In comp.arch.embedded,sci.electronics.design, larwe@larwe.com (Lewin
A.R.W. Edwards) wrote:

>Hi Tauno,
>
>> play an AM radio. I'd put small capacitors directly across the amplifier
>> inputs, with maybe some series resistance.
>
>Check, I'll try this.
>
>> There are separate signal and power ground pins on the chip. If the
>> noise comes in due to a ground loop, it's tempting to separate them
>> and take the signal ground along the input cable from the same source.
>
>I'm not sure if this will achieve much. "Signal ground" on the SBC
>goes straight to chassis ground, which is physically screwed to the
>chassis. The LCD controller board's digital ground is also in direct
>contact with this chassis ground. There is no isolation.
>
>> To test for a ground loop, I'd disconnect the input and ground it to
>> the signal ground, being careful not to disturb the DC level on the
>> input.
>
>? I'm not sure what you mean by the latter part of this sentence. I've
>capacitor-coupled the inputs to the amp, so there should be no DC
>component. There certainly is a significant DC component on the output
>from the SBC.
>
>> A second test would be leaving the circuit as it is now, but feed the
>> power from a clean supply.
>
>I tried this, and it makes some small difference but not a whole lot.
>
>There is subtle trickery afoot here. If I touch a metal object to
>random spots around the chassis, I can make the noise much worse (just
>discovered that this morning).

   This suggests the amp chip/circuit is picking up RF interference
out of the air. A metal shield would then help, OTOH an audio amp
really shouldn't be THAT sensitive. Call Phillips and ask for an FAE
to help you solve your problem. They'll want to hear that you're
thinking of ordering really large quantities of this chip.
   With such sensitivity, it might be oscillating or on the verge of
oscillation. Look at various points with an oscilloscope. Try putting
a small inductor in series with the input pin, as shown in the TI/Burr
Brown INA103 and INA163 datasheets.

>> You do have both recommended bypass capacitors? Philips' data sheet
>
>Yep.

   Make sure all such caps are as physically as close to the chip as
possible, with traces as short as possible.

>Thanks for the suggestions.

-----
http://mindspring.com/~benbradley

Stefan Heinzmann wrote:
> Lewin A.R.W. Edwards wrote:
[...]

>>... so we have an internal audio amplifier
>> based on the Philips TDA7053A (selected because it's very simple - one
>> capacitor is about the sum total of the circuit).

[...]
> If the audio connection from the SBC to the Amp creates a ground loop, 
> try to break it. If that's impossible, put a differential amplifier in 
> front of your power amp (that's a simple OpAmp circuit). Connect the 
> ground from the SBC to the negative input of the diff-amp, not the 
> signal ground of the power amp. Alternatively use an input transformer.
[...]

I just noticed that the older TDA7053 (no A) was described as having 
differential input stages already. This suggests that you could break 
the connection between power ground and signal ground on the power amp 
PCB. The data sheet of the TDA7053A is different in this respect. It is 
not clear whether the separation of grounds is ok, or how much potential 
difference is allowed between them. Still, I think it is worth trying.

If this can be done, connect the Amp's signal ground to the ground of 
the audio circuitry on the SBC, using a wire in the cable that runs the 
audio signals from SBC to Amp.

Tell us whether and when you've cured it.

Cheers
Stefan

On 14 Jan 2004 19:42:57 -0800, larwe@larwe.com (Lewin A.R.W. Edwards)
wrote:

>but I didn't want the design
>complexity of I2C-controlled amplifiers or digital pots, so we just
>control the SBC's mixer output volume and the amplifier's
>characteristics are fixed.

If you're using a fixed amplifier gain then the amount that you're
amplifying the noise will be fixed also, and it's probably fixed high.
The noise level will be high even when you feed a low level signal
from the sbc.  So you have a low level signal and high level noise,
i.e., a poor signal to noise ratio.  The reduced complexity of the
amplifier gain circuit may not have been the best option.

