The 6AU6 based mic pre I am designing has an interesting quirk. I am
testing various valves for low noise with the first stage input grid
grounded. As expected different valves exhibit different noise levels
and susceptibility to hum. To eliminate hum from the readings my
millivoltmeter has an LF cut switch. When this filter is used I noticed
a small but regular variation in the indicated noise level. It varies by
about 10% over a regular 2 second period. We are talking very small
values here as the indicated noise is only a tenth of a millivolt. But
it is very regular and I have tried half a dozen different valves all
with the same result. So the question is what is it? Both stages of the
preamp are decoupled with 470uF caps via 10K resistors (time constant
4.7 seconds).

Any help appreciated.

Cheers

Ian

In article ,
Ian Thompson-Bell wrote:

The 6AU6 based mic pre I am designing has an interesting quirk. I am
testing various valves for low noise with the first stage input grid
grounded. As expected different valves exhibit different noise levels
and susceptibility to hum. To eliminate hum from the readings my
millivoltmeter has an LF cut switch. When this filter is used I noticed
a small but regular variation in the indicated noise level. It varies by
about 10% over a regular 2 second period. We are talking very small
values here as the indicated noise is only a tenth of a millivolt. But
it is very regular and I have tried half a dozen different valves all
with the same result. So the question is what is it? Both stages of the
preamp are decoupled with 470uF caps via 10K resistors (time constant
4.7 seconds).

Any help appreciated.

Since you mention the power supply time constants, have you tried
changing them to see if the period of the "oscillation" changes, and
ruling that path in or out?


Regards,

John Byrns

--
Surf my web pages at, http://fmamradios.com/

John Byrns wrote:
In article ,
Ian Thompson-Bell wrote:

The 6AU6 based mic pre I am designing has an interesting quirk. I am
testing various valves for low noise with the first stage input grid
grounded. As expected different valves exhibit different noise levels
and susceptibility to hum. To eliminate hum from the readings my
millivoltmeter has an LF cut switch. When this filter is used I noticed
a small but regular variation in the indicated noise level. It varies by
about 10% over a regular 2 second period. We are talking very small
values here as the indicated noise is only a tenth of a millivolt. But
it is very regular and I have tried half a dozen different valves all
with the same result. So the question is what is it? Both stages of the
preamp are decoupled with 470uF caps via 10K resistors (time constant
4.7 seconds).

Any help appreciated.

Since you mention the power supply time constants, have you tried
changing them to see if the period of the "oscillation" changes, and
ruling that path in or out?



Not yet but that is probably my next test.

Ian

If there is no negative feedback circuit in the amplifier, then there is no
reason to suspect coupling via power supply as a cause of the phenomenon.

I do not know how you are measuring the noise, but if you are using a
multimeter, it is possible to have some beat frequency between the 60Hz grid
and the harmonics of the switching cycle of the dual slope ADC which the
multimeter is based on. If you have a residual 60Hz ( I am assuming you are
in the USA) on the preamp output and if the ADC in the multimeter is not
absolutely synchronised with the grid (which is not), then you will have
sort of "product detector" delivering the beat frequency.

Regards,
Alex

"John Byrns" wrote in message
...
In article ,
Ian Thompson-Bell wrote:

The 6AU6 based mic pre I am designing has an interesting quirk. I am
testing various valves for low noise with the first stage input grid
grounded. As expected different valves exhibit different noise levels
and susceptibility to hum. To eliminate hum from the readings my
millivoltmeter has an LF cut switch. When this filter is used I noticed
a small but regular variation in the indicated noise level. It varies by
about 10% over a regular 2 second period. We are talking very small
values here as the indicated noise is only a tenth of a millivolt. But
it is very regular and I have tried half a dozen different valves all
with the same result. So the question is what is it? Both stages of the
preamp are decoupled with 470uF caps via 10K resistors (time constant
4.7 seconds).

Any help appreciated.

Since you mention the power supply time constants, have you tried
changing them to see if the period of the "oscillation" changes, and
ruling that path in or out?


Regards,

John Byrns

--
Surf my web pages at, http://fmamradios.com/

Alex wrote:
If there is no negative feedback circuit in the amplifier, then there is no
reason to suspect coupling via power supply as a cause of the phenomenon.


Nothing global but there is an unbypassed cathode resistor in the first
stage and of course the CF has 100% feedback.

I do not know how you are measuring the noise, but if you are using a
multimeter, it is possible to have some beat frequency between the 60Hz grid
and the harmonics of the switching cycle of the dual slope ADC which the
multimeter is based on. If you have a residual 60Hz ( I am assuming you are
in the USA) on the preamp output and if the ADC in the multimeter is not
absolutely synchronised with the grid (which is not), then you will have
sort of "product detector" delivering the beat frequency.


Good idea but not the cause I think. I am using a purely analogue meter
(Ferrograph RTS2).

And by the way, I am in the UK.

Cheers

Ian

"Ian Thompson-Bell" wrote in message
...
Alex wrote:
If there is no negative feedback circuit in the amplifier, then there is
no
reason to suspect coupling via power supply as a cause of the phenomenon.


Nothing global but there is an unbypassed cathode resistor in the first
stage and of course the CF has 100% feedback.

