[PJ]

Hi guys. Sorry I dont post as much as I used to. I visit everyday, but I am in my final year of EE at the University of Western Australia, and am a busy boy..

For my final year project I have to build controller for a thermoelectric cooler for a IR detector. Its mostly done, but the part I thought was easy (and the stuff I am good at) is causing me trouble. Hopefully you guys might know as my supervisor is away till next week.

Basically I have a controller etc, which outputs an 8-bit control signal to an 8-bit DAC. The DAC is the TLC 7524 R-2R DAC:

http://www-s.ti.com/sc/ds/tlc7524.pdf

I have configured it as a voltage output, as per the datasheet and it swings from 0V to 4.05V, with a 22.8k output impedance (measured).

Essentially I am trying to create a voltage controlled current source, which corresponds to 0A for 0V input and 1.4A for 4.05V. Efficiency is a key consideration as this will go in a portable IR camera. Linearity is desirable, but efficiency is more important.

Any suggestions. I have tried many designs, using the Zetex853 NPN bipolar transistor.

http://www.farnell.com/datasheets/1471.pdf

I have come close, but never succeeded. My designs have essentially been along the lines of:

Two cascaded ZTX853's. The second stage has the cooler in the collector and is run from 1.2V The second stage's base is fed current from the emitter of the first stage.

The first stage is biased by a fixed resistor network, and contains a collector resistor, which ideally puts the transistor close to saturation no minimise power dissipation. Runs from 5V. Series input resistor to control input current.

The problem is for low input voltages (say 1V), current flows from the ~1.6V bias backwards to the DAC, rather than the base of the first stage transistor.

Resistance of the cooler is 0.64R.

There is a 5V and a 1.2V source available. The 1.2V can be changed if needed (prefer not too). The 5V is used for the PIC16F877 microcontroller, which runs the PID controller algorithm, and the DAC.

Its really bothering me, as I know this cant be that difficult...probably the easiest part of my total project to be honest...

Any help would be appreciated...

[bob82274]

AHHHHHHHHHHHHHHHHHHHH I'm going to go hide now!

Sorry I'm afraid I can't help ya I'm just a lowly 2nd year EE Major.  I am enjoying it which I think is a good sign but I'm afraid your problem is WAY beyond me at this point (as much as I hate to say that).  I realize that you are in Perth PJ, but it is my understanding that there is an excelent school for EE near Melbourne.  Don't know the name... may not even be true.  I'd love to know more about it though.

[AKSA]

Hi PJ,

This may not be so difficult.  First thoughts are a blocking diode and a global feedback loop to correct non-linearities in the voltage to current conversion.

However, first rule;  pictures - man, we must have pictures!

Put in a schematic, and we'll have a gander.....

Cheers,

Hugh

[ginger]

TEC "Thermo Electric Cell (Peltier Cell)" Control

Redesign of you opamp transistor circuit below

Demand voltage from the DAC goes into +ve input of an opamp via 3K9.
Add 100R resistor form +ve input to 0V. ie voltage divider on +ve input of opamp.

Output of the opamp feeds gate of a N Channel MOSFET(via series 100R for stability). Add 1000pF capacitor from Opamp Output to Opamp -ve input (also for stability)

Put a current sense resistor of say 0.1 R in the FET source to 0V.
add 470R resistor from FET source back to -ve input of opamp (stability again)

The TEC connectds between +1.2V and FET drain.

Choose a good power MOSFET with an Rds ON off say 0.1R or better. Also watch the Vgs threshold voltage. Some opamps will only drive to within about 1.4V of the rail (+5V) ie 3.5V in this case. You want to make sure you can fully turn the FET on with the available output voltage of the opamp.

Use a TO-220 or TO3 MOSFET and you may get away with not requiring a heatsink.

Reasons for changes:
The opamp will NOT be able to drive enough base current into the transistor for a 1.4A collector current. Hence change to MOSFET for the series pass device. (you could also use Darlington connected transistor pair BUT will dissipate more power due to the saturation voltage of the transistor)

The TEC Cell is highly non-linear so using it as both the load and its own "current sense resistor" will give highly non linear result. Therefore have put the TEC Cell into the "high side" i.e the FET Drain and added separate current sense resistor in FET drain.

The Opamp will control to make the voltage developed across the 0R1 current sense resistor (fed to the -ve input via the 470R) equal the voltage on the +ve input.

Notes on TEC Coolers and Control of them

TECs work best at 30% of maximum current rating and work OK up to about 50% of maximum current rating. Above this their self dissipation (I squared R ) rapidly soaks up most of its heat pumping capability and they simply "run out of puff".

