Is anyone familiar with the IR2304 MOSFET driver?

[tim9000]

Hi All,

I'm not sure if this is the correct foray to discuss my questions about the specific IR2304 IC chip.

However, I've run into some difficulties making my first inverter and I've asked a range of questions on another website:
https://forum.allaboutcircuits.com/threads/using-the-ir2304-s-for-the-first-time-to-make-a-three-phase-inverter.155886/#post-1345395

(datasheet and circuit sketch attached to the linked thread)

but I've not had a reply.

I'm wondering if anyone knows from experience if the Vs pin goes between VB and COM? And if someone can elaborate on what the functionality diagram in the datasheet is with reference to?

I believe not what Hin must be pulsed to 'bootstrap' the capacitor, so I don't think I can really test the driver's circuit operation just using a DC source and voltmeter.

Thanks

 

[Baluncore]

You must turn on the low-side mosfet first, so as to charge the capacitor, only then can you turn on and test the high-side mosfet.
Maybe you could drag and drop your circuit for one phase onto your next post.

 

[tim9000]

Thanks Baluncore,

You must turn on the low-side mosfet first, so as to charge the capacitor, only then can you turn on and test the high-side mosfet.
Maybe you could drag and drop your circuit for one phase onto your next post.

That is really strange, how long does the low-side mosfet have to be ON for before I can trigger the high-side?

The circuit is approximately this:

With Vcc being 15V, and the COM being connected to the GND of a raspberry pi, similar for Hin and Lin to the GPIOs.

So do you know if Vs goes to GND when the HO is triggered? or is Vs always Vcc?

Thanks again!P.S.
I just tried triggering the lower mosfet first, however, I COULDN'T GET IT TO TRIGGER AT ALL!
I was expecting LO to be equal to Vcc, however, it was closer to 0V.
Also I noticed that Hin and Lin went between a couple volts and zero, I was expecting they would just be zero. Perhaps I damaged the chips?

P.P.S. it really bugs me that the output to the load was Vcc, as I would like the mosfet controller to be separate, as I expected it should be.

 

[Baluncore]

That is really strange, how long does the low-side mosfet have to be ON for before I can trigger the high-side?

Just long enough to charge the Vb capacitor. Initially, maybe 10 to 100 microseconds.
When LO goes high, the low-side mosfet is turned on. That pulls Vs to ground and charges the Vb capacitor through the diode from Vcc. That then supplies the capacitor voltage needed later to switch the high-side mosfet gate between off and on.

So do you know if Vs goes to GND when the HO is triggered? or is Vs always Vcc?

You are confused. I wonder what exactly you mean by triggered. Vs goes low when LO is high. Vs is the high-side mosfet source voltage which is the half bridge power output, not a supply.

With Vcc being 15V, and the COM being connected to the GND of a raspberry pi, similar for Hin and Lin to the GPIOs.

Hin and Lin control voltage thresholds are below 0.8 for low and above 2.3V for high. Your Rpi must supply that voltage.
Vcc is selected to be sufficient to fully turn on the mosfet gates.

Also I noticed that Hin and Lin went between a couple volts and zero, I was expecting they would just be zero. Perhaps I damaged the chips?

You must fully control the voltage on the Hin and Lin inputs. If input voltages do not change it is probably because you are not fully controlling them. The internal interlock between the Lin and Hin require those inputs be fully controlled at all times.
It is possible that your power supplies are not functioning as you expect, or that you have a wiring error in your prototype.

While the bridge is idle and not driving the load, you should turn low-side mosfets on, so as to maintain Vb capacitor charge in preparation for the high-side mosfet to be turned on.

P.P.S. it really bugs me that the output to the load was Vcc, as I would like the mosfet controller to be separate, as I expected it should be.

The high voltage supply will usually be significantly greater than Vcc, and often greater than the Vgs breakdown voltage of the mosfet. For higher voltages the supplies must be separate, and the Vb voltage must stay close to, and above Vs by Vcc.

 

[Tom.G]

Take a look at pin 1, labeled LIN. It is connected to HIN.
Pin 2 has the same labeling problem.
Confusion anyone?

 

[tim9000]

Take a look at pin 1, labeled LIN. It is connected to HIN.
Pin 2 has the same labeling problem.
Confusion anyone?

View attachment 237415

Ha! You're right! I didn't notice that. This bloody datasheet is a real turkey.
Oh God, I'm so confused now.

 

[tim9000]

Just long enough to charge the Vb capacitor. Initially, maybe 10 to 100 microseconds.
When LO goes high, the low-side mosfet is turned on. That pulls Vs to ground and charges the Vb capacitor through the diode from Vcc. That then supplies the capacitor voltage needed later to switch the high-side mosfet gate between off and on.You are confused. I wonder what exactly you mean by triggered. Vs goes low when LO is high. Vs is the high-side mosfet source voltage which is the half bridge power output, not a supply.Hin and Lin control voltage thresholds are below 0.8 for low and above 2.3V for high. Your Rpi must supply that voltage.
Vcc is selected to be sufficient to fully turn on the mosfet gates.You must fully control the voltage on the Hin and Lin inputs. If input voltages do not change it is probably because you are not fully controlling them. The internal interlock between the Lin and Hin require those inputs be fully controlled at all times.
It is possible that your power supplies are not functioning as you expect, or that you have a wiring error in your prototype.

While the bridge is idle and not driving the load, you should turn low-side mosfets on, so as to maintain Vb capacitor charge in preparation for the high-side mosfet to be turned on.The high voltage supply will usually be significantly greater than Vcc, and often greater than the Vgs breakdown voltage of the mosfet. For higher voltages the supplies must be separate, and the Vb voltage must stay close to, and above Vs by Vcc.

Thanks for some more insight into the operation and clarification about Vs.
So the Vb cap won't charge unless the LO mosfet has pulled Vs to ground, I didn't realize that, because I always measured Vs as being Vcc, and I was expecting it to be 0V (floating disconnected internally). I.e. there is Vcc volts coming out of the Vs pin on the chip.

You must fully control the voltage on the Hin and Lin inputs. If input voltages do not change it is probably because you are not fully controlling them. The internal interlock between the Lin and Hin require those inputs be fully controlled at all times.
It is possible that your power supplies are not functioning as you expect, or that you have a wiring error in your prototype.

While the bridge is idle and not driving the load, you should turn low-side mosfets on, so as to maintain Vb capacitor charge in preparation for the high-side mosfet to be turned on.

