• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Push-Pull Cathode Bias Resistors

Question: Why do the KT88 datasheets emphasize that two cathode resistors are essential for push-pull operation with KT88 tubes?

Why not a single shared cathode bias resistor? (bypassed or not bypassed, depending on A/AB operation)

Also, if you feel like debating bypassing the cathode bias resistor, check out the old article just posed on Tube Cad Journal (TCJ):
“Effects of the Cathode Capacitor on P-P Output Stage.”
Excellent article.
 

ray_moth

Ex-Moderator
2004-01-27 8:55 am
Jakarta
Question: Why do the KT88 datasheets emphasize that two cathode resistors are essential for push-pull operation with KT88 tubes?
I can only presume that they must be talking about Class AB operation, which ideally requires fixed (grid) bias but can be achieved, to a limited extent, with separate cathode bias resistors, provided that they are bypassed. If you use a single, shared cathode bypass resistor, you are pretty well stuck with Class A, whether you bypass it or not.

I didn't have time to dig for the article in Tubcad but, as I understand it, bypassing a shared cathode resistor is advisable if there is any chance (which there always is) of imperfect matching between the OP tubes. Whether the cap should be between cathodes and ground or cathodes and B+ is also an interesting topic. I believe the latter has some nickname or other - 'Shining Path', perhaps? :D
 

tubewade

Member
2007-06-03 8:54 pm
That is an interesting question for which I have no answer. In my datasheets the same recommendation is made for KT66, 6550, and KT90. For the most part these have commonly been run with fixed bias but there are some amplifiers that use them and use a single cathode resistor for both valves; Altec Lansing A340A, Leslie 147, and Leslie/Vibratone 22H, 122 and 145 to name a few.
 
I can only presume that they must be talking about Class AB operation, which ideally requires fixed (grid) bias but can be achieved, to a limited extent, with separate cathode bias resistors, provided that they are bypassed. If you use a single, shared cathode bypass resistor, you are pretty well stuck with Class A, whether you bypass it or not.

I'd have to second that assumption.

Related note:
If you have them Pentode wired and sharing a G2 resistor you will still have to add a G2 capacitor in class A. But you can get away without cathode bypass cap. I prefer no bypass cap on shared resistor push pull designs.
 
tubewade, good articles. Thanks.


From http://www.aikenamps.com/TI_Other.htm:
For those of you who *still* think class AB amplifiers are "class A at lower volumes", read the difference in this next paper:

This is a nice summary paper. For the A/AB argument, the paper could have stated things a lot more clearly, but I don't think the intent of the paper was to do this.

I've pretty much talked myself into build a fixed negative bias supply, but I'm annoyed with myself for building an amp that will require potentiometer adjustments. DC servo loops would solve this problem, but that's just more circuity, making an otherwise simple amplifier into, well, a less simple amplifier.

Being from the Zen school of thought, it's hard to me to keep throwing components into an amp until I can't draw the schematic onto the back of a napkin easily. Is there a good way to have reliable auto-bias for class-AB amplifiers? It that the big hubub over the CCS threads on this forum? One CCS per valve?

Hmmmm ... maybe back to cathode resistors ... http://www.meta-gizmo.com/tri/sunschematic.html

The Japanese ego is very different, which makes possible such Zen circuits.

It's hard for westerners to understand the Japanese school of thought, but a worthwhile exercise to go through.

K.
 
In my latest guitar amp project, I've got a 12AX7 paraphase PI driving EL84's push-pull with cathode bias. I was determined to get rid of any crossover distortion that was related to blocking distortion. I tried all the methods on the Aiken site but blocking was still real bad. Pure crossover distortion (from under biased output tubes run push-pull) will give you a relatively smooth transition across zero volts, as opposed to blocking, which creates a very abrupt sharp cornered glitch that spews much high frequency spectra that's not particularly musically related to the signal.

