Modified LM324 Based 150W 72V 10A Electronic Load [COMPLETED]

[enut11]

I have just completed an Electronic Load project based on a Chinese 150W/10A/72V LM324 four MOSFET kit:

https://www.ebay.com.au/itm/150W-Constant-Current-Electronic-Load-Discharge-Capacity-Tester-DIY-Kit/173078658449?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2648

A 3.7v LiPo isolates and drives the LED panel meter. An LM317 is used as a manual fan speed control. Jiffy box measures 155 x 95 x 55mm.

I downgraded the wattage to <75 watts because of the CPU heatsink and to 33v max and 3A max based on the LED panel meter ratings.

The unit works well and is able to hold the load steady with <0.1% variation for currents over 100mA.

I now want to build the same kit with a wider range of constant current capability, up to 10A and  down to <10mA, perhaps down to 1mA.

The basic design uses an LM324 quad Op Amp to control four To220 MOSFETs through 0.22 ohm sensing resistors. The 4 current paths are paralleled at the input.

The new design is to have several input sockets and use different current sensing resistors for each range. I will aim for up to 100W max load for now.

The first MOSFET would have a 5 ohm current sensing resistor and control currents up to 100mA
The second MOSFET would have a 0.5 ohm current sensing resistor and control currents up to 1A
The 3rd and 4th MOSFETs would have 0.1 ohm current sensing resistors each, with current paths paralleled, and control up to 10A total (5A ea).

I expect to replace the 3rd and 4th To220 MOSFETs with higher rated units capable of handling about 50w continuous load each. Suggested devices?

Any fundamental reason why this approach would not work?

Is there a better Op Amp than the LM324 for this task?
enut11

EDIT: Some of the functionality above has been incorporated into the original CC Load as as result of my experiments.
Take #2 CC Load, which is discussed later in this thread (from Reply #8), will be aimed at the original advertised specs of 150W and 10A but limited to around 30v max.

[Kleinstein]

The electronic load might need an RC snubber across the load to avoid oscillation with a difficult (inductive) source. Something like 10 Ohms in series with 1 µF (maybe 10 µF as the LM324 is slow) should be OK.
It might be a good idea to have fuse in series, just in case, as MOSFETs tend to fail short if they fail.

A possible more powerful MOSFET would be an IRFP250( IRFP240.  Depending on the manufacturer (e.g. Fairchild) is might even be rated for linear use. It's kind of a compromise between low cost and suitability for high power loss - not guarantied to work, but usually will do. This types were used a lot in the 1980s when MOSFET based audio amps were popular.

[enut11]

Thanks @Kleinstein.
Would I need to add the RC snubber across all three loads, ie the 10A, 1A and 100mA inputs? If so, would the values be the same for R and C?

Is there an easy way to test the linearity of a MOSFET?

Just checked eBay and IRFP240 MOSFETs are available from $1.45 to over $10 each. Is it just pot luck or is there a way to pick the genuine ones?
enut11

[Kleinstein]

Ebay is about the second best source for fake transistors. So normally I would buy those at a special electronics company.
I Europe I found  1.20 EUR  (around $1.50) from a reasonably good source and EUR 2,20 from an official distributor. So I would consider the $10 some with high end audio voodoo, maybe a set of 4.

The problem with MOSFETs in this kind of linear application is that the forward bias safe operation area is not really good with many MOSFETs, especially modern ones made for fast switching. Very few and expensive MOSFETs are tested for a use like in a electronic load, but old types for relatively high voltage often work well. So with the IRFP240 it is not only the normal genuine vs fake but it can also be old original style and modern replacements that might be genuine from some manufacturers, but could fail the SOA testing.

Testing the MOSFETs can be kind of destructive: use it in load circuit and check that is does not brake down at something like 100 V and around 1 A. A high voltage is the more difficult case for a given power level.
Depending on the extra protection (e.g. classic fuse or extra electronic current limit), the MOSFET might literally blow up, fuse to a short or survive and work again after cool down. SOA testing is kind of tricky, especially if it needs to be non destructive (for the failing ones).

[enut11]

Thanks to @Kleinstein
1) I have now added a 5A fuse in series with the positive input. The fuse was inserted at the unused A+/A- terminals.
2) I added a snubber 10ohm/2uF across the input terminals to counteract possible oscillation with inductive leads.

