a little throttle is applied sufficient for the car
to roll uphill , but a little pressure on the brake pedal holds
the vehicle in position ......yes power is being " wasted " but
for very good reasons which the situation requires , a balancing of forces which allows the vehicle to only roll forward when the traffic light changes and pressure on the brake is released , with no chance of rolling backwards and
smashing the grille of the police car stopped behind :D

Now that's regulation ! :D

[Edited on 4-2-2006 by Rosco Bodine]

12AX7 - 4-2-2006 at 15:09

Originally posted by Rosco Bodine
or series rheostat circuits which are often used even though their behavior is counterintuitive
in that the inherent feedback to motor current increase under loading goes in the wrong direction , reenforcing the
speed decrease towards a stall.

Yeah, you can reduce the RPM of an AC motor by reducing voltage, thus limiting power by the DC resistance of the winding and running it on the stall slope (relatively constant torque), but that's a really shitty way to do it.

And then there's the buzzing created by switching harmonics of triac circuits which
cause stator hum and insulation breakdown over time .
What I propose should address those deficiencies and that
is the motivation.

Proper snubbing can prevent this, although in general triac control of any inductive load is irresponsible anyway.


Other than that, the circuit ought to work as shown, not counting the blown junctions of course (I hope you have a good way to dissipate all that wasted power).

Exactly where do you expect blown junctions and why ?

*Points at schematic* That transistor, and that transistor.

And who cares about dumping heat and " wasted power "

Well, I suppose you wouldn't understand what engineering is about, so it's a bit of a lost cause on you anyway...

from smoothly controlling a sub-fractional horsepower motor

It might be rated for 0.6A at 120V. At half voltage (assuming a resistive response), that's 0.3A at 60V = 18W in each (9W per transistor). That can be easily dissipated, but using this on any heavier motor will quickly cost you.

Oh, and don't forget short circuit protection. Transistors don't appreciate 160V peak at 50A. (Don't say it won't happen, if it can it will. A fuse is NOT fast enough to protect silicon!)

When it comes to the matter of active closed loop speed regulation , there is a method which is fairly simple to implement to provide for fine control of the speed when
there is not a great variation of the load , perhaps in the range of plus or minus 10% . The method involves eddy
current braking being applied as a parasitic load to a motor
to increase its slip and reduce its stable rpms at a given voltage.

Holy crap, I suppose you also run your heating and air conditioning at the same time, too?!

If you're going to do *breaking* in the first place, skip the whole voltage control and use the low torque output of the motor itself... it'll waste less power, too! (All shaded pole motors are rated as "impedance protected".)

If you have two iron core inductors with their
airgaps widened sufficiently that the opposite edges of an aluminum disk on the motor shaft is travelling through the airgaps during rotation , the magnetic field through the gap induces eddy currents in the disc and produces a torque loading on the motor which varies quadratically with rpms or with the strength of the field which may be varied by the current through the electromagnets.

Ahem.. force (torque) is proportional to RPM linearly, not squared. Since *power* is force times rate, the power varies as a squared law.

There is some inherent
speed regulation present with this sort of parasitic load ,
and the effect can be increased by active control of the current through the electromagnets

I wonder, do you also keep your car's gas pedal floored and control your speed down the road with the brakes?


Rosco Bodine - 4-2-2006 at 15:53

Taken to extremes of absurdity , any rationale
including engineering principles which have validity
in a qualified sense and limited application can
be made into a farce , which is what you are doing .

The small rotor mass and momentum of small motors
involves different considerations for how to do things
best , than is directly applicable to much larger motors .

You are a purist when it comes to engineering .
I am a pragmatist , get the job done with what
works best within economic reason , but not
necessarily the cheapest or dumbest POS
that can be made .

If mosfets were never intended nor designed to be
run in DC linear mode , then why in the hell do the
manufacturers data sheets specify their performance
operating in exactly that fashion as one of their
parameters ? They will not fry so long as they are
operated within limits .

