Yet another DIY Electronic Load

[OM222O]

yes, you can drop in BJTs/ Darlington pairs as direct replacement, no tuning required!
this is a terrible design for multiple reasons:

H2 (Ext.drive) is basically pointless! there is a diode (D2) with non linear AND non constant characteristics with quite horrible temp co in series with it, which creates a variable voltage drop, throwing out all accuracy all together!

Ditch the current sense amplifier weirdness all together! you can't mix signals like that. just use a proper current sense IC on VCC, instead of P2 (Ext. Current measur.)

what is the MCU that you're using? the DACs on those can be quite bad. an external DAC + ADC is always preferred.

DO NOT PARALLEL two fets from one op amp inputs, ESPECIALLY if you want to replace them with BJTs/ Darlington pairs. base threshold voltages would be different enough that one will be basically off, when the other one is carrying all the current!

10 resistors in parallel give you quite good power dissipation on those resistors, but a higher wattage equivalent resistor is often cheaper and takes less space on the PCB. if you use a proper current source IC, you can just use a high power 1ohm resistor and not care about temp co, since current is actively monitored and corrected for in the software.

why do you want to do division by 19? that makes no sense and makes all calculations very difficult? I can't see any advantage in that.

[OM222O]

I also had a look at the fets that you bought. they are awful for linear applications, even kept at 25C (that is ice cold btw!) they are only good for up to about 10 volts for your 2.5A amp goal!



add some de-rating as they get hotter and you're down to nothing!

this is where a proper linear mosfet like the IXTQ14N60P shows its advantages. just loook at the SOA for yourself. it starts at 10V and 1A! it's good for up to 200 volts at 1A DC! that's about 200 watts! to be safe just assume 100V but even that alone is 100 watts! PER FET that is! even at 2.5A you can push 70 volts! again, add some head room and say 50 volts, that's still miles better than what you have.


They cost 3.16GBP a pop though, which is quite a bit more than what you paid for. I think they're worth it however. again, in the hobbyist price range, a few parallel darlington pairs would do better, however you can't just parallel them like you did with mosfets as I mentioned in my previous reply.

[MarkF]

You might want to consider this alternate control circuit.
Note that if you select external control and nothing is connected to the control inputs, R13 will pull down the op-amp input and turn OFF the load.

[pinoccio]

Great remarks and ideas flowing here 

So to sum up recent recommendations:
- use just one op-amp per transistor
- reconsider transistor - proper linear MOSFET or BJT/Darlington
- correct package pinout, doh
- sense both branches / for average current

----

You have now changed the names of the parts to LM324, but you still have the LM339 pin numbers:

Yeah, I just changed part name and didn't pay attention to the pinout %) Will fix that, but for principle discussion, the circuit is clear...

No.   Only one MOSFET per op-amp.

ok

I would use the 2K and 18K resistor divider and adjust your sense resistors for the max current you want.

Will think about that. So to have 0.5V driving voltage for max load of about 2.5A, means just reducing the resistor array to half (0.2R).

You might want to average the sense resistor voltages to get a better load current measurement. 
The MOSFETs are not going to sink the exact same current.

I also thought something like this might be necessary. Hopefully just averaging via those two resistors will work reliably enough. Btw. is 100R not too low to avoid mutual influence?

yes, you can drop in BJTs/ Darlington pairs as direct replacement, no tuning required!

nice

H2 (Ext.drive) is basically pointless! there is a diode (D2) with non linear AND non constant characteristics with quite horrible temp co in series with it, which creates a variable voltage drop, throwing out all accuracy all together!

I'm considering the modified control circuit from previous post.

Ditch the current sense amplifier weirdness all together! you can't mix signals like that. just use a proper current sense IC on VCC, instead of P2 (Ext. Current measur.)

I'd like to read some more info on this.
Why some constructions I've seen online, and based my concept on, use this with (presumably) no problems... It's supposed to improve resolution for ADC (we're reading 1V at max load here), but frankly might not be necessary. Or I could just turn it into a buffer. Or even play with the ADC module gain, but that means tweaking all inputs.

what is the MCU that you're using? the DACs on those can be quite bad. an external DAC + ADC is always preferred.

see first post ... Arduino Nano (so 328p) with ADS1115 ADC and MCP4725 DAC.

