[Ian Bell[_2_]]

Patrick Turner wrote:

Ian Bell wrote:
Patrick Turner wrote:
Ian Bell wrote:
Patrick Turner wrote:
Ian Bell wrote:
Patrick Turner wrote:
Ian Bell wrote:
Patrick Turner wrote:
snip

If you have 50uV of noise at the output and gain is 15x, and input is
grounded, then you could have a total of 2uV grid input noise if the
input tube is a real good one. 2uV gets amplified to make about 30uV at
the output, and some of that is LF noise. where does the rest of the
noise come from? By observation you should be able to see where the
noise is being generated and how, and find ways of stopping it without
much complexity and cost.
One of the potential sources I have discovered is excess noise in the
anode resistors. This could account for 20uV or more of the total noise.
The solution seems to be to replace metal film resistors with wire wound
types. However, I am having trouble finding sources for 100K and 39K
wire wound types at anything less than 5W rating which is something of
an overkill. Do you know any good sources of high value low wattage wire
wound resistors (preferable non-inductive of course)?

Cheers

Ian
Suppose you had 30uV of noise due to the input tube at the output of
your amp with gain = 15x.

And total noise = 50uV. Say you have Ra of the input tube = 20k, and the
load resistor for dc in parallel to the following stage bias R might be
50k. The 50k 20Hz ro 20k noise can be calculated, but the its shunted by
Ra = 20k and thus attenuated. The excess noise would also be shunted,
but unlikely to be as high as the V1 tube noise, even if its a poor
quality R. Its the value of R in ohms which gives rise to noise, and
usually 90% of the noise at the output of a mic amp is due to the
broadand noise of V1, and all following R and tubes have little
contribution unless the output of V1 is a tiny voltage.

I think you should just build it and then try things to make it silent.
Wire wound would appear to be blameless, and getting a high value one
may mean you series a few. A 47k WW type will at least make the
predicted amount of noise because its the ohms than make the noise.
Any other type of R, metal film, carbon film also will make similar
noise you should find unless they are faulty.

Read up from RDH4 about noise and equivalent input resistance and noise
in R and summing of noise voltages.

Thanks for the input Patrick. I know about Johnson noise and in plate
resistors it is not much of an issue as it is swamped by other noise
sources. What I am concerned about is excess noise which is voltage
dependent. In the few places I have seen this specified it is typically
0.1uV/V so with a plate resistor with 150V across it there is possibly
15uV of noise due to this effect - enough to warrant a change of
resistor type.

Are you saying that regular metal film resistors do not suffer from
excess noise?
I've never noticed it in amps i have built. If I'd found that the noise
measurements didn't seem to add up, then I'd have been the one to reason
why.

I've usually used Welwyn or Beyshlag metal film R for most circuits.
Noise is negligible.
I'll check out Welwyn and Beyshlag specs, thanks

Cheers

Noise is mainly the product of the resistance value. It should not
matter what the R is made from. If properly made from any resistive
material the number of ohms is the main issue, then temperature. I don't
expect Welwyn or Beyshlag or Mills or Cadock or anything else from the
Vishay Resistor Mafia to be much different.
I hear what you say Patrick but Morgan Jones and the manufacturers data
sheets I have seen do not support that view.

For example the Johnson noise in a 20KHz bandwidth at room temperature
for a 39K resistor is ABOUT 3.6uV. If it is in an anode circuit with
150V across it then the excess noise of a typical metal film resistor
will be about 15uV per decade.

In low voltage circuits I agree Johnson noise predominates but I am not
sure it does in anode circuits.

Cheers

The Johnson noise values given sound about right.

In experiments when testing batches of triode tubes like 12AX7, 12AT7,
12AU7, and to view the noise voltage on a CRO, I amplify the anode noise
signal with the grid grounded with a short link. And with dc applied to
the heaters. In my case I am using cheap carbon film R,
and the noise measured appears to be the grid input noise x tube gain.
If the gain = 15 for 12AU7, 2uV of grid noise in agood sample becomes
30uV at the anode and when amplified 1,500 times by an opamp,
you have 45mV of noise which is easily displayed on the CRO. The test
load for the 'AU7 is variable, but say it was 100k.
according to you the excess noise might be about 5uV, but the 100k noise
is shunted by the Ra of the tiode at about 10k, so the excess noise
is much attenuated and would be 0.5uV, a tiny amount compared to 30uV
because of amplified grid input noise.

That is an interesting point. The current preamp circuit has an 8K2
unbypassed cathode resistor (as the feedback point) which means with a
tube mu of 35 the effective ra is nearly 300K so it does not attenuate
excess noise in the same way as you describe.


Cheers

Ian

In j-fet such as a 2sk369 with fully bypassed source, and with the same
gain as the 'AU7, load = 15 / 0.04 at 5mAdc, so load = 375 ohms.
Excess noise and Johnson noise indeed be low, and drain noise = 15 x
0.14uV = 2.2uV approx. The fet looks like it has about 1/14 of the
triode noise.

Where is any need to consider any noise of RL?

Patrick Turner.


Ian
Metal film do have low noise, and noise below old fashioned carbon
composition types. I have a box full ot NOS from the mid 1930s which are
carbon rods with brass end shells clamped on and painted. They are in
good condition but I doubt they will be noiser than later carbon
compostion types which were notorious for change of value and moisture
ingress and chemical corrosion and decomposition over time. Quad used
these horrible R in the Quad-II and 22 control units. All may have been
fine when installed. I'm sure Peter Walker would have checked all this
out. But maybe they were not so quiet after 55 years...

The circuit resistances in most old ancient tube amps including Quad
junk and others was always on the high side. EG, Quad 22 volume control
is 500k.

Using say 100k instead of 1M to bias a tube reduces noise by 1/3.16, but
is true only if there isn't anything else shunting such R.

Noise in smaller R is proportional to 1 / sq.rt of the reduction of
resistance.

Or if you had 1uV noise in 1k for a given bandwidth, then noise with 10k
= 3.16uV, 100k gives 10uV, and 1M gives 31.6uV.
10M gives 100uV, and is a typical value used to bias a tube with grid
leak biasing as in a radio.
But there is usually some other R shunting the 10M, like a 1M volume
control, so noise is less at the output.


Patrick Turner.


Patrick Turner.
Cheers

Ian
Patrick Turner.