When I use TECs I usually choose a unit that has the heat pumping capability I need at 30% of its max. current rating and then use Pulse Width Modulation Control ie On/Off control.

Cheers,
Ginger

[PJ]

Thanks so much for the help guys. Really, really appreciated.

The schematic I first mentioned , which wont work, is here:



The one I have considered today is:



I only just thought of this late this afternoon. The idea is the OPAMP essentially uses its massive gain to turn the transistor on, even for very small Vin. I am yet to really analyse it, but I thought I may as well throw it up in the air..

For the second design, the opamp I am currently considering is the Burr-Brown OPA340. Single 5V supply, with necessary current drive.

http://www.farnell.com/datasheets/3099.pdf

The transistor is the Zetex ZTX688B. High gain. Low Vce(sat), high current, enough Pd.

http://www.farnell.com/datasheets/32229.pdf

[PJ]

The DAC can be run in either current mode or voltage mode.

http://www-s.ti.com/sc/ds/tlc7524.pdf

Unfortunately at uni they love to teach us about theoretical stuff, and not so much about the real world

Ginger, with the current mirror you were talking about, how does the DAC bias the two transistors on. I dont really know what current mode means...which is why I put the DAC in voltage mode!

In case you guys are interested, here is an image my IR camera produced. It is a low cost IR camera, aimed to produce an image like this for around $1000.



Thanks for all your help.

[bob82274]

Taking a stab in the dark that what you photographed was a lighter...

Am I close?

[ginger]

I have severe;y edited what I posted earlier - modification of you opamp suggestion with comments on changes.
Cheers,.
Ginger

[AKSA]

Hi PJ,

Ian has contributed massively to this thread, and since it's more his area than mine, I'll only give a few simple thoughts.

Your circuit looks good.  It will obviously hold equal voltages on the two inputs to the opamp under all conditions.  However, since it is operating in open loop, stability might be a factor, and a 100pF capacitor between output and inverting input will pull feedback back to sub unity levels by the pole frequency, ensuring complete stability.  You might also consider a rudimentary filter at the DAC output to rid the digital artefacts, which are definitely a problem with modern opamps.  You don't want MHz crud appearing at the load;  currents are considerable, and aside from stability issues you could create massive RF emissions.

I am assuming the DAC output is solely voltage, not current.

What you need to determine is the transconductance (gm) of the combination, fixed by the ratio of input voltage (measured at the DAC output) to output current (measured by the voltage appearing at the top of the load).

Presently, one volt at the input will give you 1/0.64 amps at the output, viz 1.562A/volt.  If you wish to increase gm, then take the sensor for pin 2 from a voltage divider across the load;  if you wish to decrease it, place the voltage divider at the output of the DAC.

Of course, this is all theoretical, and you need to PSpice it, build it, then exhaustively test it, including examining oscillograms on a 100MHz CRO to guarantee that it is unconditionally stable.  For example, any phase shift in the load will either add or subtract from the expected 180 degree feedback, and could cause instability, so this is an important consideration.

In summary, the circuit brief is ideal for an opamp, but high levels of NFB are necessary  to confer stability, although NFB does greatly improve load control.

May I inquire if you built out your AKSA to 200W, PJ?  If so, I'd like to chat about the design issues with you!

Cheers,

Hugh

[PJ]

Thanks again guys.

Ginger, where about did you post the changes?

Hugh, I am just running out the door to a 9am lecture, appropriately on Electromagnetic Compatibility. I will reply to you in 2 hours, after I have finished these two lectures!

I really bothers me how hard I am finding this. I thought this was going to be easy bit of the Autotuning PID Temperature Controller I am building.....also using a PIC16F877 Microcontroller, interfaced with a MAX233 line driver.

[ginger]

PJ
Was busy trying to modify the previous post while the new ones came in. Then had a crash when I tried to paste a schematic. The edited ealier post is now there. You will see modded circuit includes the cap suggested by Hugh for stability (a bit bigger) plus two other resistors required for stability when using FETS (because of their large gate capacitance) - feel free to email me direct for any clarification.

Since I couldn't seem to paste the schematic in here I put it in a word document and have emailed it to you direct.

Cheers,
Ginger

Direct Contact:
Ian Miller
Senior Electronic Engineer
Laser Airborne Depth Sounder
Tenix LADS Corporation
email: [email protected]

[PJ]

Hugh,

No, I haven't built out my AKSA to 200W. I may in the future, but at the moment:

A) No time - my life is a shambles at the moment  
B) No money
C) Currently, the 100W has more than enough "room to breathe"

I may definately in the future, I remember being interested when I first read about it.