• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Low noise, low microphony, low hum valves

Hi,

The article "A new low-noise circuit approach for pentodes" by Frank Bloehbaum in Linear Audio volume 0 shows that classic audio input stage valves such as EF86 and ECC808 have rather high levels of 1/f noise. This applies especially to the Telefunken new old stock EF86's he measured, these were even far out of spec. High transconductance RF valves are much better, but they are not designed for low microphony and low hum with an AC heater supply.

Do you know of any valves that combine low noise at audio frequencies with low hum and low microphony?

Best regards,
Marcel
 
Low-noise? Use triodes. But I am sure u have a good reason to use pentodes so that doesn't help much.

Sorry I'm not much help, I don't know any pentodes that fill your requirements.

I guess everything is relative, and compared to the most silent SS device, even a triode is considered 'noisy' by some people. But I've never ever had any problems with noise due to the tubes themselves, and have come to believe the notion of noisy tubes is another bad rap authored by those who don't use tubes. (Or have only used pentodes).
 
Hi SemperFi,

The article I referred to is about a trick to get rid of the partition noise in pentodes, but my question is more general: do you know any valve, triode, tetrode, pentode, whatever, that combines low noise with low hum and low microphony?

By the way, the article shows that at least for audio frequencies the difference in noise between pentodes and triode-connected pentodes is not that big. For some obscure reason the author sometimes even measures a higher equivalent input noise for a triode-connected pentode that for a normally connected pentode.

Regards,
Marcel
 

kevinkr

Administrator
Paid Member
Low noise pentodes - take a look at D3A, 6688, 7788/EF810F, C3G, C3M. You should probably plan on heating with DC if low hum is an issue, and these are all variably microphonic - some samples hardly at all, whilst others can be quite microphonic.

Note that the C3G and C3M have fragile filaments and should probably be heated via a CCS and not disturbed in any way until cool.
 

kevinkr

Administrator
Paid Member
Hi SemperFi,

The article I referred to is about a trick to get rid of the partition noise in pentodes, but my question is more general: do you know any valve, triode, tetrode, pentode, whatever, that combines low noise with low hum and low microphony?

By the way, the article shows that at least for audio frequencies the difference in noise between pentodes and triode-connected pentodes is not that big. For some obscure reason the author sometimes even measures a higher equivalent input noise for a triode-connected pentode that for a normally connected pentode.

Regards,
Marcel

The manufacturer's data sheets generally show a significantly lower equivalent noise resistance (typically less than half) for a triode connected pentode as compared to that pentode in pentode connection. (D3A... etc.)

Practically speaking the difference might be a dB or so..

Direct experience with both the D3A and 7788 indicate that these types are quiet enough whether pentode or triode connected for signal levels down to a mV or so.

The EF86 in my experience is a much noisier pentode than any of the devices I mentioned in my previous post.

I think you will have to select for microphony or design to avoid exciting microphonic tubes. (It can be done)
 
Last edited:
Hi SemperFi,

...my question is more general: do you know any valve, triode, tetrode, pentode, whatever, that combines low noise with low hum and low microphony?
Regards,
Marcel

well yes indeed:) the wonderful 6922 and even the lower spec'd 6dj8. (I'm convinced all new production '6922' is actually 6dj8, still my favourite small signal tube.

The only small signal pentode I've used is the E180F, nice NOS version with gold pins and all. Still too microphonic and got swapped with my fav triode.
 
valves low noise low microphony

Hi

I have built a pile of mike preamps and phono stages and here are a couple of "secrets"

Let's talk 60 DB gain, which is where most mike preamp gain stages are when the gain is all the way up, and the phono stage for MM cartridges is the same in the bass region


A. Pentodes are much less noise than triodes at high gain as the screen grid gives pentodes like 5879, ef86, #78, 39/44 and the lost EF40, ef804 and a few other unknowns, and the great 6sj7, 6j7, and 6sk7 pentodes gain levels up to 800, while a realy good 12ax7, can get a voltage gain of 100.

So a 12ax7, running at 60 DB gain, is still much more noise than a pentode at same as the pentode is loafing at that gain level, and designed to have low noise at much higher levels.

I have built several tube mike preamps, on the right side of page one of my old photo album,

Pictures by SalesBoy - Photobucket

you will see a four tube preamp. I built several of these using a 39/44 early pentode, grid cap, and found this tube to be super quiet.

I built two models that used a 39/44 or a EF40 as the first stage, switchable by just turning on the filament of the tube you wanted to run. The sonic and noise difference between a 1939 39/44 and the EF40, a European version of the "EF86 with the right Telefunken style insides" is almost zero.


B. The "lost EF40" and the "EF86 with the right Telefunken style insides" bit is this prroblem with finding low noise tubes.

There have been various versions of any tube made over the years, and by now, most of the better, lower noise, better sonics versions have been collected and used.

I built a series of preamps for a super piezo pickup rig, for testing tube types and preamp circuits to see what version preamp they liked for sonics and noise.

We made one with a 12ay7, and my boss had a friend that had scarfed a pile of 12ay7s that were very special versions and much lower noise than any other.

