Direct audio output from FPGA pins

I'm playing around with sigma-delat ADC and DAC for audio. It's amazing
how good this works without any active components. Just Rs and Cs.

The question now is: Can I hook a speaker direct to the output pins of
an FPGA?

The idea is to omit any passive components as the speaker acts as
low-pass filter and to use two pins, feeding one with the inverted
sigma-delta stream (or PDM), to build a bridge amplifier (class D?).
With 3V3 output swing this should result (on an 8 Ohm speaker) in an
output power of:

    Peff = Ueff^2/R = (Up/sqrt(2))^2/R =
         = (3.3V/2*sqrt(2)/sqrt(2))^2/8Ohm = 340mW

First question: Is the multiplication with sqrt(2) correct? The idea
is that it is a class D amplifier with rectangular output resulting
in a sine wave that is sqrt(2) times higher than Up of the square wave.

Second question: Is the inductive load from the speaker an issue for
the output drivers?

Martin

On Thu, 30 Jun 2005 14:52:21 -0700, Martin Schoeberl wrote:

> I'm playing around with sigma-delat ADC and DAC for audio. It's amazing
> how good this works without any active components. Just Rs and Cs.
> 
> The question now is: Can I hook a speaker direct to the output pins of
> an FPGA?
> 
> The idea is to omit any passive components as the speaker acts as
> low-pass filter and to use two pins, feeding one with the inverted
> sigma-delta stream (or PDM), to build a bridge amplifier (class D?).
> With 3V3 output swing this should result (on an 8 Ohm speaker) in an
> output power of:
> 
>     Peff = Ueff^2/R = (Up/sqrt(2))^2/R =
>          = (3.3V/2*sqrt(2)/sqrt(2))^2/8Ohm = 340mW
> 
> First question: Is the multiplication with sqrt(2) correct? The idea is
> that it is a class D amplifier with rectangular output resulting in a
> sine wave that is sqrt(2) times higher than Up of the square wave.
> 
> Second question: Is the inductive load from the speaker an issue for the
> output drivers?
> 
> Martin
 
I think the thing you have forgotten is the output driver resistance (or
drive current capability). Peak current would be ~400 mA! Would probably work
ok with about 20 outputs in parallel... 

(or one of those little PMOS/NMOS SOT23 MOSFET pairs)

Peter Wallace

Martin Schoeberl wrote:
> I'm playing around with sigma-delat ADC and DAC for audio. It's amazing
> how good this works without any active components. Just Rs and Cs.
> 
> The question now is: Can I hook a speaker direct to the output pins of
> an FPGA?
> 
> The idea is to omit any passive components as the speaker acts as
> low-pass filter and to use two pins, feeding one with the inverted
> sigma-delta stream (or PDM), to build a bridge amplifier (class D?).
> With 3V3 output swing this should result (on an 8 Ohm speaker) in an
> output power of:
> 
>     Peff = Ueff^2/R = (Up/sqrt(2))^2/R =
>          = (3.3V/2*sqrt(2)/sqrt(2))^2/8Ohm = 340mW
> 
> First question: Is the multiplication with sqrt(2) correct? The idea
> is that it is a class D amplifier with rectangular output resulting
> in a sine wave that is sqrt(2) times higher than Up of the square wave.
> 
> Second question: Is the inductive load from the speaker an issue for
> the output drivers?
> 
> Martin

TI have a range of Class D amplifiers, so their data gives good info
on Filterless Class D.

You are pushing things a little doing this straight into the FPGA [ I 
hope it's a cheap one :) ]
If you work 3.3V / 8 Ohms, that's 412mA, into a bridge load, you are 
asking for - from a single IO pin.... ?!

Possibly on a whole port/bridge side ?

-jg

So you would have one pin driving a High into the 8 Ohm speaker, while
the other pin sinks a Low.
No pin would ever be negative, but you have the drive impedance plus
the sink impedance in series with the speaker. Each of the three might
be around 8 Ohm, for a total of 24 Ohm, and a peak current closer to
100 mA.
If you pick the strongest output standard, you still will have limited
aplitude.
I agree with the suggestion of a low-cost Clas-D amplifier instead.
Peter Alfke

> So you would have one pin driving a High into the 8 Ohm speaker, while
> the other pin sinks a Low.
> No pin would ever be negative, but you have the drive impedance plus
> the sink impedance in series with the speaker. Each of the three might
> be around 8 Ohm, for a total of 24 Ohm, and a peak current closer to
> 100 mA.
> If you pick the strongest output standard, you still will have limited
> aplitude.
> I agree with the suggestion of a low-cost Clas-D amplifier instead.

OK, I forgot the output resistance. So let's suppose Imax of 100mA:

    Peff = (Imax/2)^2*R = 20mW

That's perhaps enough for a head phone.

However, the question if I will burn my FPGA still remains. Two issues:
The maximum current: There is no maximum rating in the data sheet of
the Spartan-3, in the Cyclone data sheet the absoulte maximum rating is
25mA.

