[?] Audio amplifiers and GSM interference.

[David Chapman]

Hi,

Low-cost commercially available audio pre-amplifiers are usually very
vulnerable to interference from adjacent cellular phones switching, in
the UK, at around 217Hz.

When designing an audio amplifier with high immunity to such
interference, I can see the obvious value of adding suitable RF
filtering (ferrite chokes and shunt capacitors with low-Z in the UHF
band) to the input circuitry but wonder if a pre-amplifier using
differential inputs (FET or Bipolar ?) rather than single-ended would be
better. Transformer coupling of source to the amplifier input is another
possibility, of course.

To meet a current requirement that I have to meet, amplifier noise
should not be too much of a problem since the input signal levels are in
the region of 20-50mV pk-pk, with a bandwidth from around 300Hz to 5KHz,
and the output level required is around 2 volts pk.pk. into a nominal
600 ohm load.

Does anyone with practical experience of dealing with this problem
have any comments that they'd care to make about how to minimise such
interference?

TIA - Dave

David C.Chapman - Chartered Engineer. FIEE. ([email protected])

 

[Pooh Bear]

Hi,

Low-cost commercially available audio pre-amplifiers are usually very
vulnerable to interference from adjacent cellular phones switching, in
the UK, at around 217Hz.

When designing an audio amplifier with high immunity to such
interference, I can see the obvious value of adding suitable RF
filtering (ferrite chokes and shunt capacitors with low-Z in the UHF
band) to the input circuitry but wonder if a pre-amplifier using
differential inputs (FET or Bipolar ?) rather than single-ended would be
better. Transformer coupling of source to the amplifier input is another
possibility, of course.

To meet a current requirement that I have to meet, amplifier noise
should not be too much of a problem since the input signal levels are in
the region of 20-50mV pk-pk, with a bandwidth from around 300Hz to 5KHz,
and the output level required is around 2 volts pk.pk. into a nominal
600 ohm load.

Does anyone with practical experience of dealing with this problem
have any comments that they'd care to make about how to minimise such
interference?

Turn the damn phone off !

That stuff gets *everywhere*.

What kind of audio are you working with that only has 5kHz bandwidth btw ?
Telecoms ?

Graham

 

[David Chapman]

Turn the damn phone off !

Obviously, but that's not possible if there's no PSTN or alternative
means of communication where it is being used.

That stuff gets *everywhere*.
Yup.

What kind of audio are you working with that only has 5kHz bandwidth btw ?
Telecoms ?

I only need to record the audio received by the GSM phone, and since
the highest audio frequency that GSM can pass is little more than 3KHz
the amplifier doesn't need to provide a flat response out to 20KHz!

- Dave

David C.Chapman - Chartered Engineer. FIEE. ([email protected])

 

[Poldie]

Hi,

Low-cost commercially available audio pre-amplifiers are usually very
vulnerable to interference from adjacent cellular phones switching, in
the UK, at around 217Hz.

This might not help/be interesting, but every single phone I'd had
except for my current one (Nokia 6230) had had some sort of audio
effect on radios, hifis and my pc monitor. Perhaps there's less of a
problem because of the attention paid to reducing potential brain
damage?

 

[Pooh Bear]

This might not help/be interesting, but every single phone I'd had
except for my current one (Nokia 6230) had had some sort of audio
effect on radios, hifis and my pc monitor. Perhaps there's less of a
problem because of the attention paid to reducing potential brain
damage?

You mean the brain daamge resulting from talking on them too long ? ;-)

Graham

 

[Dave Higton]

In message <[email protected]>

This might not help/be interesting, but every single phone I'd had
except for my current one (Nokia 6230) had had some sort of audio
effect on radios, hifis and my pc monitor. Perhaps there's less of a
problem because of the attention paid to reducing potential brain
damage?

For any given "victim" device, the only determining factors are the
strength of the interfering field at the victim, and the frequency.
Victims are only broadly tuned, so the frequency only really depends
on which band the phone is operating in (i.e. 900 or 1800 MHz etc.)

Dave

 

[Dave Higton]

In message <[email protected]>

Hi,

Low-cost commercially available audio pre-amplifiers are usually very
vulnerable to interference from adjacent cellular phones switching, in
the UK, at around 217Hz.