And short circuit protection is optional but can be added .
Voltage sensing across a low value resistor in series with the load can be used like a meter shunt as an input to a comparator , and a reference for current limit set on the other input , with the comparator output used to drive optocouplers clamping the gates to their rails at below their minimum on voltages . But with your vast knowledge of engineering , I'm sure I don't need to explain this to you .

I know how to drive , and can power brake even a
manual transmission vehicle on a hill by using my right heel on the brake pedal and the toes of the same foot on the gas while the left foot eases out the clutch ....
but I have never found this useful while simply cruising
down the road .

What you were saying about eddy current braking being used entirely for speed control is indeed possible but it
is generally done by letting the motor run at full speed
and using an eddy current clutch to couple to the load ,
they may call it a hysteresis clutch in some circles . It
works like a fan clutch on an automobile radiator , and they
may even make some of those using an eddy current clutch ,
but most of them are miniature hydraulic torque converters
which are thermostatically engaged or disengaged or
throttled in between by a little spiral pilot valve operated
by a bimetal coil which responds to the temperature of the airflow from the radiator .

With regards to quadratic loads , the torque requirement appears as the speed squared , and the power requirement
as the speed cubed . Somehow we aren't on the same page . And the linear slope for the torque speed characteristic is the shallow slope upwards from locked rotor
to about 75% of synchronous speed of 1800 for a 4 pole motor , 1260 rpm or so . The steep part of the slope is from
about 95% of synchronous back down to about 80% , not
much of a speed range for adjustment where the motor is
working in the range of its maximum output . So the
low and gradual slope is exactly where speed must be
controlled if the purpose for such a speed control is the
drive for a magnetic stirrer , which is what I am doing .
Does my strategy now make more sense ?

[Edited on 5-2-2006 by Rosco Bodine]

12AX7 - 4-2-2006 at 16:39

Originally posted by Rosco Bodine
If mosfets were never intended nor designed to be
run in DC linear mode , then why in the hell do the
manufacturers data sheets specify their performance
operating in exactly that fashion as one of their
parameters ? They will not fry so long as they are
operated within limits .

I don't like wasting power. Linear region means volts times amps. You can certainly make a wonderful amplifier using the full linear region, but if you tack on a PWM and integrator, you can go from 33% to 95% efficiency in one fell swoop! Class D has plenty of volts and amps, but not together, so there's very little waste.

FYI: because of the favorable characteristics, many FETs are only rated for pulse or switching applications. That doesn't mean they won't work in the linear range, it just means they weren't intended for it. This isn't new: vacuum tubes designed for TV sweep or radar pulse applications have few linear operating points listed.


Rosco Bodine - 4-2-2006 at 17:09

Hey I don't like wasting power either , but there
is a time for practicality versus pure power utilization
economy , and for a small motor it isn't worth the expense and grief of hardware complexity to save a few watts
when they can be more easily just dumped as the cost of doing business in the particular application . With light
motors , cogging and hunting oscillations are inevitable
using pure brute force digital closed loop speed control
at low speeds . So it is far better to try to match the
motor to the task , and exploit every inherent speed regulation method which can be used , before ever getting
to the point of using closed loop feedback for fine control .
The idea is that the thing should run pretty stable ,
and perhaps stable enough , even without any feedback
from digital controllers . Well designed equipment where
the components are well matched can do the job very well
open loop , and many times that is exactly the way equipment of the old days was made , mechanically governed like steam engine throttles , before any digital
circuitry was ever invented . Some of these scenarios
are really quite ingenious and work very well . There is
a trap involved in getting overly technical on a simple device where it isn't necessary if there is a simpler and
possibly better way , even if it does waste half the power
of a 75 watt device at its least efficent point of operation .
For the price of 37 Watts I'll take simplicity and reliability
and economy any day .

