Tube phonostage RIAA design

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

emi2345

Active member
Joined
Oct 20, 2017
Messages
43
Hi everyone,

I have few nos tubes lying around and wanted to upgrade my £50 transistor phono amp with a home build tube one. I’m using a MM cartridge, the ortofon blue one. I’ve built a number of tube circuits, hifi, guitar and mic, but this would be my first phono. I’d love some help drawing up a design. Here are the tubes I have to play with:

Octals
5x 12SN7 = c.x15 24dB Rp 7k7 2.6mA/V
2x 12SL7 = c.x40 32dB Rp 44k 1.6mA/V

Subminis
5x 6021 twin triode = c.x20 26dB Rp 6k5 5.4mA/V
5x 6948 twin triode = c.x45 33dB Rp 39k 1.7mA/V
4x 6788 pentode = c.150 44dB Rp 1.2m 1.13mA/V; can’t find triode curves for this

I’m also happy to throw in any new tube to get the best design like 12A_7, EF86, ECC88 etc.

My initial idea is to try a DC-DC converter power supply that I already have, which can give me 300v 50mA from 12vdc, and I can do the 6v heaters in series pairs.

The 6948 is quite similar to the 12SL7, and the 6021 is quite similar to the 12SN7.

I have two ideas for topology:

1. Hi-mu triode common cathode in parallel pairs, passive RIAA network all in one, then low-mu triode common cathode and DC-coupled cathode follower for output.

2. Hi-mu triode common cathode into DC-coupled cathode follower, passive RIAA network all in one, then low-mu triode common cathode into DC coupled cathode follower.

For either idea I can use the 6948 or 12SL7 as the hi-mu and the 6021 or 12SN7 as the low-mu. Both would give me a gain of 40-45dB which is about right for the MM phonostage. I have 5x 6948 and 2x 12SL7 so for idea #1 I could get Rp down to 10k or 22k respectively.

Any guidance or advice as to which direction I should go would be much appreciated!
 
Try a D3a triode connected for the input gain stage...has a gain around 77 or so, much less noise or microphonics than a 12AX7..Observe precautions (stopper resistors, etc.) when working with these high gm tubes....The B plus is in the range of 150 volts and 15 ma or so...the RIAA eq could be done at lower impedance to avoid "squashing" the high end...
 
emi2345 said:
Any guidance or advice as to which direction I should go would be much appreciated!

What about balanced RIAA with subminies (6948+6021)?
http://www.tubecad.com/july99/page9.html
 
Is the D3A really that good? I am sure its shot noise will be very good but for audio work the 1/f noise is just as important especially for an RIAA preamp with its 20dB of LF boost.

Cheers

Ian
 
moamps said:
https://tavishdesign.com/pages/audio-tube-noise-measurments

Yes, I am aware of this work. Unfortunately, except for the D3a, the operating currents chosen are not optimum for lowest 1/f noises so it  is not clear what is actually being tested. Interestingly, the D3a is only 3dB better than a 12AX7 so you could do as well simply by paralleling both halves of a 12AX7 at half the current.

Cheers

Ian
 
IMO, D3a is excellent tube (as C3g and C3m too),  and I would never change it for two boring halves of 12AX7. :)
 
emi2345 said:
Is the implication of this that JFETs are a much better choice for low noise?
Usually yes, but you would like to listen music, not how low is the background noise.
The lowest noise doesn't mean the best music performance automatically.

 
moamps said:
What about balanced RIAA with subminies (6948+6021)?
http://www.tubecad.com/july99/page9.html

This looks great, I have enough submini 6948 and 6021 to make a balanced circuit. The problem is that I’ve always worked with single ended designs before so I’m a bit out of my depth in the design and calculations. Thank you for the suggestions of different tubes, but I’d like to use the ones I already have provided I can get a decent circuit out of them. Here are my questions about the topology:

1. Input differential 6948 gain stage with 10M45 CCS in the tail to force balance between the triodes. How is this going to perform noise-wise? Can I also use a MOSFET CCS as a load for each plate to up the gain and linearity?