Below are some more pix of the internals showing how everything fits in the small jiffy box. One of the pics shows fan speed control on one side and a 3.7v LiPo for the panel meter on the other side. At 15mA, the LiPo will power the panel meter for about 6 days. The small red plug out the back allows charging of the LiPo.
enut11

[enut11]

Details of where I connected the 3A/33v LED panel meter. There are 4 digits for the volts and amps. I have found these to be very accurate and the amps reads to 0.1mA when the load is under 1A.

I used a single cell LiPo (or 18650 cell) to power the meter. It turns out that the meter LEDs start to dim around 3.5v which is a reminder for me to charge the internal LiPo battery.

During long term tests, the panel meter can be switched off to save battery power.
enut11

[enut11]

To limit the maximum current to 3A I added a 39K resistor in series with R22 at the top of the 4.7K 10-turn pot

The 10-turn pot that sets the load current is good for fine adjustments but not very helpful in letting you know where the setting is within the range of 0-3A, especially at switch on.

I found it handy to calibrate the amps load for each turn of pot rotation. This now forms part of the specs label on top of the case.
enut11

[enut11]

Well, I found the limits of the ST P75NF75 TO220 MOSFET  . Shorted one while testing the CC load to 30v/3A. It appears that these devices are not very robust in linear mode so I would now down-rate this circuit to 50W and 25v/2A max. The heatsink did not even get warm so failure was rapid. I will try some better MOSFETs when the second kit arrives.

I also experimented with isolating one of the MOSFETs and using a 2.2ohm Source resistor in place of the 0.22ohm. Full range on the pot is a handy 0-100mA and will now become a permanent part of this project. A SPDT switch at the input is used to select either 100mA or 2.5A full range.

BTW, the MOSFET tab for the 100mA range has to be electrically isolated from the heatsink for this to work. The photo shows the replacement green 2.2ohm resistor where the 0.22ohm was. The red wire is soldered to the Drain pin which has been isolated from the PCB.
enut11

[enut11]

Take #2 for Electronic Load

After a few hiccups with the first kit, including blowing up one of the MOSFETs, it is now working well, albeit at reduced wattage and voltage, limited by the supplied 75NF75 devices.

A second kit is underway. Photo shows partially populated PCB. I have added a 7812 voltage regulator at the input stage in place of the 4 rectifiers. Power will come from a 15v plugpack.

Missing are the MOSFETs and Source resistors. I ordered some TO247 MOSFETs and will replace the kit 0.22ohm with 5W types (maybe 10W with parallel 0.47ohm 5W). The PCB has space for the larger transistors and Source resistors.

What I have learned so far is that using MOSFETs as linear controls is fraught with danger. It is easy to to be taken in by the numbers on the spec sheets, but these were intended for switching circuits. So it seems, for DC operation, the spec sheets may be of limited use. Practical advice seems to point to significant underrating of device wattage and current and to provide very effective cooling. I could use a very expensive but robust device or a number of lesser devices in parallel. I will also look at thick copper sheet as an initial heat spreader. This would be bolted to a larger fan-cooled heatsink.

As @Kleinstein pointed out, vintage high voltage MOSFETs are more suited to DC Loads but they are not easy to source.

So, I will now experiment further with four IRFP150N TO247 devices in place of the smaller 75NF75 TO220 devices.
enut11

[enut11]

I found this document from  Infineon to be easy to read and very useful. It is all about using MOSFETs in linear applications such as DC Electronic Loads. They explain the Safe Operating Area limits and how to choose the right MOSFET.
enut11

[enut11]

Another article on the linear application of power MOSFETs, this one a little easier to read.
enut11

First 2 pages...

[enut11]

Last 2 pages...

[CustomEngineerer]

Kerry Wong does a good job of explaining the benefits of a linear mosfet.

[enut11]

Thanks @CustomEngineer. The video is very informative and the IXTK90N25L2 MOSFET (960W, 250v, 90A) has very impressive performance in linear mode.
Also, the thick copper pad between MOSFET and heat-sink makes a big difference to transistor case temperature.

Just checked the price here in Australia. $32.70 AUD + delivery from Element 14. Hmmm
enut11

[sorin]

Check:

FCA35N60
SPW20N60C3

You need Mosfets that have the DC. operation specified on their "Safe Operating Area".

[enut11]

Thanks @sorin. Will also look at using the FCA35N60 MOSFET.

Take #2 Electronic DC Load
Progress with the second electronic load while I wait for the TO247 MOSFETs, heat-sink and fans to arrive. Plastic box is 200Wx70Hx174D mm.

The display is a 90x50 LCD that I modified to read down to 0v. Reads volts, amps and watts simultaneously.

https://www.eevblog.com/forum/projects/smart-panel-meter-can-it-be-modified-to-read-down-to-zero-volts/

With 4 large MOSFETs, I am hoping for the full 10A and 150 watts, up to 30v max. I have soldered copper wire to the Drain and Source tracks to reduce track resistance on the PCB.

