Hi. Does anyone know of a PWM controller where the pulses are interleaved 3-phase, and width controlled through an SPI/I2C/CAN or other serial interface. I want to build a 3-phase flybuck converter for minimum input current ripple. Ideally, it should also have 2 (or more) ADC channels. I accept that a PIC may be the best option.

 

I have a feeling you have answered your own question but there are some multi-phase digital controllers here http://www.ti.com/lsds/ti/power-management/digital-power-isolated-controller-products.page#
http://www.ti.com/lsds/ti/power-management/digital-power-isolated-controller-products.page#
Only others I can think of is LTC devices intended for multi-phase buck low voltage processor cores.

 

whats the freequency?
I ask because if its modsest, relative to your uP clock, you could simply do it in software with interupts or even a counter.

Al

 

Frequency about 300kHz. All the DC-DC converter chips that I have found have internal regulation. I want to access the secondary V & I remotely and vary the amount of energy transferred with each pulse. I mean over minutes, not pulse by pulse variation.

 

that wouldnt be easy unless the uP was crazy fast so I suspect software is a poor idea.
I dont know much about PSU design but I suspect that converter chips are available with clock sync capabilities.
Assuming that they are and use some form of PLL then you would need an external phased clock tom get 3 to work together.

They must exist as monolythic units, did that TI link not yeald anything?
it wouldnt load for me but my connection is horrible today

would a three phase pwm motor controller not work for you with an extermal feedback circuit, there are plenty of those tom be had,
( I may be talking nonesense here - not my thing, but it would be where I started looking )

Al

 

that wouldnt be easy unless the uP was crazy fast so I suspect software is a poor idea.
I dont know much about PSU design but I suspect that converter chips are available with clock sync capabilities.
Assuming that they are and use some form of PLL then you would need an external phased clock tom get 3 to work together.

They must exist as monolythic units, did that TI link not yeald anything?
it wouldnt load for me but my connection is horrible today

would a three phase pwm motor controller not work for you with an extermal feedback circuit, there are plenty of those tom be had,
( I may be talking nonesense here - not my thing, but it would be where I started looking )

Al

Any motor PWM's that I have found use the PWM to generate a sinewave, rather than repeatedinterleaved on time.

-------____________________-------___________________--
________-------____________________-------____________
________________-------___________________-------_____
is what I need, where the on-time is defined by a byte via SPI

 

I dont know enough about flyback circuits to give you a proper ansawer, although someone will and I will learn.

You are talking about 3 seperate transformers or at least an E core with 3 distinct winding pares...
I am pretty sure it will not work on a single core as a flyback is generating output as the flux colapses as far as I know.

Why flyback anyway, isnt that just adding complexity, given the requirement for a sense winding.

Would simple buck, or boost not be easier? for that matter what about push pull, there is a reason that almost all cheap inverters use this method.
whats your application?

Al
(anyone want to tell me how flyback actually works in pracgice, feel free, I will not be offended when I find I am missing some fundamental point as long as we dont go off topic and hijack the thread.)

 

I dont know enough about flyback circuits to give you a proper ansawer, although someone will and I will learn.

You are talking about 3 seperate transformers or at least an E core with 3 distinct winding pares...
I am pretty sure it will not work on a single core as a flyback is generating output as the flux colapses as far as I know.

Why flyback anyway, isnt that just adding complexity, given the requirement for a sense winding.

Would simple buck, or boost not be easier? for that matter what about push pull, there is a reason that almost all cheap inverters use this method.
whats your application?

Al
(anyone want to tell me how flyback actually works in pracgice, feel free, I will not be offended when I find I am missing some fundamental point as long as we dont go off topic and hijack the thread.)

Flyback dumps the flux energy in a short pulse, through a single rectifier, a bit like the old car ignition coils, but with the added rectifier. I want to use that sharp pulse, then read the steady state between pulses to determine how much power is needed over minutes/hours. Poly-phase reduces the input ripple current.