I do not know how you are measuring the noise, but if you are using a
multimeter, it is possible to have some beat frequency between the 60Hz
grid
and the harmonics of the switching cycle of the dual slope ADC which the
multimeter is based on. If you have a residual 60Hz ( I am assuming you
are
in the USA) on the preamp output and if the ADC in the multimeter is not
absolutely synchronised with the grid (which is not), then you will have
sort of "product detector" delivering the beat frequency.


Good idea but not the cause I think. I am using a purely analogue meter
(Ferrograph RTS2).

And by the way, I am in the UK.

Cheers

Ian

Ah Ha! My RTS2 does much the same on its lower voltage (greatest
sensitivity) setting. Even with the input shorted, the needle fluctuates by
about +- 10% full scale. I have changed all the electrolytic caps, and
cleaned every contact, but the effect's the same. I use my Levell meter for
low-noise measurements but I suppose I should sort out the RTS2 sometime.
Have you checked whether the RTS2 needle still moves with the input shorted?

S.


--
http://audiopages.googlepages.com

Serge Auckland wrote:
"Ian Thompson-Bell" wrote in message
...
Alex wrote:
If there is no negative feedback circuit in the amplifier, then there is
no
reason to suspect coupling via power supply as a cause of the phenomenon.

Nothing global but there is an unbypassed cathode resistor in the first
stage and of course the CF has 100% feedback.

I do not know how you are measuring the noise, but if you are using a
multimeter, it is possible to have some beat frequency between the 60Hz
grid
and the harmonics of the switching cycle of the dual slope ADC which the
multimeter is based on. If you have a residual 60Hz ( I am assuming you
are
in the USA) on the preamp output and if the ADC in the multimeter is not
absolutely synchronised with the grid (which is not), then you will have
sort of "product detector" delivering the beat frequency.

Good idea but not the cause I think. I am using a purely analogue meter
(Ferrograph RTS2).

And by the way, I am in the UK.

Cheers

Ian

Ah Ha! My RTS2 does much the same on its lower voltage (greatest
sensitivity) setting. Even with the input shorted, the needle fluctuates by
about +- 10% full scale. I have changed all the electrolytic caps, and
cleaned every contact, but the effect's the same. I use my Levell meter for
low-noise measurements but I suppose I should sort out the RTS2 sometime.
Have you checked whether the RTS2 needle still moves with the input shorted?


I just tried that and it sits rock solid on the -60dB range at 20uV and
that's just pick up from the unscreened part of the probe.

However, I just reconnected my scope and switched it to dc (this is on
the 1mV/division range) and the output noise sits on a couple of mV
positive dc. I am just wondering if this is leakage from the output
electrolytic. The load is a 10K resistor so 2mV represents about 20uA if
my calculations are correct. I am just going to try another one and also
a 0.47uF non electrolytic to see if there is any difference.

Well, the second electrolytic, same make (Jamicon) same value (4.7uF,
450V) is even worse - it sits at +5mV.

Now tried with 0.47uF non electrolytic and that seems to have pretty
much cured it. There is still a very small movement on the Ferrograph
meter but it is hard to tell if it is regular or just a bit of very low
frequency shot noise. And the scope shows the noise straddling 0V as it
should.

So now all I need to do is decide how to get a large enough output non
electrolytic coupling cap to avoid losing bass response.

Thanks for all the input.

Cheers

ian

"Ian Thompson-Bell" wrote in message
...
Good idea but not the cause I think. I am using a purely analogue meter
(Ferrograph RTS2).

Well, the phenomenon remains unexplained -- a mystery.


And by the way, I am in the UK.

Ooops! I did not look at your e-mail address.
And I am in Australia.
Regards,
Alex

Cheers

Ian

"Ian Thompson-Bell" wrote in message
...
I do not know how you are measuring the noise, but if you are using a
multimeter, it is possible to have some beat frequency between the 60Hz
grid
and the harmonics of the switching cycle of the dual slope ADC which the
multimeter is based on. If you have a residual 60Hz ( I am assuming you
are
in the USA) on the preamp output and if the ADC in the multimeter is not
absolutely synchronised with the grid (which is not), then you will have
sort of "product detector" delivering the beat frequency.


Good idea but not the cause I think. I am using a purely analogue meter
(Ferrograph RTS2).

Then even a crazier idea!
If you are located close to a strong TV station, you might have direct RF
rectification on the detector of your meter (RF induced on the leads of the
instrument) or even worse, on the non-linearities of your preamp, if the
input is not well shielded.

If it is the case then you might have infra-low beat between the crystal
stable 50.0000Hz frame sync of the TV station and the hum from the grid,
which might be fractions of a Herz off.

Regards,
Alex

And by the way, I am in the UK.

Cheers

Ian

In article ,
"Alex" wrote:

If there is no negative feedback circuit in the amplifier, then there is no
reason to suspect coupling via power supply as a cause of the phenomenon.

Negative feedback is not necessary for oscillation to occur because of
coupling through the power supply, although this particular circuit may
not have enough stages to invoke the problem, I missed the schematic
though.

I do not know how you are measuring the noise, but if you are using a
multimeter, it is possible to have some beat frequency between the 60Hz grid
and the harmonics of the switching cycle of the dual slope ADC which the
multimeter is based on. If you have a residual 60Hz ( I am assuming you are
in the USA) on the preamp output and if the ADC in the multimeter is not
absolutely synchronised with the grid (which is not), then you will have
sort of "product detector" delivering the beat frequency.

Something of this sort could easily cause the problem.

Regards,

John Byrns

--
Surf my web pages at, http://fmamradios.com/