High frequency PWM generation

Hi All ,

I am working with a 8 bit Atmega2560 uC, when operating at 16 Mhz
clock the maximum frequency of PWM I can generate is ~31Khz at 8 bit
resolution. Are there any external PWM generator ICs which I can use
to generate higher frequency PWM signals?

Also If I want to sin modulate the PWM with sin wave of 4khz then is
31khz (PWM Wave) enough?

Any guideline would be appreciated a lot!
BR
Rate

ratemonotonic wrote:
> 
> I am working with a 8 bit Atmega2560 uC, when operating at 16 Mhz
> clock the maximum frequency of PWM I can generate is ~31Khz at 8 bit
> resolution. Are there any external PWM generator ICs which I can use
> to generate higher frequency PWM signals?

There are quite many of the high res high speed analog and digital PWM 
ICs intended for the audio. Check the websites of ST, TI, Zetex, Freescale.

> Also If I want to sin modulate the PWM with sin wave of 4khz then is
> 31khz (PWM Wave) enough?

It depends. Generally, for the PWM the Nyquist condition applies in the 
same way as for the PCM.

> Any guideline would be appreciated a lot!

How much is "a lot of appreciation" ?


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

> Hi All ,
>
> I am working with a 8 bit Atmega2560 uC, when operating at 16 Mhz
> clock the maximum frequency of PWM I can generate is ~31Khz at 8 bit
> resolution. Are there any external PWM generator ICs which I can use
> to generate higher frequency PWM signals?
>
> Also If I want to sin modulate the PWM with sin wave of 4khz then is
> 31khz (PWM Wave) enough?
>
> Any guideline would be appreciated a lot!
> BR
> Rate
>

There are some other AVRs which has PLLs onchip.
AT90PWM3 has a 64 MHz PLL giving 250 kHz PWM frequency at 8 bit resolution.
There are some tinyAVRs with PLL as well.


-- 
Best Regards,
Ulf Samuelsson
This is intended to be my personal opinion which may,
or may not be shared by my employer Atmel Nordic AB 

ratemonotonic wrote:
> Hi All ,
> 
> I am working with a 8 bit Atmega2560 uC, when operating at 16 Mhz
> clock the maximum frequency of PWM I can generate is ~31Khz at 8 bit
> resolution. Are there any external PWM generator ICs which I can use
> to generate higher frequency PWM signals?

Yes, you could use a CPLD for this, for example

There are also some uC that allow variable-resolution PWM.
In your example you don't really need 256 steps of time precison,
if you can only get 8 pulses per sine.
You might choose 64 pulses of 32 step granularity, for a 'better' sine wave.

You can also generate Sine by step in both Axes - ie if you vary the PWM 
period per PWM cycle, yu can get higher precision sine frequencies.
What is your desired Frequency step size ?

Other uC have higher resolution PWM - in the same core you have
AT90PWM (but not with same code resource), tho this could become
your slave chip.
Look for Motor Control devices, for Infineon, TI, ADI, etc

> Also If I want to sin modulate the PWM with sin wave of 4khz then is
> 31khz (PWM Wave) enough?

Depends on what level of distortion is enough, and your step size needs.
What is the Sine output ultimately used for ?

-jg

On Nov 2, 7:07 pm, ratemonotonic <niladri1...@gmail.com> wrote:
> Hi All ,
>
> I am working with a 8 bit Atmega2560 uC, when operating at 16 Mhz
> clock the maximum frequency of PWM I can generate is ~31Khz at 8 bit
> resolution. Are there any external PWM generator ICs which I can use
> to generate higher frequency PWM signals?
>
> Also If I want to sin modulate the PWM with sin wave of 4khz then is
> 31khz (PWM Wave) enough?
>
> Any guideline would be appreciated a lot!

Lot of IC companies provide this.
Check TI, ADI, Infineon .