2. Differential passive RIAA all in one. I’m not sure quite how to calculate the values for a differential one.

3. Differential 6021 gain stage. Can I use a ccs in the load and the tail to help with balance and linearity? Can I DC couple each phase to a cathode follower to drive the line level output? This would require 2x 6021 bottles per channel but I have enough for this.

Many thanks
 
emi2345 said:
Is the implication of this that JFETs are a much better choice for low noise?

The implication is that for that particular configuration a FET will exhibit lower noise. It is well known that tubes are not the quietest devices, hence the plethora of measurements of them. What is less well understood is the variation in the several noise types with quiescent current. The prevailing 'wisdom' among the hi-fi fraternity is that high gm equals low noise so they seek hi gm tubes and run them at high currents where the gm is highest. Hence the preference for the D3a.

A deeper analysis will show that high gm equals low noise is true only for shot noise. Most tubes had their noise figure specified in the MHz region where shot noise predominates. Audio, on the other hand is far away from this region and in the lower parts of the audio band 1/f noise predominates. This can easily be seen in spectrograms of tube mic pres made by myself and several others here. 1/f noise is easy to see but it  is less easy to characterise and is thought to be dependent in part on material impurities. The ancients knew this well which is why they selected tubes  and operating points for low 1/f noise for the first stages of mic pres. Here is the basic noise formula for a triode:

noiseformula.png


The first term is the shot noise and the second is the 1/f noise. Since shot noise is inversely proportional to gm then it makes sense to maximise gm for lowest shot noise. However, 1/f noise is proportional to the square of the ratio of the quiescent current (Ia) to gm. If you look at the 12AX7 data sheet you will find that at 1.5mA quiescent current, the ratio Ia/gm is about 0.8. The square of this is 0.64.  At a quiescent current of 0.5mA this ratio is about 0.4 and the square is 0.16.  In other words the 1/f noise is a lot better at low quiescent currents.

Cheers

Ian
 
ruffrecords said:
.... The prevailing 'wisdom' among the hi-fi fraternity is that high gm equals low noise so they seek hi gm tubes and run them at high currents where the gm is highest. Hence the preference for the D3a....

The Hi-Fi fraternity is the population that buys CDs and records, goes to concerts and evaluates the work of musicians and professional technicians. So it is not clear to me why it is often underestimated and ridiculed here. Attached is a graph showing the dependence of gm of Ia for tube D3a in the triode connection. The Ia + g2 / gm ratio changes very little, from below 0.4 to 0.66 over the entire current range Ia + g2, which means that wherever this current is set, the possible “error” will be small if it exists at all. Given that the gm of the D3a tube is far larger than the gm of the 12AX7 tube, it is also obvious that the total noise will be less and the proportions of the flicker noise may be equal. Linearity and load capacity is also better with D3a.
Of course, some tubes are preferred by "professionals", but that's only because there are still factories in Russia and China that produce them and they are cheap. DIY-ers on the other hand still have the freedom to use better solutions because the higher price for just a few tubes is not a problem, nor the time it takes to get an outstanding and unique product. I will end by quoting the introductory part from Pete Millett pentode riaa preamp project:
http://www.pmillett.com/pentode_phono_preamp.htm
"I've been working on a passive RIAA pentode phono preamp. I really like the way it sounds, but as you might expect, using a pentode in the front end is not the quietest thing you can do. I find it acceptable, but others may or may not ... the noise floor is about what I get for surface noise on a brand new test record. On most of the vinyl I have (granted, not all pristine) I can't hear the noise contribution of the preamp. "
 

Attachments

  • D3a.png
    D3a.png
    75.6 KB · Views: 10
moamps said:
The Hi-Fi fraternity is the population that buys CDs and records, goes to concerts and evaluates the work of musicians and professional technicians. So it is not clear to me why it is often underestimated and ridiculed here.
the modern hi-fi fraternity is into cables:  directional and "burning" them in.    they deserved to be ridiculed.
the fraternity of 60 or so years ago appreciated designs of the ancients, including liberal use of 12AX7a and similar at low quiescent current.

