Electric Motor Secrets

[uusedman]

I am back from my trip and will start hacking away. will post new improved motor results.

[Jetijs]

Jetijs,

Here is my 2 cents on making FET's running cool. I have made many PCB's employing FETS, at first mine ran hot too, now most are only cooled by a small copper area around the transistor, as I solder the FET metal back side directly to a PCB copper area. But it of cause depend of the application and the quality of the FET.

I see 3 problems related to your hot FETs:

1. Drive of the UCC37321:
The data sheet states:
"The input stage of each driver should be driven by a signal with a short rise or fall time".

The optical forks are usually very slow (it will be with a passive 4K7 pullup), so you need to add a schmitt trigger between the fork and the UCC37321 to obtain clean fast switching.


2. Gate drive:
The gate capacitance is 4nF and you use a gate resistor of 82 ohms.
That gives a time constant of approx. 330ns which results in slow switching and lots of heat, I have seen this myself. As your driver is very fast and able of 9 Amps (good choice), you have to limit the current like you did, but the value must be much less, I would use 1,5 ohm resistor. See if this solves the problem else keep on reading.

I must admit I have not read all posts in this very long thread, but I say this anyway, as I'm not sure of the type and value of your decoupling of the UCC37321. Also the circuit build up is important when we go for very fast switching.

If you don't already employ surface mounted components I can recommend using a double sided PCB with ground plane on one side and signals on the other. Use ceramic multilayer chip capacitors very close to the power pins of the driver 100nF (loop length less than 15mm), the larger 1uf is also a ceramic, keep that within 30mm of the power pins.

Locate the driver close to the FET (less than 30mm) and use two individual traces from the output pins to the non-inductive smd resistor and gate. The design of the PCB layout is important to avoid unwanted spikes in the circuit, a PCB trace is also in most cases an inductor unless striplines are calculated. Keep current loop areas small (close to bifilar) If the PCB is OK, you don't need D1 and D2. The 12V drive is fine.

3. The FET itself:

If the FET still gets hot with correct gate drive, consider using a faster FET with a lower Rds on.

May I recommend the IPW60R045CP from infineon.
Main data:
Vds 650V
Rds on max 0.045 ohms
Id 60A
Rise time 20ns, fall time 10ns
Free datasheets on:
Datasheet archive (search, preview and download electronic components documentation) | doc.ChipFind.ru

I consider converting a step motor for some experiments, I will make the secondary winding turns 5 times the primary winding to follow the advice of Hector. This results in time compression for sure, and has potential to account to other positive effects as well.

Good luck with your continued work !

Eric

Ps. I have a question: Is the recovery optimal when only one battery is used ?

Thanks Eric
You are right about the slow optoswitch rise/fall times. Can you give me an example how to use a schmitt trigger between the optoswitch and MOSFET driver to solve this?

What do you mean with only one battery in your question?
Thanks,
Jetijs

[Tecstatic]

Use of a schmitt trigger to make slow flanks fast

To take the last part about my question first.

I have done some reading from the beginning of this thread, and got my question partly answered in post #408 and #421. What I still don't understand is the looping part, I have seen Hector state you can not loop directly back, also this setup does not result in resonant charging of the capacitor as I see it.

But I guess Peter will consider this out of topic here, I think he wants to follow a path to keep it simple and do the learning step by step.

Back to busyness:

Download the data sheet from the link I gave you:
Part: CD40106BC, Manufacturer: Fairchild

So we can refer to the same data sheet. The data sheet is only 5 pages, by reading it you can see the effect of a schmitt-trigger.

The 40000 logic series can be operated up to 15V, so you do not need an additional 5V regulator.

How to connect the IC:
use a 4K7 resistor to connect the output from the fork to pin 1 of the 40106.
connect pin 2 to pins 3, 5, 9, 11, 13.
connect pins 4, 6, 8, 10 and 12 to the input pin (2) of the UCC27321.

Short distance between pin7 (Vss) and pin 4 (AGND)
Use a ceramic 100nF to decouple the 40106 very close to pins 7 and 14.
connect pin 14 to your 12V supply.

The purpose of the 4K7 resistor is to protect the input diodes from input transients. If will not harm the timing.

I have one more advice for fast FET switching: The use of a negative PGND voltage (-3V).
This is the last touch I know for reducing switching losses in the FET.

4. Use of negative gate turn off:
If you want to do that, add a 10K pull down resistor between pin 2 and 4 on the UCC27321.

