Looking for a PWM waveform generator chip 3

[gigahurts]

Hello, does anybody know of an integrated circuit that can generate a PWM waveform of about 500 kHz with a duty cycle that can be controlled between 0...100% with a DC input voltage (or maybe even via programming with digital input) ? This is for a new product, so I am looking for modern parts and a simple and manufacturable solution (I love good old geeky 555 circuits, but this circuit is for a medical instrument and not a breadboard-wired morse code practice keyer).

I would like to use it together with a MOSFET gate driver in order to build a switched-mode output stage because the commonly available switched-mode ICs from Linear, National etc. don't fit my requirements (I don't want to build a standard closed-loop regulated power supply, I need a functional block which takes a DC input control voltage from 0...5V and drives a buck converter output stage with a PWM duty cycle between 0...100%).
The frequency should be several 100 kHz (preferably 500 kHz) so that I can use small ferrites and smoothing caps.

I've already looked into a combination of a 555-timer wired as a sawtooth generator in combination with a fast comparator. But with today's huge gamut of PWM and switched-mode stuff I can't help asking myself if there's a more modern integrated solution out there with lower parts count and better performance ???

any tips are appreciated, thanks in advance

 

[jimkirk]

Microchip MCP1630 might be interesting. Pules Width Modulator, built in gate drive, operates up to 1 MHz. Designed to build switch mode power supplies, but very versatile.

Search

http://www.microchip.com/

for PWM, plenty of hits.

Also, many processors can generate PWM signals pretty easily, maybe not to the frequency and whatever resolution you need.

 

[gigahurts]

Dear Jimkirk,
I've had a first glance at the MCP1630 datasheet, it looks pretty interesting and it sure is pretty close to what I'm looking for. Thanks for pointing that out to me !
Concerning the direct use of a microcontroller for PWM waveform generation: I'm using an 8-MHz clock on my microcontroller, so unfortunately the highest PWM frequency that I can generate at 8-bit resolution is 8MHz/256 = 31 kHz
and that would mean fairly high values of L and C for my switched-mode power stage and bulky filters for getting rid of residual ripple (low noise is an issue in this application).

 

[treez]

hi

I think exactly what you are looking for is the dsPIC range from Microchip.....eg, the dsPIC30F2020

there are eval boards such as their sync buck board etc etc.

You can have excellent duty cycle resolution,
have multiple fast ADC's
fast instruction time so you can do software PID control.
-have comparators etc etc

you can command duty cycle by serial link...or by using the ADC...whatever....

also, the oscillator is independent...ie you just set it up in code and it runs on by itself...you just tell it when to change duty cycle/frequency whatever.

you can have antiphase outputs for synchronous rectification

There are four pairs of PWM outputs available.....each can be phased with respect to one another if you wish....dead time...leading edge blanking time.

just remember not to enable anything (eg PWM module, ADC) before youve set it up in code as it won't work.

It has a huge number of other features.

 

[gigahurts]

Hello treez,
thanks for pointing out the dsPIC series. I've just checked them out and indeed they would be great for my application but somewhat too complex and packed with features that I don't need. Maybe for a future project. Right now I'm still opting for the more simple MCP1630 which jimkirk mentioned.

 

[Warpspeed]

I have been through a similar search looking for a nice linear adjustable 0-100% duty cycle PWM chip fairly recently myself for a buck regulator project. This is not as simple as it first appears.

Many of the commercial switching power supply chips have some very serious limitations. The first is that very few of them actually cover the full 0-100% range. Even the MCP1630 uses a triggered flip flop, which means it will be triggered on-then off very fast at minimum duty cycle. It can never go completely to zero duty cycle, there will always be a residual very narrow spike at minimum duty cycle.

Another problem might be the linearity of the sawtooth waveform. Some of the better PWM chips have a decent internal current source to charge the timing capacitor, while others just use the timing resistor directly. A distorted sawtooth reference may, or may not be a problem for you. But it is something to think about if you are somehow programming a particular duty cycle externally.

From the data sheet, rather surprisingly, the MCP1630 seems to require driving by an external oscillator.

The third thing to be aware of is the input common mode range of the error amplifier. Some PWM chips can have a fairly restricted range of input voltage. Many do not go right down to zero, which may be very inconvenient.

I spent a lot of time investigating all this, and testing many commonly available switching power supply chips, and eventually gave up trying to find something that was simple to apply and did what I wanted.

What I ended up doing was using a function generator chip to create a nice, very linear high amplitude triangle wave. That then went to a voltage comparator with hysteresis. This in turn drove a proper high power MOSFET gate driver.

That gave me the best of everything, without having to compromise.

It ended up being a very simple solid design without any hidden problems, and it easily covered the whole 0-100% with excellent linearity.

None of this probably matters with a closed loop switching power supply, because the duty cycle never needs to go right down to zero, and the error amplifier inputs usually operate within a very narrow specified voltage range.

 

[gigahurts]

Hi Warpspeed,

interesting, I'm glad you confirm my suspicion that there's no obvious simple solution for my problem out there.

I also came up with the idea of building my own triangle-wave and comparator circuitry. I'm thinking of using a fast dual opamp for making a triangle oscillator (the standard triangle-wave generator circuit using one opamp as an integrator and the other as a comparator whose output is fed back to the integrator) and then using a fast LT1016 comparator for generating the PWM.
There used to be some nice oscillator chips out there that produced a triangle wave output directly but the ones I found were all not recommended for new designs. Can you tell me which function generator chip you used ?

I can confirm your experiences with the commercial PWM regulator chips concerning the limited duty cycle range. My first design for this project uses an LT1374 (I used a trick which I found in a Linear Technology AppNote for overriding the error amplifier output with my own control signal) which actually works quite nicely for duty cycles down to about 30%, below which the chip goes into a horrendous discontinuous mode operation which produces output spikes at a frequency of about 1kHz for which I had to use a ridiculously large choke and electrolytic as a brute-force filter !

 

[Warpspeed]

Function generator chips such as the venerable XR2206 and ICL8038 are now considered fairly old and slow, but are still useful for some things.
Something newer such as a MAX083 which can reach 20 MHZ looks a lot more promising.

http://www.dieelektronikerseite.de/Datasheets/Sonstiges/MAX038 (Maxim).pdf

-

The rest of the design should be fairly straightforward.

 

[gigahurts]

Hi Warpspeed,
OK, excellent. Thanks for the tip concerning the MAX083 !