Cathode bias shift - crossover dist.

[EFK]

In another thread, pdf64 wrote:

"Consider limiting the max VB+ to ~400V in cathode bias, to avoid large value cathode bias resistors and consequent bias shift / crossover distortion at high signal levels."

Can someone elaborate on this a bit? I have found, through experimentation and a lot of cathode bias amp building, that when I try to run an amp in cathode bias with higher plate voltages (say up in the 420-440 V range) that often, as the volume is increased and the amp starts to break up a bit, there is an ugly underlying fizzy/static-y sound as the notes decay. Of course if you completely crank the amp it's usually masked and/or disappears, but I never seem to have this problem when the plate voltages are lower, say sub-400V as noted above. Also, cathode bypass caps seem to aggravate the situation. If I don't bypass the cathode R (shared), I usually don't hear it.

Anyway, I came across the above quoted comment and thought "aha!" maybe this is what is going on, however without a scope and all that I'm just guessing. But, the above comment seems to fit the situation as I've described it. Doesn't seem to matter what tubes are being used either - 6L6, EL34 etc.

Anyone with some commentary and explanation?

[Steve Conner]

The further from Class-A the amp is, the worse the bias shift will be. And high B+ voltages need a bias far away from Class-A to avoid excessive idle dissipation.

The cathode bypass cap adds a time lag to the bias shift, making it hang around after the loud note that caused it is over, hence making it more audible.

According to this theory, decreasing the screen voltage would have the same effect as lowering the B+: it keeps the tubes closer to Class-A by restricting the peak current.

[soundguruman]

you can only raise the plate voltage so high before you see the crossover distortion, that's correct

[EFK]

Thanks guys! Steve your post explains a lot. I've worked on a lot of old 50s amps that were cathode biased (at idle) right at the tube's 100% dissipation limit and they used these big dropping resistors between plate and screen supply, as in like 10K big. Of course they were functioning as a choke also, but they always seemed excessively large to me.

What I don't quite understand here is the "why" higher voltages require a more conservative idle than a lower voltage. I mean, I've always viewed it in a primitive and possibly incorrect manner - a degree of voltage sag (whether via PT or rectifier type, or both) seems to keep things under control and as the cathode voltage/current rises, the amp B+ sags and keeps things from over-dissipating. So looking at it this way, a really stiff power supply ought to cause problems whether the voltage is lower or higher. This may not be making any sense.

I guess what I'm asking is, how do you *know* when building the amp and scaling the cathode R that you can safely bias to class A? For example, I can go bias one of my amps now at 100% at idle, but when I crank it up and hit it with signal, if I measure B+ and cathode voltage and do all the calculations while signal is applied, it might have climbed up to 150% or 200%, which ain't good and surely can't be class A. I don't quite see how a lower B+ alone can prevent this from happening. ??????

[Steve Conner]

Well, the calculations are for idle only, they don't work when signal is applied.

The reason is that you're measuring the total DC power going into the output stage. At idle, all of the DC power is dissipated as heat in the plates, but with signal, some of it is converted to audio power and sent to the load. (About half of it at rated output, for the average PP tube amp.)

[Chuck H]

you can only raise the plate voltage so high before...

the tubes dissapation must be offset with a bias condition that, with cathode bias, biases the tubes colder when conducting than at idle. And then...

you see the crossover distortion

that's correct

SGM, That was almost correct except that it was a little vague and implies that crossover distortion is a consequence of high plate voltage. Which is misleading. I hope you don't mind the ammendment

[exclamationmark]

I guess what I'm asking is, how do you *know* when building the amp and scaling the cathode R that you can safely bias to class A? For example, I can go bias one of my amps now at 100% at idle, but when I crank it up and hit it with signal, if I measure B+ and cathode voltage and do all the calculations while signal is applied, it might have climbed up to 150% or 200%, which ain't good and surely can't be class A. I don't quite see how a lower B+ alone can prevent this from happening. ??????

Plot the load lines and all will become apparent . You can easily tell, given your OPT impedance, if it crosses the dissipation limit at any point. Usually when I design anything in class A, I'll put an anchor point at 2x the B+ voltage on the data sheet, and draw a dot on where the B+ voltage intersects the dissipation limit (ie, draw a vertical line up from the the B+ voltage on the bottom of the data sheet). I'll connect these two dots and extend the line to the Y-axis. From this I can determine the 'ideal' load impedance and bias point. For example if this magic impedance doesn't exist for my particular B+ I can settle on a different load, and adjust my B+. Or do what most people do and simply live with it. It's really no longer centre biased class A, but it is class A for a majority of the signal. When something is pushed into clipping, it's no longer amplifying both parts of the signal anyway, so discussion of class really becomes moot.

