Yet another DIY Electronic Load

[pinoccio]

This is my planned concept of a simple DIY Electronic load ...any remarks or suggestions are highly welcome.
I'm posting it here for the community to possibly check for issues I might have missed or even not thought of and for inspiration to others.
I will publish the final schematic and resources once I make at least a working proof of concept on a breadboard/perfboard (will surely take some time).

Heavily inspired by:
https://juangg-projects.blogspot.com/2019/02/arduino-based-electronic-load.html





https://www.instructables.com/id/DIY-Adjustable-Constant-Load-Current-Power/

Similar thread: https://www.eevblog.com/forum/projects/diy-electronic-load-project-thread-d/


Main specs:
- 24V max, <6A max, (withstand continuous 5A@12V; used heatsink limited)
- generic Arduino Nano controlled
- use of external ADC & DAC modules for improved resolution/precision and simplification
- temperature limit control and active (pwm) fan cooling
- simple UI - small 0.96" 128x64 OLED screen, one button and a rotary encoder


The code / UI part is not ready nor being prepared yet. But that's a relatively** easy part...
Planned vision:
- simple menu, operated by a rotary enc. + its button, and 1 extra button
- display will show: mode and current+set: A, V, W, T, +Temp.
- modes: CC, CP, Capacity, *Resistance?
- settings for respective modes: Amps, (m)Watts, Capacity, Time?
- protection settings for max limits to shut off at: A, V, W, T, +Temp.
- maybe storing last settings in EEPROM


I want to design a reliable circuit first...
So here is a draft:

https://ibb.co/zFKLnpv

Sorry for the usage of many connectors and headers instead of direct lines or specific parts and modules, but the circuit is designed directly so that a PCB can be made easily (spot the logo in the schematic ).


The main part is built around a logic level MOSFET driven by OP amps - a well-known concept
First block (U1.1) drives the fet according to sensed voltage from a shunt (array of 1R/1W --> more precise 0.1R@10W), which is meant to provide 1V@10A -read via a 10k resistor (R11) [necessary?], and a the set voltage on its other input.
C1 is supposed to improve stability [maybe someone will comment if this is necessary].

The desired current (input voltage) is controlled by the second block (U1.2) -a voltage follower with a approx. 1/9 divider (R12+R13) - that's supposed to go up to 1/9 * 5V = 0.555V --> so up to 5.5Amps max load. This is driven either from the MCU by the MCP4725 module 12bit DAC (#IO3), via a 1k resistor (R16) -no low pass filter needed because of using the DAC, or an optional external source (H2) [someone may comment on options or the protection diodes for details], selected by a switch (P4).
The current reading is obviously fed into the MCU - but first it's 4x amplified by the third block (U1.3) for improved resolution. That amplified voltage (0.555 * 4 = 2.22V @max current of 5.5A) is fed to a ADS1115 module 16bit ADC (#IO1).
A optional connector (H1) for external reading can be placed too.

And to make a use of the fourth block (not really necessary), we buffer the connected voltage sensed by a 1/5.3 divider (R17+R18) - that defines the max limit of 26.5V on the input transformed to 5V on the divider. This is then fed (#IO2) also to the ADC module, and an optional external output also (H3).

This all needs just some fine calibration in software naturally.

So to make it a complete description of the core part, the tested source connects to P1. Those optional P2 (for verification/test multimeter) and P3 (for a fuse) can be omitted.


Temperature (on the heatsink) is sensed by a simple NTC in a divider configuration (I may modify this later), and fed to the ADC module (#IO4). To simplify this part and avoid calibration + calculations, I'm planning to measure the voltage at room temperature and when dipped in water boiled to 75C, then just map() it in software to pwm %.
A digital voltage sensor can be used instead, just mind the max temp. specifics.

The heat-sink fan will be pwm driven too (about 35-75C -> 20-100%). I hope it can work in a basic Arduino pwm mode (@490Hz I belive) without too much noise.

Rest of the wiring consists of a simple 5V power input with basic protection and filtering, and the Nano with a rotary encoder + 1 button. The external modules and a display are attached to connectors, all running on I2c (with pullups if needed: R20+R21).

