For the sake of this discussion, let's assume a conventional Class-AB P-P stereo amplifier chassis. Each channel has a pair of EL84 in the opt section. For self-bias, we can use 1x 120R for both channels (4x EL84 total) or 2x 240R, one for each channel (2x EL84 per channel, 4x total).
In the latter case, with separate 240R resistors, cathode bypass would seem to be optional, with the standard trade-off between higher gain (bypass) or local degenerative FB (no bypass). Each has its fans and foes.
But in the former case, when both channels are sharing a single 120R cathode resistor, it would seem that bypass is mandatory. Not due to a need of gain, but due to the fact that lacking AC bypass there is going to be degenerative FB interaction between the two channels. I can't see any result of this that doesn't equate to some level of harmonic or phase distortion.
Is this correct? Is a robust cathode bypass cap mandatory when two (or more..) channels share a single cathode circuit?
In the latter case, with separate 240R resistors, cathode bypass would seem to be optional, with the standard trade-off between higher gain (bypass) or local degenerative FB (no bypass). Each has its fans and foes.
But in the former case, when both channels are sharing a single 120R cathode resistor, it would seem that bypass is mandatory. Not due to a need of gain, but due to the fact that lacking AC bypass there is going to be degenerative FB interaction between the two channels. I can't see any result of this that doesn't equate to some level of harmonic or phase distortion.
Is this correct? Is a robust cathode bypass cap mandatory when two (or more..) channels share a single cathode circuit?
Push pull:
Any AC imbalance between the output tubes of one channel will bleed over into the other channel. It may be very small, but it will be there. That means you might hear a signal from one channel that is bleeding over into the other channel (if there is no signal applied to the other channel). But that will also be dependent on how well AC balanced the other channel is. AC balance is not always the same at large signal levels as it is at small signal levels.
Single Ended:
It is a different case when the 2 channels are single ended, rather than push pull.
A 2 channel single ended 45 amp used AC filaments, with a pair of 25 Ohm resistors that were connected to the parallel 45s filaments. Those resistors were there to act as a virtual center of the AC filament voltage, to reduce the hum in the output. The other ends of the 25 Ohm resistors were connected to the common bias resistor and common bypass capacitor.
For the signal current, that was 12.5 Ohms which caused a voltage drop.
The resultant channel separation was -40 dB over the whole frequency range of the amplifier, caused entirely by the signal current that flowed through those resistors.
It actually did not sound bad (i.e. very few phono cartridges are ever capable of 40 dB separation at one frequency, and certainly not over the whole range of audio frequencies (no matter how carefully the cartridge and tone arm are adjusted).
Any AC imbalance between the output tubes of one channel will bleed over into the other channel. It may be very small, but it will be there. That means you might hear a signal from one channel that is bleeding over into the other channel (if there is no signal applied to the other channel). But that will also be dependent on how well AC balanced the other channel is. AC balance is not always the same at large signal levels as it is at small signal levels.
Single Ended:
It is a different case when the 2 channels are single ended, rather than push pull.
A 2 channel single ended 45 amp used AC filaments, with a pair of 25 Ohm resistors that were connected to the parallel 45s filaments. Those resistors were there to act as a virtual center of the AC filament voltage, to reduce the hum in the output. The other ends of the 25 Ohm resistors were connected to the common bias resistor and common bypass capacitor.
For the signal current, that was 12.5 Ohms which caused a voltage drop.
The resultant channel separation was -40 dB over the whole frequency range of the amplifier, caused entirely by the signal current that flowed through those resistors.
It actually did not sound bad (i.e. very few phono cartridges are ever capable of 40 dB separation at one frequency, and certainly not over the whole range of audio frequencies (no matter how carefully the cartridge and tone arm are adjusted).
One more thing about sharing bias resistors. It is not easy to get DC balance with a common cathode bias resistor. That means that the output transformer DC currents in the primary are not balanced (not a good thing).
The disadvantage of separate cathode bias resistors is that you need more resistors, and more you now must have bypass caps (and not just 1 common cap).
