12ax7/5751 - using the full tube or half in gain stage

[elsewhere]

12ax7/5751 - using the full tube or half in driver stage

I've noticed in some amp designs. 12ax7 or 5751 tube in the gain stage can be wired in 2 ways:

1. Use Two 12ax7 or 5751 tubes - bridging the dual grids of the tube so that one tube is used for each stereo channel

or

2. Use only One 12ax7 or 5751 tube where each half of the tube handles a channel

Looking at the designs, it looks like there would be better channel separation using #1. I see the advantage in terms of cost going with #2 but am wondering if this design worsens the sound quality (all else being equal)

Generally speaking I'm wondering if the sound quality is better if each channel in a stereo amp had its own driver tube or if it is no big deal in using 1 tube using a half for each channel?

Thanks!

 

[Osvaldo de Banfield]

Since my point of view, and having in mind that both halves of each tube become mixed (P,G,K, HH, P,G,K,H), it is obvious that one channel will have more mixing from the other when using the same tube for both channels. I largely prefer one bottle per channel.

 

[BinaryMike]

Paralleling the twin triodes in a single envelope is one way to obtain a single tube with characteristics best suited to the application. Transconductance doubles and anode resistance is halved, relative to a single device. Inter-element capacitance also doubles, of course.

 

[elsewhere]

Thanks! I didn't know what the term for doing this was parallel triode. Doing a search for that pulled back a lot more information that was useful to read!

Paralleling the twin triodes in a single envelope is one way to obtain a single tube with characteristics best suited to the application. Transconductance doubles and anode resistance is halved, relative to a single device. Inter-element capacitance also doubles, of course.

 

[DF96]

There is another option. Use two valves, but using opposite triodes in each channel e.g. left channel uses 1,2,3 triode and right channel uses 6,7,8 triode. You also need to ensure that only the relevant heater is connected (pins 4 or 5). This gives you good channel separation, just like using separate triodes ought to. Then when the valves are beginning to wear out and lose emission, swap them over - each channel then gets an unused triode.

Not my idea; I saw someone suggest it on here a while ago.

On the other hand, using a single valve might not give too much crosstalk. I use a single ECC81/12AT7, but that has the advantage of much smaller anode impedance so less crosstalk for a given amount of stray capacitance.

 

[elsewhere]

Thanks! I was reading through a bunch of threads and saw that idea and thought that was really cool

Does using only one triode of the pair in the glass tube affect the sound quality in any way? I was reading that some people don't like paralleling the triodes because the triodes are generally not matched in the glass and it can make the sound less clear but I've only seen a handful of these reports. The vast majority said it would be fine to parallel and you can find 5751s or 12ax7s that have their triodes matched in the glass on sale. Many tubestores have that option if you need to purchase a matched internal pair triode tube.

There is another option. Use two valves, but using opposite triodes in each channel e.g. left channel uses 1,2,3 triode and right channel uses 6,7,8 triode. You also need to ensure that only the relevant heater is connected (pins 4 or 5). This gives you good channel separation, just like using separate triodes ought to. Then when the valves are beginning to wear out and lose emission, swap them over - each channel then gets an unused triode.

Not my idea; I saw someone suggest it on here a while ago.

On the other hand, using a single valve might not give too much crosstalk. I use a single ECC81/12AT7, but that has the advantage of much smaller anode impedance so less crosstalk for a given amount of stray capacitance.

 

[DF96]

People often say that parallel triodes cause distortion. They never explain why. In most cases the difference between two triodes in the same envelope will be smaller than the differences between 'near grid supports' and 'far from grid supports' in the same triode.

 

[6A3sUMMER]

One of the definitive works written about paralleling triodes was the cover issue of Glass Audio, Volume 12, Number 5, 2000. The research with over 600 measurements and 20 dual triodes was done with a professional spectrum analyzer, professional scope, and professional meter. The article was "Paralleling Tubes Effects", written by Matt Kamna, Dave McDonald, and Ken Boehlke. Please be sure to read ALL of the article.
It started as a lark: They thought the measured distortion of a parallel pair should be greater than just a single triode; and that it could be easily measured to prove that premise.
One of the most important parts of the research was the discovery that "matched" triodes are not matched. Each "matched" triode section must have its own individual bias set (or individual self bias). Failure to do this can cause the distortion to be bad. That
probably explains why so many attempts at paralleling triodes do not to sound as good as a single triode. But with proper individual bias, the distortion does have optimal results.
Some thought he measurements were not enough for proof, so several venues were used
for listening tests. When that did not prove that paralleling was bad, the next step was to
use math to prove it good or bad.
Conclusion: Paralleling Triodes Is good, when properly implemented. That was the Opposite of what the 3 authors were originally trying to prove. But they could see
the results, and had to accept it.
When Scientists Michelson and Morley were doing experiments to prove that Ether Did exist in space, they ended up proving the Opposite, it did NOT exist.