To be clear, I am confident that these voltages were coming from the driver, not the controller. Because I think I remember measuring the voltage of a '0' gpio pin and it was 0V and when I plugged the wire for the Lin or Hin it went to 1.4V or something. (this is just hearsay because I may be mis-remembering)

I'm going to completely rebuild the dirvers and MOSFET setup with some new leaded ICs I've ordered. Because the ones I previously used were surface mount ICs and were very hard to work with. And failing that, I'll might have to sort out using a larger voltage to drive Lin and Hin than the 3.3V coming from the GPIO of the raspberry pi (hopefully not).

Much appreciated

 

[Tom.G]

Because I think I remember measuring the voltage of a '0' gpio pin and it was 0V and when I plugged the wire for the Lin or Hin it went to 1.4V or something. (this is just hearsay because I may be mis-remembering)

1.4V sounds like a floating input. Check that the GPIO pin you connected to is configured as an Output and that there is a pullup to +Supply implemented. The pullup (resistor or transistor) may be internal and programmable or an external one may be needed.

Also check that a connection was really made! Use an Ohmmeter from driver chip pin to raspberry CPU chip pin. That is tough with surface mount stuff. Just the pressure of a meter probe may complete a faulty solder joint. If you have two meters available (and enough hands), measure the GPIO pin and the driver pin voltages at the same time.

Cheers,
Tom

 

[tim9000]

1.4V sounds like a floating input. Check that the GPIO pin you connected to is configured as an Output and that there is a pullup to +Supply implemented. The pullup (resistor or transistor) may be internal and programmable or an external one may be needed.

Also check that a connection was really made! Use an Ohmmeter from driver chip pin to raspberry CPU chip pin. That is tough with surface mount stuff. Just the pressure of a meter probe may complete a faulty solder joint. If you have two meters available (and enough hands), measure the GPIO pin and the driver pin voltages at the same time.

Cheers,
Tom

Food for thought, thanks.

If you both wouldn't mind, I'll draft a new circuit layout, more professional and I'll try and make this next one be 'better', can I post it here when I'm done for both of your peer-review?
Cheers

 

[Tom.G]

Looking forward to it!

 

[Baluncore]

This bloody datasheet is a real turkey.
Oh God, I'm so confused now.

Only the external labels were swapped on the diagram. It makes no difference to the product or terminal identification. The terms are correct throughout the rest of the data sheet.

If you are still confused about the application of the IR2304 or IRS2304 half-bridge drivers you should ask a specific question.

Now you need to test the circuit by controlling the Lin and Hin signals.
Check that you have tied the drain of the low-side mosfet to Vs, the source of the high-side mosfet as shown in the diagram. If you do not do that you will be unable to charge the Vb-Vs capacitor.

When pin Lin is low, Lo will be at Gnd which will turn off the low-side mosfet.
When pin Hin is low, Ho will be at Vs which will turn off the high-side mosfet.
Do not take pins Lin and Hin high at the same time.
When pin Lin is high, Lo will be at Vcc which will turn on the low-side mosfet which will charge the Vb-Vs capacitor.
When pin Hin is high, Ho will be at Vb which will turn of the high-side mosfet.

 

[Baluncore]

And failing that, I'll might have to sort out using a larger voltage to drive Lin and Hin than the 3.3V coming from the GPIO of the raspberry pi (hopefully not).

According to the datasheet 3.3V is sufficient to drive the half-bridge controller.
Make sure the Rpi outputs you use are able to pull the output voltage up. If the outputs are open drain outputs they will require a pull-up resistor to the output logic voltage supply.

 

[tim9000]

Only the external labels were swapped on the diagram. It makes no difference to the product or terminal identification. The terms are correct throughout the rest of the data sheet.

If you are still confused about the application of the IR2304 or IRS2304 half-bridge drivers you should ask a specific question.

Now you need to test the circuit by controlling the Lin and Hin signals.
Check that you have tied the drain of the low-side mosfet to Vs, the source of the high-side mosfet as shown in the diagram. If you do not do that you will be unable to charge the Vb-Vs capacitor.

When pin Lin is low, Lo will be at Gnd which will turn off the low-side mosfet.
When pin Hin is low, Ho will be at Vs which will turn off the high-side mosfet.
Do not take pins Lin and Hin high at the same time.
When pin Lin is high, Lo will be at Vcc which will turn on the low-side mosfet which will charge the Vb-Vs capacitor.
When pin Hin is high, Ho will be at Vb which will turn of the high-side mosfet.

Okay, this is a very useful sanity reference for me, when I've built the new circuit. And I'll endevour to ask specific questions and provide concrete details if issues persist. (I'll need to read through my ad hoc notes to see if I still have confusion.)

I think the safest thing to do would be to totally scrap my existing MOSFET wiring and start from scratch.

Looking forward to it!

I've just knocked it up, please tell me what you both think, hopefully you can open the picture in full size to see all the values.

Side note, I'm quite nervous about the inductors I've made. I bought three of the largest ferrous toroids I could from ebay, I think they were about 5-10cm in diameter from memory, and wound about 80 turns on each (measured three equal lengths of silicone insulated wire). I know they saturate at a couple amps at 50Hz, but I'm hoping they'll be good at several Khz PWM.

Thanks heaps fellas!

 

[Baluncore]

I can't actually resolve any of the details in that image.
PF does tend to blur what are otherwise clear .png images. You may need to post a more detailed image, in anticipation that PF will sacrifice the quality while increasing the size. PF seems to do that when converting from a compact B&W, or 4 bit colour image, to 24 bit colour.

 

[tim9000]

I can't actually resolve any of the details in that image.
PF does tend to blur what are otherwise clear .png images. You may need to post a more detailed image, in anticipation that PF will sacrifice the quality while increasing the size. PF seems to do that when converting from a compact B&W, or 4 bit colour image, to 24 bit colour.

The DC bus is 36V.

 

[tim9000]

I can't actually resolve any of the details in that image.
PF does tend to blur what are otherwise clear .png images. You may need to post a more detailed image, in anticipation that PF will sacrifice the quality while increasing the size. PF seems to do that when converting from a compact B&W, or 4 bit colour image, to 24 bit colour.

Or maybe this will work if you don't want to splice the three together.

