EL84 transformers

[Phil_S]

According to "the book" you'd use 130R on a pair. That would be 65R on a quad. 50R is on the hot side but not out of the realistic range. That's the Vox design.

[pdf64]

Ok, so the Vox AC30 power amp uses a 50 ohm resistor on the cathodes, is that really enough?

The schematic voltages indicate that the EL84 plate dissipations are probably at, or above, the limiting value http://schems.com/manu/vox/acbass.pdf
With such shared cathode bias arrangements, how well matched the EL84 set are on plate current may affect this.
Bear in mind that the plate voltage and screen grid voltage are amongst the factors that will affect the range of resistance that's acceptable for the cathode bias resistor.

[roberto]

Of course it depends on the B+, but I personally prefer 82R on a quad, instead of 50R. Expecially when we are talking about old PT.

[Stevem]

Duh, did I really post that! Sorry I had killer tooth pain this moring until I got to a dentist this afternoon!

[gingertube]

I prefer one cathode bias resistor for each tube. Its just far safer, more reliable, less colateral damage when something does go wrong.

For a VOX AC30 I remove the single 50 Ohm resistor and fit 235 Ohms on each tube (2 x 470 Ohms 2 Watt in parallel) each 235 Ohm is bypassed with a 47uF 50 Volt or 63 Volt capacitor.

For new design and build I would tend toward Roberto's value(s) above but keeping separate bias resistors for each tube.
82 Ohms for a quad is equivalent to 328 Ohms for each tube. Use 270 Ohm or 330 Ohm 5W on each tube, 47uF bypass capacitor on each.

Why can't the bypass cap be lower with a higher value of cathode bias resistor? - Do I hear you ask?

When you look between the cathode and 0V you see the cathode bias resistor IN PARALLEL with the Impedance looking into the cathode. That impedance = 1/gm or about 90 to 100 Ohms for an EL84. That 1/gm term dominates the impedance to 0V (90 Ohms parallel 270 Ohms is not much different to 90 Ohms parallel 330 Ohms) and so it largely determines the required bypass capacitor value.

Cheers,
Ian

[pdf64]

Regarding shared c.f. independent cathode bias resistors, can anyone shed any light on this info from http://www.valvewizard.co.uk/pp.html

'If using cathode bias, each valve can be given its own bias resistor and capacitor, or they can both share the same one (of half the resistance, twice the power rating), which tends to encourage extra second harmonic distortion'

ie why would a shared resistor tend to increase 2nd harmonic distortion?

And

'A small capacitor (less than 100uF say) also increases non-linear distortion'

Again, any thoughts on what would cause that effect?

[gingertube]

With a shared cathode bias resistor then there is no balancing of operating points between the 2 tubes. If one idles at higher current then the other will back off its idle current to keep the total current relatively stable, if one tube fails the other will double it's idle current which may cause it to fail too.

This is why it is essential to use matched tubes when using a single cathode bias resistor.

Giving each tube it's own cathode bias resistor is MUCH more reliable and each tube then finds its own idle point. Matched pairs are no longer essential and a fault in one tube will not "take out" its pair in sympathy.

The push pull output stage (from an operating point of view) can be considered a differential pair amplifier. A fundamental characteristic of differential pair amplifiers (tube or SS) is that for minimum distortion you first balnce the two sides to reduce 2nd and other even harmonic distortions and what is then left is 3rd and other odd harmonic distortions due to non-linearities in the individual devices themselves. (Read the first couple of pages of Doug Self's "Blameless Amplifier" paper for an explaination of this). This applies to differential input stages (not generally seen in Git Amps), push pull output stages and differential pair phase splitters.

Now for a guitar amp we are not necessarily wanting minimum distortion particularly 2nd and even harmonic distortions. A common cathode bias resistor tends to emphasize tube missmatches (the stronger tube will hog the current). This will increase the 2nd harmonic distortion.

Now on the bypass cap thing, there are a few things going on here and we could write a paper on it, except that several people alraedy have so the following is a much simplified view.

The bypass cap allows Class AB operation.

In Class A as one tube drops its current the other increases by the same amount. The limiting condition is one tube at cut off and the other a twice idle current. The total current through the common bias resistor is constant, there is no signal induced voltage acting as a feedback signal. For Class A no bypass cap is required.

For Class AB we get to the condition where one side is cut off and we are trying to drive more than twice the idle current in the other side. The total current is now not constant any more and so there is a signal induced feedback which is level dependent. It starts when one tube hits cut off and the other goes above 2 x idle. This introduces distortion. To get rid of that distortion we need to allow the signal current to bypass the cathode bias resistor which we do by adding a "bypass" capacitor.
A small value cap still has significant impedance compared to the cathode to 0V impedance and hence not all of the signal current bypasses the bias resistor, some distortion will remain.

