Maker Pro
Maker Pro

Cell phone noise into opamps

J

Joerg

Jan 1, 1970
0
Hello Folks,

In a really sensitive sub-audio application we are experiencing cell
phone pickup. It got a lot worse when switching from the LT1112 to the
LM833 in order to lower the sub-audio noise (and yeah, reduce cost...).
Anyhow, this surprised me because the LM833 doesn't even have the input
protectors like the LT1112 so the chances of something clipping and
demodulating should be lower. How could this be? Older process with
larger patterns and thus more pickup loop area?

The challenge here is that bypassing is all nice and dandy but
capacitors always have a residual inductance and at cell phone
frequencies that begins to matter. Cell phones are particularly nasty
because anything that gets AM demodulated produces this rat-tat-tat
pulse pattern between phone and cell tower. Shielding is going to be a
major pain in the neck in this application.

In case anyone wants to look at the datasheets:
http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1010,C1764,P1364,D1714
http://www.national.com/ds/LM/LM833.pdf
 
V

Vladimir Vassilevsky

Jan 1, 1970
0
Joerg said:
Hello Folks,

In a really sensitive sub-audio application we are experiencing cell
phone pickup.

In a really sensitive sub-audio application you use LM833 despite of all
your complaints :)
It got a lot worse when switching from the LT1112 to the
LM833 in order to lower the sub-audio noise (and yeah, reduce cost...).
Anyhow, this surprised me because the LM833 doesn't even have the input
protectors like the LT1112 so the chances of something clipping and
demodulating should be lower. How could this be? Older process with
larger patterns and thus more pickup loop area?

It could be some resonance affected by the parasitics of the opamp.

VLV
 
J

John Larkin

Jan 1, 1970
0
Hello Folks,

In a really sensitive sub-audio application we are experiencing cell
phone pickup. It got a lot worse when switching from the LT1112 to the
LM833 in order to lower the sub-audio noise (and yeah, reduce cost...).
Anyhow, this surprised me because the LM833 doesn't even have the input
protectors like the LT1112 so the chances of something clipping and
demodulating should be lower. How could this be? Older process with
larger patterns and thus more pickup loop area?

The challenge here is that bypassing is all nice and dandy but
capacitors always have a residual inductance and at cell phone
frequencies that begins to matter. Cell phones are particularly nasty
because anything that gets AM demodulated produces this rat-tat-tat
pulse pattern between phone and cell tower. Shielding is going to be a
major pain in the neck in this application.

In case anyone wants to look at the datasheets:
http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1010,C1764,P1364,D1714
http://www.national.com/ds/LM/LM833.pdf


Yup. The b-e junctions think they're crystal sets. I've sold about
4000 thermocouple temperature controllers because the Ox**** units
rectified RF so well you could shut them down with an old GenRad
generator from across the room. Well, that and the fact that they lost
the source code.

Ferrite beads and small caps to ground help enormously, and won't
affect baseband performance. I've also observed that such rf
sensitivity is often associated with sharp resonances in the wiring
and packaging. A well-placed bead or figure-8 ferrite, in a cable, can
kill a resonance nicely. We picked up another 20 dB by running our
thermocouple wiring, from a connector to the pcb, through a fig-8
(dual-hole) bead.

One quick fix would be to dump in a jfet opamp. Some of them are very
good.

I saw an appnote somewhere that graded various opamps on RF
sensitivity, but I can't remember where.

John
 
J

Jan Panteltje

Jan 1, 1970
0
Shielding is going to be a
major pain in the neck in this application.

But you will have to.
Metal case
All inputs and outputs including power.
Do not blame it on the chips.
(Well there may be some chips.....)
:)
 
T

Tilmann Reh

Jan 1, 1970
0
John said:
Yup. The b-e junctions think they're crystal sets. I've sold about
4000 thermocouple temperature controllers because the Ox**** units
rectified RF so well you could shut them down with an old GenRad
generator from across the room. Well, that and the fact that they lost
the source code.

Ferrite beads and small caps to ground help enormously, and won't
affect baseband performance.
...
One quick fix would be to dump in a jfet opamp. Some of them are very
good.

Ack. However, the JFET types often tend to be more noisy.

I made the experience that it's often good to place a small capacitor
between the two inputs of the OP. That doesn't affect baseband behaviour
much, but effectively short circuits the RF before it can be rectified
by the junctions at the OPs input.

Tilmann
 
Yup. The b-e junctions think they're crystal sets. I've sold about
4000 thermocouple temperature controllers because the Ox**** units
rectified RF so well you could shut them down with an old GenRad
generator from across the room. Well, that and the fact that they lost
the source code.

Ferrite beads and small caps to ground help enormously, and won't
affect baseband performance. I've also observed that such rf
sensitivity is often associated with sharp resonances in the wiring
and packaging. A well-placed bead or figure-8 ferrite, in a cable, can
kill a resonance nicely. We picked up another 20 dB by running our
thermocouple wiring, from a connector to the pcb, through a fig-8
(dual-hole) bead.

