Author Topic: Derpy Load: Is my dummy load design any good? (Read 20083 times)

LEECH666 · « **on:** April 12, 2012, 11:52:15 am »

Hi,

I want do build an adjustable dummy load just with higher rated sink capability.

Let's say I want to be able to test a PSU that is rated at 30V / 5A (150 Watts).

It was suggested in this thread http://www.mikrocontroller.net/topic/168218 (sorry it's in German) to just parallel a few instances of the FET and OP amp dummy load circuit to increase the current sink capability.

Circuit diagramm from the mentioned thread:

This circuit is similar to the one shown by Dave in "EEVblog #102 - DIY Constant Current Dummy Load for Power Supply and Battery Testing".

Anyway I came up with the following circuit (attached PDF file) diagram for a dummy load and would like to hear your thoughts on it. It's aimed to be as simple as possible. I left out the decoupling caps and power supply for the OPs on purpose, and I am aware that 150W are a lot of power and as such would need a pretty beefy heat sink. I am thinking about using PC coolers for this task.

Oh and the components aren't final (FET, OP, etc.), I really just threw together this shematic during lunchhour.

Any reason why it wouldn't work this way?
Will it explode the world or at least melt my work bench?

Cheers from Germany,
Florian

amspire · « **Reply #1 on:** April 12, 2012, 12:42:50 pm »

Its basically OK.

For 30V, there are some things I would change. Many MOSFETS are rated at 20V gate-source voltage, and transients on the drain will drag the gate high or low, so I would limit the gate drive to between 5 and 10V (depending on the gate sensitivity).

My advice is to make it so it can run from a 9V - 12V battery, so I would have a 5V regulator and run the opamp and the voltage on the divider. Change R27 to 560 ohms.

Check the stability and increase C18 up to 1nF if necessary.

Now for the MOSFETS, to get to 150A, you will either need a very big heatsink or a good fan cooled heatsink. Many of the PC processor ones are very good. Ideally you don't want to let the heatsink get above 75 degC, so you don't burn yourself. If you only need a minute or so under full load, then a heatsink with good thermal mass will keep the temperature down. There are CPU coolers that can achieve an incredible 0.1 degC/W temperature rise and you are not going to get close to that with a convectional aluminium heatsink, no matter how big you make it.

You will probably need several MOSFETS in parallel. Look at the datasheets. A MOSFET may be 150W at a case temperature of 25 degC but check how much it can handle when the case is 100 deg C. It will be under SOA or Safe Operating Area in the datasheets. Don't exceed this number. You also have to take into account MOSFET case to heatsink thermal resistance and multiple MOSFETS will reduce the effect of that, particularly if you want the MOSFETS with thermal insulation from the heatsink. When you assemble it, remember you only use a thin smear of thermal compound on the MOSFET - thermal compound is a lousy thermal conductor so you want it as thin as possible.

Richard

LEECH666 · « **Reply #2 on:** April 12, 2012, 01:11:18 pm »

Yes, thanks for your reply.

In the original mikrocontroller.net forum it was discussed, that paralleling just the mostfets might be a bad idea because of current sharing issues One of the FETs might have a higher hFE than the other and then that FET would take the majority of the current, unless you used current share resitors. IIRC the current share resistors are deemed bad too cause of the voltage drop. That's why it was suggested to parallel the circuit up, with each fet driven by their own op amp.

I had the fear that it was a bad idea to include that picture. The picture in my first post, is not my design. It's designed by ELV (a german kit reseller). I just included it as a reference, as it was suggested to parallel up two of these circuits to achieve doubled wattage.

My question was actually about the attached PDF file, which is drawn by me. It's basically Daves dummy load paralleled up. An adding op amp is used to add up the voltages over the sense resistors to reflect the total current sinked by all four "sink modules" or stages or however you want to call that. The resulting (added) voltage is then fed to the panelmeter. It's probably a good idea to add another inverting op amp (gain = 1) to change polarity. Or maybe I could just switch the positive lead with the GND lead so that I don't get a negative readout on the panelmeter.