I have not calculated the required heat-sink, but the coarsely estimated 70W will require some decent dissipation, the fan should help to reduce its size thou.
(EDIT: not sure if the single IRLZ44 is suitable for the job and won't get a bit overloaded at that planned maximum power, might be safer not running it continuously at more than 2A in this setup)


Thanks

[pinoccio]

--- reservation for followup / updates ---

[Cliff Matthews]

Serious doubts here. IRLZ44 is likely not the device you want running in linear mode at 70w continuous. Edit: If you want 4 jelly bean devices you could modify something like this..

[pinoccio]

Could you elaborate on that a bit more? We might learn something here... What is the problem running this MOSFET in its linear/ohmic region at 5A with a solid heatsink? It's not exactly obvious from the datasheet. Will it get so extremely hot?
I'm not an expert in this

One of the goals was a simple design and construction. So I'd like to stay with a single transistor.
Are there any more suitable MOSFET options for this purpose?
Btw. the 70W was an estimate for an appropriate heatsink to be safe, at 12V 5A the FET has 11.5V * 5A = 57.5W to dissipate. Provided my calculations are right


I'll have to have a better look at your circuit to understand how it works and spreads load across 4 ...but the basic idea is obvious 
The modification would be to use several fets, also multiply the driving op amp (U1.1 in my case), and also the shunts. That complicates the current sensing too. I'm not sure it would be ok to sense at just one shunt and assume they all will be balanced without further circuitry...
For now, I still prefer the single piece configuration.


One more thing for possible improvement of my circuit might be some stabilization for the op amps, like RC network... I'll have to look up and figure something out maybe.

[Cliff Matthews]

Member Gyro wrapped it up in nutshell "Most switching mosfets get unhappy if used in their linear region and can fail due to localised hot-spots on the die. It's normal to significantly de-rate them for linear operation, regardless of heatsinking"

This thread gives the issue at little more light:
https://www.eevblog.com/forum/projects/disapating-750w-of-mosfet-heat-for-under-$100/

[OM222O]

IXYS makes linear fets, specifically for this application. have a look at their part selection.

[Jay_Diddy_B]

Hi,

I see that you have used an LM339. This is a quad comparator. A comparator is not the same as an op-amp. A comparator compares the two input signals and gives, essential a logic signal that indicates which one is more positive than the other. You need to use op-amps in this circuit.



If you want to understand how to design the analog portion of the load:

https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/

Regards,
Jay_Diddy_B

[pinoccio]

Oh damn, you're right!    I know I need op-amps, not comparators... Bought some LM324N-s too... Not sure how I managed to put those 339s in the schematic not noticing the obvious. Sorry, might be I just forgot to change the part name.

Btw. that's is a fine thread you linked ...I'll have to study it for a while

Ok, next steps will be redesigning the analog portion - correct op-amp (LM324 should be fine as this circuit doesn't need rail-to-rail output), find a more suitable mosfet (logic level; maybe I'll just try the MTP3055VL from Daves video), and at least duplicate the power part fet+op-amp combo to spread load (I can sacrifice the unnecessary voltage buffer to get another driver op-amp and not need another package).

Thanks

[pinoccio]

Found some information on Linear MOSFETs:





And a nice electronic load project: http://www.kerrywong.com/2013/10/24/building-a-constant-currentconstant-power-electronic-load/
...interesting idea those paralleled mosfets
(so I could drive 4 with 2 available op-amps in my circuit)

[Conrad Hoffman]

The earliest use of a MOSFET for an active load that I've seen was in the 1983 Siliconix MOSPOWER design catalog. Most everything is an offshoot of that circuit. Note the use of the 100 kohm resistor to provide local feedback. Some use a small capacitor there. Without that local feedback, the circuit is prone to oscillation and unpredictability, especially if something faster than a 741 is used.

[Yansi]

Lol, why you call it an off-shoot? This circuit can think up from zero almost any sensible engineer.

To OP: At least add a series resistor to gate (~100ohm).

[pinoccio]

Something like this?


Is it ok to change gate resistor to 1k in this wiring when connecting two parallel MOSFETS (like here) ? I don't know yet how to design proper RC for stabilization here :/
Not sure about the C5 cap  - connect to ground, or source?

