My amp is push-pull 6B4G. I am currently using a shared cathode resistor. The output tubes are plate current matched using a wee bit of fixed bias and operate in class A using the classic 2A3 SE operating point. Is having a shared cathode resistor a poor idea in my situation?
I understand that you want to avoid current mismatches in the output transformer that results from running class AB or from mis-matched tubes. But as I address these is there any reason NOT to use a shared, unbypassed cathode resistor?
Rod
I understand that you want to avoid current mismatches in the output transformer that results from running class AB or from mis-matched tubes. But as I address these is there any reason NOT to use a shared, unbypassed cathode resistor?
Rod
In a perfect ideal "tube" world with "cloned" factory tubes......... Class A P-P will produce a net 0 AC current in the common cathode.... Since the AC signal currents are 180 degrees out of phase, you have perfect cancellation in the common cathode circuit.... there fore, there would be no need for a cathode by-pass cap, since there is no net signal to by-pass..... LOL, in reality there is a residual signal due to no perfect cancellations...
As long as you want to put "sense" resistors in the plate leads, so that you can measure the match of the quiescent current, with a DMM set to DC millivolts across the resistors. You can use 1 to 10 Ohms, but make sure the 2 resistors are matched for measurement accuracy. Then . . . Be willing to retest the DC current match every so often as the tubes age. You will need part of your bias be provided by a low voltage, with separate adjustable fixed grid voltage, unless the tubes are matched initially, and that they age together.
Or . . .
Using separate filament windings, and separate self bias resistors, and separate bypass caps, with reasonably but not perfectly matched tubes will usually intrinsically closely match the plate currents at the beginning, and as the tubes age.
Common cathode resistor plus some individually adjustable fixed bias just seems like too much circuitry and work for me. And . . . If the transconductance of the two tubes are not matched, then the common unbypassed 'cathode' resistor will tend to help balance the signals (reduce the 2nd harmonic that would result from an imperfect match). But that only works in the class A region . . . When the amp goes from class A to class AB, it will not work as well if the common cathode resistor is unbypassed. You will get more 3rd harmonic distortion in the AB region if you do not bypass the common cathode resistor, versus with the common cathode resistor bypassed. You may be surprised how quickly you actually transition from class A to class AB. (I know I was when I did some checking by measuring different push pull designs)
Your choice.
I use individual self bias, individual bypass caps, and start with reasonably matched tubes. The tube vendors I use do a good job of matching, sometimes for free, sometimes for a small charge . . . it is worth it.
Just my preferences.
Some tradeoffs of using individual self bias:
The self bias voltage reduces plate to cathode voltage (so just increase the B+)
You have to use separate filament windings for each DHT (parts count). But indirectly heated tubes do not need that.
You have to use individual self bias resistors (parts count).
You have to use individual self bias bypass capacitors (parts count) and . . . There is lots of discussion of using electrolytics there (I will let someone else argue about that, I just go ahead and use electrolytics).
I never designed and built a push pull 6A3 amp. But one of the amps I designed and built was a push pull 2A3 amp (pretty close cousin). It did have individual filament windings, individual self bias resistors with individual electrolytic bypass caps. Sounded OK to me.
Or . . .
Using separate filament windings, and separate self bias resistors, and separate bypass caps, with reasonably but not perfectly matched tubes will usually intrinsically closely match the plate currents at the beginning, and as the tubes age.
Common cathode resistor plus some individually adjustable fixed bias just seems like too much circuitry and work for me. And . . . If the transconductance of the two tubes are not matched, then the common unbypassed 'cathode' resistor will tend to help balance the signals (reduce the 2nd harmonic that would result from an imperfect match). But that only works in the class A region . . . When the amp goes from class A to class AB, it will not work as well if the common cathode resistor is unbypassed. You will get more 3rd harmonic distortion in the AB region if you do not bypass the common cathode resistor, versus with the common cathode resistor bypassed. You may be surprised how quickly you actually transition from class A to class AB. (I know I was when I did some checking by measuring different push pull designs)
Your choice.
I use individual self bias, individual bypass caps, and start with reasonably matched tubes. The tube vendors I use do a good job of matching, sometimes for free, sometimes for a small charge . . . it is worth it.
Just my preferences.