I attached a new version with the inverting op amp added, plus I fixed a small mistake.

Cheers from Germany,
Florian

amspire · « **Reply #3 on:** April 12, 2012, 01:29:59 pm »

Florian,

I did actually throw together a dummy load recently. Some things are a bit odd - that is just the way the circuit was when I decided it was working well enough.

https://www.eevblog.com/forum/projects-designs-and-technical-stuff/switching-pre-regulator-gp-power-supply/

The important things you need from this circuit is there has to be a resistor from the output to the MOSFET gate, along with the reverse biased diode on the opamp output. A Schottky diode would be better then the 1N4004 I used, but the reason that it is there is if the voltage on the drain drops quickly, the voltage on the gate will go negative, and that can latch up the LM324. The resistor is need to protect the LM324 output, and to stop the MOSFET from wanting to oscillate.

The resistor on the input is needed as you will definitely need a compensation capacitor across the opamp, and for the compensation to work, you need the resistor.

Also i would modify your meter circuit. I am not sure why you need the opamps. Just feed the summing point of the resistors into the meter. You can make the summing resistors smaller if you want - 1K would be fine.If you want an extra opamp just to make it easier to calibrate, use it in a non-inverting configuration. The summing resistors go to the + input. The negative input has a 10K to ground and a 1K to the output to give it a gain of 1.1. Then have a pot on the output going to the meter for calibration.

Otherwise, it looks good.

Richard

LEECH666 · « **Reply #4 on:** April 12, 2012, 03:08:20 pm »

Thank you amspire, that is exactly the kind of input I expected. I am a noob with FETs, I don't really know much about them. Hell I am a noob with most electronics, I fail to see the bigger picture in most circuits.

The summing amplifier was just the easiest circuit I could think of (read as "circuit that I am most familiar with") to sum together voltages. The idea was to put as little loading as possible onto the 1R sense resistors of each dummy load subcircuit. Wheather that makes sense or not ... well I am not sure.

I will try to implement your suggestions.

I think I'll have to read your post a few more times tho. But that is more of a language thing.

Oh and I am not really set on any of the main components like the FET or the op amp. The LM324 was just the first op amp I found in the library of the CAD system I use at my work place. I think I read somewhere that a rail-to-rail op amp might let me set lower (higher?) load currents, so maybe it's wise to go for such an op amp.

Cheers from Germany,
Florian

amspire · « **Reply #5 on:** April 12, 2012, 03:26:52 pm »

The LM324 is probably the easiest to get opamp, and rail-to-rail opamps offer no advantages.

The LM324 inputs go down below the negative rail, and you do not need the inputs going to the passive rail. If you put a 4k7 resistor from the output to the 5v rail, the output will go to the positive rail.

You never needed the opamp output to go to the negative rail.

Richard.

alm · « **Reply #6 on:** April 12, 2012, 04:45:10 pm »

Quote from: LEECH666 on April 12, 2012, 01:11:18 pm

In the original mikrocontroller.net forum it was discussed, that paralleling just the mostfets might be a bad idea because of current sharing issues One of the FETs might have a higher hFE than the other and then that FET would take the majority of the current, unless you used current share resitors. IIRC the current share resistors are deemed bad too cause of the voltage drop. That's why it was suggested to parallel the circuit up, with each fet driven by their own op amp.

Are you sure you're not confusing MOSFETs with bipolar transistors?

T4P · « **Reply #7 on:** April 12, 2012, 06:06:12 pm »

Quote from: LEECH666 on April 12, 2012, 01:11:18 pm

In the original mikrocontroller.net forum it was discussed, that paralleling just the mostfets might be a bad idea because of current sharing issues One of the FETs might have a higher hFE than the other and then that FET would take the majority of the current, unless you used current share resitors. IIRC the current share resistors are deemed bad too cause of the voltage drop. That's why it was suggested to parallel the circuit up, with each fet driven by their own op amp.