Dan

"Lewin A.R.W. Edwards" <larwe@larwe.com> wrote in message
news:608b6569.0401150930.b4cd85b@posting.google.com...
> Hi Dilton,
>
> > I'd try wrapping a coil of wire around the audio cable in hook it in
series
> > with a capacitor. Maybe if you're lucky it will create a tank circuit
with
>
> You mean, take say 20 turns of wire-wrap wire, wrap it around the
> audio cable, and solder both ends to a cap?

Yeah, tell me if it works! LOL

Lewin A.R.W. Edwards wrote:
> 
>>There are separate signal and power ground pins on the chip. If the
>>noise comes in due to a ground loop, it's tempting to separate them
>>and take the signal ground along the input cable from the same source.
> 
> 
> I'm not sure if this will achieve much. "Signal ground" on the SBC
> goes straight to chassis ground, which is physically screwed to the
> chassis. The LCD controller board's digital ground is also in direct
> contact with this chassis ground. There is no isolation.

The chassis is not equipotential at this level of signals we are after.
There are surprisingly many different buzzes around the chassis. I mean
that the signal ground should be connected to ground point very near the 
circuit genereting the signal.


>>To test for a ground loop, I'd disconnect the input and ground it to
>>the signal ground, being careful not to disturb the DC level on the
>>input.
> 
> 
> ? I'm not sure what you mean by the latter part of this sentence. I've
> capacitor-coupled the inputs to the amp, so there should be no DC
> component. There certainly is a significant DC component on the output
> from the SBC.

That's OK. I mean that the input pins must not be connected directly to
signal ground, but via the coupling capacitors.


>>A second test would be leaving the circuit as it is now, but feed the
>>power from a clean supply.
> 
> I tried this, and it makes some small difference but not a whole lot.

That's good - you do not need to make big/expensive supply filtering.

HTH

Tauno Voipio
tauno voipio @ iki fi

Lewin A.R.W. Edwards wrote:

> I have an appliance built around an SBC and an LCD monitor, all in a
> metal housing. We can't direct-drive the LCD, because of
> resolution/timing issues; we use analog VGA output to drive the LCD
> via an analog-digital board. The inside is a terrifying hive of EMI;
> to get UL compliance we have to seal every hole with conductive tape.
> 
> The SBC's audio output doesn't deliver enough power to drive the
> internal speakers directly, so we have an internal audio amplifier
> based on the Philips TDA7053A (selected because it's very simple - one
> capacitor is about the sum total of the circuit). Now, we need to
> control speaker volume in software, but I didn't want the design
> complexity of I2C-controlled amplifiers or digital pots, so we just
> control the SBC's mixer output volume and the amplifier's
> characteristics are fixed.
> 
> The amp is powered from the same 12V rail that powers the LCD
> analog-digital board. The SBC runs off a separate 5V rail.
> 
> Problem is that high-contrast patterns on the LCD (e.g. the B&W
> stipple pattern XFree86 shows while starting :) cause a buzz in the
> amp output. This noise appears to be generated by the LCD controller
> board, not the SBC. So I'm working with a clean audio source; the
> noise is being picked up elsewhere.
> 
> I've:
> 
> * Decreased the input shunt resistors on the amp as far as practical
> while still maintaining a good volume range.
> * Moved the amp as far away as possible from everything else. This
> doesn't seem to make any noticeable difference.
> * Put a large bypass cap on the amp's power rails.
> * Disconnected the ground line on the input to the amp. This, plus my
> finger anywhere on the audio line, makes the unit into an AM radio
> tuned to WCBS New York.
> * Run the audio cable (from SBC to amp) through a fat ferrite bead.
> This was the best step I took so far.
> * Run the 12V line to the amp through another ferrite. This didn't
> make any noticeable difference.
> 
> * For test purposes, disconnected the internal amp and connected
> external amplified speakers. No significant noise. Still no noise even
> when I disassemble the external speakers and put the PCB inside my
> housing. That circuit is complex and has an unlabeled IC in it, though
> - I don't want to try to copy it.
> 
> The noise is still just outside acceptable despite my best efforts.
> So, I'm looking for other ideas on how to mitigate this noise. What
> else could I add to this circuit? Is there some better kind of audio
> power amp I could use?
> 
> Any suggestions appreciated. This project is kind of an interference
> nightmare, I poke it gingerly with sticks...

