Need ideas for 8-channel output

With the speeds that slow, any microcontroller can demodulate the control signal and make 8 channels of PWM by bit banging.

VLV

Doug B. wrote:

On a sunny day (Wed, 1 Jul 2009 09:21:55 -0700 (PDT)) it happened "Doug B." wrote in :

Stndard way is use 2 quad opamps as 8 comparators, audio into one input of each, the other inputs conneted together to a sawtooth wave that you make with an other opamp integrator perhaps. The output will then be 8 PWM channels with 50% duty cycle for zero audio.

ch 1 input ---- - out -------- PWM ramp --- +

8 x

. . . . . . . . . . . . ---- 0V . . . . .-.------- . .-------- slice level if audio a bit negative . .

_______ ______ |__| |__| PWM out

Do you know how to make a ramp waveform?

The opamp output can directly drive a suitable MOSFET.

No micro needed.

Maybe that'll help you out.

you could make a board that holds a single Class D IC and power driving section, out source the boards to be etched and drill and simply make a form of card rack to support the number of channels desired.

This rack would have one main power supply for it.

Just and idea..

This might work:

Generate one common triangle or sawtooth, maybe 0.5 to 4.5 volts swing, whatever PWM frequency you like.

Each channel is a comparator driving a mosfet gate, directly or maybe through a gate driver chip.

Each comparator compares an input (in the 0-5 volt range) to the shared triangle thing, and generates a 0-100% duty cycle as the input ranges from 0.5 to 4.5.

Something like that, scaled to your exact needs.

John

Vladimir,

Yes, I guess I could use a controller with 8 output pins. I was thinking of one of the smaller PICs that only has 4 or 5 output pins available and trying to Mux the output. But a bigger controller would be simpler, I guess.

Thank you, doug

John and Jan,

Your suggestions of comparator + sawtooth/triangle generator are exactly what I had originally considered, but I got stuck on how to scale the audio properly to avoid the 50% PWM duty cycle with no audio signal present.

I am trying to do this with a single-ended supply. I know how to bias the audio up into, say 2.4 V (my current prototype biases the op amp using a reference based on a 2.4V zener), but that means the output with no audio present is also 2.4 V. Wouldn't the PWM be at 50% if I centered the triangle wave around the bias level?

I sense that there is a straightforward, standard way around this that I am not understanding.

Thanks for your help!

doug

I don't see what the big deal is? Why not use bjt's? It's not efficient but who cares? How much power are we talking about here?

Then I don't understand the transfer function that you want.

And please don't top post here. See what a mess it makes?

John

e r

ls

y e o

Peak power per channel is 25-30 W. Average is much lower - I'm not sure exactly how much - but could be in the low single digits as the input tends to be periodic pulses. As I mentioned, my friend already uses two fairly bulky car stereo amps to do the job. These are BJT-based, I assume, which explains the big heat sinks. I was hoping to get the same output (minus the audio quality) from a much smaller board and enclosure. I also have to admit that part of me wants to use a switched solution simply to learn how it works.

doug

ng

Jamie,

Thanks for the idea and document. I'll have a look, but full out Class D seems overkill for this (and frankly too hard for me to tackle at this point). For my prototype board, I did the design in Eagle and had a couple of boards etched by Batch PCB

Double-side and through-hole (so a lot of soldering pins for me), but they did a nice job. I eventually want to move up to SMD parts using some of the tiny soldering techniques that have been discussed here.

doug

s

John,

First, my apologies about the top post. I think your suggestion makes perfect sense and is probably what I want. I'm just not sure how to accomplish it. I'll try to explain better how I see this problem and where I am stuck :

The incoming audio will vary around by, say +2 to -2 volts around zero. I want to use single supply, so I need to bias the input up by, say, 2.5 volts. Now my audio signal is at 0.5 to 4.5 volts, centered at 2.5 V. That goes into one of the comparator inputs. The other is a triangle wave, also centered on 2.5 v. I understand how the comparator will give a duty cycle of 0 to 100% as the audio swings above and below zero.