They might have been made for oh-silly-scopes, HP I think, and that happens that special versions are made for test gear or special equipment, and limited runs.

The EF40 deal is this. I have very super critical studio owners I fix gear for, and some of them have old Telefunken and other German mikes and preamps.

There is a certain version of the E804s and the EF86 tubes that are much less noise and have a "Magic Tone" that no other version has. For a while, these special versions were sold for up to $300 and now are I think, all gone.

The EF40 gag is that you can buy these all day for $5 from ebay, and other than the Euro 8-pin socket, they are usually the best version, plate and grid version of the best EF86.

I have 40, Amperex, Bugle Boy, made in France, NOS in box, EF40s, and I recently tried to sell some, and no interest at all. I just looked and there does not seem to be any out there now, very odd stuff.

But the point is that I have been building gear with 1935 to 1950 tubes, and EF 40s, 6sj7s etc, and building low noise preamps for very little Moo while others fight over a few special rare tubes.

Also, maybe a "C" is that you can also make very low noise preamps by going three or even for stages, running at lower gain levels, where the tubes produce low noise. I know of one great phono stage that used like three different tubes, and ran them at low gain levels, where the tubes had thier best noise region.

A 12ax7 for example, has low noise at about a voltage gain of 30

A EF86, about 100

EF86, E804s, 6sj7, and I have also used 6eh7, 6ej7, have special designed plates and grid structure, with outside shields that give low noise and low microphony

Lots of study, and you need a pile of tubes, different versions to play
 
Thanks everyone, for sharing your favourite valve types with me!

I have two related questions about the 1/f noise:

1. Is there any relation between the equivalent input 1/f noise and the biasing conditions of the valve? If so, how should I bias a valve to get the smallest equivalent input 1/f noise?

2. Do you know of any long-term degradation effects that aggravate 1/f noise in old but unused valves? The reason why I'm asking is that in the article I referred to, a new EF86 was exactly on spec while three new old stock EF86's were far above the specified noise level.
 
I recently "discovered" the 6J7 and can at least attest to the fact that is both quiet and good sounding. I'm using it in a power amplifier design, it is somewhat microphonic, but quiet on AC filament power.

For low microphonic I have some other missile-rocket-science tubes. I won't name them, because I am going to buy supply for myself, then reveal all details.
 
About the bias point dependence of the 1/f noise: I've borrowed the book "Noise" by Aldert van der Ziel from the Delft university library. Apparently there are several different mechanisms leading to 1/f noise in valves, all with their own bias dependence:

1. One of the 1/f noise mechanisms mentioned in the book is work function variations of the cathode due to atoms moving around. Ideally this gives a bias point independent equivalent input 1/f noise voltage.

2. If the cathode suffers from island effect, however, the 1/f noise due to work function variations gets worse at too low bias currents. As a rule of thumb, use at least 0.5 mA.

3. Another effect mentioned in the book is a theory about emission centres with a limited lifetime that Schottky came up with in 1926. This effect would give a bias current independent equivalent input 1/f noise voltage.

4. However, for oxide coated cathodes, 1/f noise due to variations in the conductance of the cathode coating give an equivalent input 1/f noise voltage that increases with increasing bias current. This effect gets much worse at low cathode temperatures.

5. Also 1/f noise due to a barium orthosilicate interface layer between cathode and coating gives an equivalent input 1/f noise voltage that increases with the bias current. Such an interface layer forms in aged valves that have not conducted much current.

The effect can be quite dramatic: a graph in the book shows an equivalent noise resistance of 70 kohm at 600 uA and 100 Hz for a fresh 6SJ7 valve and 6 Mohm for an aged 6SJ7. Maybe this is the reason why the new old stock EF86's in Bloehbaum's article were so much worse than the new EF86.

The interface layer can be removed by running the valve at a higher cathode temperature or a larger current over a prolonged period. By the way, the datasheets of some special quality valves specifically state that these do not suffer from the formation of an interface layer.

6. Some bright emitting valves with tungsten cathodes suffer from extra 1/f noise due to the emission of positive ions from the cathode. This effect is at it worst in the space charge region of the operating characteristic.

7. Finally, from an article abstract I found on the internet I understood that there is such a thing as partition 1/f noise, so using triodes or triode connected pentodes also helps against some of the 1/f noise.


To make a long story short, use triodes or triode-connected valves (or Frank Bloehbaum's patented circuit), do not use a filament voltage below the specified value and determine experimentally what bias current results in minimum 1/f noise. You may get a different result for an old valve than for a new valve of the same type.
 
Very interesting. It is a pity that much of this stuff is still not easily accessible on the web. It is either in old books, or charged-for journals. When I worked in a university I could access some of this, but not now.

My guess is that partition 1/f noise is small, as I think it was not discovered until the 1960s. If it was large it would have been found earlier.
 

trappeur

Member
2010-10-18 9:30 am
High gain low noise stage ??

Hello , new player interested by your concern here
Surprisingly nobody's talking about cascode ??
I remember very efficient E88CC playing that game !!
But it was compared only against EF86 and anything else ...