 

[Baluncore]

1. You do not have to insert the picture into the post, you can simply attach it.

2. The IR2304 is a low current gate driver so you should select a mosfet with a low gate capacitance. That mosfet will need a fast parallel flyback diode as it is driving an inductive load.

3. The resistor between the driver and the mosfet gate can be 22 ohm. It is there primarily to prevent a parasitic ultrasonic oscillation between the driver and the mosfet gate which might damage the gate and cause electrical noise. If the resistor value is too high it may limit the driver current to the gate during transition.

4. I would not bother with the 1 MEG resistor as the driver will clamp the gate voltage.

 

[tim9000]

1. You do not have to insert the picture into the post, you can simply attach it.

2. The IR2304 is a low current gate driver so you should select a mosfet with a low gate capacitance. That mosfet will need a fast parallel flyback diode as it is driving an inductive load.

3. The resistor between the driver and the mosfet gate can be 22 ohm. It is there primarily to prevent a parasitic ultrasonic oscillation between the driver and the mosfet gate which might damage the gate and cause electrical noise. If the resistor value is too high it may limit the driver current to the gate during transition.

4. I would not bother with the 1 MEG resistor as the driver will clamp the gate voltage.

Ah, I should have remembered about attaching it, I was thinking it attached by drag and drop, woops.

Excellent, thanks so much for the feedback.

I am currently using IRF3205 NMOSFETs, what am I looking for on the datasheet to this end? And what range should I be hoping for?

The datasheet for the 3205 says:
total gate charge: 146 nC, gate to source charge 35 nC, gate to drain charge 54 nC
input capacitance 3247 pF
output capacitance 781 pF
reverse transfer capacitance 211 pF
reverse recovery charge 143-215 nC

...?

I was using some 1N5817 1A Shottkey diodes. Could you please suggest some? (Googling suggests: 1N4001?) Does the flyback diode need to be directly on-top of the MOSFET pins, or could it be on the circuit board of the chip?

Okay, I'll go with 22 Ohm gate resistors and not worry about the 1M Ohm

Much appreciated

 

[Baluncore]

I was using some 1N5817 1A Shottkey diodes. Could you please suggest some? (Googling suggests: 1N4001?) Does the flyback diode need to be directly on-top of the MOSFET pins, or could it be on the circuit board of the chip?

All mosfets have an intrinsic body diode, some have a fast Schottky diode included in the package. If they do not include the diode you will have to place one as close to the mosfet as possible. The common 1N4001 will NOT survive the inductive application. What inductor current do you expect?

You will need to browse the selection tables provided by mosfet manufacturers. Mosfets for motor drive applications will include a fast recovery schottky diode matched to the mosfet current and speed.

The flyback diode should be schottky so it will conduct before the body diode. It must be fast so there is minimum transition overshoot, which also explains why it must be close to the mosfet. The body diode in the IRF3205 may be sufficient, but I would expect to find a better mosfet in the selection tables.

The IRF3205 has a high capacitance because it's high current requires a large area die. Select a lower current device and you will find it will switch faster. Also, consider later generation mosfets. Alternatively, select a higher current half bridge driver.

Again, browse the selection tables provided by the mosfet manufacturers. Learn to read the data sheets.
Also, see the selection filters available on suppliers websites such as digikey, mouser, element14, RS etc.

 

[Tom.G]

1) The IRF3205 data sheet shows T_rr (reverse recovery time) as 104ηS. That would seem adequate for the motor since the schematic shows capacitors across the motor windings. Depending on the motor load current, the series inductors could cause trouble if they are less than about 6μH. That would put the peak of the first ringing pulse at less than the switching time of the body diode. I don't have a 'good' guess for the inductors, but it sure feels like the inductance of 80 turns on a toroid would be substantially higher, in the range of milliHenries.

2) Do not use electrolytic capacitors for the Boost cap, their high leakage current will limit the maximum pulse width.
2a) As @Baluncore said, drop the 1M resistors in the gate circuit; they are effectively the same as leakage current. (just don't turn on the 36V power without the driver chips being functional)

If you wish to calculate the needed value of the Boost caps, there is an explanation and formula on pg.6 of:
https://www.infineon.com/dgdl/Infin...N.pdf?fileId=5546d462533600a40153559f7cf21200

They talk about a slightly different chip but the basics are the same.

Cheers,
Tom

 

[tim9000]

All mosfets have an intrinsic body diode, some have a fast Schottky diode included in the package. If they do not include the diode you will have to place one as close to the mosfet as possible. The common 1N4001 will NOT survive the inductive application. What inductor current do you expect?

You will need to browse the selection tables provided by mosfet manufacturers. Mosfets for motor drive applications will include a fast recovery schottky diode matched to the mosfet current and speed.

The flyback diode should be schottky so it will conduct before the body diode. It must be fast so there is minimum transition overshoot, which also explains why it must be close to the mosfet. The body diode in the IRF3205 may be sufficient, but I would expect to find a better mosfet in the selection tables.

The IRF3205 has a high capacitance because it's high current requires a large area die. Select a lower current device and you will find it will switch faster. Also, consider later generation mosfets. Alternatively, select a higher current half bridge driver.

Again, browse the selection tables provided by the mosfet manufacturers. Learn to read the data sheets.
Also, see the selection filters available on suppliers websites such as digikey, mouser, element14, RS etc.

Okay, I'll learn to read the MOSFET datasheet in better detail and understand how they vary by model and application. But could you give me a rough idea of what is a fast and slow (small and large) capacitance, and which of the above properties is the relevant capacitance for this? I'm guessing input capacitance 3247 pF?

Also, I appreciate you're probably getting quite sick of me at this point, but this is just my FIRST but not only foray into building an inverter (I hope to build multi-level in future). For arguments sake, could you suggest a possible overkill-over spec fast Shottky diode? (I'll mount it directly on the MOSFET)
I'm looking online, and one I see that is 25nS t_rr (the NXP BAT86) but it's really expensive, and I'd very much like something cheaper.

Note, that this is for a BLDC motor, so I expect it will be significant inductance, but cannot quantify at this point.

1) The IRF3205 data sheet shows T_rr (reverse recovery time) as 104ηS. That would seem adequate for the motor since the schematic shows capacitors across the motor windings. Depending on the motor load current, the series inductors could cause trouble if they are less than about 6μH. That would put the peak of the first ringing pulse at less than the switching time of the body diode. I don't have a 'good' guess for the inductors, but it sure feels like the inductance of 80 turns on a toroid would be substantially higher, in the range of milliHenries.