It's the transconductance not hFE

LEECH666 · « **Reply #8 on:** April 12, 2012, 06:31:00 pm »

That's probably my fault, I just wrote hFE. As I said, I don't know much about FETs.

Dave uses a logic level MOSFET in his dummy load circuit. I am not sure if I can just use any MOSFET or if it's mandatory to use a logic level one.

The IRLIZ44N is in the suggested standard parts list of mikrocontroller.net and seems to bemore than up to the task.

It's a logic level power MOSFET rated at 30A / 55V with 22mohm RDS_on in a TO-220 case. Comes in at 0,81€ per piece at my prefered parts store (www.reichelt.de).

Cheers from Germany,
Florian

T4P · « **Reply #9 on:** April 12, 2012, 07:06:27 pm »

A single big arse TO-3P/TO-247 package mosfet with a big heatsink meant for a pentium 4 would do the job ...
Like richard's dummy load . No paralleling , no need to deal with big resistors .

amspire · « **Reply #10 on:** April 12, 2012, 07:33:54 pm »

Quote from: Dave.S on April 12, 2012, 07:06:27 pm

A single big arse TO-3P/TO-247 package mosfet with a big heatsink meant for a pentium 4 would do the job ...
Like richard's dummy load . No paralleling , no need to deal with big resistors .

Not saying it is impossible, but it is pretty hard. I would use two MOSFETs at least myself.

Richard

alm · « **Reply #11 on:** April 12, 2012, 08:58:36 pm »

(Lateral) FETs can often be safely be paralleled without series resistors because of their positive tempco, several FET manufacturers have appnotes on this topic.

Note that a low RDS(on) won't reduce dissipation in this application, since the primary purpose of a dummy load is to dissipate energy. Might as well use a copper wire if you want to short the power supply with the lowest possible on resistance.

LEECH666 · « **Reply #12 on:** April 12, 2012, 09:39:16 pm »

RDS_on is probably more important when you want to switch on big loads. I was probably thinking about using it as a switch earlier.

I tried to work in some of Richards suggestions ... hope I got it right.

It's bedtime for me now.

Until tomorrow,
Florian

T4P · « **Reply #13 on:** April 12, 2012, 10:26:57 pm »

ONLY if you want to switch on big loads , but it makes no difference if it's working as a resistor .

LEECH666 · « **Reply #14 on:** December 03, 2012, 12:49:34 pm »

Blackdog suggested in this post that the IRFP064N might be a good candidate for a power MOSFET in a dummy load. I checked the datasheet for the device and VGS(th) max seems to be around 4V. I reckon from the threshold value that this FET could be used as a logic level device then? (Even tho it's not specifically called a logic level MOSFET in the datasheet).

Cheers,
Florian

LEECH666 · « **Reply #15 on:** December 05, 2012, 08:57:15 pm »

Hmm, just wondering, would any sort of polarity protection circuit make sense in this circuit?
I am not sure what would happen if I hooked up the 30 volts in reverse.

Kevin.D · « **Reply #16 on:** December 05, 2012, 09:50:54 pm »

I would just put a large fuse inline to limit reverse current .Since you already have a reverse parasitic diode in the Fet.

LEECH666 · « **Reply #17 on:** December 05, 2012, 10:24:47 pm »

I came across this video and it explained the P-FET protection in a little more detail.

P-FET Reverse Voltage Polarity Protection Tutorial

Anyway I put it into my circuit in LTspice and for some reason I didn't realize that I hooked it up back to front (drain switched with source). But now it seems to work well. Component values are not final ones. LTspice never seems to have the parts I want to use and I really can't be bothered with importing some half assed spice model that only gives me errors 'n shit. Really, importing new spice models seems to be a bit bothersome.

I attached the LTspice file if anyone wants to check it out. (Sorry had to zip it, this board doesn't allow .asc files)

One thing I noticed is, that my circuit wouldn't allow me to go down to 0 Amps without a negative supply on the opamps (at least in the simulation).
I guess I could design a -12V inverter based on the MC34063.
(Thanks Dave for showing us all this stuff!)
I planned on using the LM324 in my design. Open for suggestions tho, if this is a bad choice. Maybe a rail-to-rail opamp is more suitable and doesn't need a -12V supply?

