MOSFETs in linear operation

RE: MOSFETs in linear operation

ben0127 — Fri, 19 Aug 2022 09:56:33 GMT

Thanks for that, no rush to come back if you think of anything else, whenever you have time.

Fault conditions will be dealt with later on, the feedback is only to hold the current stable. Good points made though.

RE: MOSFETs in linear operation

michaelkellett — Thu, 18 Aug 2022 14:08:19 GMT

When everything is about right I would expect it to work OK as you describe. Might be worth checking that the currrent isn't fluctuating which might be an indication of instability. If you wind up the gain enough in your control loop you will be able to make it unstable.

My main worries would be the slow response to serious fault conditions.

(It is possible that this doesn't matter in your application.)

I would normally include a software bug as a fault condition to be protected against.

Since failures in chargers for big Lion batteries have such serious consequences (don't trust the internal protection - that's just insurance) I would like to see processor independent over voltage and over current protection.

Overcurrents will happen when some fault shorts the output to ground, on a good day this blow the MOSFET open circuit, on a bad day it will blow it short circuit and the large current into the battery may pull the supply down a bit leaving you in a very bad position.

The other overcurrent that can happen is when the processor (during development) decides to set maximum possible current (PWM output sticks high).

I'm a bit busy right now but I may be able to get back to this tomorrow morning. Let me know if you need some tips re. how to limit the MOSFET current.

For thermal protection I would glue some electro-mechnical thermal switches to the battery.

RE: MOSFETs in linear operation

ben0127 — Thu, 18 Aug 2022 13:44:13 GMT

Ive been thinking about what you said regarding linear mode without analogue feedback as Ive been able to do some testing with mosfets over the last day or two. All is running smoothly now with good heatsinking so Ive been able to see my feedback solution in operation and it seems to be working very nicely but I cant help being slightly nervous, considering your comment.

Its really crude but seemed like it might be good enough considering that the mosfet doesnt need a closed loop control system to operate (as opposed to an SMPS chip for example). Essentially the current value is read by the ADC and a simple algorithm adjusts the duty cycle to maintain the desired current.

It seems to work exactly as expected with a significant reduction in duty cycle initially as the mosfet warms up, then it settles and stays put. My tests include mosfet temperature so I can see the heating tracking with the algorithm adjustments, which presumably is down to the gate threshold reducing with heat as its operation is below the ZTC point.

Once everything settled down (within about 5 minutes) and was left for another 10 minutes or so I tested the feedback by putting my fingers over the drive mosfet (cooling it). The result was as expected, the duty cycle started increasing quickly to compensate but the current held nicely. When I let go the opposite happened and it returned.

Im really interested to know what you think, it seems to work great but am I fooling myself?

RE: MOSFETs in linear operation

michaelkellett — Tue, 16 Aug 2022 14:59:32 GMT

Thanks for explaining where you are with this.

I think you will always have trouble using power MOSFETs in linear mode without an analogue feedback loop to control them, the problem with controlling the gate precisely is that the relationship between gate source voltage and drain source current is not well defined.

Most of the modern power MOSFET specs seem to be concerned with switching them hard on or off.

Good luck with your project, perhaps you can post some more about it as it progresses.

RE: MOSFETs in linear operation

ben0127 — Tue, 16 Aug 2022 14:36:42 GMT

First of all I would like to say thankyou for putting what must have been a very significant amount of time into your reply, with constructive criticism, alternate ideas and an entire circuit design. Unfortunately, reading through the points made there seems to be quite a lot of misunderstanding of the circuit. This may well be my fault for not explaining everything but I was attempting to not draw attention to things that at the moment are not the focus.

The thread was really about making sure I was not misusing or abusing the main mosfet with my circuitry, although I dont want to sound ungrateful for the opinions and advice on the whole circuit.

I would like to go through some of the points just to clear things up:

The feedback - this was never meant to replicate a classic closed loop control system, it would be useless in that respect, it was just basic feedback to nudge the gate drive if and when it needs it. That probably sounds ridiculous but its not so much that I think its a good way of doing feedback in the conventional sense, although its perfectly adequate for this purpose, but more that I like the idea of the micro being involved, it knows how hard its driving the gate etc. The 20ms filter you mentioned is there to remove the switching of the 2N7000 drive signal. I use PWM from the micro as I get a much better resolution than that offered by the DAC. As mentioned in the post I am still working on the drive circuitry as it has other issues.