Karthik Balaguru

Hi jg,

> There are also some uC that allow variable-resolution PWM.
> In your example you don't really need 256 steps of time precison,
> if you can only get 8 pulses per sine.
> You might choose 64 pulses of 32 step granularity, for a 'better' sine wave.

Many thanks , Good suggestion about reducing the PWM resulution from 8
bit (256 steps), that is certainly a possibility. as you say ,the only
problem bieng that of the Sine wave resolution will reduce , e.g. if I
have 256 bit resolution then I can represent a better range of sin
wave values in my Look up table. A trade off.

> Depends on what level of distortion is enough, and your step size needs.
> What is the Sine output ultimately used for ?
I am actually sin modulation the PWM wave(s) , and the PWM waves will
be fed to full or half H-bridge depending on the application.

Also on the processor I am using I have found a fast PWM mode where
instead of a up-down counter(triangle), I can use a up counter(saw
tooth) and double the maximum frequency. Thats another possible
solution. I dont know the fundamentals of what the diffrence in sine
wave would be if I use a sawtooth instead of traingle.
I am in the processes of researching (googleing it!).

Thanks,
Rate

On Nov 2, 2:18 pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
wrote:
> ratemonotonic wrote:
>
> > I am working with a 8 bit Atmega2560 uC, when operating at 16 Mhz
> > clock the maximum frequency of PWM I can generate is ~31Khz at 8 bit
> > resolution. Are there any external PWM generator ICs which I can use
> > to generate higher frequency PWM signals?
>
> There are quite many of the high res high speed analog and digital PWM
> ICs intended for the audio. Check the websites of ST, TI, Zetex, Freescale.
>
> > Also If I want to sin modulate the PWM with sin wave of 4khz then is
> > 31khz (PWM Wave) enough?
>
> It depends. Generally, for the PWM the Nyquist condition applies in the
> same way as for the PCM.
>
> > Any guideline would be appreciated a lot!
>
> How much is "a lot of appreciation" ?
>
> Vladimir Vassilevsky
> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com

Many thanks, for the suggestions. Being a new engineer I literally "
appreciate a lot!" as a matter of fact you can actually feel the
appreciation if you close your eyes :-) .  Moreover I can buy you a
beer if we some day meet in our short life span.

ratemonotonic wrote:

> Hi jg,
> 
> 
>>There are also some uC that allow variable-resolution PWM.
>>In your example you don't really need 256 steps of time precison,
>>if you can only get 8 pulses per sine.
>>You might choose 64 pulses of 32 step granularity, for a 'better' sine wave.
> 
> 
> Many thanks , Good suggestion about reducing the PWM resulution from 8
> bit (256 steps), that is certainly a possibility. as you say ,the only
> problem bieng that of the Sine wave resolution will reduce , e.g. if I
> have 256 bit resolution then I can represent a better range of sin
> wave values in my Look up table. A trade off.

Correct, you can of course have a few tables, and choose more X and Y 
steps, as the Sine frequency drops and allows this.

> 
> 
>>Depends on what level of distortion is enough, and your step size needs.
>>What is the Sine output ultimately used for ?
> 
> I am actually sin modulation the PWM wave(s) , and the PWM waves will
> be fed to full or half H-bridge depending on the application.

dF Frequency Step Size ?

> Also on the processor I am using I have found a fast PWM mode where
> instead of a up-down counter(triangle), I can use a up counter(saw
> tooth) and double the maximum frequency. Thats another possible
> solution. I dont know the fundamentals of what the differnce in sine
> wave would be if I use a sawtooth instead of traingle.
> I am in the processes of researching (googleing it!).

The SqWave output does not 'know' it was generated one way or the other :)

Where triangle wave is used, is when you want multiple PWMs and some 
phase relationship between them. With triangle, two slice levels give 
symmetric dead-bands, whilst sawtooth has deadband on the ramp, and none 
on the flyback.

