dgta,
Some advantages of Paralleling triodes:
Lowers the plate resistance that drives the next stage
Doubles the current to drive the next stage.
In the case of an output stage, doubles the power out (and you can use your "favorite" tube, instead of having to select a completely different output tube type to get the extra power you want/need).
All topologies have tradeoffs. You mentioned some.
Miller Effect, individual self bias (or individually adjustable fixed bias).
Just like all other circuits, connecting more tube sections will increase the complexity,
wiring, and parts count.
If you have a line stage, and want to connect negative feedback to the cathode circuit, consider this example:
Single triode: a 1k self bias resistor (with bypass cap across it) connects from the cathode to a 100 Ohm negative feedback-node resistor to ground.
Suppose the negative feedback resistor is 10k that connects from the output transformer secondary, back to the feedback node junction of the 100 Ohm and 1k Ohm resistors.
To use Parallel Triodes, just use individual 1k self bias resistors with individual bypass caps across them.
Connect the bottoms of Both 1k resistor/bypass cap network to a 50 Ohm resistor, and the other end of the 50 Ohm resistor to ground.
That 3 resistor junction becomes the new negative feedback node.
Connect a 5k resistor from the output transformer secondary to the '50, 1k, 1k negative feedback node'.
That parallel triode stage has 1/2 the output impedance of a single triode to drive the next stage.
That parallel triode stage has 2x the drive current of a single triode to drive the next stage.
I hope that gives you an idea of how to create a practical parallel triode stage.
Remember, if you do not need 1/2 the output impedance, if you do not need 2x the drive current, if you do not need 2x the output power, then . . . Just use one triode.
Do not forget to use Separate grid stoppers, do not directly parallel the grids (the plates are paralleled).
Do not parallel triodes without one or more of those reasons above.
Keep it as simple as possible.
Some advantages of Paralleling triodes:
Lowers the plate resistance that drives the next stage
Doubles the current to drive the next stage.
In the case of an output stage, doubles the power out (and you can use your "favorite" tube, instead of having to select a completely different output tube type to get the extra power you want/need).
All topologies have tradeoffs. You mentioned some.
Miller Effect, individual self bias (or individually adjustable fixed bias).
Just like all other circuits, connecting more tube sections will increase the complexity,
wiring, and parts count.
If you have a line stage, and want to connect negative feedback to the cathode circuit, consider this example:
Single triode: a 1k self bias resistor (with bypass cap across it) connects from the cathode to a 100 Ohm negative feedback-node resistor to ground.
Suppose the negative feedback resistor is 10k that connects from the output transformer secondary, back to the feedback node junction of the 100 Ohm and 1k Ohm resistors.
To use Parallel Triodes, just use individual 1k self bias resistors with individual bypass caps across them.
Connect the bottoms of Both 1k resistor/bypass cap network to a 50 Ohm resistor, and the other end of the 50 Ohm resistor to ground.
That 3 resistor junction becomes the new negative feedback node.
Connect a 5k resistor from the output transformer secondary to the '50, 1k, 1k negative feedback node'.
That parallel triode stage has 1/2 the output impedance of a single triode to drive the next stage.
That parallel triode stage has 2x the drive current of a single triode to drive the next stage.
I hope that gives you an idea of how to create a practical parallel triode stage.
Remember, if you do not need 1/2 the output impedance, if you do not need 2x the drive current, if you do not need 2x the output power, then . . . Just use one triode.
Do not forget to use Separate grid stoppers, do not directly parallel the grids (the plates are paralleled).
Do not parallel triodes without one or more of those reasons above.
Keep it as simple as possible.
Last edited:
An important point of the "Paralleling Tubes Effects" article was the importance of using individual bias for each triode.
The spectral measurements were of up to the 7th harmonic, and as low as 90dBc.
Even moderately mismatched tubes had the worst performance in this regard.
Using a common bias resistor of 1/2 value when paralleling, had unpredictable results that were lower performance than a single un-paralleled triode.
The conclusion was to use individual separate bias resistors.
Your mileage may vary.
The spectral measurements were of up to the 7th harmonic, and as low as 90dBc.
Even moderately mismatched tubes had the worst performance in this regard.