First, have a look at your grounding arrangement. Make sure you keep the 
different grounds separate. In particular, you should have a chassis 
ground, which is separate from analog (audio) ground, which is separate 
from digital ground. They should all be connected together, but in *one* 
place only. This might be a convenient spot close to the power supply.

Connect all cable shields on external cable connections to chassis 
ground, as close as possible near the cable entry into the chassis. Do 
not connect it to signal ground! If you need a signal ground connection, 
run a separate wire for it inside the cable. If this is done properly, 
it keeps the radio stations out of your box.

Watch the layout of your amplifier PCB. Are there any wiring loops that 
could pick up stray magnetic fields? Minimize the area of those loops. 
Magnetic fields can not easily be shielded off (unless you use expensive 
mumetal shields).

If the audio connection from the SBC to the Amp creates a ground loop, 
try to break it. If that's impossible, put a differential amplifier in 
front of your power amp (that's a simple OpAmp circuit). Connect the 
ground from the SBC to the negative input of the diff-amp, not the 
signal ground of the power amp. Alternatively use an input transformer.

Use a common mode choke on the input of the power amp. Use shielded 
wiring from the SBC to the power amp. Connect the shield to chassis 
ground on both ends.

Moral: You need to make sure noise currents of whatever origin *cannot* 
flow in the audio amplifier's signal ground wiring, as this tends to 
create noise voltages along the ground wiring which will be in series to 
the input signal and hence get amplified with it. A common error is to 
confuse signal ground and chassis ground. They serve different purposes 
and need to be kept apart.

Good luck!

Cheers
Stefan

Hi Tauno,

> play an AM radio. I'd put small capacitors directly across the amplifier
> inputs, with maybe some series resistance.

Check, I'll try this.

> There are separate signal and power ground pins on the chip. If the
> noise comes in due to a ground loop, it's tempting to separate them
> and take the signal ground along the input cable from the same source.

I'm not sure if this will achieve much. "Signal ground" on the SBC
goes straight to chassis ground, which is physically screwed to the
chassis. The LCD controller board's digital ground is also in direct
contact with this chassis ground. There is no isolation.

> To test for a ground loop, I'd disconnect the input and ground it to
> the signal ground, being careful not to disturb the DC level on the
> input.

? I'm not sure what you mean by the latter part of this sentence. I've
capacitor-coupled the inputs to the amp, so there should be no DC
component. There certainly is a significant DC component on the output
from the SBC.

> A second test would be leaving the circuit as it is now, but feed the
> power from a clean supply.

I tried this, and it makes some small difference but not a whole lot.

There is subtle trickery afoot here. If I touch a metal object to
random spots around the chassis, I can make the noise much worse (just
discovered that this morning).

> You do have both recommended bypass capacitors? Philips' data sheet

Yep.

Thanks for the suggestions.

Hi Dilton,

> I'd try wrapping a coil of wire around the audio cable in hook it in series
> with a capacitor. Maybe if you're lucky it will create a tank circuit with

You mean, take say 20 turns of wire-wrap wire, wrap it around the
audio cable, and solder both ends to a cap?