2) Do not use electrolytic capacitors for the Boost cap, their high leakage current will limit the maximum pulse width.
2a) As @Baluncore said, drop the 1M resistors in the gate circuit; they are effectively the same as leakage current. (just don't turn on the 36V power without the driver chips being functional)

If you wish to calculate the needed value of the Boost caps, there is an explanation and formula on pg.6 of:
https://www.infineon.com/dgdl/Infin...N.pdf?fileId=5546d462533600a40153559f7cf21200

They talk about a slightly different chip but the basics are the same.

Cheers,
Tom

Ah, so I can use electrolytic between Vcc and COM, but use a ceramic between Vs and VB?

Yes, that is my gut feeling on the inductors too, I'm just hoping they don't saturate.

What are you envisioning to occur will happen with the 36V if one of the chips breaks? What will happen mid operation if one goes faulty? Is there a way of detecting and protecting for this scenario?

I'm looking at your infineon .pdf now, cheers.

As always, thanks!

 

[tim9000]

P.S. I can also see there is the 1N5817, 1N5818, 1N5819 range. But I'm not sure if dV/μs is related to t_rr, other than that, I can't see the reverse recovery time explicitly stated...Also, what reverse voltage does the flyback diode need to be rated for, 36V?

Cheers

P.P.S. actually I think I have found a cheaper BAT85 diode:
https://www.ebay.com.au/itm/Schottk...QgTmGQkspeRTILv27nm/7+mcmp8wCl&frcectupt=true

 

[Tom.G]

All mosfets have an intrinsic body diode, some have a fast Schottky diode included in the package... I'm looking online, and one I see that is 25nS trr (the NXP BAT86) but it's really expensive, and I'd very much like something cheaper.

Considering how cheap the MOSFETs are, I'm inclined to ignore adding a Schottky diode until it is proven they are needed.

Note, that this is for a BLDC motor, so I expect it will be significant inductance, but cannot quantify at this point.

Do you have a make & model of one you have in mind... and/or perhaps you can tell us what it will be driving at what speed.

Ah, so I can use electrolytic between Vcc and COM, but use a ceramic between Vs and VB?

Yes.

Yes, that is my gut feeling on the inductors too, I'm just hoping they don't saturate.

If they do, you can remove some turns.

What are you envisioning to occur will happen with the 36V if one of the chips breaks? What will happen mid operation if one goes faulty? Is there a way of detecting and protecting for this scenario?

That would leave the MOSFET gates floating which means they may partially turn on. If both the upper and lower ones turn on you have a short circuit across the 36V supply. A fuse would likely avoid a spectacular failure but not necessarily save any MOSFETs or the power supply.

P.S. I can also see there is the 1N5817, 1N5818, 1N5819 range. But I'm not sure if dV/μs is related to trr,

Me either, that's a new one for me. In SCRs there is a limit on dV/dT so they don't turn on and on dI/dT so they don't short. Just from the wording though, I speculate that is the fastest voltage rate-of-rise that doesn't destroy the device. If true, the reason is that conduction starts in a few tiny areas of the semiconductor and rapidly spreads throughout the body to carry the rated current. If a high current is forced thru these tiny conducting areas the temperature rises high enough to melt and electrically short the device.

Also, what voltage does the flyback diode need to be rated for, 36V?

General rule of thumb: Components should be rated at least twice what you expect them to handle. [/size]
That said, if there is a 'decent' reason for ignoring the 50% safety factor and it is for a short experiment or personal experimentation, go ahead and use what is available. Explicitly, I recommend at least a 50V rating (33% safety factor), or better yet 75V.

Cheers,
Tom

 

[Baluncore]

But could you give me a rough idea of what is a fast and slow (small and large) capacitance, and which of the above properties is the relevant capacitance for this?

It would be helpful if you specified the maximum output current that will flow in the mosfets.
You should also specify the switching frequency.

 

[tim9000]

Considering how cheap the MOSFETs are, I'm inclined to ignore adding a Schottky diode until it is proven they are needed.Do you have a make & model of one you have in mind... and/or perhaps you can tell us what it will be driving at what speed.Yes.If they do, you can remove some turns.That would leave the MOSFET gates floating which means they may partially turn on. If both the upper and lower ones turn on you have a short circuit across the 36V supply. A fuse would likely avoid a spectacular failure but not necessarily save any MOSFETs or the power supply.Me either, that's a new one for me. In SCRs there is a limit on dV/dT so they don't turn on and on dI/dT so they don't short. Just from the wording though, I speculate that is the fastest voltage rate-of-rise that doesn't destroy the device. If true, the reason is that conduction starts in a few tiny areas of the semiconductor and rapidly spreads throughout the body to carry the rated current. If a high current is forced thru these tiny conducting areas the temperature rises high enough to melt and electrically short the device.General rule of thumb: Components should be rated at least twice what you expect them to handle.
That said, if there is a 'decent' reason for ignoring the 50% safety factor and it is for a short experiment or personal experimentation, go ahead and use what is available. Explicitly, I recommend at least a 50V rating (33% safety factor), or better yet 75V.

Cheers,
Tom

It would be helpful if you specified the maximum output current that will flow in the mosfets.
You should also specify the switching frequency.

A bunch of great points here.

I appreciate that I can probably omit the flyback diodes but I want to make it uber spec'ed to be safe. I just bought 100x 1N4148 diodes. These apparently have a reverse recovery time of 8 ns and a more than ample 100V breakdown voltage.

As for the make and model of the bldc motor: it's a 48V, 1.8 kW, 4500 rpm Boma motor. So I take all of these with a grain of salt.

I hope I don't end up having to remove turns, but good point.

I am using a 20A DC circuit breaker, which I want to use the PWM reference level to limit within.

I'm hoping to have the MOSFETS output 10A but keep it less than 20A peak. As for the switching frequency, I wanted to keep it as high as possible. I was thinking anywhere from 20khz to 250khz depending on what was achievable.

Thanks and all the best

 

[Baluncore]

I appreciate that I can probably omit the flyback diodes but I want to make it uber spec'ed to be safe. I just bought 100x 1N4148 diodes. These apparently have a reverse recovery time of 8 ns and a more than ample 100V breakdown voltage.

Well at least they are cheaper than resistors. But negative Vpn temperature coefficient diodes do not work well in parallel because the first to conduct gets hot, hogs the current, then melts into a short circuit. That short circuit will then be connected across the supply when the other mosfet turns on.

