Re:load: A simple, flexible adjustable dummy load

[nickjohnson]

I've been testing Loki's power supply lately, and hurting for the lack of a good dummy load. There are plenty of flexible dummy load projects out there, but they're largely overengineered and on the expensive side. I decided to put together a really simple but flexible dummy load.

Re:load is an adjustable constant current load with the following properties:

• No external power supply required - powered by the device under test
• Wide range of input voltages, from 3.3 volts to 32 volts
• Adjustable load from 0 to 3.5 amps
• Up to 14 watts power dissipation (with design heatsink)
• Virtually indestructable: The power FET, BTS117, has built in overtemp, ESD, and overcurrent protection
• Load remains constant under different input voltages - 40 milliamp variation over input voltage range
• Screw terminal and banana plug footprints
• Low BoM cost, and easy to solder thru-hole parts
• Test points for reading current with a voltmeter

Feedback and suggestions greatly appreciated. I plan to send off a PCB order for a prototype shortly.

[matseng]

The lm358 have an offset voltage up to 7 mV if I remember correctly. This could cause the output current to be off by 0.1 amp. But since this design is set manually by a trimpot it might not matter.... If the current was set by an external voltage or a microcontroller/dac it would have been a problem.

The gain of the sense amplifier is set to 3 giving an output of 150 mv per amp - not the 100 mv that the pcb silk says. Wouldn't a gain of 20 giving 1 volt per amp be easier?

[sqkybeaver]

carefull- the LM358 does not bring the output fully to lower rail, if you plan on using this in the lower limits you may not be able to get it all the way down.

[matseng]

Sqkybeaver: I don't think that the lack of true rail-to-rail will affect this circuit. The the fet probably need several hundred millivolts at its gate before it starts to conduct in the linear region.

[sqkybeaver]

Sqkybeaver: I don't think that the lack of true rail-to-rail will affect this circuit. The the fet probably need several hundred millivolts at its gate before it starts to conduct in the linear region.

true, as long as the fet is in in the linear region no problem there. i have run into problems in where the circuit would oscillate because of using fets in low voltage applications in similar circuits.

I suggest switching to a bipolar darlington, the bts117 is a smart low side power switch, designed to replace mechanical relays.

[matseng]

I suggest switching to a bipolar darlington, the bts117 is a smart low side power switch, designed to replace mechanical relays.

I thought so too, considering all the extra internal circuity that they have put into the bts117.
But they actually say this:
Application
All kinds of resistive, inductive and capacitive loads in switching or [i]linear applications
[/i]
http://www.infineon.com/dgdl/BTS117_DS_ ... 6637844df3

It is kinda funny device.... It will pull the gate to ground to signal that one of the protection circuits has been activated so you can monitor it from a microcontroller.

[sqkybeaver]

it is definitely not your jellybean fet. it probably cost more than your average transistor.

if you plan to read current with a dmm, using the opamp as a buffer is unnecessary, switching to a 0.1 ohm load resistor would make it easy to convert. than the second opamp could be used as a comparator to pull the gate(or base) low in an over temp condition, and would only require an additional thermistor.

there should be over voltage protection for the op amp too.

[nickjohnson]

Thanks for all the feedback. In no particular order:

- The gain on the second opamp was a mistake; I'll fix that. Thanks for pointing it out.
- The output voltage range of the LM358 does turn out to matter at low input voltages, since the FET requires about 3V on its gate to conduct 3.5A. I've decided to switch to a rail-to-rail opamp, powered off the voltage reference (which eliminates the need to find one with a high voltage range).
- I picked 0.05 ohm for the shunt because I want to dissipate as little power as possible in it, to ensure that only the FET requires heatsinking. I could probably get away with 0.1 ohm here, but I'm not sure if it's worth making the change or not.
- The BTS117 was a deliberate choice because of its built in overtemperature protection, which eliminates the need to separately sense and handle overtemperature.

I also need to add reverse polarity protection for the opamp circuit; probably a reverse biased FET for the low voltage drop.

[nickjohnson]

Building in reverse protection is proving tough. Diodes have too much voltage drop; PFETs are all large, expensive, and don't support high Vgs. Using an NFET would raise the opamp ground above gnd, impeding rail-to-rail operation. I think I have a couple of practical options:
1) Leave it out. Label terminals clearly.
2) Use a schottky diode, and acknowledge that it will have limited current capability at 3.3v.

Ideas?

[nickjohnson]

Okay, a big rework. I've replaced the voltage reference with an LDO regulator, the LP2951, which has an astonishing 50mV dropout at low load. The opamp is now powered off that. I've also added a schottky diode for reverse polarity protection for the opamp, which should make this circuit pretty close to indestructible if you don't go over 30 volts.

With the diode and LDO, at 3.3v the power rail will be slightly under - somewhere between 2.85 and 3 volts - which means the maximum load will be lower, but almost certainly at least 2A, maybe higher.

[Bertho]

You may also need a way to calibrate the output current reading. The contact resistance in the solder-joints and traces may influence your reading negatively. A 1mOhm deviation on a 50mOhm sense resistor is already a 2% deviation.

You are measuring not exactly over the sense-resistor and the feedback vs. sense measurement are physically offset.

[nickjohnson]

You may also need a way to calibrate the output current reading. The contact resistance in the solder-joints and traces may influence your reading negatively. A 1mOhm deviation on a 50mOhm sense resistor is already a 2% deviation.

You are measuring not exactly over the sense-resistor and the feedback vs. sense measurement are physically offset.

That's a good point. I've literally run out of space to put a trimpot, though.

I sort of added the current readout for convenience and because there was a spare opamp, but it would still be preferable to be able to make sure it's accurate.

[Kean]

LP2951 or LP2950? The former is 8 pin DIP/SO, the latter is TO92.

Make sure you put the Schottky diode before the regulator - most regulators don't handle reverse polarity very well, or even having their output pulled above their input (sometimes caused by bulk storage caps not draining quickly enough).

A Kelvin style connection to the sense resistor is best. Even better to use a sense resistor with Kelvin connections. But just do the best you can!

Kean

[nickjohnson]

LP2951 or LP2950? The former is 8 pin DIP/SO, the latter is TO92.

You're right, I meant the LP2950.

Make sure you put the Schottky diode before the regulator - most regulators don't handle reverse polarity very well, or even having their output pulled above their input (sometimes caused by bulk storage caps not draining quickly enough).

That's how it is in the current configuration.

A Kelvin style connection to the sense resistor is best. Even better to use a sense resistor with Kelvin connections. But just do the best you can!

I don't think there are any PTH 0.05 ohm 1 watt resistors with Kelvin connections, unfortunately. And I'd like to use a Kelvin connection, but I'm not sure I can route the tracks. :)

I'm still of two minds about the trimpot. I can just about fit an SMD trimpot on the board, but that means introducing an SMD part, and in a rather cramped position. Or I can just acknowledge some inevitable error in the test point readout - it's designed for power, after all.

[Kean]

Ah, yes - I see you've updated the design above.
Don't forget to change the label of the opamp from to MCP6002. So nice that many opamps have common pinouts.