At first it looked like the Paul Ruby zener diode thing worked. Then later after changing the cathode bias R and screen grid R it appeared not to work any more. Eventually I tried removing the cathode bypass C, and boom, no crossover distortion at all, even well into clipping.

I later remembered that I had increased the cathode bypass cap from about 25uF to 200uF, which may be why the Paul Ruby method no longer worked. So there's all this talk about blocking distortion being about the coupling caps on the grids of the EL84's, but it was the removal of the cathode bypass caps on the EL84's that actually got rid of ANY blocking and crossover distortion...

B+ is 332VDC, bias is 11.8VDC, 36mA at Q per EL84, 11.8W at Q per EL84. OT is from a Fender Deluxe Reverb amp with 6.6kohm primary Z. The screen R's are 1K (which goes to B+) and the cathode bias R is 330R.

So now I'm trying to find out if I've increased distortion in any bad ways by removing this cap. Gain went down obviously but I had too much gain anyway. An article in Tube Cad Journal from 1955 that was highly recommended suggests that there's not likely a lot of difference in distortion. My belief, although not actually measured yet, is that if the output pair is operated class B (which it's not), the lack of bypass cap would suggest real time dynamic range compression, which creates more sustain, which most guitarists like. But since I'm operating the EL84's closer to class A, per tube, there are questions.

When the EL84 is pushed hard to near clipping, since it's got unbypassed cathode bias, the bias voltage on the El84 will increase as the waveshape voltage increases (in real time). So I would expect dynamic range compression. But what about when both output tubes are operating simultaneously (because they are each biased close to class A)? Is there still compression at the low or medium signal levels? Is there expansion? Is there excessive harmonic or IM distortion? The amp sounds real good (driving G10 Greenbacks), but I guess I need to go borrow my buddies spectrum analyzer and do more testing. I'd love to have some realtime compression for the sustain it gives me, but not if I've also cranked up other distortions. Any thoughts? Thx.
 
Last edited:
I believe that you are all overthinking this.

With a single shared cathode bias resistor you are entirely dependent upon the tubes being matched for them to current share properly.
If tubes are not tightly matched one tube can hog the idle current. The higher idle current in that tube means higher grid current and more disturbance (reduction of the bias) of the bias point which worsens the problem to the extent that one tube can eventually take over most or all of the idle current to the point where it will red plate an destroy itself. This is less of a problem with smaller power tubes. It is worsened by the "usual" abuse of the max Rg1 value. It is alleviated somewhat by biasing at 70% of maximum dissipation where the max Rg1 value can be X2 the datsheet value.

Using separate cathode bias resistors allows each tube to find its own idle point independent of what its pair is doing.

Bypass Caps are not required for Class A but are required for lower distortion when using Class AB Operation.

Its just a bigger power tube reliabilty thing.

Cheers,
Ian
 
I forgot to mention that I used separate cathode bias resistors on my EL84's mentioned above. I read that if both P-P tubes share a common cathode R that is not bypassed, it amounts to positive feedback between the tubes, and allegedly caused spurious oscillation in one case.

I totally agree about the current hogging with mis-matched tubes (same issue exists in differential or long-tail PI circuits, especially when they use a CCS in the tail - which is why I don't use that topology for a PI).

It also occurred to me that when you remove a cathode bypass cap, you cause the plate Z of the tube to go up significantly. Now the OT that was right for the EL84's may not be right anymore, causing dynamic range compression in that way as well... If it was a Hi-Fi amp, I wouldn't want that (actually I might), but in a guitar amp some compression causing more sustain is appreciated if attack and decay issues are good. Since this is real time compression, I don't think attack and decay times are issues at all. Any thoughts? Thx.
 
Ian's ravings:

Bob,
Unbypassed cathode resistors cause degeneration (negative feedback) of the tubes gm with a coresponding increase in rp which raises the output impedance.
With a common cathode bias resistor the feedback becomes common mode which changes its effect.