12AX7 - 4-2-2006 at 19:54

Bah... Dig out a switchmode power supply chip, a pair of MOSFETs, rig a ferrite transformer (all of ten minutes winding one) with DC restorer for gate drive, rectify the line voltage then chop it, add a filter and you've got a PWM class D amplifier, 95% efficient or better, and if you generate your own drive signal you can get any frequency, amplitude or waveform you want. :D

For something this small, pffbt... but anything over a half horsepower, this is how it's done.


Rosco Bodine - 4-2-2006 at 20:43

For those really critical range hood ventilator fan
speed settings when only the best will do ......

make this little gem the powerhandler and the manufacturer
will throw in the dedicated workstation with user friendly
interface for free :D

Then again , when it's a five dollar motor ......
there just has to be an easier way :D

Actually my first thought on this project was to open up
an inexpensive uninterruptable power supply and
see if it would be possible to easily adjust the voltage
output from the inverter .

The output would already be well filtered sinewave
so it would simply be a matter of being able to adjust the output voltage . The input side is already there for keeping the battery charged , but of course the battery isn't needed for the use to which the UPS power stage would be adapted .

[Edited on 5-2-2006 by Rosco Bodine]

densest - 4-2-2006 at 23:21

Modern circuit design has gone away from analog power circuits for a number of reasons, like size, weight, cost, and reliability. Choppers are widely used to drive motors because they are very simple and do a "reasonable" job for low cost.

There's a "simple" way to make sine waves digitally at reasonably arbitrary voltages from a fixed supply. It uses sneaky math to work, but the circuit is simple. More or less.

Get a microcontroller with one digital output and several digital inputs or one analog input. If the inputs are all digital, this requires log(2) of the number of discrete voltage steps. The microcontroller should be capable of switching the output at (say) 256X the desired output frequency while sampling the inputs at (say) 10x per second. For this application jitter is not too important.

Compute the Fourier transform of the desired waveforms, which for sine outputs is a single nonzero value. Compute the reverse Fourier transform at the desired sample rate and quantize to 1 bit using a good random number generator for dither. Burn tables into the microcontroller for the desired amplitudes. Program the microcontroller to output the table values (1 or 0) depending on the control inputs.

Connect the output of the micro to a high-voltage mosfet high/low driver pair. Invert the output and connect it to another driver pair. Make an "H" bridge of mosfets - the output is the crossbar. If desired, compute a 3 or 4 section passive LC Bessel filter for the output of the bridge. At least put an L -> C for 12dB/octave at 2X output frequency or so and some snubbers (choose your favorite).

There are a couple of options for the input (AC line).

1) Brute force: Get the DC buss from a bridge rectifier and cap. Ripple doesn't matter much. Output voltage is proportional to input voltage.

2) Sneaky: Sync the micro to the input line zero crossings and apply a sine function as a correction to the quantization during table generation. Rectify the line with a bridge rectifier. No filter cap needed at all. Output voltage is proportional to input AC voltage. This is the closest I think practical to a "just step down the input" circuit.

3) Drive the DC buss from an off-the-shelf PFC circuit. You get regulation and universal input for free.

Total parts count: 20-40, with most in the PFC version. All but the power components (4/5 FETs, diodes, inductor(s), filter caps) could be tiny little things on a tiny little PCB.

None of this addresses the problem of speed control. Dissipative speed control is possible but inelegant. Monitoring the reverse EMF of the motor is pretty easy given this approach, but it would add another (say) 10 parts and some microcontroller inputs. Response times could be < 5 output cycles or so.

Efficiency should be > 80%, RFI should be easily controllable, response quick, and cost < $30 for parts. If you want cheaper use a universal motor and an SCR.

Possibilities: for lower speeds changing the output frequency to a lower value would allow the motor to run closer to its synchronous frequency and therefore run cooler. That approach would require input version (3) since the output would no longer be synchronous with the power line. Of course this approach is limited because the motor phase shift capacitor becomes less effective with lower frequency excitation.

A three-phase version would always run the motor at its synchronous frequency, but it is probably impractical for < 200W motors unless you have a cheap supply of synchros, which would work beautifully.

 Pages:  1