The disadvantage of separate cathode bias resistors is that you need more resistors, and more you now must have bypass caps (and not just 1 common cap).
In class AB yes, you need a bypass cap if the resistor is shared. In a class A amp it's not essential.
https://www.aikenamps.com/images/Documents/cathcap.pdf
I'd strongly recommend a cathode resistor/electrolytic combination for each valve. Though almost any datasheet depicts common ones, we have to consider that in those days tube manufacturing standards were higher and tolerances closer then those we observe from today's tubes. This demands for separate biasing.
Anyway, 240 ohms per EL84 pair in class AB seems a bit high. Those datasheets call for 130 ohms, so 270 ohms per tube should be appropriate.
Best regards!
Anyway, 240 ohms per EL84 pair in class AB seems a bit high. Those datasheets call for 130 ohms, so 270 ohms per tube should be appropriate.
Best regards!
Note that the effect of omitting the bypass for a PP output is rather different from SE. For SE you get degenerative feedback, which reduces gain and increases output impedance - in most cases both are unwanted effects, although you will get some distortion reduction too.legendre said:
For PP you get little change in gain or output impedance; what you get is re-entrant distortion, but in some cases this can be of opposite polarity to the intrinsic distortion of the valve so can reduce distortion a little. I would not describe this situation as degenerative feedback.
Mandatory is too strong a term in this application. The circuit will function just fine without it. Been there, done that. No data, just ears. And I can show you a type of PP output circuit where a bypass is actually forbidden.
This is just one aspect of attention to details:
It is important to balance the DC currents in each 1/2 of the primary windings of push pull output transformers. Those transformers do not have air gaps in the laminations.
Do I need to say why you need balanced currents?
And, after you got the currents matched, how long has it been since you checked them?
After a week of burn-in, after 6 months, etc.
Tubes do not always age the same.
Matched tubes are usually only matched for 3 quantities: one plate voltage, and one bias voltage, and one current. Regardless of which 2 of these quantities is set, the 3rd quantity is measured, and must match.
Typically, other combinations of plate voltage, bias voltage, and current are not tested.
Sometimes, the transconductance is also tested (but usually only at those 3 original quantities of voltages and current).
Now, you are going to take these “matched tubes” and put them in your amplifier that uses different plate voltage, different bias voltage, and different current.
Methods to match currents in your push pull amplifier is to use either individual cathode self bias, or individually adjustable fixed bias.
Preferably, even with these circuits, you do Start with matched tubes (at least they did match somewhere on the curves, instead of nowhere on the curves).
I do not use regulated power supplies. My power line voltage varies from a low 118V to 123V (+/- 2%). So, the B+ and filaments also vary +/- 2.5%.
I use individual cathode self bias, and I check them at first, after burn-in, at different power line voltages, and after a few months.
I am satisfied if they match within about 500uA, most of them match within 300uA.
I do Not change the individual cathode self bias resistors to "make them match".
It is important to balance the DC currents in each 1/2 of the primary windings of push pull output transformers. Those transformers do not have air gaps in the laminations.
Do I need to say why you need balanced currents?
And, after you got the currents matched, how long has it been since you checked them?
After a week of burn-in, after 6 months, etc.
Tubes do not always age the same.
Matched tubes are usually only matched for 3 quantities: one plate voltage, and one bias voltage, and one current. Regardless of which 2 of these quantities is set, the 3rd quantity is measured, and must match.
Typically, other combinations of plate voltage, bias voltage, and current are not tested.
Sometimes, the transconductance is also tested (but usually only at those 3 original quantities of voltages and current).
Now, you are going to take these “matched tubes” and put them in your amplifier that uses different plate voltage, different bias voltage, and different current.
Methods to match currents in your push pull amplifier is to use either individual cathode self bias, or individually adjustable fixed bias.
Preferably, even with these circuits, you do Start with matched tubes (at least they did match somewhere on the curves, instead of nowhere on the curves).