(EDIT: minor bug in circuit - sticky 339 ;P )

[Yansi]

No. C5 shall go agains pin 4 not ground.

[OM222O]

No. C5 shall go agains pin 4 not ground.

NO! what he did is correct. to prevent oscillations, you feed back the output to the inverting input rather than smoothing the oscillations from the feedback!
I would also get rid of that 100k resistor between gate and source of the fet. it is not needed in newer designs.

I created a thread and got some very good feedback on it. if you purely want the details about stability, jump to page 2, otherwise there are some good info about electronic loads in general. you can also use the same idea for range switching.

https://www.eevblog.com/forum/projects/current-source-feedback-capacitor/

on a side note: you can use BJTs / darlington pairs as a cheaper alternative to linear fets. they tend to be better in linear applications imho.

[MarkF]

Modified circuit
Plus some alternate resistor divider choices

   

[MarkF]

No. C5 shall go agains pin 4 not ground.

NO! what he did is correct. to prevent oscillations, you feed back the output to the inverting input rather than smoothing the oscillations from the feedback!

Contradiciton! 
His circuit is wrong

[OM222O]

No. C5 shall go agains pin 4 not ground.

NO! what he did is correct. to prevent oscillations, you feed back the output to the inverting input rather than smoothing the oscillations from the feedback!

Contradiciton! 
His circuit is wrong

yes, looking at the schematic again, he has connected output to ground  my bad

also the best divide by 10 is 3k and 27k, perfect ratio.

[MarkF]

No. C5 shall go agains pin 4 not ground.

NO! what he did is correct. to prevent oscillations, you feed back the output to the inverting input rather than smoothing the oscillations from the feedback!

Contradiciton! 
His circuit is wrong

yes, looking at the schematic again, he has connected output to ground  my bad

also the best divide by 10 is 3k and 27k, perfect ratio.

2K and 18K resistor divider works too.  I like the lower total resistance. 
Gut feeling that the higher current will be more immune to noise.
I was trying to target a total resistance less then 10K but there isn't an exact match using standard values.

[OM222O]

that would be 300 ohm and 2k7 or 200 ohm and 1k8.that is an insane amount of power draw for a voltage divider! keep in mind that a dac / op amp is driving that. noise won't be an issue with 20k or 30k overall resistance.

[Jay_Diddy_B]

Hi,

quite often in these discussion the paralleling of MOSFETs comes up.

I presented a study of why one op-amp per MOSFET is (much) better. You can find the analysis here:

https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/75/

Most, if not all, commercial electronic loads used one op-amp per MOSFET.

Regards,

Jay_Diddy_B

[Jay_Diddy_B]

U4.2 should be an op-amp (LM324) not a comparator (LM339).

Jay_Diddy_B

[MarkF]

that would be 300 ohm and 2k7 or 200 ohm and 1k8.that is an insane amount of power draw for a voltage divider! keep in mind that a dac / op amp is driving that. noise won't be an issue with 20k or 30k overall resistance.

Yes.  That's why my table only goes to 620 and 5.6K.  (i.e. 0.8mA)
I considered that as the lowest overall resistance.

[pinoccio]

Thanks for lots of great info and suggestions.

Tried to implement some improvements...
Newer version:

I'd still like not to overcomplicate this project

I thought it's not good to go too low with the divider exactly because of the current. But maybe my values are now a bit too much  The ratio is even higher with 4 fets because I only need to go up to about 2.5A per fet group. So the fets should be now safe running below 2A each.

on a side note: you can use BJTs / darlington pairs as a cheaper alternative to linear fets. they tend to be better in linear applications imho.

Regarding the price, I just bought 5pcs MTP3055VL from ebay for US $2.80 shipped, so no big deal even if some die in tests 
Anyhow, would that darlington be just a drop in replacement with minor tuning? Worth the linear characteristics?

U4.2 should be an op-amp (LM324) not a comparator (LM339).

Yes, I wrote about the typo at the and of my post

[Jay_Diddy_B]

Hi,

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



The LM324 pin numbers are:




Regards,

Jay_Diddy_B

[MarkF]

No.   Only one MOSFET per op-amp.  I used the IRFP064 in my load.
I would use the 2K and 18K resistor divider and adjust your sense resistors for the max current you want.

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.