The second: A speaker is an inductive load. When we switch the output
the energy stored in the L will still drive the current resulting in
high voltage peaks until the 'current finds it's way'. Will this damage
the output driver? Shall I use external diodes to VCC and GND like
the self-induction recuperation diode used on a relay?

Martin 

"Martin Schoeberl" <mschoebe@mail.tuwien.ac.at> wrote in message 
news:42c51713$0$12384$3b214f66@tunews.univie.ac.at...
>
> However, the question if I will burn my FPGA still remains. Two issues:
> The maximum current: There is no maximum rating in the data sheet of
> the Spartan-3, in the Cyclone data sheet the absoulte maximum rating is
> 25mA.
>
I used most of the Xilinx FPGAs over the years, and never broke any of them 
with a short to ground or Vcco. They just warm up a bit!
>
> The second: A speaker is an inductive load. When we switch the output
> the energy stored in the L will still drive the current resulting in
> high voltage peaks until the 'current finds it's way'. Will this damage
> the output driver? Shall I use external diodes to VCC and GND like
> the self-induction recuperation diode used on a relay?
>
Provided you never tri-state the output pins, there's always a path for the 
current. However, as you switch, this happens:-

Immediately before switching
Output_1    =>   Vout1   --this one is driving high
Output_2    =>   Vout2   --this one is driving low

Immediately after switching (current in speaker remains constant)
Output_1    =>   -Vout2          --this one now driving low
Output_2    =>   2Vcco - Vout1   --this one now driving high

Where Vcco is the IO supply. This assumes that the output resistance of the 
positive driving and negative driving IO pins stays constant with voltage. 
You could simulate with (or just read) the IBIS files to get exact answers. 
So, I reckon an 8 ohm speaker is gonna break your pins, unless you use a 
awful lot of them so that Rout << 8 ohms.
HTH, Syms. 

Martin Schoeberl wrote:
> I'm playing around with sigma-delat ADC and DAC for audio. It's amazing
> how good this works without any active components. Just Rs and Cs.
>
> The question now is: Can I hook a speaker direct to the output pins of
> an FPGA?

FWIW, we did a pattern generator based on an FPGA.  Some of the outputs
ran through RS-422 drivers.  As a goofy test, we cooked up a pattern
that would use PWM to play music through the diff output.  We used caps
between the diff driver and the speaker as a DC block.

It sounded like an AM radio, but it worked.

-a

No need for capacitors if your digital signal is dc-balanced, i.e. on
average the same length of time High as Low. That's the beauty of the
bridge output.
Peter Alfke

Andy Peters wrote:
> Martin Schoeberl wrote:
> > I'm playing around with sigma-delat ADC and DAC for audio. It's amazing
> > how good this works without any active components. Just Rs and Cs.
> >
> > The question now is: Can I hook a speaker direct to the output pins of
> > an FPGA?
>
> FWIW, we did a pattern generator based on an FPGA.  Some of the outputs
> ran through RS-422 drivers.  As a goofy test, we cooked up a pattern
> that would use PWM to play music through the diff output.  We used caps
> between the diff driver and the speaker as a DC block.
> 
> It sounded like an AM radio, but it worked.
> 
> -a

Martin Schoeberl (mschoebe@mail.tuwien.ac.at) wrote:
: I'm playing around with sigma-delat ADC and DAC for audio. It's amazing
: how good this works without any active components. Just Rs and Cs.

: The question now is: Can I hook a speaker direct to the output pins of
: an FPGA?

Others have commented on the mismatched driver and speaker resistance which will 
limit power transfer - you might get better results from a single raw IO pin using 
a piezo sounder, as found in 'novelty' cards and crystal earpieces - I don't have 
any numbers to hand but I'm guessing these will have a higher capacitance, lower 
inductance and higher resistance, so will still provide low pass filtering, but 
will represent better power transfer from the FPGA.

After all you can build an AM reciever using one of these that is audiable where 
the only powre source is 15 meters antanna wire...

---

cds

Martin Schoeberl wrote:

>I'm playing around with sigma-delat ADC and DAC for audio. It's amazing
>how good this works without any active components. Just Rs and Cs.
>  
>
The sigma-delta works really well, and the faster you clock it the more 
equivalent bits you get.  You'll have to parallel a bunch of pins to get 
enough drive for an 8 ohm speaker, and you'll still end up with a 
sizable loss due to the driver impedance.  You could add a simple 
transistor stage to each pin to boost the current up and still keep the 
parts count low.  The sigma-delta will drive a 600 ohm headset with no 
problem (I did that with the shortwave radio demo shown on my website).  
You can also drive a set of powered PC speakers through a single pin.

-- 
--Ray Andraka, P.E.
President, the Andraka Consulting Group, Inc.
401/884-7930     Fax 401/884-7950
email ray@andraka.com  
http://www.andraka.com  

 "They that give up essential liberty to obtain a little 
  temporary safety deserve neither liberty nor safety."
                                          -Benjamin Franklin, 1759