Disconnect pin4 and 5 from GND and connect them to -3V

Connect the outputs from the schmitt-triggers to pin 2 via a 3.0V zener diode in series with a 1K resistor. Cathode towards the 40106. put a 1nF capacitor parallel to the zener.

Decouple the -3V supply with a 100nF capacitor and parallel the capacitor with a low forward voltage schottky diode eg. BAT60A, cathode towards pin5 (PGND)

I have always used a power supply to deliver the -3V, but as it is normal to bootstrap positive supplies, why not do the same for the negative supply with a charge pump. This way you can generate the -3V with only 6 additional passive components.

If this seems confusing with the descriptions, let me make a suggestion:
Why do "free energy" experimenting using non-free programs. Take a look on the free "Kicad" schematic and PCB program.

It is intuitive to use, and very close to perfect for making diagrams and PCBs. It runs on both Linux, Mac and Windows. An alternative is to offer postscript files but they are bigger than the diagrams, and you can not change it for further work like you can by receiving compact Kicad sheet files.

Eric

[lighty]

@Tecstatic

While everything you wrote is more than true I suspect that most of it is way over head of average experimenter. Also, there are some variables that may occur that could confuse people if they do happen. For example gate series resistor value can be easily calculated but that calculation will be valid only for one value of current. In practice one would have to find the value of that resistor oneself. The easiest way would be to use multiturn non-inductive trimmer, adjust it and then observe input signal on the gate until one achieve as steep rise time as possible.

Also, SMD caps and resistors will help quite a bit with fast transients but I suspect most of people won't know what to do with them and how to solder them and not destroy them. Same goes with PCB- one surely must observe the length, width and geometry of gate tracks. As you very well know there are simple ways to compensate for the tracks inductivity but even the more experienced amateurs have problems grasping that. Also, I agree that ground plane would reduce some of the problems with transients. However, there are no visible transients on the driving side of the MOSFETs so I don't see any point of bothering with it. It simply doesn't look to me as noisy environment worth the effort of all of the precautions one would employ in commercial projects.

Negative voltage on the gate would help getting steeper faling edge but it matters only when MOSFETs are getting hot and when they are passing larger currents. Also, introducing negative gate signal would complicate schematic even further.

Of course I will again sound ominous and arrogant (as some people called me in the past) but I think all of the above is way above knowledge of average amateur. What you could do if you have time is to engineer everything and then simply offer people finished solution that they could copy. If you do that there is a simple solution for most of the driving problems you mentioned- you could use advanced drivers like ST series TD35x- they have Schmitt trigger input, adjustable delay, active Miller clamp (no need for negative drive) and ever desaturation protection (can also be used as overcurrent protection). Of course they can sink "only" 1A so you may add totem pole driver stage and negative voltage drive in order to be able to controll larger MOSFETs conducting larger current. I used TD351 and TD352 in comercial projects and they're amazing at what they can do when switching fast IGBTs (and MOSFETs) that are working near their maximum power ratings.

Also, optical switch rising time can be solved by using integrated reflective optical switches rather than slotted ones. Because one can use reflective tape it's also easier to adjust timing without any additional mechanical contraptions.

[Jetijs]

Eric, I am not sure I understood you correctly. I am not very good in electronic circuits as you are, I know only the basics and it is also hard for me to draw a schematic from descriptions. This is how I understood your descriptions:



Is this correct?
Thank you!
Jetijs

[Tecstatic]

Almost got it

@Jetijs

No need to excuse for your electronic skills. A friend of mine says that a sound foundation for success is 15% ingenuity and 85% hard work. You surely qualifies for that and have my full respect. (no intended judgement on our intelligence)

I know several persons being excellent doing something although they have no formal education in the field. Again good internet search skills and hard work is the key.

And you got it almost right.

Let the optical fork be connected as it has been before, just move the resistor on pin 1 from the fork emitter to fork collector.

Add a decoupling capacitor on the 40106. thats it :-)

@Tecstatic

While everything you wrote is more than true I suspect that most of it is way over head of average experimenter. Also, there are some variables that may occur that could confuse people if they do happen. For example gate series resistor value can be easily calculated but that calculation will be valid only for one value of current. In practice one would have to find the value of that resistor oneself. The easiest way would be to use multiturn non-inductive trimmer, adjust it and then observe input signal on the gate until one achieve as steep rise time as possible.