Lewin A.R.W. Edwards wrote:
> I have an appliance built around an SBC and an LCD monitor, all in a
> metal housing. We can't direct-drive the LCD, because of
> resolution/timing issues; we use analog VGA output to drive the LCD
> via an analog-digital board. The inside is a terrifying hive of EMI;
> to get UL compliance we have to seal every hole with conductive tape.
> 
> The SBC's audio output doesn't deliver enough power to drive the
> internal speakers directly, so we have an internal audio amplifier
> based on the Philips TDA7053A (selected because it's very simple - one
> capacitor is about the sum total of the circuit). Now, we need to
> control speaker volume in software, but I didn't want the design
> complexity of I2C-controlled amplifiers or digital pots, so we just
> control the SBC's mixer output volume and the amplifier's
> characteristics are fixed.
> 
> The amp is powered from the same 12V rail that powers the LCD
> analog-digital board. The SBC runs off a separate 5V rail.
> 
> Problem is that high-contrast patterns on the LCD (e.g. the B&W
> stipple pattern XFree86 shows while starting :) cause a buzz in the
> amp output. This noise appears to be generated by the LCD controller
> board, not the SBC. So I'm working with a clean audio source; the
> noise is being picked up elsewhere.
> 
> I've:
> 
> * Decreased the input shunt resistors on the amp as far as practical
> while still maintaining a good volume range.
> * Moved the amp as far away as possible from everything else. This
> doesn't seem to make any noticeable difference.
> * Put a large bypass cap on the amp's power rails.
> * Disconnected the ground line on the input to the amp. This, plus my
> finger anywhere on the audio line, makes the unit into an AM radio
> tuned to WCBS New York.
> * Run the audio cable (from SBC to amp) through a fat ferrite bead.
> This was the best step I took so far.
> * Run the 12V line to the amp through another ferrite. This didn't
> make any noticeable difference.
> 
> * For test purposes, disconnected the internal amp and connected
> external amplified speakers. No significant noise. Still no noise even
> when I disassemble the external speakers and put the PCB inside my
> housing. That circuit is complex and has an unlabeled IC in it, though
> - I don't want to try to copy it.
> 
> The noise is still just outside acceptable despite my best efforts.
> So, I'm looking for other ideas on how to mitigate this noise. What
> else could I add to this circuit? Is there some better kind of audio
> power amp I could use?
> 
> Any suggestions appreciated. This project is kind of an interference
> nightmare, I poke it gingerly with sticks...

I saved my ass once by amplifying some of the ground
noise and feeding it back inverted into the input
of the audio amp.  It worked amazingly well.

There is historical precedent for this technique.
Most old tube amps feed a little 60Hz from the filament
transformer back into the input to cancel out the
hum.

Hi Albert,

> you look at the Philips spec sheet:
> http://www.semiconductors.philips.com/acrobat/datasheets/TDA7053A_4.pdf
> you'll see that the part is intended for mobile applications using a
> battery which has a much stiffer, cleaner, supply voltage than a

The only reason we chose this is because it worked well on the
breadboard and it's both cheap and the circuit is very very simple.
This subassembly is assembled by hand, so simplicity is important.
Actually we were originally using the TDA7053, but it seems to be
impossible to buy that part, so I respun it for the 7053A.

The amp isn't always necessary. We have two sources of SBCs. One
source uses an LM4863 "Boomer" amp directly on the SBC, so an ext amp
isn't needed. When supply fluctuates, we use a different SBC that
doesn't have an on-board power amp.

> only 34 db.  If you absolutely must use the part then clean up the supply.
>  First I'd try a linear supply if you now use a switcher.  Or even a

Our appliance has an internal open-frame SMPS, Astec SA40-1313, with
+12 and +5 rails. We can't change this. I guess I could put say a 7809
on the 12V rail and power the amp from that +9V.

> 	This is a bit of black magic since we can't see your layout or shielding
> nor do we have any idea of what other digital noise abounds.  First
> overkill to find the noise threshold then back off on the filtering.

The whole thing is black magic really :) But overkill works for me.

> 	A last resort might be to add some ferrous material between the LCD and
> the amplifier as close as possible to the LCD.  Ferrite beads are useless
> at audio frequencies, you need mass.

The LCD is separated from the amp and other electronics by a 3mm thick
steel wall. The appliance is built around a subchassis. The LCD and
its inverter are screwed to one side of this wall, the electronics are
screwed to the other side, and a two-piece cosmetic painted metal can
surrounds the resultant assembly.

The source of the noise appears to be the LCD controller board, rather
than the LCD itself (this determined by selectively powering things
down and disconnecting cables).

Thanks for the reply!