The Muscovite Mini 6ж9п (6Z9P) Phono Stage

Status
Not open for further replies.
Thanks for the post with the simulation issue as it made me go back and have another look. I have changed to a 4.7uF Erse Pulse X and am happy to report that it sounds good and there is a bit more deep bass, enough that I expect to be able to measure it..

So the official recommendation is to use a 4.7uF or larger cap at C7, but bear in mind that quality is critical and a compact construction with minimal inductance is recommended. The 4.7uF Erse Pulse X is cheap and seems a good compromise.
 

Attachments

  • 4.7uF 20 - 30k laplace mm7.PNG
    4.7uF 20 - 30k laplace mm7.PNG
    34.4 KB · Views: 243
  • 6J9P Phono Pre SCH v10 (as built).png
    6J9P Phono Pre SCH v10 (as built).png
    38.5 KB · Views: 250
Hi,

Actually I went from gas tube back to zener for a couple of reasons, one was it limited the OLG of the error amplifier and I was throwing away at least 20dB of the available gain by virtue of the dynamic impedance of the 5651 which when measured was around 1K vs 100 ohms for 100V zener. The second issue was noise, a zener can be shunted with meaningful levels of capacitance and large numbers of them can also be placed in series and since they are essentially gaussian noise sources each doubling in number increases noise by 3dB and voltage by 6dB for a net improvement of 3dB. Four in series are actually about 6dB quieter than a single zener..

We can discuss that over a nice trappist.
5651As are true reference voltage VRs but you don't really need such precise VR in the context of a preamp.
I look at the function of the series regulator more as an AC line buffer with a battery like supply behind it. This battery consists of low esr caps, the regulator is your trickle charger if you like.
O.K. Let's make that two trappists and it may become clear.

Now, what's with the infamous C7? It determines a pole in the RIAA correction? Not sure I get how this happens.

Les chefs et leur cuisine, n'est-ce pas? :drink:

Ciao, 😉
 
Last edited:
Hi Frank,
R11/C7 create an LF pole in the transfer function of the pentode gain stage which is what I originally alleged, measured and simulated. The subsequent discussion and my mis-handling of the step parameter in LTSpice temporarily led me astray - guess I am suggestible... lol The difference is subtly audible, but noticeable - as I indicated my system is not doing much useful at 30Hz and below.. I'm sure this is measurable and perhaps not as extreme as simulated, but I will have to do it to know. Next window of opportunity is next week.

I think in terms of power supply philosophy we might be on very slightly different pages. I design mine for relatively low noise, low output impedance (relatively speaking - not nearly as low as achievable with SS, but as low as I can reasonably get it, often <1 ohm) and open loop bandwidth encompassing the entire audio pass band.. There is actually a reason why I use the technically questionable choice of a 12AX7A in cascode as an error amplifier, the ECC88 seems to have a sonic signature in this application that I can't live with. There is no significant decoupling between my supplies and what they power - and this is a deliberate choice based on many years of listening and tinkering which means that the supply is quite audible since my designs generally have really terrible PSRR.. I have some really strange aversion to RC decoupling in audio and avoid it where possible.. (Something about LF poles too close to the audio range, and dielectric effects in large, reasonably price film caps, not to mention electrolytics. Yeah guess I am loony.. 😀 )

I would most definitely relish chatting with you about anything audio over a good Trappist Ale when I do finally make it to Belgium. (One of my new goals)
 
Hi,

I think in terms of power supply philosophy we might be on very slightly different pages.

Not as much as you seem to think.

For series regs I also use a cascoded ECC83 (You could even use a penthode for even greater error correction if you must, adds noise though).
The Z of the PS is however dominated by the massive cap bank that is put behind. mere C's, not RC's.

With well over a thousand microF at 300VDC per stage, per channel that constitutes an enormous Joule content the valve can draw from.
A valve stage that will draw 10 to 25 mA peak?

Enjoy the music, 😉
 
A couple of additional minor changes, I have shunted the 4.7uF screen bypass with a 1.0uF - both are Erse Pulse X types.. In addition I am now returning the screen bypass capacitors to the cathode, in theory this should result in slightly lower distortion as the cathode is not quite at ground potential although the dynamic impedance of the LED is rather low. (Under 10 ohms)

The 6N14P arrived this week-end while I was away at Son et Image in Montreal listening to lots of vinyl. I have some ideas for another design if these perform sufficiently well. Moving on to the next design shortly...
 

Attachments

  • 6J9P Phono Pre SCH v11 (as built).png
    6J9P Phono Pre SCH v11 (as built).png
    35.2 KB · Views: 212
Hi,

In addition I am now returning the screen bypass capacitors to the cathode, in theory this should result in slightly lower distortion as the cathode is not quite at ground potential although the dynamic impedance of the LED is rather low.

Incidentally I was going to suggest that to you.
Either that or doing away with the caps altogether and regulate the screen grid with a string of (dare I say it) zener diodes....

Good to see you enjoyed your stay in Montreal. Love that city. 😎

Ciao, 😉
 
Hi Frank,
I actually started with zener diodes, but got rid of them because it appeared that running the screens on a fixed voltage supply resulted in significantly different transconductances sample to sample and made gain match between the channels difficult to achieve without careful matching. (I'm lazy and I was concerned that many potential builders who do not have the groovy test hardware I have would be put off by this issue.) The current feedback the screen resistors provide actually helps to provide consistent performance over varying tube samples which is why I have suffered over the cap conundrum.. 😀

Montreal was fun, good food, plenty to look at, and an even worse climate than here. (The weather sucks here so... 😛 )
 
Hi Kevin,

I have no idea about this but when the caps are returned to the cathode this normally allows for lower value compared to a return to ground.
Not sure if this is the case here with LED bias.

