electronic load

[tasty]

Fine in theory, difficult in practice when the FETs aren't fully on or off. Because of production spreads, even FETs from the same batch are unlikely to have exactly the same Vgs/drain-current characteristic. The result is that they won't share current equally and will probably self-destruct sequentially, domino style. Adding source resistors helps in current-sharing, but isn't a guaranteed cure.

Yes I thought the same but adding gate resistances instead source will do the same and I have a larger option like that, right?

 

[Alec_t]

adding gate resistances instead source will do the same and I have a larger option like that, right?

No. It won't do the same because there is virtually no voltage drop across the (series) gate resistor except when Vgs is actually changing and the gate capacitance is being charged/discharged. The FET gate is insulated. This is unlike a BJT where current flows through the base-emitter junction.

Edit: I suggest you re-read the Wong article in the first link in post #20, regarding SOA and 'linear' MOSFETs. Hot-spotting is the likely reason for your fireworks display.

 

[WHONOES]

Have a look on the datasheet of your chosen device for a graph entitled "Maximum Safe Operating Area". This defines the maximum voltage and current combinations for your choice. Don't be tempted to load you device to the maximum permitted. Keep it to 50% or less. Also, have a look at the "Transient Thermal Impedance" graph which can be quite illuminating. Plus,TO220 devices are limited to 50W of power dissipation. If you need to dissipate a lot of power then you need a device with a very large thermal contact area (footprint).

 

[tasty]

I made a circuit like in this link : https://learn-cnc.com/wp/wp-content/uploads/2018/04/ConstantCurrentExample.png but with 6 mosfets and op amps separatly. The result is prety good for the mosfets and th resistances, they not heating a lot. But there is an issue with the op amps. One of them exploded. It works well with 8V as DUT, I can increase the current untill 5A without problem. But when I tested with the 55V, the op amp exploded at about 2.6A. How it can be possible? In theory, there is not a big current in the gates when I increase slowly. And the negative input of the op amps have a huge impedance. From where this overload of the op amp can come from? May be I have to add some gate resistance to increase the stability?

 

[WHONOES]

What power supply voltage is applied to the opamp?

 

[tasty]

What power supply voltage is applied to the opamp?

12V limited to 1.5A

 

[WHONOES]

Jus noticed that the LM358 is a dual opamp. If that is indeed what you are using, what are you doing with the unused half?

 

[tasty]

Jus noticed that the LM358 is a dual opamp. If that is indeed what you are using, what are you doing with the unused half?

I use 4 IC of LM358 wich makes 8 op amps.I made the same circuit as in this link 8 times : https://learn-cnc.com/wp/wp-content/uploads/2018/04/ConstantCurrentExample.png
So I'm using all the op amps.

 

[WHONOES]

How are the grounds to the fet and the opamp shared.

 

[tasty]

How are the grounds to the fet and the opamp shared.

Exactly like this : http://www.kerrywong.com/2017/01/15/a-400w-1kw-peak-100a-electronic-load-using-linear-mosfets/
but not 2 times but 8 times.

 

[WHONOES]

That doesn't tell me a lot. Have you daisy chained the grounds or connected them all back individually to one point?

 

[tasty]

That doesn't tell me a lot. Have you daisy chained the grounds or connected them all back individually to one point?

All the ground points are connected together and all the drain pins too, like the schematic in the preview link.

 

[tasty]

For example if I want to sink 8A from the DUT in total with the circuit, each combination of op amp and MOSFET will drain 1A. But there is no problem with low voltage about 8V? I think it's an oscillation issue in high voltage. I will try to add capacitor between negative inputs and outputs of the op amps. and resistances between out of op amps and the gates.

 

[WHONOES]

As already mentioned, you need to INDIVIDUALLY connect all ground connection back to a single point. Failure to do so will inevitably result in catastrophic oscillation.
Also, throw away the Fet's you are using and use a single device in a TO227 package which will be capable of much higher dissipation.
Further, avoid long leads to the Fet's at all costs, particularly for the gate connection. If using a gate resistor, it must be connected physically as close to the gate as possible this means 1/4inch or less. A ferrite bead on the gate wire will help as well.
As you suggest, a small capacitor from the output of the opamp back to its - input might help but, beware that to large a value could in its self cause the opamp to oscillate. This is usually caused by the current limiting circuit in the opamp. You should also place a 10μf capacitor across the supply legs of your opamp, as close as you can get, to decouple it from the supply wire inductance.

 

[hevans1944]

... use a single device in a TO227 package

Did you perhaps mean a TO247 package instead of TO227? And, BTW, where does the DUT connect? Between the V_cc supply and the MOSFET drains or somewhere else?