I do not use regulated power supplies. My power line voltage varies from a low 118V to 123V (+/- 2%). So, the B+ and filaments also vary +/- 2.5%.
I use individual cathode self bias, and I check them at first, after burn-in, at different power line voltages, and after a few months.
I am satisfied if they match within about 500uA, most of them match within 300uA.
I do Not change the individual cathode self bias resistors to "make them match".
Yes ....
This be a problem, particularly with spread in tube characteristics found on the market these days.
One thing not mentioned is that with separate bypassed bias resistors an extra time constant (pole) is added to the phase picture, which might affect the NFB stability if not checked (and it often does).
If I may I would like to add another factor I learned a long time ago, concerning the matter of music output power (no, NOT PMPO ... rubbish or whatever). It has been shown that using a large cathode bypass capacitor (in addition to relatively high value Cs in the power supply) will simulate fixed bias conditions for most cases of music reproduction owing to the low mark-space ratio of such. I found a definite improvement in performance with say 6,800µF cathode bypass in such cases. Caps being physically comparatively small these days, this is quite practical.
I do sometimes 'sin' by using a common cathode resistor, but then mostly with a bias equalisung feature (pot) somewhere. (Separate cathode resistors alleviate the dc-in-OPT problem, it does not fix it.)
This be a problem, particularly with spread in tube characteristics found on the market these days.
One thing not mentioned is that with separate bypassed bias resistors an extra time constant (pole) is added to the phase picture, which might affect the NFB stability if not checked (and it often does).
If I may I would like to add another factor I learned a long time ago, concerning the matter of music output power (no, NOT PMPO ... rubbish or whatever). It has been shown that using a large cathode bypass capacitor (in addition to relatively high value Cs in the power supply) will simulate fixed bias conditions for most cases of music reproduction owing to the low mark-space ratio of such. I found a definite improvement in performance with say 6,800µF cathode bypass in such cases. Caps being physically comparatively small these days, this is quite practical.
I do sometimes 'sin' by using a common cathode resistor, but then mostly with a bias equalisung feature (pot) somewhere. (Separate cathode resistors alleviate the dc-in-OPT problem, it does not fix it.)
Take a tube that has a plate resistance of 1k Ohms, and a 3k plate load. The power supply B+ looking into that 'sees' an impedance of 4K Ohms. We need a power supply impedance to be much lower than 4k Ohms.
Now take that same tube that has a transconductance of 5000uMhos. That is 200 Ohms.
And say the cathode self bias resistor is 400 Ohms.
Ground looking up into an un-bypassed self bias resistor to the cathode 'sees' 600 Ohms.
Ground looking up into a properly bypassed self bias resistor 'sees' 200 Ohms.
You need to Bypass the self bias resistor, but you also need to bypass 1/transconductance of the tube.
Failing to properly bypass the self bias resistor and the cathode impedance allows there to be a voltage that degenerates the stage gain, and raises the stage output impedance.
Remember, this is a Common Cathode circuit, common to both input and output signal.
Because the cathode circuit impedance is lower than the plate circuit impedance, it stands to reason that the bypass capacitor in the cathode circuit needs to have much more capacitance, than the filter cap for the plate B+.
I will have to try a little more capacitance in the cathode circuit, i.e. the 6800 uF suggested above (the most I have used is 2000uF).
I use tubes that are at least balanced at one set of plate voltage, bias voltage, and current. I may be using a different set of plate voltage, bias resistor (to create bias voltage), and different quiescent current, than the they do when they match the tubes (I do not know what setup they use), but it is close enough that with self bias, the currents are very well matched in my amplifiers (300uA out of 2x 60mA, is that close enough?
Now take that same tube that has a transconductance of 5000uMhos. That is 200 Ohms.
And say the cathode self bias resistor is 400 Ohms.
Ground looking up into an un-bypassed self bias resistor to the cathode 'sees' 600 Ohms.
Ground looking up into a properly bypassed self bias resistor 'sees' 200 Ohms.
