2
\$\begingroup\$

I'm trying to design a circuit that allows a micro controller (arduino) to control a voltage to a load. The load voltage needs to be bipolar in nature (+- 3 to 5 V) for use in controlling a TEC.

My current design idea is to take the output of the micro controller as PWM and lowpass filter it to obtain a voltage between 0 and 3.3 V. From there, I plan to level shift it to +- 3 V via input to a BJT as shown below:

schematic

simulate this circuit – Schematic created using CircuitLab

When this portion of the circuit is tested alone, it appears to deliver the correct characteristics (50% PWM duty cycle gives 0 V, the average of many switches between + and - 3 V, varying the duty cycle increases/decreases this voltage up to the supply rail voltages in a proportional manner).

From there, the input is fed into a simple push-pull amplifier topology feeding off the same bipolar supply as the level shifter, to drive the load (load current in the region of 1 Ampere desired).

schematic

simulate this circuit

When the level shifter and low pass filter is connected to the push-pull amplifier, it becomes loaded and does not give the desired proportional voltage control I am after. I've tried moving the lowpass section to before and after the level shifting transistor, but in that configuration I have difficulty driving enough current through the base of the transistor to get it to switch (current amplifier needed?).

How would I best go about resolving these issues with the given (rough) topology? I've considered using an op-amp as a buffer for the low-pass stage, but I've only access to 741s which won't do for the +- 3 V power supply (Maybe I could increase the supply voltage - the driven load is a TEC device).

Is it possible to buffer the lowpass section output with a transistor simply? Or do I move the lowpass section to before Q1 and somehow build a current amplifier?

Any input or suggestions on overcoming this problem would be much appreciated, thank you.

I have access to only a single bipolar bench power supply and BJTs (no FETs, unfortunately) (along with some LM741 op-amps, and a selection of resistors, capacitors and inductors, but nothing fancy or integrated (might be able to squeeze some 7400 series logic AND gates for doing cheap level shifting if required).

Please note, the component values used in these schematics are for rough description only.


I've implemented Spehro's buffer suggestion, and it appears to be working as desired. Can anyone make any suggestions as to how I could further improve this circuit? (Q1 current seems a little high, ~100 mA, but only 75 Ohm on the collector seems to bias the RC lowpass output to around 0 V at 50% PWM duty)

The circuit is drawn up in LTSpice for simulation here.

Below is the circuit drawn up in CircuitLab, if that's easier to see.

schematic

simulate this circuit


Update: I've added some biasing resistors to the emitter-follower section and tweaked stuff a bit in LTSpice in order to get the system as linear as possible from the PWM input to output. I'm not quite sure if what I'm doing is correct, but it seems to be better than what I've been doing before.

LTSpice circuit linked before has been updated to match if that's easier to simulate.

schematic

simulate this circuit

\$\endgroup\$
  • \$\begingroup\$ Why such low values for R2,R3? You should be able to increase them by an order of magnitude at least to reduce Q1 current (and scale the load, i.e. the LPF, appropriately if needed. In one version R4=200k and wouldn't need scaling) \$\endgroup\$ – Brian Drummond Oct 8 '14 at 9:00
3
\$\begingroup\$

You can try buffering the output of the filter with a pair of darlington emitter-followers, which will replace the four diodes with four transistors.

schematic

simulate this circuit – Schematic created using CircuitLab

You can try buffering the output of the filter with a pair of darlington emitter-followers, which will replace the four diodes with four transistors.

Check that the no-load bias current in the output transistors is not too high.


It might be easier to use an op-amp though, if you can get some that are suitable.