The diode must carry the entire mosfet current. I would not expect a 1N4148 to survive the first kick. Over specified in some dimensions is not good engineering if you ignore the critical parameter. In motor control you should look at the repetitive surge current specification for the diodes. Sometimes the best guide will be the non-repetitive surge current specification. That tells you how fast the fuse must blow.

 

[tim9000]

Well at least they are cheaper than resistors. But negative Vpn temperature coefficient diodes do not work well in parallel because the first to conduct gets hot, hogs the current, then melts into a short circuit. That short circuit will then be connected across the supply when the other mosfet turns on.

The diode must carry the entire mosfet current. I would not expect a 1N4148 to survive the first kick. Over specified in some dimensions is not good engineering if you ignore the critical parameter. In motor control you should look at the repetitive surge current specification for the diodes. Sometimes the best guide will be the non-repetitive surge current specification. That tells you how fast the fuse must blow.

Oh dear, I didn't realize it needed to take the full load current. I have misunderstood the situation when the inductive flyback current occurs.
So does this take place during the dead-time? If so, the only current path (loop) I can see for it to circulate through the flyback diodes and load is also through the DC supply source. Is this correct?

Also, what would be a respectable reverse recovery time for a 30A diode, how about 90 ns?

Thank you

 

[Baluncore]

So does this take place during the dead-time? If so, the only current path (loop) I can see for it to circulate through the flyback diodes and load is also through the DC supply source. Is this correct?

Also, what would be a respectable reverse recovery time for a 30A diode, how about 90 ns?

That depends on the current being switched and the value of the output inductance into the 3uF capacitor. Faster diodes capable of switching greater currents cost more than you would be prepared to pay. They would probably be fabricated, not as a diode, but from a huge array of parallel transistors on one die.

These days people do not make decisions based on detailed specifications, they simply buy the brand that is acceptable to their peer group. Some maybe spend a little more to get better recognition in their social group. Unfortunately that attitude cannot be used to engineer a reliable circuit. I suspect you do not yet understand the complexity of the task you have undertaken or the process that must be followed to achieve that outcome.

The aim must be to design a reliable circuit that can be built at a reasonable cost. To do that, each component must be selected based on specifications. For each parameter there is an upper and a lower bound. Changing one parameter will influence the bounds of other parameters. Selecting optimum values is done by calculation. You have an output inductor, but with unspecified value. Why have you not specified that value? Is it because you do not understand how to apply V = L * di / dt, or do you not know why you need to do that?

There is a reason earlier generation parts are cheap. It is usually because there are significantly improved parts now available. You are basing your design on part number recommendations and low costs. That is not a good approach because it prevents adjacent components working well together. The cost of recovering from early poor decisions is the cost of meltdowns, which is greater than the cost of getting it right the first time.

As I understand it you have a 3 phase PM motor and will sense the position of the rotor while generating three phase drive voltages. But you have no direct control over current or torque because you are hiding your motor inductance behind an LC low-pass filter. That will make it inefficient and difficult to control. I believe the motor controller and drive switching needs to be more closely integrated with the inductance of the windings in the motor.

You have selected a high-voltage half-bridge controller with low output current for a motor control application where fast transitions are needed. You have selected a high current mosfet, with high capacitance, where a faster, lower current mosfet, with a fast integral diode would be an easier choice. Your interest in the 1N4148 signal diode as a flyback diode demonstrates that you need to study electronic engineering for at least a year before attempting to design a prototype switching motor controller. I am sorry, but I believe you are still in a social design mode, looking for happy choices, rather than understanding the engineering calculations that must be your focus if you are to assemble a controller.

Designing a complex motor controller from scratch is well beyond a thread on a forum. Rather than seeking part number and parameter recommendations, you need to browse hundreds of data sheets and digest the detail in scores of application notes. If you need help understanding why something needs to be done, or how to calculate a component value, then we can probably help you.

While studying electronic engineering you might halve your motor voltage and consider using a prefabricated and tested controller such as an H-Bridge PWM module based on the BTS7960. They should handle 40 amps at 28 volts, and cost less than $10 per channel, search on the web. That will give you a chance to get some hardware working so you can test your software and control systems.

 

[tim9000]

That depends on the current being switched and the value of the output inductance into the 3uF capacitor. Faster diodes capable of switching greater currents cost more than you would be prepared to pay. They would probably be fabricated, not as a diode, but from a huge array of parallel transistors on one die.

These days people do not make decisions based on detailed specifications, they simply buy the brand that is acceptable to their peer group. Some maybe spend a little more to get better recognition in their social group. Unfortunately that attitude cannot be used to engineer a reliable circuit. I suspect you do not yet understand the complexity of the task you have undertaken or the process that must be followed to achieve that outcome.

The aim must be to design a reliable circuit that can be built at a reasonable cost. To do that, each component must be selected based on specifications. For each parameter there is an upper and a lower bound. Changing one parameter will influence the bounds of other parameters. Selecting optimum values is done by calculation. You have an output inductor, but with unspecified value. Why have you not specified that value? Is it because you do not understand how to apply V = L * di / dt, or do you not know why you need to do that?

There is a reason earlier generation parts are cheap. It is usually because there are significantly improved parts now available. You are basing your design on part number recommendations and low costs. That is not a good approach because it prevents adjacent components working well together. The cost of recovering from early poor decisions is the cost of meltdowns, which is greater than the cost of getting it right the first time.

As I understand it you have a 3 phase PM motor and will sense the position of the rotor while generating three phase drive voltages. But you have no direct control over current or torque because you are hiding your motor inductance behind an LC low-pass filter. That will make it inefficient and difficult to control. I believe the motor controller and drive switching needs to be more closely integrated with the inductance of the windings in the motor.

You have selected a high-voltage half-bridge controller with low output current for a motor control application where fast transitions are needed. You have selected a high current mosfet, with high capacitance, where a faster, lower current mosfet, with a fast integral diode would be an easier choice. Your interest in the 1N4148 signal diode as a flyback diode demonstrates that you need to study electronic engineering for at least a year before attempting to design a prototype switching motor controller. I am sorry, but I believe you are still in a social design mode, looking for happy choices, rather than understanding the engineering calculations that must be your focus if you are to assemble a controller.