Some tradeoffs of using individual self bias:
The self bias voltage reduces plate to cathode voltage (so just increase the B+)
You have to use separate filament windings for each DHT (parts count). But indirectly heated tubes do not need that.
You have to use individual self bias resistors (parts count).
You have to use individual self bias bypass capacitors (parts count) and . . . There is lots of discussion of using electrolytics there (I will let someone else argue about that, I just go ahead and use electrolytics).
I never designed and built a push pull 6A3 amp. But one of the amps I designed and built was a push pull 2A3 amp (pretty close cousin). It did have individual filament windings, individual self bias resistors with individual electrolytic bypass caps. Sounded OK to me.
Unbypassed common cathode resistor will add 3rd order distortion. In some cases this could partially cancel existing 3rd; in other cases it will add to it. That is why some circuits have it and some do not. At the same time, it will reduce 2nd order distortion.
If you use a shared K resistor you need to bypass it because the degenerative FB signal created by the resistor on each cathode is then mixed and shared as crosstalk. That has to be removed by bypassing. In a PP channel the FB on one K is in opposite phase, being positive to the opposite tube. The distortion and noise component of the FB will be amplified or cancel the good FB affect on the opposite tube. And if the bypassing is not done on a 2-channel amp with 2 or 4 cathodes common, then you'll get channel cross FB, too. Gotta bypass common K resistors.
DF96,
Except for unusual (non-monotonic) curves of the push pull output stage tubes, third harmonic signals do not cancel that way. If you want the proof, do an example math exercise: look at a set of triode, pentode, or beam power tubes curves in push pull and apply the Taylor series math. The Even orders harmonics will cancel, but the Odd order harmonics will not.
Without the common cathode bypass cap, when the push pull stage gets a large enough signal that tries to bring the stage from class A to class AB, you will find that the plate current no longer follows the change in grid voltage, because with one tube off, now the other tube cathode is driving the common cathode resistor and so the cathode voltage starts following the grid voltage. That causes a fore-shortening of the plate signal of the only tube that is turned on. In the other direction, with the output tubes exchanging roles, the first tube turns off, and the other has a fore-shortened plate swing. That is the beginning of clipping. If you clip both the plate swings that way just as you leave class A and go to class AB, you will do what is defined as the very beginning of a square wave (all odd order harmonics).
You can use negative feedback to cancel 3rd harmonic distortion.
Thats all.
Except for unusual (non-monotonic) curves of the push pull output stage tubes, third harmonic signals do not cancel that way. If you want the proof, do an example math exercise: look at a set of triode, pentode, or beam power tubes curves in push pull and apply the Taylor series math. The Even orders harmonics will cancel, but the Odd order harmonics will not.
Without the common cathode bypass cap, when the push pull stage gets a large enough signal that tries to bring the stage from class A to class AB, you will find that the plate current no longer follows the change in grid voltage, because with one tube off, now the other tube cathode is driving the common cathode resistor and so the cathode voltage starts following the grid voltage. That causes a fore-shortening of the plate signal of the only tube that is turned on. In the other direction, with the output tubes exchanging roles, the first tube turns off, and the other has a fore-shortened plate swing. That is the beginning of clipping. If you clip both the plate swings that way just as you leave class A and go to class AB, you will do what is defined as the very beginning of a square wave (all odd order harmonics).
You can use negative feedback to cancel 3rd harmonic distortion.
Thats all.
kodabmx,
I believe there was both lower than normal voltage fixed adjustable bias, and common cathode bias in his amp.
Both fixed adjustable bias and self bias have their strengths and weaknesses (and tradeoffs).
As you can see in my post # 3, and also in post # 8, there are some reasons to do each method properly.
In my opinion, common cathode un-bypassed self bias for push pull is not a proper way.
It may help reduce 2nd harmonic distortion in some cases, but it will increase 3rd harmonic distortion.
When you play back dynamic music through a push pull amplifier, how do you know when you are in class A, and when you are in class AB?
I believe there was both lower than normal voltage fixed adjustable bias, and common cathode bias in his amp.
Both fixed adjustable bias and self bias have their strengths and weaknesses (and tradeoffs).
As you can see in my post # 3, and also in post # 8, there are some reasons to do each method properly.
In my opinion, common cathode un-bypassed self bias for push pull is not a proper way.
It may help reduce 2nd harmonic distortion in some cases, but it will increase 3rd harmonic distortion.