 

[DF96]

It was always a mystery to me why people thought that paralleling triodes was bad. Nice to hear that, provided they are reasonably matched, it does no harm.

 

[Sheldon]

Looking at the designs, it looks like there would be better channel separation using #1. I see the advantage in terms of cost going with #2 but am wondering if this design worsens the sound quality (all else being equal)

In my option, channel separation is overrated. For example, in vinyl the best you can do from the cartridge is about 30dB -usually less. I haven't tested the AX7 varieties, but I have tested SN7's. If I recall correctly, the separation was well over 50dB. In an ordinary room, the reflections would swamp the ability to detect the difference between perfect separation and something well under 50dB. I have some old recordings, when stereo was new and the channel separation was intentionally wide in order to emphasize the effect. They sound to me like a right speaker and a left speaker - not a well integrated stereo image.

Sheldon

http://www.diyaudio.com/forums/tube...-triodes-dual-triode-tubes-5.html#post4047234

 

[Sheldon]

In pictures: http://www.diyaudio.com/forums/tube...-triodes-dual-triode-tubes-6.html#post4175795

 

[6A3sUMMER]

Sheldon said:
"It was always a mystery to me why people thought that paralleling triodes was bad.
Nice to hear that, provided they are reasonably matched, it does no harm."

Be careful about what "is reasonably matched". A pair of Linear 6SN7 triodes
was 'matched' to within 12% (maybe not very well matched).
Then those same 2 triodes were paralleled with the same
cathode resistor (not individually biased).
After paralleling, the current in the 2 tubes was now Very Badly Miss-matched,
with one section hogging 2.14 times the current of the other section!
Wow!
And now, the higher harmonics were present, when they were not present
on the individual tubes.


It is recommended that even with matched triodes, that individual bias
be used for each triode.

 

[DF96]

That was me, not Sheldon.

I am puzzled how a '12% mismatch' could result in a 2:1 current ratio. 2:1 current ratio at the same bias is unmatched, in my book. That would be expected to create problems when paralleled.

 

[6A3sUMMER]

DF96, I'm sorry for not giving the credit to you for your quote.

Each 6SN7 section had a 1 k Ohm self bias cathode resistor and bypass capacitor;
and a 25 k Ohm plate resistor.
The cathode to plate voltage was about 244 Volts. The cathode was at about 8V,
and there was about 8 mA current.

When the triodes were separated the currents were 8.24 mA and 7.36 mA (12%).

Now, lets talk about how the two triodes were combined.
The cathodes were joined together, now the common bias resistor was 500 Ohms.
The plates were joined together, now the common plate resistor was 12.5 k Ohms.
The 12% difference when separate, was now accentuated to about 33% and 66% current in the two triodes.
The result was a degradation of the harmonics.

With closer matched triodes, the harmonics were essentially the same.
The advantage is 2 times lower plate resistance, and 2 times more power (to a load of 1/2 the original 25 k Ohms, without degrading the harmonic structure.
Just use separate bias, and get good results. This applies to power triodes as well.

 

[nigelwright7557]

I found using the 2 halves of a tube in series cancels hum.
The hum gets into both channels but as each half is out of phase the hums cancel.
Having said that if the layout is good you shouldn't get hum anyway.

I have just built a tube mixer for USB and a mic.
I rushed the layout and am now paying the price.
I had the mic input miles from the valve and it is picking up hum from a transformer and a 160VAC track close to it.
I also didn't put the grid stoppers close to the valve.
I also forgot to add a copper pour to each side of the layout.

So on to version 2 which should be much better.

 

[DF96]

Each 6SN7 section had a 1 k Ohm self bias cathode resistor and bypass capacitor;
and a 25 k Ohm plate resistor.
The cathode to plate voltage was about 244 Volts. The cathode was at about 8V,
and there was about 8 mA current.

When the triodes were separated the currents were 8.24 mA and 7.36 mA (12%).

I would not call that 12% matching, as they had different bias: one had 8.24V bias and 8.24mA current; the other had 7.36V bias and 7.36mA current. There would also have been a difference in anode voltage. As this sort of self-adjustment tends to halve difference, I would say that they were at least 25% different.

Given that they were poorly matched, poor performance when paralleled is unsurprising.

 

[6A3sUMMER]

I would not call that 12% matching, as they had different bias: one had 8.24V bias and 8.24mA current; the other had 7.36V bias and 7.36mA current. There would also have been a difference in anode voltage. As this sort of self-adjustment tends to halve difference, I would say that they were at least 25% different.

Given that they were poorly matched, poor performance when paralleled is unsurprising.