But won't the duty cycle sit at 50% in the absence of any input signal? And if it does, how do I get rid of that, short of filtering it out, a la Class D? I should also say that I would prefer to use a single output transistor, with the load on the high side. No H bridge or half bridge. I guess this is where I'm stuck.

I had thought of the microcontroller solution because I can subtract out the bias in software and simply generate PWM output at the proper duty cycle. But I agree with you (and Jan) that I probably don't need to go that route. It's just that my analog skills are pretty basic.

Thank you for your suggestions.

doug

o t

ice

ut

a l

s,

1

wer

nals

an

as

way

-

the

t t

to

es

WM

r/

be

ut

Then what about

Basically you use switching. The mosfets are on or off. The duty cycle is varying so that the filtered output is the signal. The wiki page gives a graphic of exactly how it works for sinusoids. This is your best bet IMO. The only problem here then is the filtering as the pulsing is very simple and done with a simple uC. Because of the low frequencies it should be quite easy to do very accurately. As far as the filtering is concerned you might need to use a switching capacitor do reduce the size or potentially use some other method.

Also note that your limited by the maximum voltage. If you want a varying amplitude you gotta change the max amplitude of the mosfets or try some other techniques or use a variation of the PCM that is more complex(essentially reducing the on time of the mosfets so the peak power).

What's the transfer function? Do you want to rectify the audio and have the duty cycle be zero at zero input signal level, and increase as the rectified level increases?

I still don't understand how you want it to work.

Maybe you could post a chart of a few input and output cases.

John

John,

First, my apologies about the top post. I think your suggestion makes perfect sense and is probably what I want. I'm just not sure how to accomplish it. I'll try to explain better how I see this problem and where I am stuck :

The incoming audio will vary around by, say +2 to -2 volts around zero. I want to use single supply, so I need to bias the input up by, say, 2.5 volts. Now my audio signal is at 0.5 to 4.5 volts, centered at 2.5 V. That goes into one of the comparator inputs. The other is a triangle wave, also centered on 2.5 v. I understand how the comparator will give a duty cycle of 0 to 100% as the audio swings above and below zero.

But won't the duty cycle sit at 50% in the absence of any input signal? And if it does, how do I get rid of that, short of filtering it out, a la Class D? I should also say that I would prefer to use a single output transistor, with the load on the high side. No H bridge or half bridge. I guess this is where I'm stuck.

I had thought of the microcontroller solution because I can subtract out the bias in software and simply generate PWM output at the proper duty cycle. But I agree with you (and Jan) that I probably don't need to go that route. It's just that my analog skills are pretty basic.

Thank you for your suggestions.

doug

A class D amplifier IC is probably the easiest route.

Bob

On a sunny day (Wed, 1 Jul 2009 17:05:10 -0700 (PDT)) it happened "Doug B." wrote in :

Well if you only want output proportional to the absolute amplitude of the audio drive signal, that means zero for 0 V, 100% for -V and +V, then you could use a full wave rectifier before the comparator, and DC couple to it.

You need to use a bridged output if you plan on using single ended supply.

o t

ice

ut

a l

s,

1

wer

nals

an

as

way

-

the

t t

to

es

WM

r/

be

ut

Your questions and Jan's comments led me to realize that what I want is indeed a fully rectified audio signal, scaled into a 0-5V solution as you suggested. Something like this (view in courier or other fixed- width font):

in out || |-| |--| |--| |-| || || |-| |--| |--| |-| || || | | | | | | | | || || | | | | | | | | || . || | | | | | | | | || || | | | | | | | | || . . || | | | | | | | | || || | | | | | | | | || . . || | | | | | | | | || || | | | | | | | | ||

-----.-----. =3D> ||___| |__| |__| |__| |___||______||___| |__| | __| |__| |___|| . . .

scaling is off input vertical range is audio 0.5 - 4.5 V output vertical range is 0-13 V output horizontal range matches one full input cycle

What frequency PWM would you suggest, given the low input frequencies? I have read that 20kHz or higher is recommended to avoid the audible range, but is it really necessary to go that high?

doug

Sorry, I still can't understand what you want to do.

Jonh

."

e audio drive signal,

Yes, I think that is just the ticket. Thank you for your suggestions!

doug