When you play back dynamic music through a push pull amplifier, how do you know when you are in class A, and when you are in class AB?
Last edited:
As usual DF96 has chosen his words carefully. Under some circumstances, an optimum value for a common unbypassed cathode resistor can substantially reduce third harmonic distortion. This is demonstrated by Morgan Jones (Western Electric Harmonic Equaliser pg 186 "Valve Amplifiers" fourth edition) where a 28 dB reduction in third harmonic is seen.
Mr Jones also indicates that a reduction isn't always seen and that the optimum resistor value is load dependent.
Mr Jones also indicates that a reduction isn't always seen and that the optimum resistor value is load dependent.
Monotonicity is not the issue, as we can assume that any competent output stage will be monotonic. The issue is the sign of the 3rd order term in the valve response. In most cases re-entrant distortion (for that is what it is) will add to this. However, if in some region of its characteristic a valve looks a bit like a variable-mu (remote cutoff) device then the 3rd could partly cancel.6A3sUMMER said:
I assumed it was obvious that if Class AB is possible then a common cathode resistor cannot be used, whether bypassed or not.
Some designs use a common unbypassed resistor, and claim that it reduces distortion. Of course, it may simply be that they are reducing 2nd by more than they are increasing 3rd - below a certain signal level this is more or less guaranteed by the algebra.
Well, that's simply not true. There are plenty of commercial AB amps using 1 resistor for 4 tubes, bypassed. This thread was requesting advice about a stereo SE with a common, un-bypassed, K resistor. So it will be Class A but have crosstalk channel FB on the K.
Using a common cathode resistor for Class AB is simply poor engineering. It guarantees poor bias control, and as a minimum means either that the quiescent bias must be much hotter than would be needed for good bias control or that the amplifier cannot be tested (or used) for full amplitude signals for more than a brief time. If an amplifier is almost Class A except for rare musical peaks then you can get away with it but it is still poor engineering. Cathode resistor bias for AB is bad; common cathode resistor bias is worse.
You can have 4 independant fixed bias setting controls to make all tubes have identical quiescent currents but you will have 4 bifferent bias voltages on the grids creating 4 different clipping and cuttoff points. Is that any better? A sufficiently bypassed K has stable bias voltage. No scheme is without its tradeoffs.
Last edited:
At least separate bias means you can get the bias right. Common bias means that it must be wrong, unless the valves are very well matched and stay matched as they age.
Anyway, you cannot avoid clipping but you can avoid common cathode resistors.
Anyway, you cannot avoid clipping but you can avoid common cathode resistors.
Only one would be right, the most conservatively set one. It's the first one who's limits cannot be be exceeded without clipping, limiting the others then.
Take a look at the Western Electric 275A data sheet at Frank's electron Tube Data sheets
The Western electric 275A was designed to be a 3 Watt Audio Output Triode. Presumably it is reasonably linear.
Look at the very bottom graph. You will see very sharp, and very deep nulls of the third harmonic distortion as the load impedance changes.
That tube "curve" is extremely non-monotonic.
These 4 curves have nulls of up to 40dB. They merely changed the grid bias volts, and plate load impedance.
Try and find a combination of output transformer and loudspeaker that has a flat enough impedance from 20Hz to 20kHz to take advantage of that null.
Resistor testing of vacuum tube amps only goes so far.
Just my observation.
The Western electric 275A was designed to be a 3 Watt Audio Output Triode. Presumably it is reasonably linear.
Look at the very bottom graph. You will see very sharp, and very deep nulls of the third harmonic distortion as the load impedance changes.
That tube "curve" is extremely non-monotonic.
These 4 curves have nulls of up to 40dB. They merely changed the grid bias volts, and plate load impedance.
Try and find a combination of output transformer and loudspeaker that has a flat enough impedance from 20Hz to 20kHz to take advantage of that null.
Resistor testing of vacuum tube amps only goes so far.
Just my observation.
Last edited:
How many of you have ever put 1 Ohm sense resistors in the cathodes of a push pull output stage . . . and then looked at the cathode currents on a 2 channel scope to see at what signal level you go from class A to class AB?
And yes I have.
And yes I have.
I use a current probe in the plate circuit...ie placed around the Output Transformer lead ....
Haven't done these types of measurements in many years... did these things back when I was young, eager and liked to experiment....LOL
- Status
- Not open for further replies.