The 4 power transistors will be mounted on a 3mm thick copper heat spreader which in turn will be bolted to a large aluminium heat sink cooled by 2 fans.

The fans controller is manual and based on a cheap eBay LM317 unit. I have ordered a heat sensing fan controller to automate the process.

I decided to go with 0.1ohm 5W Source resistors in place of the 0.22ohm kit items. This should lower the minimum voltage that the CC Load will work at with high currents.

Remote voltage sensing is only available on the positive line due to the relatively simple design for this CC Load.

The LCD display will be powered by 2S 400mAh LiPo pack. At <3mA current the battery will power the display for a very, very long time. I have incorporated a display switch.
enut11

[VEGETA]

I am doing a similar project and reached nearly the same result (check last page for v0.2): https://www.eevblog.com/forum/beginners/dc-dummy-load-circuit-calibration/

The only problem left for us is that current flows in potentiometer is very very small (12 uA or so) thus will be affected by EMI or so.

This is because we choose 0.2R to be shunt for each mosfet (you have 0.22 which is the same) so we need 100mV before it to achieve 2A total output current.

I want to ask: what is the voltage at each op-amp input? also, what is the current passing through it?

We concluded that we should ground the potentiometer metal body to ground in order for this project to work nicely. So kindly give me your information and if you suffer this problem or not. This has to do with zeroing the output too.

[enut11]

Hi @VEGETA.
In the original circuit for the kit that I purchased, the maximum op amp input was 4.7k/(22k+4.7k) x 2.5v=0.44v
When this voltage is present across the 0.22ohm Source resistor, the maximum current would be 0.44/0.22=2A per MOSFET.

The easiest way for me to limit the total current is to change R22. When I increase this resistor, the total current is reduced.

The pot I am using has a plastic body so no easy way to earth it but I do not appear to have a problem with noise pickup.

I can zero the output. In fact I need about a 1/2turn on the 10turnpot before any current flows.
enut11

[VEGETA]

Hi,

but he current flowing is gonna flow through 10k + 22K + 4.7K + 1K which is gonna be so low. Meaning 2.5v through 22K + 4.7k + 1k = around 90uA if my calculations are correct. My original design was around 12uA which is pretty bad.

I changed it a big with a proposed circuit and now it is around 0.2mA.

[enut11]

The 10k stabilises the TL431 (IC1) at 2.5V. The 22k + 4.7k pot form a variable voltage divider.
There is negligible current into the op amp.
So, the op amp positive input sees 0v-0.44v at the pot extremes. The current flowing through the pot makes little difference.
The op amp then drives the MOSFET so that the inverting input sees the same voltage as set on the pot.
This voltage is generated across the 0.22ohm Source resistor.
So, if the pot is set to the middle position, for example, the MOSFET will allow 0.22v/0.22ohm=1A load.

[VEGETA]

How about if you put the mosfet on to-220 heatsink?

[enut11]

If you expect a meaningful reply you need to be more specific with your question.
enut11

[VEGETA]

If you expect a meaningful reply you need to be more specific with your question.
enut11

I meant, instead of having one big heatsink for 4 mosfets... how about putting one per mosfet? One of those famous to-220 heatsinks. Will it work?

If your device is about 100 watts that means 25 watts per mosfet, so my question will such heatsink be able to tolerate these 15-25 watts? with or without fan cooling.

[enut11]

In this project all the MOSFETs will benefit by being on the same heat-sink, provided it is adequate to dissipate the total watts generated.
If all the MOSFETs are at the same temperature they tend to share the heat load.
It is beneficial to get the heat out of the MOSFET case and into the heat-sink as quickly as possible to reduce the junction temperature. For this I am going to use a 3mm thick copper plate between the transistors and the aluminium heat-sink. Of course you need good thermal transfer between the copper and aluminium otherwise you are simply creating another heat barrier. Both surfaces need to be flat with heat transfer paste in between.
enut11

[JS]

I meant, instead of having one big heatsink for 4 mosfets... how about putting one per mosfet? One of those famous to-220 heatsinks. Will it work?

If your device is about 100 watts that means 25 watts per mosfet, so my question will such heatsink be able to tolerate these 15-25 watts? with or without fan cooling.

You need to specify which heatsink you are thinking, a pict would do.

Usually they are kind of small and good for LM137 when you need a bit more umpf out of it without them.

JS