If the output goes short the system will not rely on the feedback system, the battery voltage will be seen as 0V (or extremely low) and the mosfet will be switched off, assuming that the battery PCM doesnt get there first and disconnect the battery.

Regarding the battery voltage I mentioned (48V) I can see where this caused confusion and slightly regret mentioning it, but it wasnt meant to draw attention to itself. Just for clarity the battery is Lithium with a range of 33.6V to 50.4V.

Low side current sensing, indeed low side anything, causes problems which are unacceptable in this system. Ground must be ground.

The gate-source resistor on the main mosfet is as low as it is for improved resolution. Controlling the main mosfet over a wide range of charge current and with accuracy isnt easy and as I mentioned earlier its still being improved.

LTspice is great and Ive used it quite a bit but it doesnt tell you that your mosfets going to blow up :)

Youve gone to a lot of trouble designing a circuit which is above and beyond, thankyou. Even if I dont change my approach it shows me more established methods which are good tools to keep in the box. My circuit was not intended to be a reinvention of the wheel, Im part of an innovation centre which is investigating new smarter charging methods.

My initial reason for reaching out with this thread was that the mosfets kept blowing but the thermal issues causing the problem are understood now so thats sorted out. Thanks to you and all others that have replied for your help.

RE: MOSFETs in linear operation

scottiebabe — Tue, 16 Aug 2022 03:16:27 GMT

For a one off prototype, I would just use the classic op-amp servo'd current source only in a high-side topology such as:

Source: https://www.ti.com/lit/an/slaa867/slaa867.pdf

OP-amp A1 can be powered off 5 V (or similar).

OP-Amp A2 would need a supply referenced to the Vin. The can be done with a simple shunt regulator, such as a zener diode or a TL431.

RE: MOSFETs in linear operation

michaelkellett — Mon, 15 Aug 2022 11:32:43 GMT

First thought, you have a feedback loop where you are sensing the current with theresistor and INA282, measuring the signal with the processsor ADC, working out the feedback required, driving the processor DAC and then controlling the 2N7000 via a filter with a 20ms time constant.

I was going to say that this is a very bad way to do it, but I'll settle for observing that it is a very unusual way.

The feedback loop will alwas be very slow because the micro's ADC and DAC conversion are inside it. This means that if you short the output the MOSFET will alwas blow !

A much neater way to do it is use an anlogue voltage feedback loop and control the reference voltage with the processor.

I'm suspicious of the 48V charge voltage - real '48V' batteries charge between roughly 40 and 52V (lead acid) and maybe 35 - 54V (lithium ion). The range can be bigger than this.

So I think your PSU will need to provide maybe 56V and deliver full charge current with the load at 40V.

There a a million and one different ways of doing this.

You can buy controller chips that manage a switching power supply (eg LTC 4000) but I think you might find this very hard to work with.

It's much easier to use low side current sensing, but there is a disadvanatage that your controller 0V and the batteries load 0V are not the same.

You don't need such a low gate source resistor on the MOSFET so you can reduce the power in the level shifter.

This type of design work is vastly facilitated by using a ciruit simulator.

I suggest that you donwload LTSPice from www.analog.com. It's free and very widely used.

I've thrown together (very quickly) a simple all discrete part, linbear regulator. This is nowhere near a final circuit but it might help you get started.

V2 and V4 between them are the control input or reference.

Q1 and Q2 make a diff amp which compares a proportion of the output with the reference and adjustes the drive to M1 via the emitter follower Q3.

V3 and R1 represent the battery.

R3 senses the battery current (not used at the moment, this is waht your micr can measure and decide how to tweak the drive voltage to adjust the charge current.

It's easy to add hardware current limit with an NPN transistor driven by the voltage drop on R3, but there are other better ways.

If you get this design into LTSpice and spend soem time playing both with the tool and the circuit you'll get a much better idea of where you are going.

But you really need to pin down the spec:

range of input volts, range of output volts, max current, max voltage before battery explosion, must it stand shorts to ground as well as across the battery, operating temperature range, accuracy of current and voltage measurement and must the battery 0V be a system common rail , to name just a few things you need to know.

RE: MOSFETs in linear operation

ben0127 — Mon, 15 Aug 2022 09:59:34 GMT

Hi, yes it needs to be linear as opposed to PWM to protect the battery. Thanks anyway.

