Swede
Hazard to Others
Posts: 491
Registered: 4-9-2008
Member Is Offline
Mood: No Mood
|
|
Data logging
With notebook PC's and the USB bus as a now ubiquitous feature owned by just about everyone, I'm wondering why more advantage is not being taken by
experimenters with data loggers. In the old-school labs, we had paper chart recorders, usually single channel, sometimes two or three, with different
pen colors. They were mechanical, clockwork monstrosities.
With modern, digital/analog sampling and software, data to track and log experiments can now be collected easily, but for what is essentially $30
worth of silicon chips, manufacturers are asking too much $$ for this sort of device, IMO.
I'd like to be able to track voltage, temperature, current, and maybe another channel or two for electrochemical experiments, yet I am having a hard
time finding affordable, PC-based data loggers. Does anyone have any suggestions besides "roll your own?" It could be done with the PIC or similar
series of chips, but I really don't have the time to do it, or the motivation, frankly. Thoughts?
|
|
|
not_important
International Hazard
Posts: 3873
Registered: 21-7-2006
Member Is Offline
Mood: No Mood
|
|
What does "low cost" mean
look at these guys http://www.j-works.com/ and http://www.accesio.com/go.cgi?p=../cat/usb.html and tell me if that is low cost. Their prices don't seem outrageous, being about the markup you
need to stay in business considering all overhead and that. That $30 worth of chips is going to cost several times that to turn into a PCB, these are
low volume goods so the development costs have an important impact on pricing.
Places like these http://www.aapcb.com/prototypeservice.asp?gclid=CJXE_azVt5gC... let you roll your own, and will give you an idea of costs.
But, no, I've no idea where to get such much cheaper. Sound cards don't do DC, standard PC consumer digital I/O doesn't handle the levels generally
needed for lab interface.
Look at places like these for LIMS
http://www.chemicalanalysis.com/news.php?extend.16
http://www.open-lims.org/
http://www.plausibleaccuracy.com/2008/04/02/open-source-lis-...
|
|
|
Ozone
International Hazard
Posts: 1269
Registered: 28-7-2005
Location: Good Olde USA
Member Is Offline
Mood: Integrated
|
|
From a previous post:
"Radio Shack was selling a very cool multimeter (self scaling, too) which compared well with my Fluke, AND, has a serial output. It comes with a very
intuitive software package called "Meter View"tm which seems to work on win 95, 98, 2000, and XP (have not tried XP-64 or Vista beta). It gives you a
nice graph, and outputs to ASCII wich is simple to import into Excel as a CSV. It cost me about $70 US imagine that, a Radio Shack product that is
worth every penny."
It is only a single channel, but it is cheap and versatile. If you can log voltage, your good to go. SFAIK, almost everything outputs in either volts
or current (20mA full-scale is common; just add a resistor to measure current). Add a thermocouple, thermistor, etc. to give T (you will have to
calibrate it).
I have also seen quite a few cut-rate multichannel DAQs that come with software (or code to compile yourself, >$200 US for 4-8 channels).
Cheers,
O3
[Edited on 31-1-2009 by Ozone]
-Anyone who never made a mistake never tried anything new.
--Albert Einstein
|
|
|
JohnWW
International Hazard
Posts: 2849
Registered: 27-7-2004
Location: New Zealand
Member Is Offline
Mood: No Mood
|
|
I thought that Honeywell, based in the U$A, was the leading manufacturer of data recorders and data loggers, used in large analytical and research
laboratories, and in chemical process plants and oil refineries. They also make flight data recorders and cockpit voice recorders ("black boxes") for
large aircraft. I heard recently that they have suffered large job losses.
|
|
|
Swede
Hazard to Others
Posts: 491
Registered: 4-9-2008
Member Is Offline
Mood: No Mood
|
|
After a ton of googling, I found a company that seems to have a nice selection of inexpensive (and by that I mean <$300) data loggers:
http://www.dataq.com/index.html
Ultimately the problem is "What do you want to measure?" Typical chemistry data would be things like voltage, temperature, current, pressure, pH.
Each of them has unique signals, often very low voltages that are prone to interference, especially in electrochemistry.
Voltage is easy - you can always divide the voltage in half (or whatever) with precision resistors. Current - can be trickier. You'll need a shunt,
and now your signal is typically 0 to 100 mV, still very easily readable by the datalogger.
An unamplified thermocouple, on the other hand, can be MUCH trickier. RTD - needs a power source. Same deal with pH, plus a super-low signal to
boot.
Those dataloggers that come "ready" for things like pH and perhaps an RTD suffer a high price tag. I think the answer for inexpensive is to obtain
the basic dataloger device, which is almosty always voltage, and roll your own interfaces that would bring things like a thermocouple signal into a
more practical range.