In an EL84 HiFi Amp I actually use separate bypassed bias resistors for each tube but with a small unbypassed common resistor (equal to 12 to 15% of what is the normal individual cathode bias resistor value of 270 Ohms) under those for the link back to 0V. So 33 to 39 Ohms common unbypassed resistor on the 0V side (for a pair of EL84).That means that about 25% of the bias is developed across that common resistor (since there is current from 2 tubes through it) and the individual bypassed resistors are reduced accordingly (to 220 Ohms) to generate the rest of the bias. This generates a bit of common mode feedback which supresses some odd order harmonic distortion but is really effective at supressing Intermodulation distortion. On a spectrum analyser it was remarkable to see that the side bands generated at the test frequency +/- the residual power supply ripple frequencies virtually disappered. This may also be part or wholy caused by the AC balancing action of that unbypassed common resistor.

For a guitar amp, for more compression in the output tubes push up the screen resistor value.
For EL84 I have seen values up to 10K used but I generally use 1 K or 2K2.

Nigel,
I recently did a restoration of a very valuable "Classic" 1963 Copper Front panel VOX AC30. Its power stage was totally destroyed (plus quite a few ather problems). I fitted individual bias resistors and bypass caps for each of the 4 EL84s. I had some concerns that this may "destroy" the characteristic AC30 "sound" but this was not the case and the owner was totally happy.

Cheers,
Ian
 
Last edited:
Ian's ravings:
Nigel,
I recently did a restoration of a very valuable "Classic" 1963 Copper Front panel VOX AC30. Its power stage was totally destroyed (plus quite a few ather problems). I fitted individual bias resistors and bypass caps for each of the 4 EL84s. I had some concerns that this may "destroy" the characteristic AC30 "sound" but this was not the case and the owner was totally happy.

Cheers,
Ian

If I remember right I just lifted my EL34 output stage.
I am not an expert on valves so assumed what I took was OK.

I have 2 Marshall amps, a chip amp version and a "valvestate" version with tube pre amp. To be honest its hard to tell the difference in sound.

The "valvestates" seem to go second hand for about £200.
I got the chip amp Marshall for £100.
The full blown valve amps seem to go for £300 upwards.
You have to pay for the better sound.
 
Eventually I tried removing the cathode bypass C, and boom, no crossover distortion at all, even well into clipping.

I later remembered that I had increased the cathode bypass cap from about 25uF to 200uF, which may be why the Paul Ruby method no longer worked. So there's all this talk about blocking distortion being about the coupling caps on the grids of the EL84's, but it was the removal of the cathode bypass caps on the EL84's that actually got rid of ANY blocking and crossover distortion...


Exactly what I found out in my computer simulations of the VOX AC30. !!!
Remove the Bias Cap , and the crossover distortion is just very little and smooth and nice, not the fuzzy gap, its just a change of slope in the crossing section, and that sound NICE. Gives a warm compression, without the high frequency fuzz spectrum that sounds like a cheap transistor output stage.
 
Last edited:

Aug

Member
2005-07-24 11:30 am
London
Splitting Cathode Bias Resistor.

Hi, I know this thread is a bit old, but came across it while searching for relevant info.
I have an Amplifier I am repairing, it uses 2 x KT88s per channel in push pull cathode bias. It has a common cathode resistors of 300 Ohms, bypassed with 100uF capacitor. After reading a few articles I have decided to use separate cathode resistors and caps for each tube. Would I just double the value of present resistor, is use 600 ohms for each tube? Should I stick with 100uF for cap values. Thanks Aug.
 

Ketje

Member
2012-12-19 7:24 pm
Flanders
It's not because the splitting of the resistors but when you make changes do it for the best.In the old days those capacitors where expensive and fet.So they went for the smallest possible.
The bias is the voltage drop on the cathode resistor made by the mean current though the tube.Then the C holds that voltage stable.With a big one the variations in bias caused by peaks in the output (when the amp leaves classA) have not so mutch impact.
Mona