Ian Bell wrote:

Eeyore wrote:
Ian Bell wrote:

So apart from the possible variation of gm with anode
current, the noise added by a triode seems to to fairly independent of
anode current. This seems strange to me since shot noise is caused by
random fluctuations in anode current so I would expect higher anode
currents to produce higher noise.

Not because of the shot effect for sure.

A higher anode current means the relative size of the unit electronic
charge is actually *smaller* in influence.

Er, surely it means there are more of them to create noise?

You've misunderstood how shot noise is generated.

Graham

In article , Ian Bell
wrote:
Jim Lesurf wrote:

In article , Ian Bell
wrote:
In designing a semiconductor preamp you usually aim for a low
collector current in the first stage as one method of minimising
noise.

Which "you" do you mean?... :-)

The noise level from a "semiconductor" preamp may be lower with a
higher current. Depends entirely on the device type, impedances, etc,
etc.

Slainte,

Jim


Jim, I expected better of you. Sometimes I wish people would just read
the question and try to answer it rather than critique it.

I was not doing a "critique" on your question. Just pointing out that what
you'd said above - which was not your question - was incorrect. :-)


I have had several responses to this post in rec.audio.tubes and not one
has actually addressed the question. I am looking for some guidance on
low noise tube preamp design and particularly the effect of anode
current. Can you help?

Not really. So far as I know if you require low noise then the sensible
approach is to use solid-state devices designed for such purposes. I'd
expect the noise level of a valve to be higher if only because of the
higher levels of thermal noise. So I'm afraid I've never even tried
designing a low noise valve amplifier myself.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html

In article , Ian Bell
wrote:
Don Pearce wrote:




I have no problem with that. However, what I *want* to do is design and
build a tube mic pre. I would like it to be as noise free as possible
but I accept that will not be the lowest noise that can be achieved by
using other means. It does not need to be the ultimate in low noise
design but I don't want it to be any noisier than it has to be.

A fairly standard approach is to use a step up transformer followed by a
couple of tube stages of amplification and possibly an output
transformer.

What you really need to do is to build a valve stage and measure its In and
En values. i.e. measure the output noise with a short and then an o/c (or
large resistor value) at its input. From this you can work out the source
impedance that will give optimum SNR. Knowing the actual source impedance
for your microphones this then allows you to decide the turns ratio of the
transformer.

With a short on the input the noise only comes from En. With an open it
comes from both En and In through the effective input/source impedances in
parallel. The ratio of the two gives you the source/input resistance in
parallel that provides maximum SNR.

If you wish, do the above for different bias resistor values, etc. If you
were using a solid state device I could advise on that. But afraid I don't
have the experience with valves, so you'll need advice from someone who
has. Ideally someone who understands the engineering rather than someone
who tweaks but never measures or analyses. :-)


In principle, noise in the first tube stage will not be a significant
contributor to overall noise if the equivalent input noise resistance of
the tube input Req is a half or preferably one quarter the resistance of
the source reflected into the secondary of the transformer (see RDH).


I'd suggest you use the above approach is it will tell you the actual
optimum rather than giving you a vague 'rule of thumb'. :-)

According to RDH, you can express the shot noise of a triode in terms of
an equivalent noise resistance connected between grid and cathode of a
noiseless triode. For triodes this resistance is approximately 2.5/gm at
the operating point. So apart from the possible variation of gm with
anode current, the noise added by a triode seems to to fairly
independent of anode current. This seems strange to me since shot noise
is caused by random fluctuations in anode current so I would expect
higher anode currents to produce higher noise. Indeed, RDH notes that
shot noise is produced in the anode circuit and refers it to the more
convenient grid simply by noting that the incremental grid voltage and
anode current are related by gm so that Eg = Ia/gm. The leap from this
to Req =2.5/gm for a triode is not explained.

Comment as above. In this context I'd avoid modelling the noise in terms of
a 'resistance' as the In and En approach is easier to measure and use.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html

In article , Eeyore
wrote:


Ian Bell wrote:

So apart from the possible variation of gm with anode current, the
noise added by a triode seems to to fairly independent of anode
current. This seems strange to me since shot noise is caused by random
fluctuations in anode current so I would expect higher anode currents
to produce higher noise.