\$\endgroup\$
  • \$\begingroup\$ "It might be easier to use an op-amp though" Yes, I agree with this most of all! A dual-rail op-amp seems like the best idea \$\endgroup\$ – KyranF Oct 7 '14 at 8:24
  • 1
    \$\begingroup\$ @KyranF I fear this is one of those questions that is predicated on the availability of components in the one shop just outside the wet market in some town in Arunachal Pradesh. \$\endgroup\$ – Spehro Pefhany Oct 7 '14 at 8:33
  • 2
    \$\begingroup\$ haha well there is nothing wrong with making amazing scratch built circuits with random parts \$\endgroup\$ – KyranF Oct 7 '14 at 8:33
  • \$\begingroup\$ This will probably work, but it's really not the best idea if you want to be efficient. If you're starting with PWM signal, you might as well drive the high current output transistors with the PWM signal directly as opposed to using linear mode. Otherwise, you will need a VERY large heatsink. \$\endgroup\$ – alex.forencich Oct 7 '14 at 8:34
  • 1
    \$\begingroup\$ Yes, Sphero. I'm somewhat limited by the components available to me (BC55X NPN/PNP signal transistors and some TIP41/42 power transistors, LM741s and a selection of resistors, capacitors and inductors, but nothing fancy). And that's an amazingly specific description of the source of the components (I'm located quite a distance from there though :P) \$\endgroup\$ – Lucas Oct 7 '14 at 8:51
0
\$\begingroup\$

If you want to drive a TEC from a microcontroller, I would actually suggest a slightly different technique. Take a look at the MAX1968/MAX1969. This chip is a TEC controller. It drives a TEC with two DC to DC converters. It's basically an H bridge with LC lowpass filters on both sides of the TEC, along with current sense and control circuitry. Duplicating the drive circuitry that this chip uses should work well in your application. It is quite power efficient due to the converters, and it does not require a split rail design. The circuit that you have in mind is going to be horribly inefficient and it will require an absolutely gigantic heatsink to deal with the power dissipation in the output transistors.

Take a look at he datasheet here: http://datasheets.maximintegrated.com/en/ds/MAX1968-MAX1969.pdf

Page 10 has a schematic that shows the integrated drive transistors and how they are connected to the TEC. You should be able to duplicate this configuration without too much difficulty. The chip uses FETs internally as drive transistors, but BJTs should work just as well. And the power dissipation will be far less as the output stages will be either on or off, not in linear mode.

\$\endgroup\$
  • \$\begingroup\$ Thanks for the contribution, unfortunately I'm limited by the parts available to me (A selection of small signal transistors, E96 range (roughly) resistors, some capacitors and inductors and a few LM741s). I took a look at p. 10 of the data sheet for the MAX1969, but I don't think I'll be able build anything that complex for this project. Efficiency isn't super critical for this project (only ~4 W TEC needs to be driven, but I do have heatsinks on hand and some beefier transistors for the output stage). \$\endgroup\$ – Lucas Oct 7 '14 at 8:46
  • \$\begingroup\$ Well, don't duplicate the whole thing, just the output stage. If you unwrap it, it is four transistors, two inductors, two capacitors, and a current sense resistor. Delete the resistor if you don't need current feedback. You can also delete the inductors and capacitors and just drive the TEC with a simple H bridge like a DC motor if you really want, though it will not work quite as well as with the filtering. \$\endgroup\$ – alex.forencich Oct 7 '14 at 8:51
  • \$\begingroup\$ Would you mind adding a schematic to explain how this could be done? As for using a H-bridge configuration, how easy would it be to get (roughly linear) proportional output from the PWM? \$\endgroup\$ – Lucas Oct 7 '14 at 8:55
  • \$\begingroup\$ The basic idea is just to build an H bridge and stick the TEC where you would put the motor. Adding LC lowpass filters in series with both sides of the TEC will improve the TEC performance as I don't believe they like being driven at high frequencies. I belive the output will be close to linear, but it might be a good idea to add some sort of feedback, perhaps by reading the voltage on both sides of the TEC or the current through the TEC, feeding this to an ADC on your micro, and then closing the loop in software. \$\endgroup\$ – alex.forencich Oct 7 '14 at 9:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.