One quick fix would be to dump in a jfet opamp. Some of them are very
good.

I saw an appnote somewhere that graded various opamps on RF
sensitivity, but I can't remember where.

Basically, if you hit the NPN inputs of a bipolar op amp with more
than about 26mV of ptp noise at a frequency too high for the output
put to follow, the inputs follow the upper edge of the envelope of the
noise, because you've got out of the region of linear response.

For FET input op amps, the corresponding level is about a volt, which
is what John is talking about. Low noise FET-input op amps are not
cheap.
 
J

Joerg

Jan 1, 1970
0
Vladimir said:
In a really sensitive sub-audio application you use LM833 despite of all
your complaints :)

Which complaints? Yes, we might go discrete at some point by I think the
LM833 is a nice amp. Just somewhat fussy when RF is around.
 
J

Joerg

Jan 1, 1970
0
John said:
Yup. The b-e junctions think they're crystal sets. I've sold about
4000 thermocouple temperature controllers because the Ox**** units
rectified RF so well you could shut them down with an old GenRad
generator from across the room. Well, that and the fact that they lost
the source code.

Hmm, maybe I could brain-wash the b-e junctions ...

Ferrite beads and small caps to ground help enormously, and won't
affect baseband performance. I've also observed that such rf
sensitivity is often associated with sharp resonances in the wiring
and packaging. A well-placed bead or figure-8 ferrite, in a cable, can
kill a resonance nicely. We picked up another 20 dB by running our
thermocouple wiring, from a connector to the pcb, through a fig-8
(dual-hole) bead.

We are pretty much capped and beaded out already on this one. It helps,
but it ain't enough.

One quick fix would be to dump in a jfet opamp. Some of them are very
good.

As Tilmann pointed out those are too noisy at low frequencies.

I saw an appnote somewhere that graded various opamps on RF
sensitivity, but I can't remember where.

That would be great to have. If I find it via Google I'll post. However,
in our case we are pretty much stuck because only a handful of opamps
fulfill the noise criteria. Now I wish there was more LF noise data for
JFETs because they don't have an engaged b-e junction but are a few
volts away. Or MOSFETs but there are definitely no LF noise specs for
those. Maybe I'll start a separate thread, see which FETs people think
are the best contenders, then buy all those and try one after the other.
 
J

Joerg

Jan 1, 1970
0
Basically, if you hit the NPN inputs of a bipolar op amp with more
than about 26mV of ptp noise at a frequency too high for the output
put to follow, the inputs follow the upper edge of the envelope of the
noise, because you've got out of the region of linear response.

For FET input op amps, the corresponding level is about a volt, which
is what John is talking about. Low noise FET-input op amps are not
cheap.

Not cheap would be ok right now, for this application. We can always
worry about going discrete later. Got one in mind that is in the
<10nV/rtHz class and has a farily low 1/f knee?
 
J

Joerg

Jan 1, 1970
0
Jan said:
But you will have to.
Metal case
All inputs and outputs including power.
Do not blame it on the chips.
(Well there may be some chips.....)
:)


It's not completely possible in this app. Maybe we can find a JFET or
MOSFET that's low noise instead.
 
J

Jan Panteltje

Jan 1, 1970
0
It's not completely possible in this app. Maybe we can find a JFET or
MOSFET that's low noise instead.

Joerg, I find this difficult to answer.
It seems to me however, if I make something with say -80dB noise level,
that that should stay -80dB, even when somebody puts a cellphone on top of it.
Else what good is a noise spec?
We will only see more and more use of the spectrum, Wimax is coming,
the higher bands will be used too, EU is working on an "EU wide' spec
that would leave much more space for unlicensed applications, and give complete freedom
for the modulation type... Expect at lot of RF.
http://www.heise.de/newsticker/meldung/98895
 
J

John Larkin

Jan 1, 1970
0
We are pretty much capped and beaded out already on this one. It helps,
but it ain't enough.

In the signal path?

John
 
J

Joerg

Jan 1, 1970
0
Jan said:
Joerg, I find this difficult to answer.
It seems to me however, if I make something with say -80dB noise level,
that that should stay -80dB, even when somebody puts a cellphone on top of it.
Else what good is a noise spec?
We will only see more and more use of the spectrum, Wimax is coming,
the higher bands will be used too, EU is working on an "EU wide' spec
that would leave much more space for unlicensed applications, and give complete freedom
for the modulation type... Expect at lot of RF.
http://www.heise.de/newsticker/meldung/98895


I know. That's just the problem, here we have to deal with cell phones
that are just a foot or two away from the amp and shielding is a major
challenge on this one. So a JFET that could reliably deliver around
5nv/rtHz at 20Hz would be great :)
 
J

Joerg

Jan 1, 1970
0
Jan said:
Joerg, I find this difficult to answer.
It seems to me however, if I make something with say -80dB noise level,
that that should stay -80dB, even when somebody puts a cellphone on top of it.
Else what good is a noise spec?
We will only see more and more use of the spectrum, Wimax is coming,
the higher bands will be used too, EU is working on an "EU wide' spec
that would leave much more space for unlicensed applications, and give complete freedom
for the modulation type... Expect at lot of RF.
http://www.heise.de/newsticker/meldung/98895


P.S.: Somehow, whenever I am looking up the old faithfuls that could
have done it such as the 2SK662 they are discontinued. Man, that thing
rocked when it came to audio noise.