10 resistors in parallel give you quite good power dissipation on those resistors, but a higher wattage equivalent resistor is often cheaper and takes less space on the PCB. if you use a proper current source IC, you can just use a high power 1ohm resistor and not care about temp co, since current is actively monitored and corrected for in the software.

Just a quick search on ebay - you get 100pcs of 1R/1W/1% metal film resistors for less than US $2.5, or 1pcs 0.1R/50W/?% for less than US $1.5 ...so there's no point considering price Just other factors like maybe tolerance, construction/PCB design, personal preference, etc.

why do you want to do division by 19? that makes no sense and makes all calculations very difficult? I can't see any advantage in that.

To set the max. driving voltage (thus load current) to 0.265V -> respective amps (sensing at 0.1R shunt), while having good resolution using the full 5V scale (or most of it, that op-amp won't go rail-to-rail).

I also had a look at the fets that you bought. they are awful for linear applications...

Yep, I realize that... well, maybe not that it is actually so bad   But I just picked the same one as in Daves video example (see first post).
Seems like I will have to use something else in the end.

You might want to consider this alternate control circuit.

I definitely will


++
Something to consider using BJTs: https://www.eevblog.com/forum/beginners/bjt-vs-fet-in-dummy-load/

[OM222O]

to get around the base current error, you would use a darlington pair. they are usually rated for a gain of 1000 at minimum which is about 1mA per 1A that they pass. that's 0.1% error! plenty good enough in my books, especially given that you are using LM324 with up to 3mV input offset voltage. 3mV divided by 0.1R (your shunt resistors) is 30mA error  much worse than the base current error. that is also another reason to use a higher value resistor such as 1R. then 3mV offset error will be 3mA instead (3mV/1R = 3mA). if you desperately want low voltage operations (not quite sure why you would want to load something less than about 1V?) or very low currents (less than 10mA,at those low currents the op amp will supply all the current, not the source that you are testing, since the darlinton pair might not turn on) you could use a seperate channel using a mosfet (doesn't have to be linear since it's only used for low voltage / low current ranges) but if you want to do anything reasonable (at least 100mA load / more than 1V supply) that won't be needed.

regarding the current sense circuit you used:
a proper mixer must be inverting:

all sorts of weird things happen when you do a non inverting mixer. In theory it works fine, but I'm not comfortable doing it, just in case murphy gets me with that specific setup. all in all, a current sense amplifier (if you want to thinker the least, something fully digital like the INA219, but there are tons of different versions with / without integrated shunt and special features, just search T.I. current sense amplifiers section for example) will give you a better reading, since the 0.1R shunts that you use are going to heat up and decrease in value, therefore your current value can be off by that amount. no sense in using a 16 bit ADC when your numbers are all over the place by design. a current sense amplifier will also give you a total current value, rather than some odd averaging method. it's just better in every way than your current method (probably cheaper too!).

A side note on the ADC: If you haven't bough the ADS1115, order an ADS1219. 1$ price difference but much much better in a lot of ways. you should also consider some filtering for the ADC inputs, I don't see any on your design.

To set the max. driving voltage (thus load current) to 0.265V -> respective amps (sensing at 0.1R shunt), while having good resolution using the full 5V scale (or most of it, that op-amp won't go rail-to-rail).

please tell me you're not powering the DAC from the same 5V supply as the arduino 
1) if you use the usb supply, it's actually closer to 4.75v rather than 5V
2) if you use an external supply (and LDO or similar) they can have quite bad accuracy too. also any switching noise anywhere on the 5V line will change your voltage reference which is bad!
that is why you must use DACs with proper voltage reference ICs.
There still isn't a point in doing weird stuff like division by 19 to get exactly the full range. you can do division by 20 for example and lose just a small amount of the maximum range, but making the math a lot easier.

[pinoccio]

Some improvements possibly...


Never worked with op-amps really, so the summing amp is just a first blind shot for now...

[OM222O]

much better! you also won't have an issue with the inversion since the ADS1XXX family are actually bipolar (ADS1115 is effectively 15 bits, 16th bit is the sign bit. same for ADS1219: it's a 23 bit ADC + 1 sign bit). you just need to be careful at the software stage about positive and negative being inverted.