 

OK that makes sense, and it is colapsing flux then.
I get the polly phase, more overlapping pulses = less ripple but I dont see how that would work on a single coil... no collaps no output or perhaps less collaps less outtput.

I may be being thick but if you are using the pulse, for whatever, what is the steady stare telling you?
Is it that you are using some, variable amount, of the pulse and the steady stare represents the residual, if any, energy?

Al

 

OK that makes sense, and it is colapsing flux then.
I get the polly phase, more overlapping pulses = less ripple but I dont see how that would work on a single coil... no collaps no output or perhaps less collaps less outtput.

I may be being thick but if you are using the pulse, for whatever, what is the steady stare telling you?
Is it that you are using some, variable amount, of the pulse and the steady stare represents the residual, if any, energy?

Al

3 transformers. Surprisingly to some people, flyback transformers have air gaps, and most of the energy is actually stored in the gap.

 

Ahhh yes, the black art of magnetics, well beyond my ability to comprehend/calculate without help.

I have used them, even have a basic understanding of what they do in practical applications but when it comes to a design I would have to go with something proven that was doing a similar job and then tweeking it, poorly usually.

I am still blown away with the concept that magnetism cant be screened/blocked and by some of the odd stuff it will do, check out PMH on youtibe Perpetual Motion Holder.

Now clearly this is just an induced flux being maintained untill it is interupted mechanically or by an opposing field.
I Expect that the same effect was exploited in very early memory cores.
But what I find most interesting is the fact that it only works with iron/steal cores and not ferrite.
Most of the videos show a horseshoe core with one or more coils and a free bar/keeper.
Pulse the coil and the bar sticks, break the magnetic flux, by pulling off the bar, and a voltage is induced in the coil.
One guy shows that this will not work using a ferrite E core, and then entirly misses the point when adding a second coil.

My point is that magnetics are anything but intueative. I would accept that once created a field would exist untill it was disipated and I know that holding something in one place isnt work so you couldnt reasonably call that perpetual motion but I dont understand why it doesnt work with a ferrite core. I can however see that the air gap you describe may actually be a way to limit flux, causing a collaps that would otherwise not occur and that its dimensions would shape the decay. It also suggests that core material choice is critical given that a ferite core probably wouldnt need a gap if my asumptions are correct, or even close.

Sorry back to your question/project....
Have you considdered going oldschool, using a simple waveform, levels and delays?
3 astables in a loop would give you sequental pulses which could then be extended using more astables to give you interlived pulses.
The loop would use short asymetrical pulses with the negitive edge triggering the next pulse whilst the output pulse would be triggered by the positive edge of each respective loop pulse.
Freequency and phase would be fixed but output duty cycle could be varied.
In fact now I think about it you could probably do it with nothing more complex than 3 astables, 3 caps and 3 comparitors.
each pulse would charge a cap. when thew cap was above some threshold that output would be on.
The discharge rate would be constant which would mean that the charge voltage would covern the discharge time giving yiou a method to controlling all three pulse widths with a single voltage.

Just a thought

 

Reason not to go discrete: I will have several of these circuits, each of which will need to have their duty cycle varied by a central controller, almost certainly an Arduino. The duty cycle will be varied according to the V/I profile from the isolated load.

 

Any motor PWM's that I have found use the PWM to generate a sinewave, rather than repeatedinterleaved on time.

-------____________________-------___________________--
________-------____________________-------____________
________________-------___________________-------_____
is what I need, where the on-time is defined by a byte via SPI

This looks a lot like a speed controller on a brushless DC motor.

An Arduino Teensy can run at 70 MHz so can easily be bit-banged into 300 kHz PWM.

 

This looks a lot like a speed controller on a brushless DC motor.

An Arduino Teensy can run at 70 MHz so can easily be bit-banged into 300 kHz PWM.

It is, but the feedback that I want, isn't. BLDC controllers need position sensing, be it with discrete sensors, or back-EMF monitoring.