Using a common bias resistor of 1/2 value when paralleling, had unpredictable results that were lower performance than a single un-paralleled triode.
The conclusion was to use individual separate bias resistors.
Your mileage may vary.
Thanks, that is the answer I was looking for.
I have seen several commercial and quasi-commercial designs with paralleled tubes that didn't make a lot of sense to me. In an EICO 87 topology, for example, the load on the first stage is a dc coupled LTP so you don't need Zout, gm or current.
Yes, but was this better than using a single tube? Lower distortion? Noise? Any measurements on that?
I mean, aside from the gm/Zout issue.
dgta,
The advantage of Parallel tubes was not only to lower the distortion.
But actually, it often does.
Consider a parallel tube that drives a difficult load impedance, versus a single triode of the same type driving that same difficult load.
The parallel rp of two triodes is lower than the rp a single triode.
That means the difficult load is shared by two triodes instead of one.
The effective slope of the load line for each tube is less for the parallel triodes.
The distortion is lower both for a resistive load, and for an elliptical load, when the parallel tube is properly implemented.
That can apply to a line stage, and can apply to an output stage.
What that also means, is for a pair of triodes in an output stage, you can get the same distortion at 2x the power out, versus using a single triode.
one 300B X Watts.
two 300B 2X Watts
All at the same distortion.
You do have to choose, 2x power, or less distortion; or a little more power and a little less distortion.
The advantage of Parallel tubes was not only to lower the distortion.
But actually, it often does.
Consider a parallel tube that drives a difficult load impedance, versus a single triode of the same type driving that same difficult load.
The parallel rp of two triodes is lower than the rp a single triode.
That means the difficult load is shared by two triodes instead of one.
The effective slope of the load line for each tube is less for the parallel triodes.
The distortion is lower both for a resistive load, and for an elliptical load, when the parallel tube is properly implemented.
That can apply to a line stage, and can apply to an output stage.
What that also means, is for a pair of triodes in an output stage, you can get the same distortion at 2x the power out, versus using a single triode.
one 300B X Watts.
two 300B 2X Watts
All at the same distortion.
You do have to choose, 2x power, or less distortion; or a little more power and a little less distortion.
Yes, I understand all that. It is all related to loading and gm/Zout.
What I'm trying to understand is if/how you can get lower distortion in a stage where that is not a factor, i.e. load impedance is very large and non-capacitive. In order to get lower distortion in a parallel setup you would need some partial cancellation of distortion components. With noise it's possible, as the noise in the 2 tubes is uncorrelated. Distortion, on the other hand, is the same in the 2 tubes and there is no cancellation.
What I'm trying to understand is if/how you can get lower distortion in a stage where that is not a factor, i.e. load impedance is very large and non-capacitive. In order to get lower distortion in a parallel setup you would need some partial cancellation of distortion components. With noise it's possible, as the noise in the 2 tubes is uncorrelated. Distortion, on the other hand, is the same in the 2 tubes and there is no cancellation.
That’s a rabbit hole- you’re now into signal convolution, and subsequent deconvolution. Phase networks and all other fun. Easier in digital than in tube analogue.
dgta,
Parallel tubes never were intended to Reduce the harmonic distortion.
Of course if you get 2x the power, and only use 1x or 1.5x of that new larger power, the distortion is reduced, just like when the single tube is at 0.5 or 0.75 of its 1x power, respectively.
The original Glass Audio article already covered that.
Part of the proofs was the calculations using child's law, and calculating according to the Taylor series.
So, the calculations showed that:
Individual even order harmonics simply adds to become the average of the 2 tubes.
For example, upper order even harmonics of -70dBc one tube, and -72dBc the other tube, becomes (approximately) -71dBc for the two tubes.
You want it closer than that, exponentiate both dB, add, and re-log.
Parallel tubes is not intended to cause cancellation of even harmonics, and is not intended to cause cancellation of odd harmonics.
Part of the measurements showed that without using individual bias for each triode, the odd order harmonic distortion could increase.
That sounds like a real good reason to use individual bias for each triode (unless you want more odd order distortion, see).
There are no miracles in electronics. It follows the math.
Push Pull is not a miracle either, it simply follows the math (but in that case, the 2nd harmonic is partially cancelled).