The cap impedance needs to be small compared to the circuits cathode to 0V impedance. This impedance is the cathode bias resitor in parallel with the impedance "looking into" the cathode which for a power pentode is 1/gm which is around 90 to 100 Ohms for EL34, EL84, KT88 etc and about double that for 6V6, 6L6 etc.

That 1/gm will always dominate the cathode to 0V impedance and will therefore determine the required bypass capacitor value.

Ity seems to me that this sort of distortion will always be symetrical and so it will be mostly odd harmonic distortion. I think this is why the "Wiz" said that a small bypass cap increases "non-linear" distortions.

Ians rave for today.
Cheers

[roberto]

Ian, I add a citation from an old post of yours, seems very useful here:

Gingertubes experience and opinions (feel free to debate the later). Remember that when switching out the cathode bypass cap the amp output stage is geting some signal level self balancing being applied via "common mode" feedback. I actually like that sound. In my HiFi Tube Amp designs which are all fixed biased I also add a common cathode resistor for each pair of a value of around 15% of the value what each individual cathode bias resistor would be (Example EL84 push pull would normally have a 270 Ohm cathode bias resistor on each tube - I used a 39 Ohm common resistor + fixed bias). This was intended to reduce 3rd and high odd harmonic distortions (via common mode local feedback) BUT what showed up on a spectrum analyser was not just a reduction in odd order harmonic distortions but an almost complete elimination of Intermodulation distorton, in particular the "sidebands" due to power supply ripple were reduced significantly. So there is more happening here than you might expect. I have NOT tried this on a guitar amp - YET. (You could say that this method applies 30% cathode bias + 70% fixed bias). I built a guitar amp for a friend which has 4 off 6V6 output tubes in parallel push pull with switchable fixed bias / cathode bias for each pair (that is a single "common" cathode bias resitor for each pair). With bypass caps on the shared cathode bias I could not really tell any sonic difference between fixed and cathode bias. With the bypass caps deleted there WAS a sonic difference and I then preferred cathode bias. Big cathode bias bypass caps can also cause blocking distortion problems when overdriven - another good reason to leave them out entirely. HiFi boffins (like where I came from) will tell you that unless you are running Class A Output Stage you MUST use cathode bypass caps (from a minimum distortion perspective) - my experience with Guitar Amps says you can leave them out and that extra output stage self balancing offers worthwhile improvement. Having said that (the counter arguement) - I did a full restoration on a VOX AC30 (the old copper front panel model). On that, I rebuilt the output stage with 2 x 470R 2W in Parallel (235 Ohms 4 Watt) with 47uF bypass on each of the 4 Output Tubes. In that case I was (of-course) trying to retain the VOX "sound". That also was stunning so you can't make hard and fast rules to cover every situation. Cheers, Ian

http://forum.ampage.org/forum.php?cmd=vt&tid=26018

[gingertube]

Roberto,
Thanks for digging that out. I do remember that Guitar Amp. That amp had a VERY carefully matched quad of 6V6G (the old ST "coke bottle" shape tubes). I picked out the 2 best matched quads from about 40 tubes I tested, one set went into the amp and the second set were boxed up for him to put away on his shelf for the future.
So the summary is that the common cathode resistor, unbypassed adds some signal self balancing but does nothing for idle point balancing (the DC condition), in fact the idle point balancing is better with separate cathode bias resistors. Idle point imbalance will show up as HUMMM! If the power supply design is "cheap and nasty" it will show up as BUZZ because the residual power supply ripple has lots of harmonics of the AC Supply frequencies in it.
Cheers,
Ian

[roberto]

Thanks for the explanation Ian,

What are the results of four resistors+caps, going into one single small common unbypassed resistor, then to ground?

Something like four 270R//47u going into a common 10R then to ground.

[gingertube]

Roberto,
I haven't tried that (yet) but it would be the natural next thing to do. Best of "both worlds".
I looked on the "donor" amp shelf and see that I have an old PA Amp chassis which was a 4 x 6BQ5/EL84 - so I have power and output trannies to give it a try.
Cheers,
Ian

[roberto]

Thanks in advance Ian, It would be nice to know your impressions about it.
I will give it a try too, next time I'll have a cathode biased PP on my bench!

[pdf64]

Thanks for your thoughts on my queries Ian.
I'm a bit stuck on why the small bypass cap for the shared cathode resistor tends to increase distortion on the class AB (guitar) amp but on the hifi amp, leaving it totally unbypassed seems to reduce distortion?
Is that because the hifi amp tubes are very well matched, it's in class A, or something else?

[roberto]

Pdf64, I think you'll find this useful:
http://www.tubecad.com/2005/March/02/Wh ... %20Use.pdf