You can do the same thing (slightly in reverse) with your OPT and B+ voltages if you want to check whether the bias is 'safe' or not. The tricky part is that you have to work out the gradient of your slope, which is determined by your OPT impedance. Usually what I do is pick an arbitrary voltage of 100V, and determine how much current runs through the impedance at this voltage. For example if we pick a 6.6k OPT, the current is ~15mA. Make a line between these two points on the graph. Now, take a ruler and keep it parallel to this line, while 'sliding' it up. You can slide it so the endpoint of the ruler is up to 2x the B+. Usually your ruler will run into the dissipation limit before this. Stop when any point touches the dissipation line, and draw a line. Now make a vertical line from your B+ up to intersect your new load line. This is your bias point! More often than not the designers knew what they were doing and this bias point ends up exactly where the max dissipation limit occurs. Shifting the bias simply moves this load line up and down (as you did with the ruler - it doesn't change the gradient). Note the bias point is ALWAYS at the B+. If you're at all confused (you should be ) The Valve Wizard -Single Ended is a really helpful link that helped me get my head around this stuff.

Also...I'm not even sure if it's mathematically possible for momentary overdissipation to occur in centre biased class A, as I've never seen it. If you look at the datasheets you can see the dissipation limit as a hyperbola/asymptote type function, and the load line is a straight line. Similarly to how you can only touch a straight line to one point on a circle, I have a hunch that over dissipation cannot occur in a properly designed centre biased class A amplifier under signal conditions. However, in class B or AB amps, transitions above the dissipation limit can frequently occur, but it doesn't seem to matter much as it's not an idle condition.

[Joe L]

I just finished my first cathode bias amp - an 18 watt Marshall. It suffered badly from bias shift due to grid current when driven hard and had really bad crossover distortion even though it was biased at around 110 percent max dissipation at idle. I hit the books and dragged out the scope and signal generator and tried a few things until I settled on the Paul Ruby mod. I had a -10v shift on the grids before and -.3v after using 16v zeners with a 10.5v drop across the cathode resistor.

It completely knocked out the crossover distortion, eliminated the high frequency buzz, added nice midrange and the amp sounds louder. I have it on a switch so there is no question of "tone memory".

The other things that helped was upgrading the OT transformer from a Magnetic Components to a GDS. Also I tried 100K grid leak resistors which helped a bit without the PR mod but made no difference with it.

And with prices of good used Tektronix scopes on Ebay, there is no reason to not have one. I was using a 2246A from work but picked up a 2247A last week for $111 + 28 shipping! The 2247A is outrageously good for amp work, with builtin voltmeter and frequency counter. Except for a slightly bent knob, it works perfect.

[Steve Conner]

Exclamationmark, you're right, the max dissipation in a Class-A amp is at idle.

Proof: the current draw from B+ doesn't change, so all of the power going to the load must have come out of the dissipation.

Momentary overdissipation doesn't matter for tubes, because the plate has a thermal time constant of tens of seconds. It's a silicootie thing. A silicon die can be so small that the temperature can rocket in a half cycle at 20Hz.

Joe L, that's a good point, the Paul Ruby mod is very relevant as it's essentially converting a cathode bias amp to fixed bias at high signal levels. Which version did you do, the Zener across the cathode resistor, or from grid to ground? Or both?

[Chuck H]

The series diodes from the grid to the bias source is the Paul Ruby circuit. The diode across the cathode bias resistor is an idea I first saw proposed by Ken Gilbert and later by Dale Van Zile (though I can't find his page now) I use the two together to minimize crossover distortion and avoid continued bias shift once the power tubes start clipping.

Suppose you start with 10V across the cathode resistor on your pair of el84's. And you find that they start to clip at a voltage rise of 4V. You can use a 14V zener across the cathode resistor to stop the bias shift at clipping, effectively "fixing" the bias once it reaches that point. So, knowing your tubes will be driven into cutoff beyond that drive voltage you can then use this known figure to choose the zener value for the PR mod. In this case a 15V zener gives maximum discharge allowance to the grid circuit and you're sure to be in cutoff whenever the PR zeners are conducting.