In your application, what does the H Bridge drive, and to what Power.
You may find the PWM frequency is 'self selected' by the Bridge power 
loss requirements.

You can also make a more complex Table, that stores both PWM% and 
Timebase, and that gives more precise frequency control, and
more precise voltage definition - if those matter.

-jg

On 4 Nov, 03:43, Jim Granville <no.s...@designtools.maps.co.nz> wrote:
> ratemonotonic wrote:
> > Hi jg,
>
> >>There are also some uC that allow variable-resolution PWM.
> >>In your example you don't really need 256 steps of time precison,
> >>if you can only get 8 pulses per sine.
> >>You might choose 64 pulses of 32 step granularity, for a 'better' sine wave.
>
> > Many thanks , Good suggestion about reducing the PWM resulution from 8
> > bit (256 steps), that is certainly a possibility. as you say ,the only
> > problem bieng that of the Sine wave resolution will reduce , e.g. if I
> > have 256 bit resolution then I can represent a better range of sin
> > wave values in my Look up table. A trade off.
>
> Correct, you can of course have a few tables, and choose more X and Y
> steps, as the Sine frequency drops and allows this.
>
>
>
> >>Depends on what level of distortion is enough, and your step size needs.
> >>What is the Sine output ultimately used for ?
>
> > I am actually sin modulation the PWM wave(s) , and the PWM waves will
> > be fed to full or half H-bridge depending on the application.
>
> dF Frequency Step Size ?
>
> > Also on the processor I am using I have found a fast PWM mode where
> > instead of a up-down counter(triangle), I can use a up counter(saw
> > tooth) and double the maximum frequency. Thats another possible
> > solution. I dont know the fundamentals of what the differnce in sine
> > wave would be if I use a sawtooth instead of traingle.
> > I am in the processes of researching (googleing it!).
>
> The SqWave output does not 'know' it was generated one way or the other :)
>
> Where triangle wave is used, is when you want multiple PWMs and some
> phase relationship between them. With triangle, two slice levels give
> symmetric dead-bands, whilst sawtooth has deadband on the ramp, and none
> on the flyback.
>
> In your application, what does the H Bridge drive, and to what Power.
> You may find the PWM frequency is 'self selected' by the Bridge power
> loss requirements.
>
> You can also make a more complex Table, that stores both PWM% and
> Timebase, and that gives more precise frequency control, and
> more precise voltage definition - if those matter.
>
> -jg
 hi jg,

>dF Frequency Step Size ?

I am wanting to generate two frequencies of sinus waves : 200 hz and
400hz for DQPSK modulation. Thus the Hbirdge is actually acting a a
radio transmitter with
an antenna attached to it as load. The step size of the sin LUT
parsing is dependent on the frequency at which I update the
comparator.

> Where triangle wave is used, is when you want multiple PWMs and some
> phase relationship between them. With triangle, two slice levels give
> symmetric dead-bands, whilst sawtooth has deadband on the ramp, and none
> on the flyback.

I do have two PWM signals , but I dont think there needs to be any
phase correctness between them , as I am using one PWM to produce the
+ve part of a bipolar sine wave and the other one to produce the -ve
part of the sine wave , I am doing this by using a Hbridge
controller , with the high side pulled high always and just driving
the low side fets with the PWMs. The H-Bridge controlled manages the
prevention of shoot-thorugh by swithching off the high side
apporpriately. Do you think that my assumption is correct?  I think I
might still have to have consideration for deadband to reduce power
loss?

> You can also make a more complex Table, that stores both PWM% and
> Timebase, and that gives more precise frequency control, and
> more precise voltage definition - if those matter.

Yes , the more and more I am reading about H-Bridge control and class
D amlifiers , the more your suggestions are getting more clear in my
head.

BR
Rate

ratemonotonic wrote:

<snip>
> 
> I am wanting to generate two frequencies of sinus waves :  ...

<snip>

Nasal or cardiac? ;-)

Michael