30 kHz / 3 Phase signal 1

[beerbaron]

Hi, I'm trying to generate 3 x 30kHz signal each phase seperated by 120 degrees. These 3 phase will be used to turn On/Off FETs that will power some high frequency transformers.

What I'm looking for is the lowest cost effective solution to generate the signals. I was thinking of using 3 Op-Amps but I'm a bit worried about the reliability/accuracy for a production product. I'm sure there are plenny of cheap PWM ICs that could do this more effectively, I just don't know them.

Here are my specs:
3 phase 30 kHz signals, seperated by 120 degrees
Accuracy of ±5%
Supply voltage 10V
Waveform: Square, Triangle or Sin, I just want the reference
Power Output: Very small as it will trigger FETs.

Anybody had anything to suggest?

 

[ScottyUK]

Square wave has to be the easiest, this is an almost trivial task using standard TTL logic. A simple counter and some simple logic would be a good starting point. Sine is possible but a lot more complex and not worth the hassle if you can use a square wave.


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[fsmyth]

Sort of related to something I am working on now.
Will post more recent efforts soon, but there are
some old reference drawings at:

http://hmin.tripod.com/als/fsmyth/pages/stepdrv3.html

<als>

 

[Skogsgurra]

Re "Power Output: Very small as it will trigger FETs"

Some gates need lots of charge and that translates into a high load if you run at 30 kHz. It depends on your FETs.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[geekEE]

What about the Linear Technology LTC6902 multiphase oscillator? I've used it for synchronizing switching regulators. It would make generating 3-phase outputs trivial, although you'd need some sort of gate driver if you need 10V output.

Glenn

 

[Warpspeed]

The LTC6902 has four phases, three phases are a bit more tricky to do with a counter and decoding.

How about a low end PIC microcontroller ?
One chip, low cost and simple.
And you could do neat things, like building in dead time.

 

[geekEE]

Warpspeed, the LTC6902 can be configured for 2, 3, or 4 phase operation, depending on the state of the PH pin. For 3-phase operation, you leave the PH pin floating.

However, I do agree that if you need things like deadtime, using an MCU would give much more flexibility.

Glenn

 

[sreid]

I like ScotyyUK's Idea. A three bit state counter:

Ph3 Ph2 Ph1

1 0 1
0 0 1
0 1 1
0 1 0
1 1 0
1 0 0

 

[beerbaron]

Wow thanks a lot guys, I think I'll go with a state counter as it's much cheaper then what the LTC6902 seems to be retailing at. I even thing I can make some kind of dead time with a state counter.

A uC would have been neat too but since it's a volume item I would have had to make a mask on the IC... which I try to avoid.

Anybody has recommendations on a cheap CMOS gate that could do the job?

BeerBaron

 

[Warpspeed]

It is going to require a clock, a counter, and whole bunch of logic gates or decoders to do this. Many chips, and a lot of board space. That is the problem.

A best way to build a state machine is to use either an EPROM or a programmable logic array chip, (or use a microcontroller). All of which will require individual programming.

 

[geekEE]

BeerBaron, would you really need to mask program the MCU? Even on the high-volume consumer electronics that I've worked on, the price difference didn't make up for the convenience of a flash-based MCU. And it's nice to be able change code in case of an "oops" that's found later.

Glenn

 

[sreid]

To include short dead times, the counter would have to be longer. Then a CPLD from, say, Altera would be my choice. Free design tools [really easy to use] and on board programming thru a serial link (two pins on the board, free programming software too).

http://www.altera.com/

 

[sreid]

I couldn't get to sleep. An 8 bit binary counter short cycled to 255 counts [0 to 254] is divisible by 3 [85 x 3] and has one illegal state [256]. Clock at 7.65 MHz [30 kHz x 255] and decode. Allows increments of 130 ns "Dead Bands."

 

[ScottyUK]

Have a little pink star sreid. It won't help you sleep but at least your effort is rewarded!


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[DPolak]

Hello.
TTL counter is the simplest way to go. Or just do PLD if you know how to program.
Yes, "dead" time is very important.

From TTL to 10V FET gate - use any low side FET driver from IXYS or Micrel. Do not use IXYS "4 series"!!!!
They have "5" now - improved.
Depending how big is your FET - need to use "big" gate driver.
For 30kHz and average 50A FET - 2 A driver should be OK.
"dead time" - start in 100nS-200ns range.
If you don't know what "dead time" is - Google.
"dead time" can be easy done with 1R, 1C, and diode.

This whole design is kind of strange- you will be driving 3 phase SQUARE(???) signals into one transformer? Or few transformers?

Are you sure you don't need 3 sine signals 120deg shifted? then PWM them with some higher frequency?

 

[Warpspeed]

Pulse transformers work just fine with square waves, particularly with a 50/50 duty cycle as this will be.

Saturated switching of the driver is also going to be a lot more efficient than some kind of linear power amplifier.