The combination of the two circuits has allowed me to bias near class A and the tubes pretty much behave as such until the cathode voltage rises and shifts the bias condition into AB1, where the zener across the cathode resistor fixes the bias. So the circuit is basically class A cathode biased when clean and AB1 fixed bias when overdriven, BUT, since there IS plate voltage sag once the bias shifts into AB1 the clean vs. clipped output isn't too dissimilar. So the circuit also allows for excellent clean/distortion control with the guitars volume knob.

PS, I know my "technical" explainations are pretty coarse. I formally appologize and differ to the greater minds for corrections and refinements

[Joe L]

Joe L, that's a good point, the Paul Ruby mod is very relevant as it's essentially converting a cathode bias amp to fixed bias at high signal levels. Which version did you do, the Zener across the cathode resistor, or from grid to ground? Or both?

Grid to ground.

What I found significant was that it is recommended to use a zener value a couple of volts higher than the bias. I carefully tuned it so the grid would start conducting on the top half of the wave just before the zener would clamp the lower peak. 16v zeners did the trick and prevented the bias shift with no bad effect on the tone because the EL84s were well beyond cutoff. Now I can beat the outputs senseless without crossover distortion.

The advantage is now I can increase the bias resistor to effect a reasonable idle dissipation without worry that the bias will be cold at full tilt. Seems like a win-win to me.

[Chuck H]

The advantage is now I can increase the bias resistor to effect a reasonable idle dissipation without worry that the bias will be cold at full tilt.

Except that increasing the bias resistor will also increase the voltage across it, and therefor the drive voltage required to run the tube into cutoff. Without another adjustment to the zener value you may start to hear it conducting before the tube is in cutoff. And round she goes.

I always figured the PR mod to be more of a way to allow the grid circuit to discharge rather than a bias control method. So I figure the PR mod circuit values would be designated by the bias. If you change the bias you need to change the zener value. Unless a zener V higher than the new cathode resistor voltage is already in place.

EDIT: Actually, it occures to me now that if the grid is more negative wrt the cathode as a result of cathode voltage rise that will only increase the amount of drive needed to saturate the tube. In fact the cutoff voltage could decrease instead of increase.

I still don't see the PR mod has being a bias control. But it's likely that if a ZV is working in an amp already, then making the grid more negative (or the cathode more positive) should be fine with the only exception being that your cathode voltage at idle should not exceed the ZV for the Paul Ruby mod.

[Joe L]

Except that increasing the bias resistor will also increase the voltage across it, and therefor the drive voltage required to run the tube into cutoff. Without another adjustment to the zener value you may start to hear it conducting before the tube is in cutoff. And round she goes.

I always figured the PR mod to be more of a way to allow the grid circuit to discharge rather than a bias control method. So I figure the PR mod circuit values would be designated by the bias. If you change the bias you need to change the zener value. Unless a zener V higher than the new cathode resistor voltage is already in place.

EDIT: Actually, it occures to me now that if the grid is more negative wrt the cathode as a result of cathode voltage rise that will only increase the amount of drive needed to saturate the tube. In fact the cutoff voltage could decrease instead of increase.

I still don't see the PR mod has being a bias control. But it's likely that if a ZV is working in an amp already, then making the grid more negative (or the cathode more positive) should be fine with the only exception being that your cathode voltage at idle should not exceed the ZV for the Paul Ruby mod.

Ok, the results are in. I did increase the cathode resistor to 150 ohms from 125 ohms and did not need to adjust the 16v PR Mod Zeners. The voltage at the cathode went from 10.5 to 11.5 volts.

Next I checked the cathode voltage under load and it was shifting up by almost 10 volts!. I added a 13v Zener across the cathode resistor with a switch and it was a night and day difference. There is not a hint of the buzz/fizz and the amp is quite a bit louder.

My friend that I built the amp for has never played a tube amp before. I told him to play the amp and choose which setting on the two switches he liked best. He quickly ended up with both zener mods switched on, saying "it is warmer and louder".

He left with amp in hand and a smile you couldn't slap off.

Oh yeah - he is my bosses boss.

[Chuck H]

Pretty cool eh?

The way I do it (as outlined above) is to use a scope to identify that point when the power tubes start to flatten out on the wave form. I measure the cathode voltage under these conditions and use a zener of that value across the cathode resistor. Then I use PR mod zeners with a ZV one volt higher. This way the amp is basically cathode biased when clean and fixed bias when clipping.