When designing an audio amplifier with high immunity to such
interference, I can see the obvious value of adding suitable RF
filtering (ferrite chokes and shunt capacitors with low-Z in the UHF
band) to the input circuitry but wonder if a pre-amplifier using
differential inputs (FET or Bipolar ?) rather than single-ended would be
better. Transformer coupling of source to the amplifier input is another
possibility, of course.

To meet a current requirement that I have to meet, amplifier noise
should not be too much of a problem since the input signal levels are in
the region of 20-50mV pk-pk, with a bandwidth from around 300Hz to 5KHz,
and the output level required is around 2 volts pk.pk. into a nominal
600 ohm load.

Does anyone with practical experience of dealing with this problem
have any comments that they'd care to make about how to minimise such
interference?

1) Put it in a screening case (i.e. conductive - preferably metal)
and take care that seams can under no circumstances fail to make
contact over a length exceeding about 40 mm. Pay particular
attention to paint or any other surface finish that will prevent
conduction, and to the fact that the edges aren't flat. Assume
there will be no contact except at points that are pressed together.

2) Take inputs and outputs via screened cables. Make the screens
off such that the RF interference remains on the outer skin of the
case. Once you take the interference to the inner skin, it's all
over.

3) Lay the PCB out with a complete ground plane.

4) Pay great attention to the spurious components, most notably the
inductance of all tracks. This means keep everything as small as
possible. Surface mount is a great help in this respect. The
placement of suppression components is as important as their values.

5) You are unlikely to need ferrites, which is a good thing as they
are relatively expensive, if you've done everything else right.

Dave

 

[Chris Jones]

Hi,

Low-cost commercially available audio pre-amplifiers are usually very
vulnerable to interference from adjacent cellular phones switching, in
the UK, at around 217Hz.

When designing an audio amplifier with high immunity to such
interference, I can see the obvious value of adding suitable RF
filtering (ferrite chokes and shunt capacitors with low-Z in the UHF
band) to the input circuitry but wonder if a pre-amplifier using
differential inputs (FET or Bipolar ?) rather than single-ended would be
better. Transformer coupling of source to the amplifier input is another
possibility, of course.

To meet a current requirement that I have to meet, amplifier noise
should not be too much of a problem since the input signal levels are in
the region of 20-50mV pk-pk, with a bandwidth from around 300Hz to 5KHz,
and the output level required is around 2 volts pk.pk. into a nominal
600 ohm load.

Does anyone with practical experience of dealing with this problem
have any comments that they'd care to make about how to minimise such
interference?

TIA - Dave

David C.Chapman - Chartered Engineer. FIEE. ([email protected])
--------------------------------------------------------------------------
CHAPMAN ASSOCIATES is a Consultancy offering practical expertise and
design skills in the fields of counter-surveillance, electronic protection
and security. Visit our Web site at http://www.minda.co.uk
--------------------------------------------------------------------------

As someone else said, use a metal box for the amplifier. You must prevent
the RF from getting inside the metal box which means as the other poster
suggested that the seams must be well connected along their length, and
also you must pay attention to the wires or connectors going into or out of
the box. The ideal situation would be to use 100pF feedthrough type
capacitors on all of the signal lines going in or out of the box, but these
are expensive and a similar effect might be achievable by using the
groundplane of a PCB as one face of the shielding box, with the circuitry
surface mounted on the side of the PCB which is inside the box, and then
feeding the signals through vias from the other side of the PCB, with 4
surface mount caps mounted radially connected between the via and the
ground plane.

If extreme protection is not required, then you could try just building the
amplifier on a PCB with a good ground plane (none of this star grounding
stuff except near high power stages) and putting a PI type filter with 47pF
then 47 Ohms then 47pF again in series with each input wire. If you can
tolerate higher value resistors then that would help. The caps should be
surface mount with less than 2mm of trace in series with them if you want
the best results.

Chris

 

[Mark]

I have found differenital circuits are not an effective way to reduce
UHF interference because you can never maintain adequate amplitude and
phase balance of the UHF signal on the two leads...