<sigh>
 
D

Didi

Jan 1, 1970
0
I know. That's just the problem, here we have to deal with cell phones
that are just a foot or two away from the amp and shielding is a major
challenge on this one. So a JFET that could reliably deliver around
5nv/rtHz at 20Hz would be great :)

I would try the OPA627 of Burr-Brown (ok, TI). They don't get any
better
than that - the speed should be fine for you, noise too.
But it is not dual...
A similar - but somewhat noisier and a bit more than somewhat slower,
but dual is the OPA2107, perhaps it would do.
However, the way this sounds I would be pretty sceptic that much can
be achieved without shielding. The cellphone here drives the power
supply
of the TV-set mad from 1 ft... what 80+ dB.

Dimiter
 
J

Joerg

Jan 1, 1970
0
Didi said:
I would try the OPA627 of Burr-Brown (ok, TI). They don't get any
better
than that - the speed should be fine for you, noise too.
But it is not dual...


Says 40nV/rtHz max at 10Hz. That's almost like the noise of the toilet
that I installed a couple years ago (where one of our dogs runs away the
instant I press the flush button).

Also, right now I am trying to stay clear of BB parts because they
appear to have some major issues in getting product out the door.
Digikey has a few old TO cans left for the OPA627 but I assume when they
are gone that's it. Not good for something that has to go into
production soon.

A similar - but somewhat noisier and a bit more than somewhat slower,
but dual is the OPA2107, perhaps it would do.


Similar noise but Digikey has absolutely no stock on this one.

However, the way this sounds I would be pretty sceptic that much can
be achieved without shielding. The cellphone here drives the power
supply
of the TV-set mad from 1 ft... what 80+ dB.

It can be done but not with anything bipolar. Has to be something where
all the junctions are well biased, like JFETs. We'll go discrete but
unfortunately JFET datasheets are rather skimpy. Often they don't even
contain a noise versus frequency graph.
 
D

Didi

Jan 1, 1970
0
Says 40nV/rtHz max at 10Hz. That's almost like the noise of the toilet
that I installed a couple years ago (where one of our dogs runs away the
instant I press the flush button).

The graph says something like 15 at 10 Hz.
The graph underneath indicates at Rs > apr. 5k the noise is all R...
If the 627 is too noisy for this application, don't waste your time
looking for anything better - there is nothing even very close to it
(meaning fast FET opamp).
Also, right now I am trying to stay clear of BB parts because they
appear to have some major issues in getting product out the door.

Now that would be surprising. I have been using them since 1991 or
so and they are still in production...

Dimiter
 
J

Joerg

Jan 1, 1970
0
Didi said:
The graph says something like 15 at 10 Hz.
The graph underneath indicates at Rs > apr. 5k the noise is all R...
If the 627 is too noisy for this application, don't waste your time
looking for anything better - there is nothing even very close to it
(meaning fast FET opamp).

It doesn't have to be fast. Looks like this board has to be migrated
over to a discrete design. Oh well, it ain't the first time.

Now that would be surprising. I have been using them since 1991 or
so and they are still in production...

Have you tried to buy any lately? I mean actually ordering a few
thousand? I don't know what the issues are but I have met a lot of
people who were in a bit of a frantic mode because their purchasing guys
cannot get parts.

[...]
 
D

Didi

Jan 1, 1970
0
Now that would be surprising. I have been using them since 1991 or
Have you tried to buy any lately? I mean actually ordering a few
thousand? I don't know what the issues are but I have met a lot of
people who were in a bit of a frantic mode because their purchasing guys
cannot get parts.

Now from surprising that gets depressing. I wish I had orders to need
the few thousand 627s, though... :).

I am not familiar with the slow parts (not many FET opamps too slow
anyway, IIRC), but I would still be surprised if there were any
FET opamp to match the noise figure of the OPA627.

Dimiter

------------------------------------------------------
Dimiter Popoff Transgalactic Instruments

http://www.tgi-sci.com
------------------------------------------------------

The graph says something like 15 at 10 Hz.
The graph underneath indicates at Rs > apr. 5k the noise is all R...
If the 627 is too noisy for this application, don't waste your time
looking for anything better - there is nothing even very close to it
(meaning fast FET opamp).

It doesn't have to be fast. Looks like this board has to be migrated
over to a discrete design. Oh well, it ain't the first time.
Now that would be surprising. I have been using them since 1991 or
so and they are still in production...

Have you tried to buy any lately? I mean actually ordering a few
thousand? I don't know what the issues are but I have met a lot of
people who were in a bit of a frantic mode because their purchasing guys
cannot get parts.

[...]
 
Top