You need to Bypass the self bias resistor, but you also need to bypass 1/transconductance of the tube.
Failing to properly bypass the self bias resistor and the cathode impedance allows there to be a voltage that degenerates the stage gain, and raises the stage output impedance.
Remember, this is a Common Cathode circuit, common to both input and output signal.
Because the cathode circuit impedance is lower than the plate circuit impedance, it stands to reason that the bypass capacitor in the cathode circuit needs to have much more capacitance, than the filter cap for the plate B+.
I will have to try a little more capacitance in the cathode circuit, i.e. the 6800 uF suggested above (the most I have used is 2000uF).
I use tubes that are at least balanced at one set of plate voltage, bias voltage, and current. I may be using a different set of plate voltage, bias resistor (to create bias voltage), and different quiescent current, than the they do when they match the tubes (I do not know what setup they use), but it is close enough that with self bias, the currents are very well matched in my amplifiers (300uA out of 2x 60mA, is that close enough?
Not quite true. You also have to add 'anode load'/mu to these resistances. This will be negligible for a pentode, but could be significant for triode or UL.6A3sUMMER said:
DF96,
You are correct.
I did not account for the factor you mentioned:
plate load impedance/mu of the tube.
(You made me think again)
I usually use triodes or triode wired pentodes / beam power tubes.
I most often use Individual self bias resistors, with bypass caps on each.
One system has the following characteristics:
Push pull, 6000 Ohms plate to plate; mu = 3.85; rp = 800 Ohms, Gm = 5,500 uMhos.
The 1/2 primary winding is 1500 Ohms, but effectively driven by two rp’s in parallel.
We have 400 Ohms parallel driving 1500 Ohms.
The interleaved lams and the windings of the primary, has enough inductance not to be a factor versus the reflected loudspeaker load impedance.
But the closed baffle “8 Ohm loudspeaker” that is connected to the 8 Ohm tap has a DCR of 3.5 Ohms. The impedance at frequencies well below the woofer resonance is about 3.5 Ohms.
1500 * (3.5/8) = 656 Ohms reflected impedance.
656 / 3.85 = 170 Ohms.
1/ 5,500 uMhos = 182 Ohms.
Looking up into each cathode, we have 170 + 182 Ohms = 352 Ohms.
In this case, at low frequencies, we need to bypass 352 Ohms (that is also in parallel with the self bias resistor Ohms).
If the self bias resistor is 1000 Ohms, the 352 Ohms effective cathode impedance is the major factor we are bypassing, it is about 1/3 of the self bias resistor (which is the factor I was trying to call attention to).
In this case should be bypassing 260 Ohms.
I see schematics that bypass the 1000 Ohm self bias resistor, but that do not bypass the effective cathode impedance (which I now know is a little larger than I originally was calculating,
due to the added Rload/Mu factor).
Bypassing 260 Ohms with Xc = 26 Ohms will allow about 0.9 of the voltage across the effective
cathode impedance, and about 0.1 of the voltage across the bias resistor. That is the -1 dB low frequency rolloff.
I hope I got all of it right this time.
Please let me know.
You are correct.
I did not account for the factor you mentioned:
plate load impedance/mu of the tube.
(You made me think again)
I usually use triodes or triode wired pentodes / beam power tubes.
I most often use Individual self bias resistors, with bypass caps on each.
One system has the following characteristics:
Push pull, 6000 Ohms plate to plate; mu = 3.85; rp = 800 Ohms, Gm = 5,500 uMhos.
The 1/2 primary winding is 1500 Ohms, but effectively driven by two rp’s in parallel.
We have 400 Ohms parallel driving 1500 Ohms.
The interleaved lams and the windings of the primary, has enough inductance not to be a factor versus the reflected loudspeaker load impedance.
But the closed baffle “8 Ohm loudspeaker” that is connected to the 8 Ohm tap has a DCR of 3.5 Ohms. The impedance at frequencies well below the woofer resonance is about 3.5 Ohms.