[MarkF]

The summing op-amp configuration you now have for the current sense is INVERTING.
This WILL require a +/- power supply for the op-amp.  A single positive power supply will not work.

You might also consider the IRFP064 or IRFP250 MOSFETs.  Much bigger package (TO-247) and power handling.

Edit-  And since you raised your max sense voltage to 2.5V, you will need a higher op-amp supply voltage than +5V.
Probably +/-12V would be a standard.

[OM222O]

Can you please make a copy of your schematic and make it public? I think there is still a lot left to be desired in terms of component selection.
Just let me know:
Are you willing to do SMD soldering?
Are you using the atmega 328p chip or an off the shelf aruino nano?

[JeanLeMotan]

I ended up using the IXTQ42N25P mosfet in my load. It has a SOA for DC and I tested it with 150W so far (30V 5A) without any issue.
https://www.distrelec.biz/en/mosfet-250-42-300-to-3p-ixys-ixtq42n25p/p/17111086?queryFromSuggest=true

My first tests were with a IRFZ44N and the much beefier (according to the datasheet) IRFP3006 but both kept popping like balloons above 12-15V and 2-3A, way below their "spec" according to the datasheet. The issue was that I was using them in the linear region which is not covered by the datasheet. So not good for a load.

Do you intend to go down to 0A? Due to the opamp offset voltage of the LM324 (3mV max, 2mV typical) and your shunt value (0.5 Ohms) you'll be able to get down to a min of 6mA.

Your choice of shunt resistors will probably mean that they will share a lot of the load with the MOSFET.  At 5A and 10V they will dissipate 25W, half of the total load power - so you have to consider 50-100W resistors and keep them really cool.
I really recommend going down to 0.1 Ohms for the shunt - but in this case the LM324 might be too noisy and the offset voltage will be amplified by 10 (so for a 3mV offset voltage, you get a min current of 30mA through the load).

[OM222O]

That 5A will be shared between 4 resistors! That's just over 1amp per resistor. So less than 2 watts power dissipation per resistor. No need to go with 100 watt resistors! Even the most basic smd resistors can handle that without an issue. I've usually use the 1ohm vishay resistors (2515 package, sorry I don't remember the exact part number, but it's something lile WCS or similar). That will also help the fets to some extent with higher loads. Some air flow and large ground / power planes keeps them cool enough.

[pinoccio]

Originally I meant to keep this project simple 

Found some of the recommended IXYS mosfets on ebay for good money, like IXTH75N15 for 2 bucks, so probably that will be the next swap in my circuit.

Powering will be another question, might have to inevitably go 12V, but to hold on to simplicity, I hope I can still use single supply & virtual ground.
Btw. my plan was using a small SMPS module. Yeah, module everything ...same for the ADC and DAC, and the Nano of course. That's why I designed it with those connector headers. Plugin and go, simple, cheap, quick construction.

Anyhow, reading all this, maybe the op amp is not the best choice too...

Can you please make a copy of your schematic and make it public?
...
Are you willing to do SMD soldering?

Yes, I can make a public easyeda project copy soon.

I'm ok with SMDs as long as it's not very small, like 0805 is still manageable by hand.

I really recommend going down to 0.1 Ohms for the shunt ...

I went from 0.1 to 0.5 shunt because of previously posted suggestions... Still not sure it is the best for this project, but hey, we have a nicely evolving concept
Actually I think I messed up this part in the last circuit version and the divider should be 1/4 for 5A limit.

[OM222O]

WSC25151R000FEA is the resistors that I was talking about  , unfortunately it's only a 1W resistor, so its not suitable for this application, however you can use the larger package which is the WSC45271R000FEB. 2watts and plenty good enough for this application.

also a very versatile op amp is the MAX4238.

please send me your schematic and I will apply the changes.

Also for current sensing, you can use the INA286 with a 10m\$\Omega\$ current sense resistor which will be pretty cheap at the end.



you can also buy high side current sensing modules if you want to go that route.

[Yansi]

Why would you add this overpriced sense-amp, if you can already measure the current from the bank of shunts in the source electrode of the mosfet?