NotImportant, Ozone, and JohnW, thanks for the replies and the links. I will definitely look into them.
|
|
|
watson.fawkes
International Hazard
Posts: 2793
Registered: 16-8-2008
Member Is Offline
Mood: No Mood
|
|
Signal conditioning is now the most expensive part of any data collection system, given how inexpensive A/D converters and data manipulation is.
(@Swede: I'm sure you know most of this.)
Voltage and Current: Precision resistors in Wheatstone bridge configuration allow higher precision for certain applications. The trick is to use a
slightly unbalanced bridge rather than a balanced one (as in a nulling configuration). Insertion losses (for current) can be quite low, since it's
resistance ratios that are important. Likewise, such a configuration can have the effect of magnifying the input impedance of a voltage sensor.
Thermocouples measure ΔT physically. You need a "room temperature" sensor for absolute cold temperature as well as precision voltage
reference. RTD's require the voltage reference and, preferably, a table to correct non-linearity. pH, all the same kind of circuitry, with more
attention to averaging noise from the sensor.
One project I'm planning on doing this year (or so) is a thermocouple conditioning board that also digitizes and feeds data back over an
I<sup>2</sup>C or SPI interface. The target is to feed into the open source hardware crowd including Arduino and RepRap. My own interest
is microscale industrial chemical plants, but the same device is also applicable to electrochemistry. I've mused over doing an volt-ammeter board
also; it seems of a piece with the other.
I'd also like to figure out how to hack a Bourdon tube pressure gauge to digitize, but that's not next on the list.
|
|
|
Swede
Hazard to Others
Posts: 491
Registered: 4-9-2008
Member Is Offline
Mood: No Mood
|
|
In my quest to find the ultimate TC chip, I think I found one, but at a price... the AD595. It has built-in cold-junction compensation, and off the
top of my head, all the chip required was 5V, and the type K TC would output 10 mv/C through the chip. It had all sorts of fancy features, and worked
very well. If you wanted to measure below 0 degrees, you needed +/- 5VDC. But the thing was truly monolithic and worked perfectly, with accuracy.
The only problem was the price. In the 1990's, when I was working with it, they were running about $20 each.
A current shunt would require probably a good amplifier circuit to get good results. So yes indeed, signal conditioning and if necessary
amplification seems to be the killer for data logging.
Now RTD using a PT-100 sensor is not something I'm familiar with, but many temperature measurement devices are switching to RTD. The range is
narrower than a TC but the accuracy is superior, and there are fewer troublesome issues with leads and such. I do know it requires a precision
voltage or current through the probe, adding yet more complexity to the system.
WF - how about hacking one of those $10 digital tire pressure gauges? Somewhere in there is an analog or digital signal off the pressure sensor,
which could be tapped and fed into a logger. The only problem is the limited range of the device. But the price is right.
[Edited on 31-1-2009 by Swede]
|
|
|
watson.fawkes
International Hazard
Posts: 2793
Registered: 16-8-2008
Member Is Offline
Mood: No Mood
|
|
| Quote: | Originally posted by Swede
In my quest to find the ultimate TC chip, I think I found one, but at a price... the AD595. |
Thanks for the
pointer; I hadn't started researching chip. I just checked Octopart; Jameco has the A part for $15 (A => +/- 3%, C => +/- 1%). It looks like
just the thing for prototyping, since if you feed it a 15V supply it will read a K-type junction to 1250° C. I'm looking at doing control of
the contact process, and I'll need some of that range to regulate catalyst activity. I can use a regular old AVR chip for digitizing and
communication. Not the cheapest in parts cost, but definitely cheaper on the engineering front.
There may be cheaper ways of doing it these days. Some of the TI MSP430 chips have internal temperature sensors, built-in op amps and analog
comparators. The Analog Devices chip sums the thermocouple and the reference point voltages in analog. The TI chip could do that digitally, especially
since the purpose is digitization of the signal. These TI chips would likely be good for RTD's, too.
| Quote: | | A current shunt would require probably a good amplifier circuit to get good results. So yes indeed, signal conditioning and if necessary
amplification seems to be the killer for data logging. |
Even with voltage, you want an op-amp voltage
follower in the circuit, simply as a cheap sacrificial part to blow under failure conditions.
| Quote: | | WF - how about hacking one of those $10 digital tire pressure gauges? |
As you point out, the range is too
limited for what I'm looking for. My current idea is to use a capacitative coupling (one fixed, one on the tube), an oscillator, and a lock-in
amplifier and get a signal voltage vs. distance relationship. The lock-in can be pretty cheap, I'm thinking, because you're generating the sensed
signal, so you can cheat on the modulation and still get all the nice noise reduction. Assuming this works, the mechanical part is simplicity itself:
a pair of copper pads punched out of PCB material, glued in place, and two lead wires.
|
|
|
chief
National Hazard
Posts: 630
Registered: 19-7-2007
Member Is Offline
Mood: No Mood
|
|
The usb-stuff should coast nothing, max 5$ ! More is unacceptable . Else tho good old parallel port can be used, which also has input 8 pins, that can
be connected to some A/D-converter, coupled to the measuring-circuit, and isolated via optocouplers .