Many people have a favorite tube. But They want 2x the power of a single tube.
Some want to do that with push pull (and in some cases can get even more than 2x the power).
But some want 2x the power in single ended, with their favorite tube; that means Parallel single ended.
To each his own preference.
I hope that clears it up for everyone.
PS I am tired of seeing that statement that adding 2 random noise powers cancels.
Correct me if I am wrong, but if you add 2 Gaussian noise powers (like from 2 tubes), the total noise power increases, versus that of one tube.
That is one of the laws of conservation of power, and Thermal Dynamics.
Yes?
No?
Parallel tubes never were intended to Reduce the harmonic distortion.
Of course if you get 2x the power, and only use 1x or 1.5x of that new larger power, the distortion is reduced, just like when the single tube is at 0.5 or 0.75 of its 1x power, respectively.
The original Glass Audio article already covered that.
Part of the proofs was the calculations using child's law, and calculating according to the Taylor series.
So, the calculations showed that:
Individual even order harmonics simply adds to become the average of the 2 tubes.
For example, upper order even harmonics of -70dBc one tube, and -72dBc the other tube, becomes (approximately) -71dBc for the two tubes.
You want it closer than that, exponentiate both dB, add, and re-log.
Parallel tubes is not intended to cause cancellation of even harmonics, and is not intended to cause cancellation of odd harmonics.
Part of the measurements showed that without using individual bias for each triode, the odd order harmonic distortion could increase.
That sounds like a real good reason to use individual bias for each triode (unless you want more odd order distortion, see).
There are no miracles in electronics. It follows the math.
Push Pull is not a miracle either, it simply follows the math (but in that case, the 2nd harmonic is partially cancelled).
Many people have a favorite tube. But They want 2x the power of a single tube.
Some want to do that with push pull (and in some cases can get even more than 2x the power).
But some want 2x the power in single ended, with their favorite tube; that means Parallel single ended.
To each his own preference.
I hope that clears it up for everyone.
PS I am tired of seeing that statement that adding 2 random noise powers cancels.
Correct me if I am wrong, but if you add 2 Gaussian noise powers (like from 2 tubes), the total noise power increases, versus that of one tube.
That is one of the laws of conservation of power, and Thermal Dynamics.
Yes?
No?
Last edited:
Yes and no. The correlated signal power goes up by 6 dB while the uncorrelated (noise) goes up by 3 dB. So signal to noise improved by 3db. This is true as long as the total bias current is 2x that of the single tube case. For thermal noise doubling the current in a single tube would have the same result if the tube works correctly at that current level. 1/f noise behaves a bit differently. Next chapter if I get in the mood.
Total noise power across the spectrum.. Yes.
Common signals then reinforce (signal, distortion and noise). Their power is specific to those frequencies and not spread thinly.
The average noise vs signal drops - so SNR goes up (assuming average noise is taken)
The noise doesn’t magically cancel.
You can renormalise but the noise is still there.
Last edited:
Thank you. Yes, that was my line of thinking. If the 2 tubes in one bottle are significantly different, the best you can do with paralleling is to average, not improve beyond the best tube. So maybe a good strategy for tubes known to have poor manufacturing variability. For tubes nearly the same, the advantages of paralleling seem minimal at best.
As to the noise, I believe Potentiallyincorrect is Potentiallycorrect. S/N goes up 3dB every time you double the number of tubes. But that's another discussion altogether. I'm not worried about noise in a line stage.
Use one 300B to get 8 Watts under a standard WE operating condition.
Then double the power supply current capability, and double the driver tube current capability (to drive two 300B grids capacitive loading).
Now, parallel two 300B tubes, but use the same standard WE operating condition for each tube.
Then use a different output transformer with 2 times the laminations, and 0.5 the primary impedance, and 0.5 the DCR of the original transformer.
The power out will be 2 x 8 watts = 16 Watts.
That is +3dB versus a single tube.
It is not +6dB versus a single tube.
The distortion will be the same.
The damping factor will be the same.
And, you get a single ended parallel 300B amp with 2 times the power, which was the original intent, to double the power while using your favorite tube (300B for example) in single ended, not push pull.
Similarly, take a pair (2) of your favorite tubes and build a push pull amplifier.