1500 * (3.5/8) = 656 Ohms reflected impedance.
656 / 3.85 = 170 Ohms.
1/ 5,500 uMhos = 182 Ohms.
Looking up into each cathode, we have 170 + 182 Ohms = 352 Ohms.
In this case, at low frequencies, we need to bypass 352 Ohms (that is also in parallel with the self bias resistor Ohms).
If the self bias resistor is 1000 Ohms, the 352 Ohms effective cathode impedance is the major factor we are bypassing, it is about 1/3 of the self bias resistor (which is the factor I was trying to call attention to).
In this case should be bypassing 260 Ohms.
I see schematics that bypass the 1000 Ohm self bias resistor, but that do not bypass the effective cathode impedance (which I now know is a little larger than I originally was calculating,
due to the added Rload/Mu factor).
Bypassing 260 Ohms with Xc = 26 Ohms will allow about 0.9 of the voltage across the effective
cathode impedance, and about 0.1 of the voltage across the bias resistor. That is the -1 dB low frequency rolloff.
I hope I got all of it right this time.
Please let me know.
I think you are right again, thanks.
Leaving +1 out was my 'error'.
Pun intended!
What was the reason for a mid course correction for a trip to the moon?
If if was not for 3 place accuracy of slide rules, was it for 3 place accuracy
of rocket motor burns, and the variable wind in the beginning of the trip at low altitudes?
Leaving +1 out was my 'error'.
Pun intended!
What was the reason for a mid course correction for a trip to the moon?
If if was not for 3 place accuracy of slide rules, was it for 3 place accuracy
of rocket motor burns, and the variable wind in the beginning of the trip at low altitudes?
Actually it was my error; I assumed higher mu, where the difference between mu and mu+1 is of no consequence. I should have given the correct formula with mu+1, and said that you can use mu as an approximation if it is sufficiently high.
Why would you share anything between the two channels (L and R)? You can share max. the mains transformer and the chassis.
Oh yes! I thought the OP was talking about sharing a resistor between two EL84s in a push pull pair, versus giving each valve it's own resistor, not one resistor for the entire stereo amp! 😱
I was discussing two difference scenarios - each channel (2 tubes) with its own cathode resistor, or both channels (4 tubes) sharing the same cathode resistor. One resistor per tube never entered into the discussion - as that behavior is well-understood for the sake of this discussion.
Well, I for one wouldn't necessarily do that.. but a number of manufacturers have done just that. You see it now & again in console amps. If we can assume that the tube quad is well-matched, and the cathode is properly bypassed, the only interactions between the channels should be at the DC level - as in, idle currents.
With proper bypass, that common 4-point cathode connection is at "ground", so far as AC is concerned - or so I was taught. So the only interaction would be due to DC imbalances, where the tubes would all have to settle for a middling compromise in Ik. No?
Well, I for one wouldn't necessarily do that.. but a number of manufacturers have done just that. You see it now & again in console amps. If we can assume that the tube quad is well-matched, and the cathode is properly bypassed, the only interactions between the channels should be at the DC level - as in, idle currents.
With proper bypass, that common 4-point cathode connection is at "ground", so far as AC is concerned - or so I was taught. So the only interaction would be due to DC imbalances, where the tubes would all have to settle for a middling compromise in Ik. No?
Well, you did start out asking about class AB amps. Cathode current increases at the transition out of class A, which pulls cathode voltage up at a rate determined by the bypass cap. So you end up with your bias wandering, perhaps rhythmically, at some rate which might be defined as AC or DC according to your point of view. This phenomenon seems to be one of a complex collection of artifacts that make up 'tube sound.' The effect isn't quite bad enough to rule out use in budget amp designs, but I see no reason to live with it in quality DIY projects. Fixed bias isn't that difficult.
There is no excuse for sharing cathode resistors between channels. It merely transfers cost from the manufacturer (a few resistors and decoupling caps) to the user (shorter useful valve life). It shows either incompetence or contempt for customers.
- Status
- Not open for further replies.