Designing a complex motor controller from scratch is well beyond a thread on a forum. Rather than seeking part number and parameter recommendations, you need to browse hundreds of data sheets and digest the detail in scores of application notes. If you need help understanding why something needs to be done, or how to calculate a component value, then we can probably help you.

While studying electronic engineering you might halve your motor voltage and consider using a prefabricated and tested controller such as an H-Bridge PWM module based on the BTS7960. They should handle 40 amps at 28 volts, and cost less than $10 per channel, search on the web. That will give you a chance to get some hardware working so you can test your software and control systems.

Not meaning to sound defensive or obstinate, but I am aware that constructing a professional inverter is above my current skill-set. However, it is something I feel strongly about doing so I'm extremely keen to learn. The fact that this inverter (discussed here) is for a bldc motor is incidental and I'll try to make mention of it a minimum. I appreciate that you might feel like you're holding my hand on a fools errand.

You are correct about the motor being difficult to control torque and speed, but the aim of the project is a crude application, where the PWM reference level is controlled by user. So I don't see it really mattering. I have actually written the code for directional control of the rotation, but I will only be using it in one direction.

I have actually spend many many hours working on the PWM controller for the BLDCM. I am at a stage where I have the hall sensors working, such that the controller outputs the switching pattern in response to the state of the Halls. I think it is about 400 lines of code from memory.

Because I want to get the motor working as soon as possible, I did want to start by using 3 H-bridges connected in WYE configuration, to drive the motor. However, the integrated controller on the bridges didn't like it. The model I was using was: https://www.ebay.com.au/itm/DC-50A-Double-IBT-4-Stepper-Motor-Driver-H-Bridge-PWM-Semiconductor-Cooling-AU/264020749641?epid=16025462981&hash=item3d78dced49:g:RzsAAOSwfplb3R35:rk:6:pf:0
Similar to the BTS7960, I think.

I really feel I must say however, that it is not my character to consider branding as a prime factor in choosing something. I tend to avoid the 'latest and greatest', I'd never buy an apple product, I drive a 2006 toyota camery and I've never fetishised what is popular with everyone else.

Yes, you'll think I'm unwise to jump into something like this without having read hundreds of datasheets, but have two young children and a job, so I don't have years to spend focusing on this project.

The reason why I have not specified the inductance is because although I could get out Grover's Inductance Calculations book and approximate the permeability and calculate what I expect the inductance to be. I'd not be confident in it, so I'd rather just build it and see if they saturate, and modify as necessary. Although I have an oscilloscope, and so could maybe measure di/dt, I don't have a high frequency power supply, so I don't think I could measure the inductance that way...actually come to think of it, maybe if I flicked a switch ON and OFF and triggered the oscilloscope on the current rise, that would give me di/dt? To be honest, I hadn't thought of doing L = V / (di/dt). I actually do need more practice using the oscilloscope, I've not set a trigger for many years, when I was still a student. Do you think I should try this?

I understand that engineering decisions are about being fit-for-purpose at a reasonable cost. However in this case, when I'm confident that I have a good enough handle on implementing the theory into construction, for my next project (which I won't go into) I'll be more happy to make a gold-plated version which is over-spec, which because this is a one-off and not a mass produced product, I can do. After I've learned from enough failures, I'll be willing to pay more for individual parts. The cost of the failures using cheap components is in a way subsidising the cost of learning timeframe. I actually have a quite a few high-ish voltage, high current MOSFETS which I reclaimed from industrial UPS system chargers where a capacitor failed etc. But I am saving them for when I have a much better idea of what I'm doing, and another project.

so with these faster switching, more expensive diodes, the form factor is different to a small signal diode. That is interesting. Does this mean that for a commercial domestic PV inverter, that it is highly probable that the MOSFET they would employ would either have a the diode built-in, or would just use the parasitic internal diode of the MOSFET?

I guess I'll have to see empirically if 90ns is fast enough. But regarding "why something needs to be done", so what is the path of the freewheeling current? It has to traverse the DC supply source doesn't it?

Cheers

 

[Baluncore]

But regarding "why something needs to be done", so what is the path of the freewheeling current? It has to traverse the DC supply source doesn't it?

Yes, the magnetic energy E = ½⋅L⋅I² becomes current, so it can get back to the supply capacitor, where it will be stored as energy E = ½⋅C⋅V².

The reason why I have not specified the inductance is because although I could get out Grover's Inductance Calculations book and approximate the permeability and calculate what I expect the inductance to be.

You have the Engineering all back to front. What value of inductance is required by the circuit to function correctly. What maximum current will flow in that circuit. What component is available that will meet those specifications.

 

[tim9000]

Yes, the magnetic energy E = ½⋅L⋅I² becomes current, so it can get back to the supply capacitor, where it will be stored as energy E = ½⋅C⋅V².You have the Engineering all back to front. What value of inductance is required by the circuit to function correctly. What maximum current will flow in that circuit. What component is available that will meet those specifications.

Ah, so that's the significance of the supply capacitor. I didn't realize that. So I suppose once I calculate the Energy stored on the Inductor, that defines the minimum size of my supply capacitor. I have heard that electrolytic capacitors have a higher ESR, do you think this makes them inappropriate for this application? (Since they need to charge fast?)

With respect, I appreciate that I have the construction method back-to-front, but I did say that I'd refine the necessary design, such as inductance, as I go, and that I do expect to have some teething issues (like blown MOSFETS) along the way.

Thanks again

P.S. What did you think of the ON/OFF switch methodology I thought of (previous post) for measuring di/dt of the inductors? I assume there is a better method though?

 

[Baluncore]

P.S. What did you think of the ON/OFF switch methodology I thought of (previous post) for measuring di/dt of the inductors? I assume there is a better method though?

I do not see the point. The inductance is specified in the data sheet, and you know the voltage. That only requires arithmetic.

 

[tim9000]

I do not see the point. The inductance is specified in the data sheet, and you know the voltage. That only requires arithmetic.

As I said, I purchased the large ferrite torroids and I wound them myself from silicone insulated wire. They do not have a datasheet.

 

[Baluncore]

They do not have a datasheet.

The torroid core will have a data sheet with an A_L value.
If you counted the turns, you can calculate the inductance. L = A_L ⋅ n²

 

[tim9000]

The torroid core will have a data sheet with an A_L value.
If you counted the turns, you can calculate the inductance. L = A_L ⋅ n²

One would like to think so, however I did ask multiple sellers, none of whom could provide the datasheet.
This may not be the exact one (as I just typed it in the search) but it will give you an idea:
https://www.ebay.com.au/itm/Power-T...bF8TqAALjW0K2FFuLU4y5MV7BstRA=&frcectupt=true

 

[Tom.G]

P.S. What did you think of the ON/OFF switch methodology I thought of (previous post) for measuring di/dt of the inductors? I assume there is a better method though?