DF96, That is a good point that I overlooked.
It is at least 24% mis matched.
The triode sections should be much more tightly matched, before paralleling.

Of course there were some very closely matched triodes used in the research, and some very badly mis matched triodes.
There were 20 new old stock 6SN7s, and 10 used 6SN7s. At 2 sections per
tubes, that is 60 triodes.

But even with closer matching, separate biasing is suggested for each triode section.

There are other items to consider when paralleling triodes that were listed in the Glass Audio article. Capacitance, drive requirements, load requirements, power supply, etc.
An output transformer primary could be able to have 29.3 % less turns (1/2 inductance and 1/2 impedance), but the current would be 2 times that of a single tube. And the core would have to handle 2 times the power.

One more thing to consider for those who doubt that parallel triodes sound good (when properly implemented).
Just pick your favorite triode. Perhaps a 45, 2A3, 300B, 6BX7, 6CK4, etc.
Is the spacing equal over all parts of the filament (or cathode) to the grid?
Is the spacing equal over all parts of the grid to the plate?
If not, you have a mis matched "set" of triode(s) in that single tube.
Can it sound any good, of course it can.

Some radial triodes have a cathode that is not exactly circular, instead it
has ripples in the circle that match the bird cage grid. The most outside
ripple points match up with the spaces between the bird cage wires.

Eimac produced a 1500W radial triode for linear class A service.
The cathode was radial, the "bird cage" grid was radial, and the plate was
radial. I do not know if it used the rippled cathode or not. But it was a
very linear triode.

The 416 planar lighthouse triode had a flat cathode (round but flat), a flat grid
(round but flat), and a flat plate (round but flat). The closest distance between cathode to grid were all equal; and the closest distance between grid and plate were all equal. What could be a more equal "set" of matched triodes?

 

[DF96]

But even with closer matching, separate biasing is suggested for each triode section.

Yes. Maybe a compromise: separate cathode resistors, but common anode resistor.

One more thing to consider for those who doubt that parallel triodes sound good (when properly implemented).
Just pick your favorite triode. Perhaps a 45, 2A3, 300B, 6BX7, 6CK4, etc.
Is the spacing equal over all parts of the filament (or cathode) to the grid?
Is the spacing equal over all parts of the grid to the plate?
If not, you have a mis matched "set" of triode(s) in that single tube.
Can it sound any good, of course it can.

Yes. This was my basic objection to those who said that paralleled valves cannot be good: all valves are already 'paralleled' internally anyway. Of course, some are better than others; the ideal is to have all electrodes exactly the same shape (e.g. perfect planar structure, or perfect concentric cylinders) with no grid support rods anywhere near the electron streams. Small outward deviations in anode shape can be tolerated as the electric field inside will tend to smooth these out; such deviations are useful for mechanical rigidity.

I have been doing a bit more thinking about matching valves using the method you described: same resistors at cathode and anode, record the current (or, equivalently, the cathode voltage). My hunch is that the raw measured difference (12% in your case) needs to be multiplied by something like
1 + gm x Rk + Ra/(Rk x mu) - in many cases this will be approximately 3
to get the actual difference when biased with identical voltages (i.e. direct parallel connection).
This takes account of how the test circuit will be self-adjusting, but the simple paralleled circuit will not be.

 

[6A3sUMMER]

You can not only parallel small signal triodes, you can also parallel larger triodes.
Here is an example of one way to do it, and the details to keep in mind when you design
such a circuit.

300B parallel single ended
(one way it has been implemented)

Using the same voltage and current on each 300B versus single ended
Power supply 2 times the current, and 2 times the plate B+ filter capacitance

Parallel single ended has 2 times the output power of single ended

Separate self bias (or can use separate fixed bias)
(Separate self bias is recommended for single ended, parallel single ended,
and also for Push Pull)

Driver output tied to separate 300B Grid stopper resistors

The driver must drive 2 times the capacitance:
Single ended Grid to plate capacitance 15 pF x (Miller + 1) = ~ 60 pF;
Parallel single ended = ~ 120 pF
Driver voltage swing is the same for parallel single ended and for single ended

Output Transformer:
Single ended 3000 Ohms : 8 Ohms (turns 19.4 : 1)
Parallel single ended 1500 Ohms : 8 Ohms (turns (13.7 :1)
(3dB less transformer gain reduction; 3 dB more total amplifier gain, that is 2x
the power with the same voltage input to the driver)

Parallel single ended transformer with 2 times the current, 2 times the output power
requires larger primary wire size, and larger core size

Plates tied directly in parallel to the transformer primary

Damping factor is the same for single ended and parallel single ended
(if the transformer losses are equal, and the primary impedance is 1/2 for parallel single ended)

The parallel single ended distortion is the same for 2x the power of the single tube