Also, to the OP: 5V is very insufficient as a supply voltage for the OPamps that control the mosfets.

[OM222O]

Why would you add this overpriced sense-amp, if you can already measure the current from the bank of shunts in the source electrode of the mosfet?

Also, to the OP: 5V is very insufficient as a supply voltage for the OPamps that control the mosfets.

over priced? it's only about 2$ ! and it's way more accurate than the current method he is using. it also eliminates the requirement of a dual supply which adds more than 2$ cost for sure. same can be done with a low offset amplifier like the MAX4238 or OPA333 and an external 4x100k laser trimmed resistor (you can't use generic resistors for an accurate difference amplifier). which will be more complex AND more expensive. there is a reason these parts exist. unless you don't care about accuracy and you're happy about 12mV accuracy (3mV offset x gain which is 4). in which case upgrading the op amps or the resistors makes little to no sense 

INA220 is a suitable / cheaper but less accurate alternative.

[JeanLeMotan]

That 5A will be shared between 4 resistors! That's just over 1amp per resistor. So less than 2 watts power dissipation per resistor. No need to go with 100 watt resistors! Even the most basic smd resistors can handle that without an issue. I've usually use the 1ohm vishay resistors (2515 package, sorry I don't remember the exact part number, but it's something lile WCS or similar). That will also help the fets to some extent with higher loads. Some air flow and large ground / power planes keeps them cool enough.

My math was off, you're right. At 5A and 0.5Ω that's 12.5W total going through the shunt resistors. So 3.125W per resistor. That's still a lot for this application.

At 5A in my load the shunt resistors go up to 55-65°C in a 20°C ambient. I had to drop the max current from 5A to 4A for this reason.
With a 50ppm temp coefficient, that's a change of ~0.28% over the whole range which doesn't sound that bad - but the bigger issue is that it heats up other components around including the copper traces from/to the shunt and those have a much much higher temp co. (0.4% per °C)
That's with 0.1Ω total, 4x series parallel 5W 4527 resistors: https://lcsc.com/product-detail/Low-Resistors-Current-Sense-Resistors-Surface-Mount_Ever-Ohms-Tech-MA452750FR100FZ_C252911.html

For 0.5Ω you would get 5 times the power dissipated and 5 times the heat so the temp will go much much higher.

Btw - for the WSC45271R000FEB the datasheet shows a temperature rise of 120°C at 2W dissipated!
On top of that there's power derating: at 120°C you can only dissipate 1W.

I don't see how you can do 5A with 0.5Ω and such small resistors unless I'm missing smth obvious. Someone else from this forum recommended having a shunt resistor rated at least 5-10x the power dissipated which for this load would mean a 60-100W shunt.

I'll definitely go for a smaller shunt in my next rev and I'll get a proper manganin 0.01Ω one.

[Yansi]

$2 is quite expensive, for the rest of the circuit compared and actually, quite unnecessary here.

For what would you need a dual supply? The current sense resistors are already present! And no, you won't improve accuracy with the INA.

If the thing would be properly designed with a current sense voltage amplifier within the current regulator loop, then the extra divider-buffer whatever opamp would not be even needed and current already available for direct measurement with ADC.

Making the shunt very small is also not great, as it degrades accuraccy (offset and tempco errors).

Instead of slapping there completely unnecessary part with another heating element to sense current, use a suitable precision opamp, such as MCP6072 or any similar cheap one to do the job. And please be that kind an do not lie about the INAxxx prices.

[OM222O]

jean, your math is still off  he is passing a maximum of 2.5A per branch (i.e mosfet). each mosfet has 2 resistors in parallel, so 1.25A per resistor. using ohms law we have P= I²*R = 1.25²*1 = 1.5625 watts 

Yansi, then please explain why literally any commercial product uses a huge low value resistor with a differential amplifier (much more expensive than the INA ones I mentioned) in order to achieve this? they often have 6 or more branches, each with a separate op amp driving them and a set of shunt resistors. for example this one (BK Precision 8500):



and here are the multiple mosfets each with their own driving stage:


you can clearly see the small SOIC (or maybe TSOP) package next to each fet  you can also see quite beefy power resistors soldered vertically (green package)

the reason this is done is is the "shunt" resistors are optimized for power and help the fets with power dissipation as take some load off the fet. however they get warm and they usually have about 50 PPM/C temp co which is quite bad, therefore the main low value shunt gives you the correct power draw with great accuracy and no need for summing amplifiers etc.