Yes, that could be a practical way of measuring the inductance.
Other methods:

• Borrow or purchase an Inductance meter. An LCR meter that measures Inductance, Capacitance, Resistance is in the $150USD range. Depending on what type of circuitry you tend to work on they may not be used much, but when needed are a huge time saver. You can even measure the characteristic impedance of a transmission line with them!
  
• If you have a variable frequency Audio signal generator available.
  • Connect the inductor the the Audio generator thru a series resistor (100 to a few 1000 Ohms)
  • Connect a known value capacitor across the inductor. NOT an electrolytic, their tolerance is often -50% to +100% and are not stable.
  • Connect your 'scope across the inductor/capacitor pair to measure voltage
  • Vary the Audio generator frequency to find the peak reading on the 'scope
  • This is the resonant frequency of the LC tuned circuit, from which you can calculate the inductance using L= (1/(2⋅π⋅f))²/C
    Where L is in Henries, F in Hertz, C in Farads (oh, and π of course, 3.14...)

The core you specified is being sold for power transformer usage, meaning it is a low frequency core. At the switching frequencies you are considering, that core will likely have high losses and poor magnetic performance. Don't expect it to do much more than waste energy.

Cheers,
Tom

 

[Baluncore]

Tom.G is correct.
But there is no point measuring the inductance of a low frequency inductor at low frequencies, if you are going to operate it at higher frequencies where it will have significantly lower inductance, with a much higher loss. Iron powder cores will get too hot when PWM, driven by an H-bridge.

To uneducated sellers of cores, the word “ferrite” is closely associated with the term toroid. The inclusion of the word “ferrite” on those web sites get a greater hit rate and increase sales. It does not mean the material really is a high frequency ferrite.

The frequency characteristics of a core is determined by composition. Different composition materials have different standard colour codes. Green is not ferrite, it identifies the composition as iron powder. It has electrically conductive iron particles with a low frequency response in a non-conductive ceramic or resin binder. For higher frequency applications you need to specify a more appropriate electrically non-conductive ferrite based material.

It is not for us to recommend or guess what material you should use. It is for you to work out why you need an inductor in that circuit in the first place. If you need an inductor, a design engineer must select a material and core dimension based on design specifications.
I would throw out the inductors and drive the inductance of the motor directly with the H-bridge.

 

[Tom.G]

I would throw out the inductors and drive the inductance of the motor directly with the H-bridge.

If throwing out the inductors, watch out for those 3uF caps across the motor windings. At 20kHz they have 2.6Ω impedance, 14Amps from a 36Volt supply. At 250kHz, 0.21Ω, 170Amps. Ouch! Better turn off your smoke alarms.

Cheers,
Tom

 

[tim9000]

Yes, that could be a practical way of measuring the inductance.
Other methods:

• Borrow or purchase an Inductance meter. An LCR meter that measures Inductance, Capacitance, Resistance is in the $150USD range. Depending on what type of circuitry you tend to work on they may not be used much, but when needed are a huge time saver. You can even measure the characteristic impedance of a transmission line with them!
  
• If you have a variable frequency Audio signal generator available.
  • Connect the inductor the the Audio generator thru a series resistor (100 to a few 1000 Ohms)
  • Connect a known value capacitor across the inductor. NOT an electrolytic, their tolerance is often -50% to +100% and are not stable.
  • Connect your 'scope across the inductor/capacitor pair to measure voltage
  • Vary the Audio generator frequency to find the peak reading on the 'scope
  • This is the resonant frequency of the LC tuned circuit, from which you can calculate the inductance using L= (1/(2⋅π⋅f))²/C
    Where L is in Henries, F in Hertz, C in Farads (oh, and π of course, 3.14...)

The core you specified is being sold for power transformer usage, meaning it is a low frequency core. At the switching frequencies you are considering, that core will likely have high losses and poor magnetic performance. Don't expect it to do much more than waste energy.

Cheers,
Tom

I didn't expect the seller to fully understand the specific application of the toroid, but I shouldn't have taken "ferrite" for granted. And I didn't realize there was such a colour code. However, what confuses me, is that I bought 5, and out of the 5, two of them I wound as isolation transformers. Then I wired it up in such a way that there was one diode on the secondary of each core, and I used this to make a full wave DC rectifier. So one core would rectify one half of the wave, and the other core would rectify the other half of the wave. Now, I had 80 turns on each of the cores, yet they still seemed to saturate at about 1.5A, and would audibly hum. That is why I concluded they must be used for high frequency applications (not the 50hz I was rectifying).

 

[Tom.G]

two of them I wound as isolation transformers. Then I wired it up in such a way that there was one diode on the secondary of each core, and I used this to make a full wave DC rectifier. So one core would rectify one half of the wave, and the other core would rectify the other half of the wave.

AIEE! (look it up).
You would do better by using one transformer and a bridge rectifier.

 

[tim9000]

AIEE! (look it up).
You would do better by using one transformer and a bridge rectifier.

I know, I have a ton of bridge rectifiers, I am going to make a proper DC supply using darlington transistors. But the point of this was I wanted to see if it would work (I wanted to just use two diodes for a rectifier). And my point now is that I have no idea how these bloody cores are behaving.

 

[Baluncore]

And my point now is that I have no idea how these bloody cores are behaving.

You can get an introduction to the uses of magnetic saturation from the first chapter of “Magnetic Amplifiers” By Paul Mali, 1960. http://mirror.thelifeofkenneth.com/lib/electronics_archive/Magnetic_Amplifiers_Paul_Mali_1960_text.pdf
Those green cores might make a good magnetic amplifier if you put two next to each other sharing a common winding.

 

[tim9000]

You can get an introduction to the uses of magnetic saturation from the first chapter of “Magnetic Amplifiers” By Paul Mali, 1960. http://mirror.thelifeofkenneth.com/lib/electronics_archive/Magnetic_Amplifiers_Paul_Mali_1960_text.pdf
Those green cores might make a good magnetic amplifier if you put two next to each other sharing a common winding.

I know magnetic amplifiers quite well, but I'm not sure there is much practical modern application for them?