Also, SMD caps and resistors will help quite a bit with fast transients but I suspect most of people won't know what to do with them and how to solder them and not destroy them. Same goes with PCB- one surely must observe the length, width and geometry of gate tracks. As you very well know there are simple ways to compensate for the tracks inductivity but even the more experienced amateurs have problems grasping that. Also, I agree that ground plane would reduce some of the problems with transients. However, there are no visible transients on the driving side of the MOSFETs so I don't see any point of bothering with it. It simply doesn't look to me as noisy environment worth the effort of all of the precautions one would employ in commercial projects.

Negative voltage on the gate would help getting steeper faling edge but it matters only when MOSFETs are getting hot and when they are passing larger currents. Also, introducing negative gate signal would complicate schematic even further.

Of course I will again sound ominous and arrogant (as some people called me in the past) but I think all of the above is way above knowledge of average amateur. What you could do if you have time is to engineer everything and then simply offer people finished solution that they could copy. If you do that there is a simple solution for most of the driving problems you mentioned- you could use advanced drivers like ST series TD35x- they have Schmitt trigger input, adjustable delay, active Miller clamp (no need for negative drive) and ever desaturation protection (can also be used as overcurrent protection). Of course they can sink "only" 1A so you may add totem pole driver stage and negative voltage drive in order to be able to controll larger MOSFETs conducting larger current. I used TD351 and TD352 in comercial projects and they're amazing at what they can do when switching fast IGBTs (and MOSFETs) that are working near their maximum power ratings.

Also, optical switch rising time can be solved by using integrated reflective optical switches rather than slotted ones. Because one can use reflective tape it's also easier to adjust timing without any additional mechanical contraptions.

I'm not sure I agree that working solutions should be avoided just because they maybe take more than 2 minutes to understand and implement. We are here to learn something new. Considering the mechanical work Jetijs has accomplished, I think this is manageable also. But I will follow the wishes of Peter if he thinks it is too much.

I have seen many circuit layouts in my time from terrible to excellent. Imho we have a hen-egg problem here, I can't say whether Jetijs has made a good layout.

The FET does spend too much time in the linear region to get this heating problem, so presently the circuit has no sharp pulses.

With a 1,5 ohms gate resistor they will become much sharper and circuit noise may become a problem, this is the reason for the more complicated stuff presented in steps. For me it is the removal of gate charge thats the issue, in this case the faster the better as long as we stay within specifications. When the gate charge has been removed, the resistor value is not significant anymore. And yes you can overdo everything, but simply changing a resistor to hopefully get rid of the heating problem is not the worst problem I can think of.

And may I add I have gone from a small heat sink to using a PCB copper area as "heatsink" by using negative gate turn-off although the driver data sheet states it is not necessary. And the cost of the negative turnoff is less than the heat sink, not to speak of reduced consumption during the lifetime of use.

The document AVR040 located here explains some more in case someone is interested.

Atmel Products - Application Notes

No need to dis-pare if some parts are difficult, just browse the document, every understood detail is a step forward.

Having said that from a world I have worked in for many years, I often feel a bit depressive about all the new stuff I have been though the last 18 months.

The suppression that makes most people make incomplete or hard to understand descriptions, and the internet harassment from which I have had my fair share, makes it a difficult subject to learn fast.

I wish I had all the insights of you Lighty and a lot of other experienced and gifted people in this forum. Again hard (and fun) work is part of the key.

Eric

[lighty]

@Tecstatic

As I said, you're completely correct in your suggestions. And you're also completely correct in your wishing for everybody to learn something new. However, complex methods of MOSFET and IGBT switching do require some prior knowledge. People here are not stupid, they will learn about it eventually but if you look through all of the posts you will notice that people are not patient enough to spend time learning specialized topics like fast switching. In fact you would be surprised how many EE professionals are unwilling to do so.

That being said- I still doubt that driving of MOSFETs is main problem. To me it seems like the inherent problem of coils and armature acting as a choke not allowing for the fast rise of current. One can drive switching semiconductors as perfectly as possible and there will still be a problem of slower current rise. I cannot be 100% sure since troubleshooting things on a distance is somewhat tricky.

[Tecstatic]

@lighty

"To me it seems like the inherent problem of coils and armature acting as a choke not allowing for the fast rise of current."

Please explain how that can create a FET heating problem if the gate drive is OK ie. FET operating at the specified rise and fall times.

[uusedman]

guys, i got my motor to spin work with the electricity been collected on the back to a 12 V batter.

My reed switch is buring out when i am NOT collecting the electricity. When I open circuit, the power supply injects MORE power to the motor. I think i am on the right track.