I also need to find out how much current my fan units (CPU coolers) are drawing to see if my 12V/1A wall wart is powerful enough to drive all 4 fans. But that will have to wait till later. For now it's bedtime in Germany.

Cheers, Florian

Spikee · « **Reply #18 on:** December 06, 2012, 12:21:46 pm »

Quote from: LEECH666 on December 03, 2012, 12:49:34 pm

Blackdog suggested in this post that the IRFP064N might be a good candidate for a power MOSFET in a dummy load. I checked the datasheet for the device and VGS(th) max seems to be around 4V. I reckon from the threshold value that this FET could be used as a logic level device then? (Even tho it's not specifically

In my IRFP064N mosfet scatter measurements you can see that the fet will let trought about 1A close to 4V at the gate.
https://dl-web.dropbox.com/get/Public/mosfet%20scatter.pdf?w=c21fdda0
To reach the max i use ~11V at the top rail of my LM324N to achive that. I will mount one of those fets on my big coolermaster z600 heatsink to see the temps i will achieve using one mosfet at 90W. To see if it is viable using pc heatsinks.

I probably have to lasercut a bracket to hold the resistor and mosfet into place on the heatsink.

LEECH666 · « **Reply #19 on:** December 06, 2012, 12:41:11 pm »

Well as you can see from my circuit diagram my approach is a little bit different. I am aiming for a modular / expandable system. Each "unit" will only have to sink ~1.25 ampere to keep the generated heat low (30V * 1,25A = 37,5W). I am planning to use the fairly cheap Arctic Cooling Alpine 11 Plus CPU coolers. They are around 12€ here and are rated for CPUs up to 100W (that's what they claim, no datasheet available :s). Probably not the most cost effective approach but meh, I am a noob with FETs and it's only for hobby use in my case.

I found this page very interesting. Sorry, once again it's in German, but maybe you can run it through google translate to get a generic idea what's going on on that page. Or maybe you even speak German.

http://dareal.info/test/

Take a look at his pictures, they are fairly interesting.

Cheers,
Florian

Spikee · « **Reply #20 on:** December 06, 2012, 12:57:10 pm »

If i'm correct the Alpine 11 PLUS does not have heatpipes what will limit the power dissipation because you can't transfer the heat from the mosfet fast enough into the heatsink making which will result in a mosfet that will be much hotter than the heatsink.
The freezer 7 series seems a better fit , it is larger but it has good heat pipes. They go second hand for around 5 euros and new for 15 euros.
Ill have to experiment myself whit some coolers and i will order a freezer 7 pro soon to do some experimenting.

gr. spikee

LEECH666 · « **Reply #21 on:** December 06, 2012, 01:04:54 pm »

Yeah the Alpine 11 Plus doesn't have heat pipes at all.

http://www.arctic.ac/en/p/cooling/cpu/461/alpine-11-plus.html?c=2181

I don't mind letting the magic smoke out of some FETs in the process of building this up

.
Building a test circuit and testing the heat sinks is on my todo list.

Cheers,
Florian

Spikee · « **Reply #22 on:** December 06, 2012, 01:07:55 pm »

i just ordered an artic cooling 7 pro to test some fets . Only 5 euro =) shipping is 7 euro lol ...

LEECH666 · « **Reply #23 on:** December 06, 2012, 01:18:56 pm »

Let me know the results.

Cheers,
Florian

LEECH666 · « **Reply #24 on:** December 06, 2012, 07:34:09 pm »

M'kay, fans draw about 200 mA each @12V.

With the initial planned 4 sink units this will leave me with ~200mA for the op amps.
Should be enough ...

T4P · « **Reply #25 on:** December 06, 2012, 07:36:48 pm »

Sliver TIM again? That stuff is quite electrically conductive no matter what they say

skipjackrc4 · « **Reply #26 on:** December 06, 2012, 11:15:15 pm »

Arctic Silver Ceramique is supposedly non-conductive, so you may look into that. I am eagerly awaiting your results as I plan to do something similar in my upcoming load design.