RE: MOSFETs in linear operation

ben0127 — Mon, 15 Aug 2022 09:53:10 GMT

Morning folks,

This is my circuit, essentially a battery charger but the point is that the charge rate is fully controlled. The supply will be a DC-DC with some level of output control for the purpose of reducing voltage across the mosfet. There will be other circuitry in there as well for detection of faults etc but the main design is shown. The 1.3k resistor is several in series to handle the power, this is to drop most of the remaining voltage (protecting the 2N7000) as the supply is around 50V (for 48V battery). This drive circuit still needs work but the focus for now is on the main mosfet.

Im thinking about maybe using this as a module which can be paralleled to get more current, the mosfets could be individually monitored that way rather than just assuming that each mosfet is taking the same current as the others.

RE: MOSFETs in linear operation

dougw — Fri, 12 Aug 2022 22:58:20 GMT

Hopefully the comments here don't sound like flames - that is not a motive on this site. Many of us have similar stories to yours. Shy about publishing ideas and not into social media. I also burned my finger on my first linear power supply design. What we learn from these experiences is more valuable than the components we smoke.

RE: MOSFETs in linear operation

Andrew J — Fri, 12 Aug 2022 17:45:22 GMT

What Javagoza says: you'll find that Element14 isn't like other sites and you won't get any grief at all for asking questions or posting things up. You'll get as much help as you ask for, given politely and with encouragement.

RE: MOSFETs in linear operation

javagoza — Fri, 12 Aug 2022 14:22:55 GMT

You're welcome. I have learned a lot with your question, the element14 community is great, always willing to help those of us who are just starting out.

From the element14 Participation Guidelines and Code of Conduct

"The element14 Community is intended to be a safe and inclusive space, free from violence, harassment, hate speech or exclusionary behaviour."

Don’t judge – instead share your knowledge! Knowledge is not a competition – instead of judging people for what they don’t know (or seemingly haven't google searched for), help them. Elitism is not welcome, even unintended. Remember that different people have different approaches and may have valid reasons for why they're asking.
- Similarly 'back handed' assistance which criticizes the person asking and then provides the answer requested is also not tolerated and can be seen as a form of emotional abuse.

RE: MOSFETs in linear operation

ben0127 — Fri, 12 Aug 2022 14:10:01 GMT

Thanks thats really helpful. Absolutely, I wont be blowing any more components up unless I make some other mistake. I will do the calculations needed to make this work. The price isnt really prohibitive at the moment so not concerned about that.

As you say, getting the approach right and then scaling up is what I had in mind. I want to know that fundamentally it works, then I will see how far I can take it. If in the end I can only manage 10A max, thats ok, at this point anyway.

As several very helpful folks here have asked for some details on circuitry (understandably!) I will put something up but Im afraid I wont be able to do that until monday now so please bare with me. In all honesty I have been rather shy, I dont do social media so posting anything let alone ideas or designs is quite daunting. The circuit I have is very simple but its purely my own design/idea and I half expected to be pelted with rotten tomatoes by any experienced engineers looking at it.

RE: MOSFETs in linear operation

michaelkellett — Fri, 12 Aug 2022 11:21:24 GMT

I'll offer you a couple of suggestions that I hope will help you.

The first is that 20A is a lot of current for a linear power supply of any kind. If you can adjust the source to allow only a very small drop, then you didn't really need an additional regulator. If you can't control well enough without a series regulator you should be thinking in terms of at least 2V and probably a lot more dropped across it. So thats at least 40W but maybe a lot more.

The second thing is that you really should have a spec, right now, before you start thinking about how to solve the problem. Good engineering starts with knowing where youwant to end up. (Google "requirements specifying". A lot of the stuff you'll find is over the top for where you are but you should have something.

My father, (now 94) told me when I was a kid that is usually a waste of time to measure something if you don't know what you expect the result to be.

You can extend from that to work out why you can't do a good design if you don't know what you expect it to do when you are finished.

As AJ said, don't be shy about posting an incomplete or vague spec, its way better than none. Often the problem on E14 or other such groups is that people ask component level questions when they should have explained the system level problem.

RE: MOSFETs in linear operation

ben0127 — Fri, 12 Aug 2022 10:34:21 GMT

Thankyou very much, thats solved the mystery. The maths makes it perfectly understandable why they are blowing up. Now I know what Im dealing with.

Im afraid I dont really have specs to give you at the moment as its more about the concept - can I use one of these devices to control current. There might be different setups for different power levels but Im going to want to get up to around 20A ultimately, if I can. The idea was to reduce the Vds as much as possible to make it workable, power wise. Im wondering if mosfets in parallel will help, the issue there is the relationship between the drive voltage (Vgs) and the current. I may have to control parallel mosfets individually, as opposed to just sticking them on top of each other (circuit wise not physically of course).