Just a thought....

Montreal was fun, good food, plenty to look at, and an even worse climate than here. (The weather sucks here so... )

Last time I was in Montreal it was -20C. It didn't even feel all that cold and we had a marvellous deep blue sky most of the time.
Today, en Belgique, +21C. (Not to make anyone jealous, just reporting the facts)....😛

Later, 😉
 
Hi Frank,
I looked at the cap values and in simulations which have correlated well with measurements the trend even with them returned to the cathode is that bigger is still better in terms of bass response. It did seem to sound a bit better with the change, before I switched to SACD to listen to some organ music recordings I acquired in Montreal during the week-end.
 
Hi,

@Kevin: I suspected as much. I guess that pretty much closes the chapter of the penthode Muscovite then. 🙂

@Piano3: For some obscure reason (probably its lowish mu) this double triode and its equivalents have been a sleeper so far.
Contrary to its equivalents (European and US) the 6N14P sports a gold treated grid which assures low grid leakage.
As a matter of fact I've also been eyeballing it for use in a starved anode voltage circuit myself.
Haven't even seen a single one up close yet though.😀

Ciao, 😉
 
Last edited:
The interesting design and apparently very good linearity is what attracted me to the 6N14P for another cascode front end based design. I am not sure how good its noise performance actually is, but I believe it still may edge out the 12AX7A in this regard. (The equivalent input noise resistance is about 600 ohms, compared to other types I have found to be quiet which generally range from 60 - 240 ohms.. As you know this is an RF noise measurement and may or may not correlate well to actual audio range noise performance depending on type.)

There are a number of interesting English language Russian websites that I troll from time to time out of curiosity. Here is a veritable goldmine of information:Russian Tube Tester Files - Instrumental Curves and more
 
Hi,

As you know this is an RF noise measurement and may or may not correlate well to actual audio range noise performance depending on type.)

IHME experience there's a fine line between maximizing Gm and plate current.

See also Hooper and Cherry's book entitled: "Amplifying Devices and Low-Pass Amplifier Design".

And the MC stage I offered for discussion earlier is also a fine example where shot noise and flicker noise are reduced to the limit (again IMHE) by running each triode at rather low idle current. Naturally you need to look at this within context.
People likely to have the most experience in this thorny field are often those working in the studio world building mic amps and such.

Fortunately we have some of these as contributing members of this forum which is why I always try to read these thread on mic amps etc.
Thinking of guys like Ruff, MerlinB etc.

Ciao, 😉
 
FFT Hardware Improvements

I finally got all of the bits and pieces together and modified my FFT based measurement setup tonight. I wired up a pair of WE REP111C repeater coils (land line broadcast isolation transformers) and installed them between my 24192 IO and my modified Pete Millett sound card interface.

The transformers isolate the balanced inputs and outputs from the interface from the ac currents flowing in the chassis of the computer.

I think the results speak for themselves.. Levels with the coils are 5mV and the reference from a couple of weeks back is 6mV without isolation. In both cases the interface is at DUT ground. I used an HP attenuator and run the source at a couple of volts rms. The measured fundamental is at 5mV or 6mV respectively..

The gist of this is that I should be able to get some significantly more meaningful measurements with low signal levels...
 

Attachments

  • REP 111C Modification of FFT Analyzer.png
    REP 111C Modification of FFT Analyzer.png
    16.8 KB · Views: 192
  • noise 6mVin single external ground.png
    noise 6mVin single external ground.png
    18.2 KB · Views: 188
The interesting design and apparently very good linearity is what attracted me to the 6N14P for another cascode front end based design. I am not sure how good its noise performance actually is, but I believe it still may edge out the 12AX7A in this regard. (The equivalent input noise resistance is about 600 ohms, compared to other types I have found to be quiet which generally range from 60 - 240 ohms.. As you know this is an RF noise measurement and may or may not correlate well to actual audio range noise performance depending on type.)

There are a number of interesting English language Russian websites that I troll from time to time out of curiosity. Here is a veritable goldmine of information:Russian Tube Tester Files - Instrumental Curves and more


FYI - that Russian page mentions 6N14P as being ECC42 'clone', but my Russian data book on Russian/Western tube equivalents shows it as exact equivalent of ECC84. These were used in Soviet Union in cascode front ends of TV sets prior to swithching over to 6N23P (ECC88).
I just opened one 'virgin' box of 6N14P and pulled out the leaflet. Indeed, the tube contents 0.555 mg of gold.
Apparently some of these were also used by non civilian agencies - mine has the "rhombus' stamped on it (= military QC).
 
Hi,

The gist of this is that I should be able to get some significantly more meaningful measurements with low signal levels...

Looks much better.

FYI - that Russian page mentions 6N14P as being ECC42 'clone'

Does this ECC42 even exist?
What does exist is the Tesla 6CC42 though which is close to a 2C51, WE396.
I would take all these so called equivalents mentioned on sellers' sites with a grain of salt.

Ciao, 😉
 
I will start working on a front end using the 6N14P shortly, and am thinking about a 6N23P in an odd variant of the SRPP that will perform more like a mu follower for the output stage. I will probably start a new thread to discuss this design as it is not intended to be a replacement for the mini, but another theme altogether.

The remaining questions to be answered is will it be quiet enough, and can I get sufficient gain out of it.

Touching on Frank's comments I've never seen an ECC84 or anything vaguely equivalent to the 6N14P on this side of the pond - I probably have not looked hard enough though.
 
Status
Not open for further replies.