View Single Post
  #2   Report Post  
Patrick Turner
 
Posts: n/a
Default



Fabio Berutti wrote:

Dear RATS,

I just finished a phono preamp using a 6J7 as first amplification stages,
followed by a passive RIAA, followed by a 6F5 as a second amp stage, and
finally by a DC coupled 6J5 CF as "output" tube. It sounds great (who ever
said "my project is bull.."?) but it is noisy, even with DC filaments. OK,
it has a lot of gain, but something is definitely wrong.
I swapped different pentodes and it seems that most of that noise depends on
tubes.
If anybody is interested in using these tubes in low noise application,
here's what I found:

- the less noisy are the metal envelope ones (6J7M, Svetlana)
- a bit worse are the European painted-glass shouldered bulb type (Brit
military CVsomething, supposedly Brimar)
- the more noisy so far are the the painted-glass, big bulb, straight-sided
EF36 (US/Canada type)

I suppose I have to work a bit on the "grounding" scheme of my unit, but the
noise I found is not only 50Hz hum.
Conclusion: 6J7 is surely a great tube in other applications (see Leak
"point one" gain stage) but not in this case. In fact, it was not described
as a low noise pentode like the EF86.

Suggestion appreciated.

Ciao to all

Fabio


The pentodes you mention are not suitable tubes for input stages of phono amps
because they have screen partition noise in addition to the grid input noise.

The noise you hear is probaly a spitty rumbly sound, typical of tube phono
stages unless the output of the cart is highish, and the tube is a low noise
triode.

I have found the 12AX7 with about 0.6 mA as set up in the phono stage at
http://www.turneraudio.com.au/htmlwe...tubepreamp.htm
to be the quietest using normally available triodes.
This mu-follower stage isn't quiet enough for MC carts which have 20 dB less
output than
the Shure V15 I was using with the above amp.

However, EF86 were very commonly used for phono input tubes but usually
only with shunt feedback applied, which also reduces the noise of the tube
to some extent.
Quad 22 and Leak used them.
But the stage in the above schematic of mine is quieter.

I now use a 2SK369 in cascode with a 6EJ7 frame grid pentode wired as a triode,
and then a second stage mu follower using a 12AT7, and that gives plenty
of gain with an Denon 103D MC, with output some 22 dB lower than the Shure.
The MC sounds a lot better.

The noise of the amp is around 20 dB lower than the 12AX7 input.
This is what a j-fet does when you use it.

The 6C45pi is a small 9 pin russian triode which may interest you because when
its set up
with 20 mA of anose current, its equivalent to maybe 15 12AX7 in parallel, which
in
theory would reduce the noise about 4 times.

The input noise for triodes is proportional to the square root of 1 / Gm.

If a twin triode is paralleled, then the grid input noise will be 0.7 times that
of 1/2 the tube.

So if 10 are paralleled, there is 1 / 3.2 times the noise of one, or -10 dB,
which isn't much for the touble you have to go to, and the shot noise might
defeat you.

I have not tried the 6C45pi, but I hear it makes a good input tube for phono
with a lot less noise than a 12AX7, let alone a pentode without FB.

I have tried 6DJ8 instead og 12AX7, and found them no quieter,
despite the theory suggesting otherwise, because 6DJ8 has a lot higher Gm
compared to
AX7.
Also 12AT7, 6201, 6AQ8 are also supposed to be lower noise.

I had to pick through the pile of 12AX7 to get the quietest one,
since the average of what I had was 6 to 10 dB noisier.

If you have a gain of 90 for V1, then 2 mV at 1 kHz becomes 180 mV.
The passive RIAA filter will reduce this to 18mV,
and the noise of V1 is also largely attenuated by the RIAA filter.
But LF rumble and flicker noise which is LF will not be attenuated much,
and that's why RIAA amp noise is a low throated roar when amplified.
So to get good SNR, you need the equivalent noise input voltage to be less than
2 uV.
This is -60 dB.
The use of 6J5, 6SN7 is fine for V2, but the input noise is likely to be 4 uV,
but compared to the 18 mV of signal from the RIAA filter, the 4 uV won't
contribute too much more noise than what's already in the signal.

Phono stages were regarded as passable if an unweighted figure of -55dB for
noise was achieved.
With MM carts, this means the amp noise is just less than the noise of the
stylus
scraping along in the little vinyl valleys.
-55dB SNR isn't a real great achievement, and -60 is good, and -66 is very good
for a triode
front end without resorting to a fet input, or some other special device.

-60dB means that if you had 1 volt of output signal there would be 1 mV of hum
and hiss type of noise.

Many cheap or poorly made phono stages have -50 dB, and you can hear
the amp in the background during quiet passages in the music.

The noise at the input of the amp determines how quiet the outcome is,
and the MM cart has a fairly high impedance, so it creates quite some noise.
The noise from the resistance in series with the grid is varies in proportion
to the square root of the resistance, so if the cart was 20k, then this makes
3.2 times the noise of a cart with 2k.
An MC might have a resistance of 100 ohms, and the noise from the cart is
extremely
low.

I just set my Denon 103 D up with my high gain preamp with a cascode fet
input stage and there is only the slightest little hiss when the gain is turned
up full
with the stylus off the record.
Fets have thd character like tubes, lots of even order stuff, but at the level
it operates in my circuit,
it only has to make few mV, so the thd is less than 0.05% at the highest level
of signal.
They are so much better than any bjt, IMHO.
BJT inputs ned low current to be quiet, 0.3mA is typical, so thay are paralleled

to suit a load, and then Rin is low, and the thd is poor, and I just don't like
bjt phono
inputs much.

The 2SK369 is a small TO92 package device, and it likes to operate with about
10v from drain to source,
and with 5 mA of current, when Gm = 35 to 45 mA/V.
2SK147 is the same device with a higher dissipation rating, but may be more
difficult to find.

Allen Wright at vacuumstate.com has some interesting fet+tube phono stages.

I was going to set my amp up with a balanced input, but the unbalanced amp uses
1/2 rthe devices, and there is *no* hum, so no need to balance the preamp for
the MC input.

Fet input stages in cascode with triodes *do* require great attention to short
leads and
careful bypassing or else the tube on top of the fet in the cascode will tend to
oscillate
at 200 MHz or so without you realizing it.
The fet input circuit and first triode is a little prone to magnetic fields,
so locate the PS well away on a separate chassis unless you can pot the
trannies, and
build the PS in a steel box at one end of a 500 mm long chassis.
The dimensions of the phono input circuit should fit on an area no bigger than
50mm x 35mm.

After the input circuit, you can layout the line stage parts in more space.

Patrick Turner.