The reason why Ive gone down the P-channel route is because they make the most sense for the circuit. I use the stable input voltage on the source as a supply from which to drop the Vgs over a resistor. Supplying and controlling the Vgs for an N-channel (in the high side) adds another complexity and after a look at it yesterday I decided I would rather stick to P-channel and make that work, I do appreciate the advice though.

Thanks again.

RE: MOSFETs in linear operation

ben0127 — Fri, 12 Aug 2022 09:39:58 GMT

Thanks for that, I will take a look.

RE: MOSFETs in linear operation

ben0127 — Fri, 12 Aug 2022 09:37:24 GMT

Magic smoke is one thing but I think some of the little men fell out as well.

Anyway, I have feedback on the Vgs so it is adjusted to hold a stable current and as I have been working above the ZTC point the heat should reduce the current. I did actually see this but it was too little too late as moments later it blew, so not a useful property. As others have suggested the ZTC point is not the critical detail to focus on so I will ignore it going forward. Heatsinking and overrating seems to be the direction to move in.

RE: MOSFETs in linear operation

dougw — Thu, 11 Aug 2022 21:48:14 GMT

+1 on the ZTC - I never use it.

RE: MOSFETs in linear operation

dougw — Thu, 11 Aug 2022 21:40:42 GMT

Any mosfet will work well in its linear region far below 4 Amps.

4 amps through a FET with 4 volts across it implies 16 Watts - which will fry just about any transistor without a big heatsink.

RE: MOSFETs in linear operation

jc2048 — Thu, 11 Aug 2022 18:18:18 GMT

Does your 'charging circuit' actually need to be a linear solution or could it be a switching circuit?

RE: MOSFETs in linear operation

Andrew J — Thu, 11 Aug 2022 17:57:50 GMT

It's a shame he doesn't post anymore. Those videos were great when I went hunting about parallelising MOSFETs.

RE: MOSFETs in linear operation

scottiebabe — Thu, 11 Aug 2022 16:22:11 GMT

Great video. You can ask Scullcom for your YT subscriber commission cheque, they got a subscribe from me! lol

RE: MOSFETs in linear operation

Andrew J — Thu, 11 Aug 2022 16:13:15 GMT

Good points! I put a link to a video below to explain about MOSFETs and thermal calculations, but you may also want to check out this one from the same chap which talks about two other datasheet charts which explain what ScottieBabe has succinctly put. Watch from pretty much the beginning.

RE: MOSFETs in linear operation

scottiebabe — Thu, 11 Aug 2022 16:00:14 GMT

I congratulate you on observing the magic smoke hidden inside electronic components. If you are running your fets without a heatsink eventually they will get too hot and give up the ghost.

If you are applying a constant Vgs to your FETs and expecting them to provide a constant current, your FETs will succumb to thermal runaway. The threshold voltage of a FET trends negative for increasing junction temperatures. So if you apply a fixed VGS the fet heats up, resulting in a higher overdrive voltage (VGs-Vt) and sources more current, which results in more power dissipation, and the positive feedback cycle continues....

RE: MOSFETs in linear operation

Andrew J — Thu, 11 Aug 2022 15:56:23 GMT

I had to look up ZTC as well. Michael is making some useful points here: before blowing any more MOSFETs or spending money on buying more to blow in a similar manner, you should do a bit of research on power in MOSFETs and how to calculate the thermal resistance - i.e. how much it is going to heat up with the amount of power you are putting through it. If you watch this video, (YouTube) from 10:05 onwards, the chap explains how to select MOSFET based on a datasheet and how to calculate the thermal rise from the power consumed. It's very understandable and you'll get a feel on what to do in order to experiment further, without destroying your MOSFETs. Hint: at the sort of power you are talking about it won't be cheap as you'll need a big heatsink! However, you may be able to tailor your experiment down to manageable levels to at least work out that your approach is fundamentally sound and could scale up.

The other thing: don't get disheartened, we've all been there! Don't worry about putting your ideas down here, a picture of your circuit as you've envisaged it so far, what your experimenting with etc. If you want to do your own research to discover stuff/learn stuff etc, just say and we can give you a direction to look in rather than specific answers.

EDIT: by "not cheap" I mean you'll have to lay out a few pounds for heatsinking and maybe even a fan to keep experimenting at that power level. We're not talking hundreds of pounds (or equivalent) but maybe £20 pounds or so.