Now parallel the tubes (4 total), make the necessary similar changes, and get 2 times the power (still using your favorite tubes).
When paralleling triodes, the signal power goes up by +3dB, and noise power goes up by +3dB, Well . . .
Perhaps that is why there are not very many phono preamps that use 2 Parallel 12AX7 triodes for the input stage.
Then double the power supply current capability, and double the driver tube current capability (to drive two 300B grids capacitive loading).
Now, parallel two 300B tubes, but use the same standard WE operating condition for each tube.
Then use a different output transformer with 2 times the laminations, and 0.5 the primary impedance, and 0.5 the DCR of the original transformer.
The power out will be 2 x 8 watts = 16 Watts.
That is +3dB versus a single tube.
It is not +6dB versus a single tube.
The distortion will be the same.
The damping factor will be the same.
And, you get a single ended parallel 300B amp with 2 times the power, which was the original intent, to double the power while using your favorite tube (300B for example) in single ended, not push pull.
Similarly, take a pair (2) of your favorite tubes and build a push pull amplifier.
Now parallel the tubes (4 total), make the necessary similar changes, and get 2 times the power (still using your favorite tubes).
When paralleling triodes, the signal power goes up by +3dB, and noise power goes up by +3dB, Well . . .
Perhaps that is why there are not very many phono preamps that use 2 Parallel 12AX7 triodes for the input stage.
Last edited:
IMHO you should reconsider this one more time.6A3sUMMER said:
Many phono preamp use paralleled tubes (or FETs) in first stage to improve S/N.
I suspect there's a lot of structured noise (ie non-gaussian). Interesting experimental observation.
Could you explain the reasoning behind this?
With two triode ra (or rp) in parallel the averaged or composite ra is half, and I would expect the damping factor to be doubled, all else equal.
What about a less common situation : a double triode with a common cathode (like,say,the 6N7),used as a CF driver for 2A3 or 300B?
The cathode itself is internally tied together on both triode sections so nothing to discuss here.
What about the grids and anodes?Also wired in parallel (on the sockets) or using individual anode resistors as well as grid stoppers?
The cathode itself is internally tied together on both triode sections so nothing to discuss here.
What about the grids and anodes?Also wired in parallel (on the sockets) or using individual anode resistors as well as grid stoppers?
The grids in parallel, with lower input impedance to whatever is driving it. Capacitance add, so Miller becomes doubled, compounding the situation.
mondogenerator,
One 300B, rp, 700 Ohms; RL 2800 Ohms lossless output transformer, damping factor = 4
Two Parallel 300B, rp 350 Ohms parallel; RL 1400 Ohms lossless output transformer, damping factor = 4.
Twice the power as the single 300B; equal damping factors; equal distortion @ 2X power; versus the distortion of the single 300B @ 1X power.
The above characteristics, and given the designer loves the 300B more than any other tube, are the only reasons to build the parallel 300B single ended amp.
Pick your favorite (non 300B) output tube, calculate the necessary circuit for one output tube, and then parallel two output tubes. Adjust the other circuit parameters to work with the parallel configuration (B+ current, output transformer characteristics, double driver drive current capability and output impedance, etc.).
I already said this 2 or 3 times in this thread.
Can I make it any clearer?
One 300B, rp, 700 Ohms; RL 2800 Ohms lossless output transformer, damping factor = 4
Two Parallel 300B, rp 350 Ohms parallel; RL 1400 Ohms lossless output transformer, damping factor = 4.
Twice the power as the single 300B; equal damping factors; equal distortion @ 2X power; versus the distortion of the single 300B @ 1X power.
The above characteristics, and given the designer loves the 300B more than any other tube, are the only reasons to build the parallel 300B single ended amp.
Pick your favorite (non 300B) output tube, calculate the necessary circuit for one output tube, and then parallel two output tubes. Adjust the other circuit parameters to work with the parallel configuration (B+ current, output transformer characteristics, double driver drive current capability and output impedance, etc.).
I already said this 2 or 3 times in this thread.
Can I make it any clearer?
This is true of OTL amps. With an OPT, the assumption is that when going from 1 to 2 tubes you change the OPT ratio to match. So from the speaker's perspective looking back into the amp, Zout has not changed.