I suggest you add series impeadance i.e beads and chokes and shunt
impeadance like small c's

many times a series 1 kOhm resistor directly at the suseptable device
i.e at the transistor base or at the IC input pins will work very well.

The series 1 kOhm and the internal shunt c of the device will form a
low pass filter that has little or no effect on the audio but
attenuates the UHF.. You need to attenuate the UHF to the point that
it no longer creates non-linearity or rectifies. thats when it gets
converted to audio and you can hear it....if it is just a low level UHF
signal on audio, it won't hurt anything, only when it is large enough
to cause non-linearity and rectification and is then converted to
audio, does it create audible interference...

Mark

 

[Joseph2k]

In message <[email protected]>


1) Put it in a screening case (i.e. conductive - preferably metal)
and take care that seams can under no circumstances fail to make
contact over a length exceeding about 40 mm. Pay particular
attention to paint or any other surface finish that will prevent
conduction, and to the fact that the edges aren't flat. Assume
there will be no contact except at points that are pressed together.

2) Take inputs and outputs via screened cables. Make the screens
off such that the RF interference remains on the outer skin of the
case. Once you take the interference to the inner skin, it's all
over.

3) Lay the PCB out with a complete ground plane.

4) Pay great attention to the spurious components, most notably the
inductance of all tracks. This means keep everything as small as
possible. Surface mount is a great help in this respect. The
placement of suppression components is as important as their values.

5) You are unlikely to need ferrites, which is a good thing as they
are relatively expensive, if you've done everything else right.

Dave

Ferrites expensive?? not in my experience. if you are willing to buy a
thousand about us$2 total.
you can use any small value miniature inductors (including surface mount) in
their place also (replacing the 47 ohm resistors).

 

[Dave Higton]

In message <[email protected]>

Ferrites expensive?? not in my experience. if you are willing to buy a
thousand about us$2 total.

I mean relatively expensive in that they are many times the price of
resistors and capacitors; your figures bear that out as against
resistors at maybe $0.01 and capacitors maybe $0.02.

you can use any small value miniature inductors (including surface mount)
in their place also (replacing the 47 ohm resistors).

Beware. Ferrites for interference suppression are (usually) deliberately
lossy, so they dissipate the interference. Inductors are deliberately
as little lossy as possible, so you can end up with resonances.

All these problems are soluble by careful design done by someone who
knows what he's doing, of course.

Dave

 

[Walter Harley]

[...]
When designing an audio amplifier with high immunity to such
interference, I can see the obvious value of adding suitable RF
filtering (ferrite chokes and shunt capacitors with low-Z in the UHF
band) to the input circuitry but wonder if a pre-amplifier using
differential inputs (FET or Bipolar ?) rather than single-ended would be
better. Transformer coupling of source to the amplifier input is another
possibility, of course.

Here's my understanding; maybe someone more knowledgeable will correct me:

Differential amplifiers reduce interference by cancelling the common mode
signal. For that to work, the interference has to be a common mode signal.
In general, this is achieved by balancing the impedance between the two legs
of the signal and the interference source (which is notionally
ground-referenced), so that the interference couples equally to both legs.

If your signal is already unbalanced (meaning that the impedance of the two
legs to ground is not the same), then a differential amplifier will not help
cancel interference. (Interference rejection in balanced transmission has
nothing to do with whether the "cold" leg is carrying an opposite-polarity
voltage, as should be obvious if you consider the case where the signal is
0V.)


Something that has always confused me; maybe someone can explain:

How can ferrites help, when input impedances are > 1k, as in much audio
circuitry? I thought ferrite impedance was around 100 ohms or so, max. I
wouldn't think it would be able to create enough voltage drop to make a
difference. Is there something else going on?

 

[John Larkin]

In message <[email protected]>


For any given "victim" device, the only determining factors are the
strength of the interfering field at the victim, and the frequency.
Victims are only broadly tuned, so the frequency only really depends
on which band the phone is operating in (i.e. 900 or 1800 MHz etc.)

Dave

Actually, victims often have internal resonances that can be very
sharp. I've seen thermocouple gear whose quantifiable RF sensitivity
peaked at over 20:1 in a few very narrow frequency regions.