Thanks Ian for that enlightening post.
 
moamps said:
Attached is a graph showing the dependence of gm of Ia for tube D3a in the triode connection. The Ia + g2 / gm ratio changes very little, from below 0.4 to 0.66 over the entire current range Ia + g2, which means that wherever this current is set, the possible “error” will be small if it exists at all.
You forgot the square term.
0.66 squared is 0.4356
0.4 squared is 0.16

0.4356 divided by 0.16 is 2.7225. This is not a small error.

Cheers

Ian
 
moamps said:
Given that the gm of the D3a tube is far larger than the gm of the 12AX7 tube, it is also obvious that the total noise will be less and the proportions of the flicker noise may be equal. Linearity and load capacity is also better with D3a.
It is not obvious to me. Both our figures demonstrate that the Ia/gm ratio for both tubes spans a very similar range. This means that the 1/f noise in both tubes is likely to be very similar. The D3a shot noise may be lower but the 1/f noise will not be and, I say again, this is important for an RIAA preamp with its 20dB of LF boost.

Whatever you choice of tube, in this application, it is important to select its operating point with more care than simply aiming for maximum gm.

Cheers

Ian
 
emi2345 said:
Hi everyone,

I have few nos tubes lying around and wanted to upgrade my £50 transistor phono amp with a home build tube one. I’m using a MM cartridge, the ortofon blue one. I’ve built a number of tube circuits, hifi, guitar and mic, but this would be my first phono. I’d love some help drawing up a design. Here are the tubes I have to play with:

What's wrong with your transistor phono amp? a tube one will probably not be any better. But then again I'm not a big fan of those semaring, distorting, noise adding things you people call tubes.
 
user 37518 said:
What's wrong with your transistor phono amp? a tube one will probably not be any better. But then again I'm not a big fan of those semaring, distorting, noise adding things you people call tubes.

DIY is fun, tubes are fun. This is probably a personal attack but you are not fun.
 
ruffrecords said:
You forgot the square term.
0.4356 divided by 0.16 is 2.7225. This is not a small error.
You overlooked word entire.
This ratio (squared Ia/gm) for entire range for 12AX7 is more than 16.
In a usual range of Ia+Ig2 of 10 to 25mA for D3a, this ratio is below 2,
unlike in usual range for Ia=0,5 to 1,5mA for 12AX7 this ratio is above 4.

This is in line with what Merlin Blencowe told you seven years ago on diyaudio.com:

ruffrecords:
For lowest flicker noise you need low current and for lowest shot noise you need high current. Somewhere between these two extremes there should be an optimum noise operating point.
MerlinB:
Yes that is exactly true, especially for low-gm tubes. However, for for high-gm types the 'optimum' becomes quite broad, so you can be quite relaxed about the exact operating current and still be fairly confident that the total noise will be almost as good as can be hoped for, and almost certainly better than any low-gm type you might otherwuise have used.
(Flicker noise tends to be less troublesome to the ear, too, so reducing shot noise might be a higher concern anyway).

It is not obvious to me. Both our figures demonstrate that the Ia/gm ratio for both tubes spans a very similar range. This means that the 1/f noise in both tubes is likely to be very similar. The D3a shot noise may be lower but the 1/f noise will not be and, I say again, this is important for an RIAA preamp with its 20dB of LF boost.
2,7225 isn't small error for you but difference in gm of more than 20 may (or may not?) make a difference. Interesting.
 
Please guys, behave..!

This is what happens every time we start looking into hifi thingys - quickly, get back into pro audio before it sticks to you forever..!  :eek:

The very, very important rule is, as mentioned above: No personal attacks. We mean it.

/Jakob E.
 
emi2345 said:
Can I also use a MOSFET CCS as a load for each plate to up the gain and linearity?
Not a very good idea IMO. The two CCS will fight each other until one goes out of its intended operating area.

2. Differential passive RIAA all in one. I’m not sure quite how to calculate the values for a differential one.
Just same than single-ended, but with a double source impedance.

3. Differential 6021 gain stage. Can I use a ccs in the load and the tail to help with balance and linearity?
Already answered.

Can I DC couple each phase to a cathode follower to drive the line level output? This would require 2x 6021 bottles per channel but I have enough for this.
What benefit do you expect from it? That's always what you should ask yourself. Evaluate the pros and cons and make your own opinion.
 

Latest posts

Back
Top