I already posted a link for the schematic. If you read ALL the posts you will understand better.

Sorry, I somehow missed your schematic of the op-amp negative feedback driving the MOSFET gate, based on the voltage drop across the current shunt resistor in series with the MOSFET source. I hope you took to heart the comments by other posters regarding op-amp rail-to-rail operation and low offset voltage requirements. Some "simple" unity-gain op-amp circuits can be a real bitch kitty to implement correctly in hardware. It becomes almost impossible without adequate test equipment and experience.

I did read (and continue to read) ALL the posts, but understanding eludes me. If you suspect oscillation is causing op-amps and MOSFETs to explode, an oscilloscope can quickly confirm this and perhaps help to troubleshoot the cause. However, it appears from this side of the keyboard that "hot spots" as described by the Wong site I linked to in post #18, that you apparently re-discovered and re-posted in your post #20 are the real contributing factors. As for your statement:

The issue is not the heat like a said a lot of times. There is no time for the MOSFET to get warm...

If the MOSFET case "explodes" as shown in the image you uploaded, there was certainly enough time for the MOSFET "to get warm" enough. So, on top of other lack of knowledge you appear to exhibit, let's add ignorance of thermodynamics.

Reading this thread feels like a visit to one of the Twilight Zone television series episodes... so maybe I am not in the good domain.

 

[WHONOES]

Did you perhaps mean a TO247 package instead of TO227? And, BTW, where does the DUT connect? Between the V_cc supply and the MOSFET drains or somewhere else?

No, I meant TO227. It has a much larger footprint than the TO247. It's dimensions are 38mm x 31.5mm.

 

[tasty]

As already mentioned, you need to INDIVIDUALLY connect all ground connection back to a single point. Failure to do so will inevitably result in catastrophic oscillation.
Also, throw away the Fet's you are using and use a single device in a TO227 package which will be capable of much higher dissipation.
Further, avoid long leads to the Fet's at all costs, particularly for the gate connection. If using a gate resistor, it must be connected physically as close to the gate as possible this means 1/4inch or less. A ferrite bead on the gate wire will help as well.
As you suggest, a small capacitor from the output of the opamp back to its - input might help but, beware that to large a value could in its self cause the opamp to oscillate. This is usually caused by the current limiting circuit in the opamp. You should also place a 10μf capacitor across the supply legs of your opamp, as close as you can get, to decouple it from the supply wire inductance.

I'm already using the TO227 package. I will see what it will happen thank you for your help.

If the MOSFET case "explodes" as shown in the image you uploaded, there was certainly enough time for the MOSFET "to get warm" enough. So, on top of other lack of knowledge you appear to exhibit, let's add ignorance of thermodynamics.

Reading this thread feels like a visit to one of the Twilight Zone television series episodes... so maybe I am not in the good domain.

In my last posts I said that the op amp explode and not the MOSFET. Its a proof of what i'm saying is more logicaly correct and the last time when the MOSFET exploded like in my picture, I was under the SOA. So this explosion is not because of slow overheat because it happens almost instantly. There is another parameter that causes this explosion as an involuntary exitation on the circuit.

Perhaps I'm not an engenieer but since your first post you nerver gave me a real practical solution like others. Do you really want to help me or you simple want to show that you know a lot of thinks except electronics?

 

[hevans1944]

... No, I meant TO227. It has a much larger footprint than the TO247. It's dimensions are 38mm x 31.5mm.

Agreed it is a much better package for heavier power dissipation. I just thought it was a huge departure from the TO-220 and TO-247 packages, but it probably doesn't make any difference since no one can agree on what's wrong, much less offer a troubleshooting procedure to correct it. Where is Rod Serling or his ghost when ya need him?

 

[WHONOES]

Do you have access to an oscilloscope?

 

[hevans1944]

.

Do you have access to an oscilloscope?

Are you asking me or @tasty?

If I had more time, I would purchase a handful of MOSFETs and try to duplicate the failure modes seen by @tasty. I have a lot of experience blowing parts up, all from my earlier years as an electronics bench technician. As I became more experienced, I destroyed fewer parts. I haven't had any catastrophic failures for several years now, but there is always the opportunity to smoke a new PIC (my latest obsession), like I thought I did before moving to Florida. It turned out that the "smoke" was just solder flux, and the six-pin PIC worked just fine after I got it soldered to an 8-pin header adapter more suitable for breadboard work. My bad. This particular PIC is also available in a 8-pin DIP package, but the person I was trying to help needed the tiny surface-mount package.

... I'm not an engenieer but since your first post you nerver gave me a real practical solution like others.

So... one of those "real practical solution" you got from others solved your problem? Which one was it?