John

 

[John Larkin]

Hi,

Low-cost commercially available audio pre-amplifiers are usually very
vulnerable to interference from adjacent cellular phones switching, in
the UK, at around 217Hz.

When designing an audio amplifier with high immunity to such
interference, I can see the obvious value of adding suitable RF
filtering (ferrite chokes and shunt capacitors with low-Z in the UHF
band) to the input circuitry but wonder if a pre-amplifier using
differential inputs (FET or Bipolar ?) rather than single-ended would be
better. Transformer coupling of source to the amplifier input is another
possibility, of course.

To meet a current requirement that I have to meet, amplifier noise
should not be too much of a problem since the input signal levels are in
the region of 20-50mV pk-pk, with a bandwidth from around 300Hz to 5KHz,
and the output level required is around 2 volts pk.pk. into a nominal
600 ohm load.

Does anyone with practical experience of dealing with this problem
have any comments that they'd care to make about how to minimise such
interference?

TIA - Dave

David C.Chapman - Chartered Engineer. FIEE. ([email protected])
--------------------------------------------------------------------------
CHAPMAN ASSOCIATES is a Consultancy offering practical expertise and
design skills in the fields of counter-surveillance, electronic protection
and security. Visit our Web site at http://www.minda.co.uk
--------------------------------------------------------------------------

What I know:

Fet front-ends are much less likely to rectify RF than bipolar. Jfet
and CMOS opamps can be tens or hundreds of times less sensitive than
bipolars.

Multilayer boards, with a solid ground plane, help a lot.

Ferrite beads are magic.

The smaller the better.

Wiring resonates.

John

 

[John Larkin]

[...]
When designing an audio amplifier with high immunity to such
interference, I can see the obvious value of adding suitable RF
filtering (ferrite chokes and shunt capacitors with low-Z in the UHF
band) to the input circuitry but wonder if a pre-amplifier using
differential inputs (FET or Bipolar ?) rather than single-ended would be
better. Transformer coupling of source to the amplifier input is another
possibility, of course.

Here's my understanding; maybe someone more knowledgeable will correct me:

Differential amplifiers reduce interference by cancelling the common mode
signal. For that to work, the interference has to be a common mode signal.
In general, this is achieved by balancing the impedance between the two legs
of the signal and the interference source (which is notionally
ground-referenced), so that the interference couples equally to both legs.

If your signal is already unbalanced (meaning that the impedance of the two
legs to ground is not the same), then a differential amplifier will not help
cancel interference. (Interference rejection in balanced transmission has
nothing to do with whether the "cold" leg is carrying an opposite-polarity
voltage, as should be obvious if you consider the case where the signal is
0V.)


Something that has always confused me; maybe someone can explain:

How can ferrites help, when input impedances are > 1k, as in much audio
circuitry? I thought ferrite impedance was around 100 ohms or so, max. I
wouldn't think it would be able to create enough voltage drop to make a
difference. Is there something else going on?

At RF frequencies, trace impedances and IC input impedances are low.
At 1 GHz, an opamp input impedance will be ballpark 50 ohms.

And a ferrite bead is a lossy inductor, so it kills resonances, and
resonances are a major culprit here.

You can get surface-mount beads that are 600 ohms at 100 MHz, more at
higher frequencies. A bead followed by a capacitor is an excellent
lowpass filter.

John

 

[Dave Higton]

In message <[email protected]>

You can get surface-mount beads that are 600 ohms at 100 MHz, more at
higher frequencies. A bead followed by a capacitor is an excellent
lowpass filter.

ONLY IF the trace lengths to the capacitor are short.

Dave

 

[Terry Given]

In message <[email protected]>

and it wont oscillate

ONLY IF the trace lengths to the capacitor are short.

Dave

maybe, maybe not. depends on how you do it. a "T" connection to the cap
(with a nice long stalk) is terrible, the stalk inductance is
detrimental, whereas a "V" subsumes the trace inductance into the filter

plus the cap inductance means you can only asymptotically approach some
minimum L. hence 0508 caps, interdigitated leads etc.

cheers
Terry