diyAudio (
-   Tubes / Valves (
-   -   Overload recovery (

lndm 5th May 2006 06:45 AM

Overload recovery
While I have my amp on the bench looking at overload, the output stage bias point slowly shifts during clipping. I especially noticed this when changing the value of CRk.

I have heard and liked the effect of a grid choke. Is it the same to provide a low Z path to ground for low frequencies in the grid circuit as the cathode?

I have read of SY's using multiple LED's in the cathode of an output stage. Will this eliminate the problem (I estimate I will need 80 of them)? Maybe standard diodes would be more practical?

Will fixed bias for the output stage (and grounding the cathode) cure all these issues at once?

I am also concerned about driving my driver into class A2 when driving the output deep into cutoff. Maybe this shoud be a separate question?

arnoldc 5th May 2006 07:02 AM


(I estimate I will need 80 of them)?
Then you'll have the second "red light district" after SY :D

sorry, nothing useful in my post :clown:

analog_sa 5th May 2006 07:52 AM


Is it the same to provide a low Z path to ground for low frequencies in the grid circuit as the cathode?

You can't really separate the grid and cathode connections as the only impedance of interest is between the grid and cathode. A grid choke combined with fixed bias will keep the dc part of this impedance very low. Only an interstage transformer can match this but a decent grid choke will still provide a higher impedance at audio frequencies.

Sorry, no amount of leds will achieve anything similar :)

lndm 5th May 2006 11:55 AM


Originally posted by analog_sa
the only impedance of interest is between the grid and cathode.
I'm thinking the impedance between the grid and cathode is Rk in series with Rg(or the impedance of the choke at low frequencies), all in parallel with the internal impedance between G and K, which is infinite during class A1 operation and finite during overload.

Does the grid current need to find a path from the grid, through the grid and cathode resistances and back to the cathode, and that any resistance here will impede its dissipation?

SY 5th May 2006 12:14 PM

I hate to contradict my good friend from SA, but I have an amp in front of me that disagrees with him. IMO, overload recovery is the single greatest factor in audible differences between otherwise good measuring amplifiers.

There are two forms of blocking overload- one is from the bypass capacitor in the cathode circuit, one is from the RC coupling in the grid circuit. They're both explained thoroughly in several articles by Norman Crowhurst in "Audio" magazine and in the book "Understanding Hifi Circuits", and again by Morgan Jones in "Valve Amplifiers".

The LED biasing scheme eliminates the first form totally. The second form is a bit trickier, but can be done in a brute-force manner by using cathode followers direct coupled to the output tubes' grids (this is the approach that Jones took for the Crystal Palace amplifier). If you use output tubes in pentode mode, you can do nearly as good a job by raising the value of the output tube grid stoppers to a high value (I use 47k in my LED-biased amp).

Proof is in the real amp- with overload, the bias doesn't shift, the idle current stays stable, and the recovery is faster than I can measure. The amp sounds like it's a lot more powerful than it is.

Merlinb 5th May 2006 01:09 PM

Thought I'd throw this one out, a way of providing negative grid clipping beyond the range of cutoff, to prevent blocking distortion:

analog_sa 5th May 2006 08:16 PM


Originally posted by SY
I hate to contradict my good friend from SA, but I have an amp in front of me that disagrees with him.

He,he. But i fail to see a contradiction. And I still maintain that a grid choke or interstage tr + fixed bias provides the best solution. And without involving sand noise sources and non-linearities.

SY 5th May 2006 09:00 PM

"My father's house has many mansions.":D

lndm 6th May 2006 02:26 AM

Aha :D

Simming a single ended CRk, it charges up when there is assymetrical clipping (with a DC component). There seems no ideal size, they all seem to have issues. It appears as if I will want either a CV bias of some description, or plain cathode bias.

I can see that the coupling capacitor could be charged by grid current. Perhaps it could be made small.

Aside from eliminating the cap - Is it true that a low DCR from ground to coupling capacitor (like the choke) shorts the grid current path negating it's effect?

-Wouldn't it be true that a low Zo conventional common cathode driver would inadvertently assist the charging effect?

I was once surprised to find that paralleled drivers actually made distortion worse, and I now suspect this issue. If this is in fact true, maybe cathode bias and a CCS would be a good choice for the driver (but maybe not).

BTW, Crowhurst seems knowledgeable, what I read of his was helpful. Is there a different article that would apply specifically to non-NFB SE?

lndm 6th May 2006 02:29 PM

Putting things into practice. I shifted some EQ to the driver stage. 12nF working into 100k. I also biased the outputs with a string of zeners. I needed to bypass with a small film to short out the spikes.

Impedance calculates to 15 ohms. Overload seems nonexistant :cheerful:

Can't listen just now, too late to do overload testing.

All times are GMT. The time now is 08:17 AM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright 1999-2017 diyAudio

Content Relevant URLs by vBSEO 3.3.2