Best regards , Patrick.
 
The article I referred to is about a trick to get rid of the partition noise in pentodes, but my question is more general: do you know any valve, triode, tetrode, pentode, whatever, that combines low noise with low hum and low microphony?

Regards,
Marcel

I have used:

Sovtek 6SL7
Sovtek 6SN7
Westinghouse JAN 6J5
6FQ7 (NOS, GE)
Motorola 6BQ7

All of these types proved to be very low in the microphony department, and hum-free.

The main problem with microphony occurred with Sylvania 6BQ7s with the series connected heaters. These rang like bells. The Motorola and Sylvania 6BQ7s with parallel connected heaters are not microphonic at all.

There is also a problem with some NOS 6SN7s, in that some of these can be nastily microphonic.

Surprisingly nobody's talking about cascode ??

I've used a cascoded LTP design to eliminate an additional gain stage. It works just great, low microphonics and hum if you use the right VTs (parallel heater 6BQ7s, as opposed to the ones with series heater connections). Sounds excellent.
 
Last edited:
For whoever is interested: I've done some noise measurements on four triode-connected EF86's:

Two EF86's of the brand Amperex, bought second-hand on Koninginnedag (Dutch national holiday), complete box with all kinds of valves for one euro with no guarantee of any kind

Two EF86's of the brand Trigon, bought at a shop (Radio Service Twenthe in The Hague)

For all four I listened to the noise while checking the A-weighted noise by reading off quasi peak-peak values with an oscilloscope with and without 47 kohm//1 Mohm connected between grid and ground.

The first Amperex EF86 showed very erratic behaviour: very strong low-frequency noise that sounded like a broken contact below a certain bias current level (of the order of 1 mA), reasonably normal sounding noise at higher current, grid currents up to 1 uA flowing out of the valve.

The second Amperex EF86 had no measurable grid current, but it also had strong low-frequency noise that sounded like a broken contact below a certain bias current level, reasonably normal sounding noise at higher current (1 mA or higher).

The Trigon devices were more well-behaved: the noise sounded normal at all current levels I tried.

For all four EF86's the A-weighted equivalent input noise voltage was at its lowest at the highest current I tried: 4 mA anode current.

For the second Trigon EF86 I did more accurate noise measurements. Using a home-made 100 times amplifier with A-weighting filter and a laptop, I recorded the noise of the valve, of valve plus 47 kohm//1 Mohm between grid and ground, and only of the 100 times amplifier and computer. Using GoldWave, I applied a comb filter and a steep 20 Hz high-pass to get rid of any remaining hum and subsonic supply voltage variations. I then measured the RMS values (using Volume Match in GoldWave) and calculated the equivalent input noise from the results.

The results are (expressed as an equivalent white input noise voltage that would give the same integrated noise):

Second Trigon EF86:
flat (except for comb) 20 Hz to 20 kHz:
20.02 nV/sqrt(Hz) at 298 uA
9.15 nV/sqrt(Hz) at 907 uA
7.41 nV/sqrt(Hz) at 2073 uA
7.7 nV/sqrt(Hz) at 3960 uA

A-weighted (plus comb):
17.32 nV/sqrt(Hz) at 298 uA
8.82 nV/sqrt(Hz) at 907 uA
6.53 nV/sqrt(Hz) at 2073 uA
6.34 nV/sqrt(Hz) at 3960 uA

Bandpass 1 kHz to 1.2 kHz (plus comb):
9.89 nV/sqrt(Hz) at 907 uA
9.1 nV/sqrt(Hz) at 2073 uA
9.65 nV/sqrt(Hz) at 3960 uA

A-weighted noise for three valves measured with the inaccurate quasi peak-peak method:
Second Amperex EF86, the one without grid current:
228 uA: infinite (that is, valve noise is so high that the effect of the 47 kohm//1 Mohm is not visible)
1 mA: 8.26 nV/sqrt(Hz)
4 mA: 6.19 nV/sqrt(Hz)

First Trigon EF86:
220 uA: 12.08 nV/sqrt(Hz)
362 uA: 11.58 nV/sqrt(Hz)
2.16 mA: 6.44 nV/sqrt(Hz)
4.05 mA: 6.26 nV/sqrt(Hz)

Second Trigon EF86:
201 uA: 16.21 nV/sqrt(Hz)
347 uA: 14.76 nV/sqrt(Hz)
2.11 mA: 8.76 nV/sqrt(Hz)
3.89 mA: 7.64 nV/sqrt(Hz)
 
Sorry, could you clarify these results? Are these with or without 1M, or the difference between the 47K readings and 1M readings? A 1M resistor would give 126nV/sqrt(Hz) of thermal noise on its own; a 47K 27nV/sqrt(Hz). Or did you have 47K in parallel with 1M? Have you subtracted off the thermal noise?

Although I am not clear exactly what you did, the results do show flicker noise at low frequencies getting worse at lower anode currents, while mid-frequency noise seems to be independent of current. Both results are surprising, although I am not clear what the contribution of grid current noise is.