Not because of the shot effect for sure.

Erm... the shot noise level rises with the current level. The reasons for
this are well explained by theory.

A higher anode current means the relative size of the unit electronic
charge is actually *smaller* in influence.

For an individual charge, yes. The snag is that there are many more of them
when the current is higher, and it is the statistics that matter. The
result is that with a 'Normal' distribution of statistics the shot noise
level rises with the mean current level. Standard result regardless of
device type. Also comfirmed in general observations. Indeed, I think the
theory for this, and the measurements, we established long before solid
state devices gain became common.

So, yes, I'd expect the shot noise contribution to rise with bias current.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html

In article , Ian Bell
wrote:
Don Pearce wrote:


I plan to. I had hoped some of the leading lights on this group would
have done this already. I find it hard to believe I am doing pioneering
experiments. Crikey, I might even publish the results on a web site ;-)

I doubt you are 'pioneering'. If you look back at journals and texts
published before about 1960 you may well find what you require. However I
suspect that engineers lost interest in what you are asking about some
decades ago. Simply easier to get low noise with solid state devices in
most applications. I doubt many modern texts discuss 'coherers' as
alternatives to Shottky diodes, either. ;- But you can find them in
older books.

FWIW If I get a chance I'll do a search through my old JAES papers and see
if any deal with what you want to know. They may cover it in the early
volumes.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html

Jim Lesurf wrote:

Eeyore wrote:

Ian Bell wrote:

So apart from the possible variation of gm with anode current, the
noise added by a triode seems to to fairly independent of anode
current. This seems strange to me since shot noise is caused by random
fluctuations in anode current so I would expect higher anode currents
to produce higher noise.

Not because of the shot effect for sure.

Erm... the shot noise level rises with the current level. The reasons for
this are well explained by theory.

Rises as the root of the current AIUI.

Therefore a doubling if Ia will result in an increase in In of 3dB.

However since it can normally be safely presumed that the signal level increases
with anode current too (ok not quite 100% directly with tubes), then actually,
the relative noise will *reduce* by 3dB i.e the s/n ratio will improve.

Graham

"Jim Lesurf"

What you really need to do is to build a valve stage and measure its In
and
En values. i.e. measure the output noise with a short and then an o/c (or
large resistor value) at its input. From this you can work out the source
impedance that will give optimum SNR. Knowing the actual source impedance
for your microphones this then allows you to decide the turns ratio of the
transformer.


** Shame it has got almost nothing to do with that.

It is only possible to do so much with a STEP UP input transformer before
serious problems arise with frequency response. A step up voltage ratio of
1:10 or 1:15 is the limit where flat response over the whole audio band is
required.
A step up ratio no more than 1:5 is the norm for studio grade equipment.

Attempting to go for high ratios ( ie for better s/n) results in the HF
response becoming a function of the mic's actual impedance in an alarming
way as the optimum output side RC loading network becomes very source
impedance dependant.

EG: With a low mic impedance, say 50 to 100 ohms, HF response will peak at
14kHz by many dBs. With a higher mic impedance, say 600 ohms, response
droops in the upper audio band by many dBs. Only with a 200 to 250 ohm mic
will flat response be maintained, +/- 1 dB, as intended.

Using a step up ratio of 1:10 means the input noise voltage of the valve is
effectively diminished by that same factor - making EIN figures of circa
0.3 uV ( rel 200 ohms) possible with valves like the 12AX7. This compares
fairly well with good SS designs.

The wise engineer knows to use high output condenser mics with valve input
stages for a top class result and to use dynamics only where the SPL is high
enough to overcome system noise.



........ Phil

In article , Eeyore
wrote:


Jim Lesurf wrote:

Eeyore wrote:

Ian Bell wrote:

So apart from the possible variation of gm with anode current, the
noise added by a triode seems to to fairly independent of anode
current. This seems strange to me since shot noise is caused by
random fluctuations in anode current so I would expect higher
anode currents to produce higher noise.