The parallel-port output-pins can be used to control relays (small reed-ones),. to switch between the different data-sources (A/D)-converters ...
But this really is yesterdays stuff ... ; there just has to be some parallel->usb-converter-chip, from some manufacturer. This would then be
connected to the A/D-converter ... ; definately it just has to be out there ...
|
|
|
not_important
International Hazard
Posts: 3873
Registered: 21-7-2006
Member Is Offline
Mood: No Mood
|
|
| Quote: | Originally posted by chief...
But this really is yesterdays stuff ... ; there just has to be some parallel->usb-converter-chip, from some manufacturer. This would then be
connected to the A/D-converter ... ; definately it just has to be out there ... |
PIC18F65J50, PIC18F66J50, PIC18F66J55, PIC18F67J50
Overview
http://embedded-system.net/pic18f87j50-usb-microcontrollers-...
Buy `em here
http://www.microchipdirect.com/Default.aspx
| Quote: |
Communications: The PIC18F87J50 family incorporates a range of serial and parallel communication peripherals, including a fully featured Universal
Serial Bus communications module that is compliant with the USB Specification Revision 2.0. This device also includes 2 independent Enhanced USARTs
and 2 Master SSP modules, capable of both SPI and I2C™ (Master and Slave) modes of operation. The device also has a parallel port and can be
configured to serve as either a Parallel Master Port or as a Parallel Slave Port.
CCP Modules: All devices in the family incorporate two Capture/Compare/PWM (CCP) modules and three Enhanced CCP modules to maximize flexibility in
control applications. Up to four different time bases may be used to perform several different operations at once. Each of the three ECCPs offers up
to four PWM outputs, allowing for a total of 12 PWMs. The ECCPs also offer many beneficial features, including polarity selection, programmable dead
time, auto-shutdown and restart and Half-Bridge and Full-Bridge Output modes.
10-Bit A/D Converter: This module incorporates programmable acquisition time, allowing for a channel to be selected and a conversion to be initiated
without waiting for a sampling period, and thus, reducing code overhead. It has an 8 or 12 channel MUX, depending on family member, connecting it to
I/O pins.
Flash program memory (six sizes, ranging from 32 Kbytes for PIC18FX5J50 devices to 128 Kbytes for PIC18FX7J50). The family includes 3904 bytes of
RAM.
|
Front it with that MAX6603 for thermocouple interfacing, and Bob's your uncle.
For RTDs the Analog Devices AD7711 or AD7730L work nicely, and output on a serieal data buss; they are 24 bit converters for high resolution/accuracy
applications. The AD7730L is targeted at bridge sensors, but can be used with an RTD plus fixed resistor; it uses AC excitation to improve noise
rejection, reduce RTD self-heating, and eliminate error from parasitic thermocouples in the wiring.
|
|
|
dann2
International Hazard
Posts: 1523
Registered: 31-1-2007
Member Is Offline
Mood: No Mood
|
|
Hello Swede,
Some link here may or may not be useful. The first one is interesting.
http://cgi.ebay.co.uk/low-cost-data-logger-15-channel-USB-10...
http://cgi.ebay.com/Fluke-model-2240c-Data-Logger_W0QQitemZ1...
http://www.vincenzov.net/eng/design/max147.htm
http://www.hobbycity.com/hobbycity/store/uh_viewItem.asp?idP...
Dann2
|
|
|
watson.fawkes
International Hazard
Posts: 2793
Registered: 16-8-2008
Member Is Offline
Mood: No Mood
|
|
| Quote: | Originally posted by chief
But this really is yesterdays stuff ... ; there just has to be some parallel->usb-converter-chip, from some manufacturer. |
The integration available is even higher than that. Atmel has the AT90USB chip series with on-board USB support. The AT90USB646, for
example, has 8 channels of A/D (10-bit), completely eliminating the need for an off-chip A/D converter for many applications. For process control,
this is usually plenty of accuracy. If you're trying to perform precise single measurements, it might not be enough, but in that situation you're not
going to get to really-cheap anyway. The AT90USB646 in single unit quantities is about USD 11.
If you have a definite purpose for your sensors, then it makes sense to use a single, highly-integrated chip and make sure than everything fits on it.
