THS3121: Improved driver for peltier (TEC)?

Hi Daniel,

I was able to look over and compare the devices, and I am not sure if the replacement would increase the current drive as they are very similar in their output capabilities. Are there any clues as to why the output current was limited to 140mA? The device is capable of much more. For the device specifically, the minimum voltage supply required is 10V. The model might not show adverse effects when the supply range is violated, which is why you are still seeing good results. For the internal power dissipation, I attached a link to a great blog that captures how to calculate this. There could be a way to directly drive some kind of discrete transistor stage, however I do believe driving the load with ~280mA should be possible as some kind of discrete stage brings its own concerns.

Best Regards,

Ignacio

(+) Top questions on op amp power dissipation – Part 2 - Precision Hub - Archives - TI E2E support forums

Hello Ignacio Vazquez Lam1,

Thank you for your looking into my question. 

I'd be happy if we could find a solution to directly drive the TEC from the op amps without an additional transistor stage. 

I have no idea why the THS3121 provides more current than the OPA2677. Maybe its (THS3121) output driver impedance is lower? I could not find the value in the datasheet, though. Regarding the output limit of the OPA2677 of about 140 mA, the measurements on the bench match the simulation quite well. Maybe you want to have a look at the simulation using OPA2677? Please find it attached below.

I will also check your link regarding op amp power dissipation. Thanks for sharing.

Best regards,
Daniel

10 BridgeInvertedOPA2677 V3 Uc(max)=3,3V.TSC

Hi Daniel,

I was able to look at the OPA2677 model, I believe the claw curve might not be captured properly. However, looking at the THS3121 output voltage versus output current, I believe the device would also run into a headroom concern when pulling such large amounts of current. Figure 19 shows that when the device is driving currents in the ~400mA, the headroom required is significant. One potential option is the OPA2675. This device should have enough output drive for your application. Figures 8-4 and 8-5 highlight its output drive capabilities.

Best Regards,

Ignacio

Hi Ignacio Vazquez Lam1,

Thank you for suggesting an alternative part. I would not have considered OPA2675 because of its package. I guess I will learn how to solder that. So, I gave it a try and downloaded the reference *.tsc file. Then I set up the circuit in Tina-TI with two inverted amplifiers in bridged configuration and a 2 Ohm load. I sticked with single supply Vcc = 5V and consequently VBias = 2.5 V (created by a voltage divider from Vcc).

The control voltage Vcontrol is from (Vbias - 1.5 V) to (VBias + 1.5V). To make calculations easier to understand, I only plotted the positive branch, e.g., Vcontrol = 2.5 V .. 4 V.

It seems this device can easily dump 100 mW (and more) into the load: At Vcontrol = 4V, the voltage across the load is 440 mV at  220 mA. See the simulation, file is attached.

Note, my calculation for the gain is a bit off. I wonder why this is. To reach a voltage of Vload = 0,44 V across the load at Vcontrol = 4 V, I calculated a gain of:

G= Vload/(2 * (Vcontrol - VBias)) = 0,22 V / 1.5 V ~ 0.147 

However, the simulation shows that Vload = 0,33 V is reached. Note that the diagram states UPeltier instead.

I had to increase the gain (set through Rg) to about G= 0.2 to reach Vload = 0,44 V at Vcontrol = 4 V in the simulation.

Can you explain this difference?

Thanks
Daniel

45 BridgeInverted OPA2675 Vcc=5.0V Vcontrol=1 .. 4 V - 4TI.TSC

Hello Ignacio Vazquez Lam1,

while your comments basically answer my question, one more challenge came to my mind. That is the power-up situation.

During power-up, I would like to avoid having uncontrolled high currents running through the load (TEC). 

To achieve this, I'm considering using  OPA2675's "bias mode control" pins. As I understand it, if they are LOW, the amplifier outputs are disabled. 

I plan to switch on VCC and set VControl to mid-scale first before switching "bias mode control" pins to HIGH (which in turn enables the amplifier outputs).

Does that makes sense to you? Will that work? Other solutions?

I wanted to make a transient analysis of this scheme using TINA-TI but the OPA2675 model does not provide the bias mode control outputs (note that the additional pins in the model are named "PD" and "REF" and are already connected to GND during above simulations).

Thanks.
Daniel

Hi Daniel,

I believe powering on the device (Vcc) and letting things settle before turning on Vcontrol would work without having to switch bias modes. However, biasing the device from off to on in between those steps should not be a problem. A good precaution would be to have this option in your layout that allows you to switch bias modes and see if this is ultimately needed.  We share the transient behavior of this switching from offline to full bias on Figure 7-20 in the datasheet. I would like to add that high is powering off the amplifier and low enables the amplifier. Section 8.4 has the logic table for this feature.

Best Regards,

Ignacio

Hello Ignacio Vazquez Lam1,

Thanks for pointing out the correct "polarity" of the bias control signals. Good catch! I appreciate your attention to details. I immediately changed my schematic accordingly. Do you know whether and how the two pins in the "OPAx675" TINA-TI model  "PD" and "REF" relate to the bias control pins "A0" and "A1"?

Regarding powering up the amplifier,  I'd rather not switch on VCC "before turning on Vcontrol" as that means that Vcontrol is likely at GND potential and that will drive the maximum current through the load until Vcontrol settles at mid scale. The load is a tiny TEC connected to an even tinier $15k laser diode. I'd rather not stress it with uncontrolled heat or cooling. With that said, I added a MOSFET to switch Vcc and also connected a GPIO to the bias control signals. So I can try different schemes for starting up the circuit.

Thanks for your continuous support.
Cheers
Daniel

Hi Daniel,

I was able to look through the information we have on the model, and unfortunately, I could not find any information regarding the use of these two pins. Therefore, I am not sure these pins model the real devices behavior when in 50% and 75% bias mode. In general, with most devices that feature different biasing configurations that affect the performance, we recommend customers to test this with the actual device as our models would likely not capture these performance changes as accurately. I do see your point about the startup condition, trying different startup schemes would be ideal in this case. 

Best Regards,

Ignacio