Not because of the shot effect for sure.

Erm... the shot noise level rises with the current level. The reasons
for this are well explained by theory.

Rises as the root of the current AIUI.

Therefore a doubling if Ia will result in an increase in In of 3dB.

Thus as I pointed out, the shot noise will rise with bias current.

However since it can normally be safely presumed that the signal level
increases with anode current too (ok not quite 100% directly with
tubes), then actually, the relative noise will *reduce* by 3dB i.e the
s/n ratio will improve.

That is an interesting point. I don't know enough about valve stages to
comment upon it in detail. However it is a quite distinct reason from the
one you gave previously. This was:


On 17 Jun in uk.rec.audio, Eeyore
wrote:

A higher anode current means the relative size of the unit electronic
charge is actually *smaller* in influence.

Your point now is that the gain should rise more swiftly than the current
noise level. Not the above.

I don't know the noise dependencies of valves on bias. The only info I've
found to hand is that the effective shot noise can be represented by a
resistance in series with the grid whose value varies as 2.5/gm and that
you then work out the En of this at the grid (i.e. input) to model the shot
noise. Given gm rising with bias as you say that does indeed seem to
imply an apparently falling noise level at the input due to shot noise.
Although the book I found this in does not explicitly allow for the source
impedance which makes me wonder as I'd expect In at the input to have to
pass through the source in parallel with the input bias resistors.

I'm also interested in what you say as it implies that the slope of
the transfer curve is essentially square-law (in order for the gain to
rise essentially linearly with bias). I had assumed that it was more
like to the power 3/2 than 2. I would also have expected some shot
noise from leakage currents via the grid, but as I say, I don't
know enough about valves...

For a bipolar the In tends to rise with current as you'd expect from the
above, but the En tends to fall with rising bias current. The result is
that you tend to use high currents for low source resistances, and vice
versa. H&H have an example of this for a 2N4250. This indicates you'd use
around 1 mA for source impedance of 1k or less, but would be better off
with below 0.01 mA for sources of the order of 100k.

I've not seem similar values or plots for any valves, so far as I can
recall.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html

Jim Lesurf wrote:

In article , Ian Bell
wrote:
Don Pearce wrote:




I have no problem with that. However, what I *want* to do is design and
build a tube mic pre. I would like it to be as noise free as possible
but I accept that will not be the lowest noise that can be achieved by
using other means. It does not need to be the ultimate in low noise
design but I don't want it to be any noisier than it has to be.

A fairly standard approach is to use a step up transformer followed by a
couple of tube stages of amplification and possibly an output
transformer.

What you really need to do is to build a valve stage and measure its In
and En values. i.e. measure the output noise with a short and then an o/c
(or large resistor value) at its input. From this you can work out the
source impedance that will give optimum SNR. Knowing the actual source
impedance for your microphones this then allows you to decide the turns
ratio of the transformer.


Agreed. I am expecting the rest of the parts this week. I am really looking
forward to this. Last time I built anything with valves in it was about 40
years ago.


Ian

Jim Lesurf wrote:

In article , Ian Bell
wrote:
Don Pearce wrote:


I plan to. I had hoped some of the leading lights on this group would
have done this already. I find it hard to believe I am doing pioneering
experiments. Crikey, I might even publish the results on a web site ;-)

I doubt you are 'pioneering'. If you look back at journals and texts
published before about 1960 you may well find what you require. However I
suspect that engineers lost interest in what you are asking about some
decades ago. Simply easier to get low noise with solid state devices in
most applications. I doubt many modern texts discuss 'coherers' as
alternatives to Shottky diodes, either. ;- But you can find them in
older books.


Agreed, but valve preamps are surprisingly popular.

FWIW If I get a chance I'll do a search through my old JAES papers and see
if any deal with what you want to know. They may cover it in the early
volumes.


That is very kind. Many thanks.

Ian
Slainte,

Jim