What would everyone's opinion be on sandwiching the transistors between 2 CPU coolers?

tom66 · « **Reply #27 on:** December 07, 2012, 12:43:03 am »

Tests of various thermal pastes:
http://dansdata.com/goop.htm

You may be surprised

LEECH666 · « **Reply #28 on:** December 07, 2012, 02:08:29 am »

Always used the standard white goop for semiconductors so far (Amasan brand, probably not that well known outside of Gemany). Seems to work fine for CPUs and other power semiconductors. The test from the page tom66 linked seem to confirm that the standard white goop is sufficient.

Man why am I awake ... :S

dr_p · « **Reply #29 on:** December 12, 2012, 02:22:32 pm »

Quote from: LEECH666 on December 03, 2012, 12:49:34 pm

Blackdog suggested in this post that the IRFP064N might be a good candidate for a power MOSFET in a dummy load. I checked the datasheet for the device and VGS(th) max seems to be around 4V. I reckon from the threshold value that this FET could be used as a logic level device then? (Even tho it's not specifically called a logic level MOSFET in the datasheet).

Cheers,
Florian

Gate-Source Threshold Voltage (VGSth) is the minimum voltage that allows the FET to conduct ( some!! ) electricity. The datasheet says 2...4V and I think they mean it varies from part to part, but no more than 4V and no less than 2V. With only 2 to 4V it has just crossed in the linear region (not fully on). When VGSth rises some more (rise = the Overdrive Voltage), the FET reaches the fully-on state.

Notice how RDSon is given on VGS=10V, not at 4V?

added on EDIT:

Quote from: LEECH666 on December 06, 2012, 07:34:09 pm

[...]With the initial planned 4 sink units this will leave me with ~200mA for the op amps.
Should be enough ...

I would expect 20-30mA tops.

LEECH666 · « **Reply #30 on:** December 12, 2012, 05:01:16 pm »

Updated schematic ...

say · « **Reply #31 on:** January 05, 2013, 11:04:11 am »

It's been a while LEECH. Have you produced a prototype yet?
I have been struggling to build one myself but the heatsink issues didn't let me proceed further. Would love to hear from you soon.

LEECH666 · « **Reply #32 on:** January 06, 2013, 06:46:14 pm »

Uh there is always some stuff (xmas, new year, health, motivation, etc.) that keeps interrupting me, and I am lazy sometimes. Sometimes I need a push in the right direction.

These two threads are related to this project. Just in case you haven't read them alread and in case you're wondering what I've been doing.

https://www.eevblog.com/forum/beginners/ltspice-frustration/
https://www.eevblog.com/forum/beginners/trouble-adjusting-op-offset-voltage/

Cheers,
Florian

Spikee · « **Reply #33 on:** January 06, 2013, 07:02:21 pm »

Cmon leech finish that thing already

LEECH666 · « **Reply #34 on:** January 07, 2013, 01:20:08 pm »

I knoooooooow, xD.

LEECH666 · « **Reply #35 on:** August 27, 2013, 12:25:31 am »

Okay so I finally* managed to build up a (sort of) prototype on breadboard and it seems to work reasonably well. I've been able to draw 3A at 30V from my Manson NRP-3630 PSU (36V/3A), however I am a bit puzzled by the discrepancy of the 'set' voltage and the current actually drawn from the PSU. For example - at the moment the set voltage is set to roughly 0.33V. If I measure the voltage across my 0.100R shunt resistor, I get the same voltage.

Nevermind, looks like I have some heavy losses on my bread board that probably explain the discrepancy.

Measuring between the GND near the 0.100R shunt and my main GND "node" (a few centimeters away) shows a 90mV drop ...

Guess it's perf board time again ...

(* - Damn you Borderlands 2 and your addictive gameplay!

)


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: Derpy Load: Is my dummy load design any good? (Read 20083 times)

alm

alm

Share me