I need something better than the reed switch, your suggestions is highly welcomed and asked.

[lighty]

Please explain how that can create a FET heating problem if the gate drive is OK ie. FET operating at the specified rise and fall times.

I got mixed up trying to follow this thread, remembering scope shots Jetijs provided while at the same time writing an article for magazine and getting pissed at my cats. You're absolutely correct, if MOSFET is fully opened then it doesn't matter what input current curve looks like.


@Jetijs
Could you please take scope shots of your gate driving signal? Just connect scope probe between gate and source and take screenshots in various situations. We still haven't seen any driving signal curves and I think it may give us better understanding of what's going on in your system.

[Jetijs]

uusedman, can you post the exact circuit that you are using right now?

Lighty, this is the scope shot across the gate and source of one of the driving MOSFETs:



This is at about 8k RPM and scope set to 0,5 mS. It is a perfect square wave at all the time from low RPM's to high RPM's. I used the V3 circuit with one of the V3 motors and the coils were wired in parallel (each coil has it's own MOSFET). The input current wave from stays the same, something like this:



This is with no recovery circuit attached.
Also what is interesting, if I load the motor down, the current waveform does not change, it stays in the same proportions only gets wider and higher. Thats odd. So I guess that the problem is not in the duty cycle or RPM's.

[lighty]

Driving signal seems OK. If voltage/div is 10 that makes it 12V signal which is more than OK.

Hmmm, what MOSFETs do you use and what is its On Rds? Also, do your protection diodes (transils) get warm as well as MOSFETs? And when you're saying MOSFETs are warm what temperature are we talking about? Is it mildly warm, can you easily touch it?

[Jetijs]

Lighty, I am using IRFP360 MOSFETS, they are rated 400v, 28A, 410W. ON Rds is 0.2 Ohms. You know, this time when I took the gate signal scope shots, the circuit was consuming about 5A and the MOSFET's did not heat up even after 3-4 minutes, they became just warm. I did not check the transil diodes, but one thing sure did heat up - the isolation diode that separates the power supply from the capacitor. I use four 1n5408 diodes in parallel for isolation, this should handle currents up to 12A. It is late, I will do more tests tomorrow. If the results will still be confusing, I will just make another circuit board and verify everything step by step. This can be done easily because almost all the parts from the current circuit board are removable and I can use them again in the new circuit.

[lighty]

Hm, so you have about 5W of power dissipation. Of course your signal is pulsed so it's even less. If you ask me everything works fine and the only thing to do would be to find MOSFET with smaller ON Rds.

As for recovery diodes heating up- you shouldn't connect diodes in parallel. They are not ideally same so one of them will always conduct before the other ones and thus dissipate more energy. What you could do to somewhat compensate is to put some small value resistor in series with each of the diodes. For example 0.1-0.2 Ohm will do fine, maybe even values of up to 1.5 Ohms (depending on the current because you need to get voltage drop of 0.4-0.5V). Of course since you're not dealing with sine wave but with steep rising time impulses you shouldn't use wirewound resistors because of their reactive inductivity. What you could do is to use non-inductive resistors (carbon, metal-film or specially wounded wire resistors).

Or better yet always use a single larger diode instead of trying to parallel smaller ones.

P.S.
I re-checked your schematic and realized you're not talking about recovery diodes but about isolation diode between (-) of battery and main collector capacitor. In this case you could use wirewound resistors of small value to equalize load on smaller diodes but you would unnecessarily dissipate energy on them and it would still not be ideal.

Use one big diode. The bigger the diode the slower it is but in this case its function doesn't require it to be very fast so you can use one larger rectifier diode rated for higher currents- they will have much less conducting resistance so they will dissipate less power.

Have you tried measuring current going through that diode? Use shunt and measure RMS current on your scope. Then you'll know better what you're dealing with there. If you're going to use series shunt resistors the voltage drop on them should be at least 0.4-0.5V so it would be good to know what current goes through diode in order to select proper resistance shunt.

But again- use single high current diode.

[Jetijs]

Lighty, I was not talking about the recovery diodes, for this test I did not use any recovery circuit. I was talking about the isolation diode that separates the power supply from the input capacitor. Anyway, your suggestion with resistors in series of the diodes will help also in this case. I will see if I can get a single powerful diode instead

About the MOSFETs, In V2 circuit I used IRFP450 MOSFETs and they worked just fine despite that their ON Rds (0.4 Ohms) is even higher than that of IRFP360 (0.2 Ohms).