Re: MC PHONO PREAMP

Patrick Turner wrote:

Paul Crawford wrote:
Patrick Turner wrote:
Yes, the site isn't fully operational after a move to the new ISP.

I am working on restoring links right now.

Seemed to work when I looked today :)
Paralleling 6DJ8 ECC88 etc is a hard way to get low noise, and I bet you
never get what the theory books say you ought to get. Equivalent input
noise resistance for a triode = 2.5 / Gm where Gm is in amps per volt,
so if gm was say 0.0012A/V for 1/2 a 12AX7, EINR = 2.5 / 0.0012 = approx
2,000 ohms and the noise 20Hz to 20kHz for a 2k resistance is about 2uV
at room temp - in theory. so if the signal is 2mV, then SNR = -60dB but
2mV is a high level signal quoted for an MM cart and average signals
might be a lot less to SNR isn't so great. Obviously with MC with a

I was just using the two halves of a ECC88 and came to much the same
conclusion, it was just usable with a 'good' tube for MM input and
really not a match (sadly) for a decent op-amp.
Now using just one single 2SK369 at only 5mA of idle, you get noise at
-20dB below a 12AX7, and reliably so, and noise spectra is less bassy so
the noise should sound like a high pitched hiss with gain turned up
+20dB above normal use levels and not at all noticeable at normal
volumes. The noise of an unmodulated vinyl track will then be much
greater than the amplifier noise.

J-fet noise theoretically = 0.7 / gm. 2SK369 at only 5mA has gm =
0.04A/V so EINR = 0.7/0.04 = 17.5 ohms.
The noise of 17.5 ohms = 0.2uV approx. The j-fet noise formula is from
the UK amateur radio handbook and the triode formula from RDH4.

I was looking at this:

www.linearsystems.com/datasheets/LSK170.pdf

That has about 1/2 the gm of the 2SK369...but a good noise figure.
as a possible choice. Given the need to buy a lot (at least, from
someone) then paralleling them, each with a gate stopper, makes some
sense prior to a high dynamic range valve system..

Paralleling j-fets is very doable; using 4 will reduce noise -6dB, ie,
halve noise voltage.
Gain into a given load will 4 x more, or if you want, RL can be lower
than for one j-fet.

Mosfets are not much good despite their higher gm than a j-fet. Its
because mosfets have "popcorn noise", ie, pops and sploshes like tubes.

The reason they are unloved for RF oscillators - too much 1/f noise :(


Maybe a better tube for MC is the 645pi russian twin triode. The gm per
two parallel halves appraches that of the 2sk369, but you need quite
high Ia.

I did wonder if the planned new UK made vales from here:

http://www.techtubevalves.com/index.php

Might include an E88CC alternative. They claim to have higher emission
cathodes, and I guess the cloud of surplus electrons might help reduce
noise if they did, not sure if it would be enough for the very low
frequency lumpiness though..

I emailed this crew for samples i'd test independantly for them.

But so far they are wanting to make only equivalents of 3 staple tubes,
the 12AX7, 12AT7, and 12AU7.

The gm of such tubes is mainly determined by the distance between grid
and cathode and the fine wire coil pitch; the problem for tube makers
who wanted to make high gm tubes was that normal helical grids close to
a cathode were hard to produce.

So they turned to making tubes with the frame grids as used in the 6DJ8
(i think?) and definately in quite a few pentodes such as 6EJ7.

Now if ever anyone wanted to have a high gm input and hence low EINR,
one could have say 4 x 6EJ7 in triode and get 50mA/V and thus theoretically EINR = 50 ohms, and hence should be OK for MC. Now
the µ of the 6EJ7 = about 60 in triode, so such an input stage would
have good gain with a CCS load.


I have yet to come across any work that verifies the old 2.5/gm equivalent input noise resistance for a tube. I have yet to see better than 2uV from a single tube.


Cheers
.

Quantcast