The reason a dual supply would be needed is summing amplifiers must be inverting like I explained before, if you don't want weird things to happen which implies dual supply requirement. it is also quite bad in terms of accuracy when used with cheap op amps. if you want to spend more on op amp and a precision resistor network, again it would be pointless. just use a proper chip like the INA series at that point.

Since you brought up me lying about the INA series prices as well, here is the prices on mouser 


I'm not paid by T.I to promote their products  it would be best to not spread misinformation if you have no idea what you are talking about. giving false advice is really bad, so be sure to do your research before recommending something 

[Jay_Diddy_B]

Hi group,
Let me share my idea of what the BK8500 load schematic would look like, based on studying the photographs:



There is one op-amp per MOSFET. This is essential to make sure each MOSFET shares the current. Each MOSFET has its own source resistor.

There is a low value shunt and a low drift differential amplifier that is used to measure the total current.


HP 6060 STYLE



This is a partial schematic from an HP 6060 load. It uses 8 MOSFETs in 60V 60A 300W load.
There are two op-amps per MOSFET. The OP270 was chosen for low noise and low offset, 75uV. The MC34072 was chosen because of its ability to drive capacitive loads. Each MOSFET has its own current sense resistor.

A summing amplifier U5 is used to obtain the total load current.


Dual Rails

Both of these commercial loads used dual supply rails.

Regards,
Jay_Diddy_B

[OM222O]

thanks john! I wasn't sure if they are doing high side or low side current sensing, but I just assumed high side. both are equally valid methods but the idea still is that the low value shunt is used for actual current monitoring, not the summing the the voltage across source resistors and assuming they stay at 1\$\Omega\$ or whatever their value is 

[Jay_Diddy_B]

Hi
Here is a picture from the BK8601 teardown:

 

The source resistors are 0.05



This is a 60A load with 8 MOSFETs. 7.5A per MOSFET
Maximum dissipation is 7.5 x 7.5 x 0.05 = 2.8W



The op-amps are NE5532. These are low noise. The offset specification isn't supper good 0.5mV typ 4mV max.

MOSFETs are IRFP250N

Regards,
Jay_Diddy_B

[OM222O]

the IRFP250N is surprisingly common with commercial loads, despite lacking any DC SOA specifications.
OP plans to go upto 2.5A per fet, so the source resistors can be a bit larger.
In terms of op amps chopper,rail to rail op amps seem best for these applications, but understandably they are more expensive.

[MarkF]

the IRFP250N is surprisingly common with commercial loads, despite lacking any DC SOA specifications.
OP plans to go upto 2.5A per fet, so the source resistors can be a bit larger.
In terms of op amps chopper,rail to rail op amps seem best for these applications, but understandably they are more expensive.

I used a TLC272ACP op-amp in my DC Load
   https://www.eevblog.com/forum/beginners/a-load-off-my-mnd/msg2151265/#msg2151265

[pinoccio]

public project: https://easyeda.com/pinoccio/Electronic-Load-DIY-open

[Yansi]

the IRFP250N is surprisingly common with commercial loads, despite lacking any DC SOA specifications.
OP plans to go upto 2.5A per fet, so the source resistors can be a bit larger.
In terms of op amps chopper,rail to rail op amps seem best for these applications, but understandably they are more expensive.

OM222O, could you please kindly stop pushing your bullshit with those high side current sense amplifiers? They don't even have enough bandwidth to begin with. They are anything but useful for designing a good stable load with good transient characteristic.

IRFP250 from ST for example specifies DC SOA. But please be aware, that even specifying it does not mean it is fully characterized and no secondary breakdown will happen at higher voltage.

Also, specifying DC SOA for 100W does not mean one can get 100W from it, proper thermal derating has to be accounted for. The typical practical limit for a TO247 package with isolation pad is about 25W  (10W for TO220), and about 40-50W when screwed directly on large enough heatsink with good thermal compound.