What I'm thinking about, on the other hand, is a one-processor-per-sensor design, with a single communications-and-coordination processor at the
center. The idea is to have a drawer full of various sensors that will simply plug in to a bus, automatically identify themselves, and start gathering
data. It's a design for tinkerers and experimenters.
|
|
|
not_important
International Hazard
Posts: 3873
Registered: 21-7-2006
Member Is Offline
Mood: No Mood
|
|
That's why I talked about the PIC18F6xJ5x line. Cheap, US$ 3 to 5 for singles. Enough I/o functionality that the same chip can be used in a number of
applications.
The PIC24HJ32GP302/304, PIC24HJ64GPX02/X04, and PIC24HJ128GPX02/X04, which is a 16 bit processor with 12 bit A/D might be a better choice for the lab.
It has a CAN bus interface, which supports longer cable runs than USB and uses twisted pair wires which can be had in insulation with better heat
resistance than USB cables. Runs less than US$ 6 in single quantity.
|
|
|
watson.fawkes
International Hazard
Posts: 2793
Registered: 16-8-2008
Member Is Offline
Mood: No Mood
|
|
| Quote: | Originally posted by not_important
The PIC24HJ32GP302/304, PIC24HJ64GPX02/X04, and PIC24HJ128GPX02/X04, which is a 16 bit processor with 12 bit A/D might be a better choice for the lab.
It has a CAN bus interface, which supports longer cable runs than USB and uses twisted pair wires which can be had in insulation with better heat
resistance than USB cables. Runs less than US$ 6 in single quantity. |
As for CAN-vs.-USB, I'm with you on CAN
for an intra-lab network. For all the applications I'm thinking of, I'll see either zero or one USB interface, for connection to a PC host for overall
control and monitoring. For the very simplest projects, USB is also appropriate, if there's a host machine that's being used for direct control. For
device-to-device, though, CAN is definitely the winner, in part because of the asymmetry between USB host and device implementations.
As for CAN-vs.-I<sup>2</sup>C, that's a different issue. There's a cost-vs.-performance trade-off here. All of these small
microcontrollers do I<sup>2</sup>C; only some do CAN. CAN is far more robust against noise, while I<sup>2</sup>C has serious
limitations about cable length and bit rate regarding noise. I<sup>2</sup>C is just barely appropriate for a fume hood area network
(FHAN), and even there you have to pay some attention to using twisted pair, termination, and noise levels. Having said that, I'd never run even a
small process completely unattended with I<sup>2</sup>C anywhere but a stripline on a PCB (or the like). For those applications, I likely
will use CAN for all cabled interfaces.
I'm planning to use I<sup>2</sup>C first for the very simple reason that the Arduino can speak it without further modification. My target
audience is the kinds of tinkerers that are already using this platform. Arduino is without a doubt the slickest platform I've come across for
learning embedded systems.
I should also state that my preference for the AVR comes entirely from the fact that there is a version of gcc for it. I've done a lot of software,
and the long-term win is generally with open-source tool chains. When Microchip and TI get on the bandwagon on this one, I'll gladly select chips
based on purely hardware considerations. Until then, though, software wins for me.
|
|
|
Swede
Hazard to Others
Posts: 491
Registered: 4-9-2008
Member Is Offline
Mood: No Mood
|
|
I know this thread is old, but I have learned a lot in the last two months and have set up a robust, simple, and reasonably inexpensive
data-acquisition and logging system using 5B modules.
5B modules are hot-swappable signal conditioners that plug into a bus card; once there, they isolate and if necessary amplify the signal. They are
made for voltage, millivoltage, TC, RTD, frequency, and a few others. Analog Devices (the company) designed the specifications of the system quite a
while ago, and they are currently offered by not only AD, but Omega, Dataq, and National.
They perform very well. The problem is the cost, at $150 to $200 each when new. The answer: they are available surplus all over the place,
especially eBay, for $20 to $60. With a bit of patience, I picked up 4 or 5 that would cover all of my needs, and built a "bus" for them:
I bought a Dataq DI-158UP AD converter, USB-based. What has me really impressed is the highly configurable and flexible software that comes free with
the unit. Pretty much every imaginable display, limits, and units parameter can be modified. The system is ready to go now, and I look forward to
putting it to use.
With care, I get the feeling that the DI-158U can be used without isolation. The differential inputs are good to plus/minus 64V, pretty hefty. It
depends upon how much noise is in the system, and whether the system is subject to spikes or other voltages that could conceivably fry the DI-158.
So there you go, not as inexpensive as the "roll your own" method, but at the same time, done on a reasonable budget. 
|
|
|
|
![[*]](images/xpblue/default_icon.gif){kind=icon}