Realistically, it looks like I might have to sideline the existing green toroids until I get some time to do some testing on them and figure out what the go is, with them.

Okay I've done some very cure research, and it seems:
Gray 175 50 Khz to 500 Khz
Blue 100 500 Khz to 5 Mhz
Red 57 2 Mhz to 30 Mhz
Yellow 47 10 to 50 Mhz
Black 32 30 to 100 Mhz

as a rule of thumb?

Ideally I think:
https://www.ebay.com.au/itm/Magneti...283326415854?_trksid=p2385738.m4383.l4275.c10

would suit me best for 250khz. However, when factoring in the shipping, that is way too expensive for me. But what about:
https://www.ebay.com/itm/NEW-1PC-T300-2-imported-magnetic-ring-iron-core-magnetic-ring/112550397278?_trkparms=aid%3D222007%26algo%3DSIM.MBE%26ao%3D2%26asc%3D20131003132420%26meid%3D4e15f26d6f044a0c91fba72d2c7de169%26pid%3D100005%26rk%3D1%26rkt%3D5%26sd%3D283326415854%26itm%3D112550397278&_trksid=p2047675.c100005.m1851

I assume they mean 35 relative permeability between 250khz-10Mhz, but they say it can handle half a kW. Is 'Carbonyl E' a suitable material for such an application?

Thanks

P.S. Is there a simple setup for measuring the permeability of a core? Or you need to make a BH curve and figure it out.

 

[Baluncore]

P.S. Is there a simple setup for measuring the permeability of a core? Or you need to make a BH curve and figure it out.

You will be able to answer that yourself when you have read sufficient technical literature on magnetic circuits and materials.

 

[tim9000]

You will be able to answer that yourself when you have read sufficient technical literature on magnetic circuits and materials.

Okay, and did you have any thoughts on:

https://www.ebay.com/itm/NEW-1PC-T300-2-imported-magnetic-ring-iron-core-magnetic-ring/112550397278?_trkparms=aid=222007&algo=SIM.MBE&ao=2&asc=20131003132420&meid=4e15f26d6f044a0c91fba72d2c7de169&pid=100005&rk=1&rkt=5&sd=283326415854&itm=112550397278&_trksid=p2047675.c100005.m1851

 

[Baluncore]

Okay, and did you have any thoughts on:

Electronic engineering is not like playing with children's construction blocks.
You appear to be quite unaware that engineering involved calculations.
I cannot give my approval for specific components in an unspecified circuit.

 

[tim9000]

Electronic engineering is not like playing with children's construction blocks.
You appear to be quite unaware that engineering involved calculations.
I cannot give my approval for specific components in an unspecified circuit.

Okay, I'll try some quick maths, so taking: https://www.ebay.com.au/itm/Magneti...283326415854?_trksid=p2385738.m4383.l4275.c10

for example, because they give better data: "Toroid Core Size: OD = 61mm (2.4 inch) x ID = 36 mm (1.4 inch) x HT=12.7 mm (0.5 inch)Material Grade: Permeability of 10,000, Max Permeability 20,000
Saturation Flux density 4300 gauss at 10 oersted, 25 deg C
Residual Flux Density 800 gauss, Volume resistivity 15 ohms-cm, Curie temp 125 deg C.
Le = Effective Magnetic Path Length: 14.5 cm
Ae = Effective Cross Section Area: 1.57 cm2
Ve = Volume : 22.7cm3
Al = Inductance Factor : 13,690 nH/t or mH/1000turns
RoHS Compliant
Application / Uses :
W material with permeability of 10,000 is used as Common Mode chokes for 100 Khz to 1 Mhz.
It is also used in resonant circuit in 1 Khz to 250 Khz. "

Note, I don't understand if they're saying it can work for frequency range 100 k to 1 Mhz, but I'll assume so.
I believe 4300 gauss is equal to 0.43 T, which is my maximum flux density.

for my application I'll be using maybe 15A rated current, I can specify the number of turns I want, they give the relative permeability (I think it's relative) and effective length of the magnetic path 0.145m. If I take the permeability to be 10,000*4Pi*10^-7 = 0.0126 (H/m) and the number of turns is 80, therefore: B = μ*N*I/L = 104 T. So yeah, clearly something isn't right. I noticed this yesterday, but I put it down to hasty calculation.

P.S. I just did the T300-2 core too:

outside diameter - diameter = 28.2 cm

L = 2*Pi*(28.2/2) = 88.6 cm

u = 35*4*Pi*10^-7 = 4.4*10^-5 (H/m)

N = 80

I = 15 A

B = 4.4*10^-5*80*15/0.886 = 0.0596 T

Which looks okay, but they don't give a maximum gauss, so I don't know.

 

[Baluncore]

for my application I'll be using maybe 15A rated current

Maybe they will work, maybe they won't.

 

[tim9000]

Maybe they will work, maybe they won't.

Well it's only the T300 I would expect to work without saturating, but good to get another opinion anyway (especially as I am so inexperienced).

 

[jim hardy]

Then I wired it up in such a way that there was one diode on the secondary of each core, and I used this to make a full wave DC rectifier. So one core would rectify one half of the wave, and the other core would rectify the other half of the wave. Now, I had 80 turns on each of the cores, yet they still seemed to saturate at about 1.5A, and would audibly hum. That is why I concluded they must be used for high frequency applications (not the 50hz I was rectifying).

And my point now is that I have no idea how these bloody cores are behaving.

A transformer supplying a nontrivial half wave rectified load will saturate.
That's because during the half cycles that the load conducts, primary current is higher than the opposite half cycle..
That means - during those half cycles the primary IR drop is more, leaving less voltage across the transformer's inductance.
So the voltage across the inductance has a DC component.
and that drives flux up the BH curve and the core saturates .

see
http://support.fluke.com/find-sales/download/asset/2103547_a_w.pdf

Theory and analysis
Some older electrical devices use half wave rectifiers to reduce power consumption. An example would be an early hair dryer with a “high/low” switch. At low speed, a series diode allows the circuit to draw current on only half of the voltage cycle. At high speed, the switch shorts out the diode — to allow current to flow during the full cycle. These devices wreak havoc on ac power distribution systems, because they generate dc current in the half wave configuration. The dc current will unbalance the magnetic flux in the transformer and push the transformer core into saturation on one half of the current cycle. The process of going in and out of saturation will produce strange noises from the transformer core.