Biased diode clipping Q

[daz]

Which works and which doesn't aside, could one of you explain to me what this paul R mod is used to accomplish?

[Malcolm Irving]

I’ll have a shot:

A cathode-biased class AB push-pull output stage with relatively sensitive tubes, such as EL84s, experiences large bias shifts when overdriven. Part of the bias shift is due to the coupling caps from the PI stage charging up, and part is due to the cathode bypass cap charging up.

The bias shift results in ‘blocking distortion’, in which a large transient signal prevents smaller signals, following within a certain time period, from being amplified.

For continuous overdriving input signals, the bias shift means that the output stage shifts into class C (‘push’ and ‘pull’ sides are simultaneously cut-off for part of the input cycle).

The Paul Ruby mod limits the charging up of the coupling caps, while the Chuck H mod limits the charging up of the cathode bypass cap.

A lot of folk who have built variations of the Marshall 18W circuit (which was actually a copy of the WEM 17W ? Dominator circuit) have complained of 'the 18W buzz' when heavily overdriving the output stage. The two mods (often used together) reduce or eliminate this infamous buzz.

[daz]

I’ll have a shot:

A cathode-biased class AB push-pull output stage with relatively sensitive tubes, such as EL84s, experiences large bias shifts when overdriven. Part of the bias shift is due to the coupling caps from the PI stage charging up, and part is due to the cathode bypass cap charging up.

The bias shift results in ‘blocking distortion’, in which a large transient signal prevents smaller signals, following within a certain time period, from being amplified.

For continuous overdriving input signals, the bias shift means that the output stage shifts into class C (‘push’ and ‘pull’ sides are simultaneously cut-off for part of the input cycle).

The Paul Ruby mod limits the charging up of the coupling caps, while the Chuck H mod limits the charging up of the cathode bypass cap.

A lot of folk who have built variations of the Marshall 18W circuit (which was actually a copy of the WEM 17W ? Dominator circuit) have complained of 'the 18W buzz' when heavily overdriving the output stage. The two mods (often used together) reduce or eliminate this infamous buzz.

So it would do nothing on a non cathode biased A/B amp? Mine started it's life as a 18 watt but it's now 6V6 and no longer cathode biased.

[Chuck H]

So it would do nothing on a non cathode biased A/B amp? Mine started it's life as a 18 watt but it's now 6V6 and no longer cathode biased.

Like Malcolm explained, this is caused by bias shift at the cathode in cathode biased amps AND grid loading which can happen in any push pull amp. Fixed bias amps CAN also suffer from bad sounding crossover distortion but it's especially COMMON in cathode biased EL84 amps. Sometimes crossover distortion can actually sound good. Like a phasor or touch wah. A little is better than a lot though. With EL84's it usually sounds like a can of bees. Sooo...

Is there a problem with your amp? If not then you shouldn't need the PR mod circuit.

[daz]

Like Malcolm explained, this is caused by bias shift at the cathode in cathode biased amps AND grid loading which can happen in any push pull amp. Fixed bias amps CAN also suffer from bad sounding crossover distortion but it's especially COMMON in cathode biased EL84 amps. Sometimes crossover distortion can actually sound good. Like a phasor or touch wah. A little is better than a lot though. With EL84's it usually sounds like a can of bees. Sooo...

Is there a problem with your amp? If not then you shouldn't need the PR mod circuit.

No problem, but not having ever used this amp for a jam or gig i have no idea what to expect when i do. I'm used to 50 watters with el34's so i figure this thing's outputs are gonna be working pretty hard.

[trobbins]

Yes it's an intriguing situation.

For a typical fixed bias, the bias voltage would vary somewhat proportionally to mains AC, and so an inverted bias voltage would make a good (but complex) clipper to constrain coupling cap DC voltage change during overload (driver stage B+ sag would be an issue to appreciate). If a zener was used, then different times/locations could possibly see +/- 5 to 10% change in mains AC. The margin included in the choice of zener voltage would take up some of that, especially for lowish cathode bias voltages. That margin could be quite high in vintage hi-fi - the Williamson ran about a 35-40V cathode bias - I will add to my list a workbench test of cathode voltage versus AC mains voltage.

The zener itself, is not normally conducting, and when it does then the current level is limited by the driver circuit and the voltage shift on the coupling cap, and could possibly increase towards a mA, so the rated zener voltage (normally speced at a much higher current, even for 0.5W types) will be somewhat higher than the voltage aimed at for clipping. The higher voltage zeners have sharp current knee, which I think is a better target than a soft knee (normally achieved by using series zeners of rated voltage below about 4V7). I've only tested low voltage zeners for knee - a 6V8 changed from 6.45 to 6.59V going from 0.01 to 1mA.

Typical cathode bias class AB amps then bring in another big and independent variable of cathode voltage variation with the degree of class B. Perhaps the only tangible measure of when grid conduction is starting to be an influence (and hence to choose a zener voltage) would be to measure grid stopper current on an oscilloscope, or as an average voltage with a meter.

A pot in series with the zener (eg. 10k) could be useful for experimenting. A modern high-efficiency LED in series with the zener could also be of benefit.

[teemuk]

The Paul Ruby mod works by limiting the extent that the power tube side of the .022 cap can go negative.

This. The "Pal Ruby Mod" introduces a limit to how much grid bias voltage can shift, and thus consequently limits how much crossover / blocking distortion such effect causes. One thing important to note is that transformer coupled inductive loading of tube power amps usually alleviates "cross overing" effects because abrupt seize of plate current in cut off causes spurious oscillation in a coil (in this case, output transformer winding), the amplitude of which can be remarkable. (In an overdriven tube power amp the very same effect kills tube sockets, circuit boards, output transformers, etc. due to HV arcing). Anyway, if load is made less inductive (e.g. use a Zobel network / "conjuctive filter) the oscillatory effects tend to be significantly damped.

The other mod cannot limit how far that side of the cap can go negative, but just adds a parallel path to the grid conduction path, increasing the clamping.

At least certain Peavey amps incorporate similar circuitry. I think the whole point in this case is totally opposite to first example: It is to preserve the grid clipping / bias shift -effect since it is regarded as one of those important dynamic characteristics that tube power amps feature. The very same characteristic is even incorporated to several solid-state circuits that emulate behaviour of tube power amp circuits (including Peavey's "T-Dynamics" power amps).

Anyway, though these designs do not really try to eliminate grid clipping they do introduce an important feature, which often gets overlooked: When grid circuit is overdriven its impedance naturally drops (due to diode clamping effect) and associated voltage dividers of driving low impedance load from a high impedance source clipping distort the signal at grids. Ok, this far everything has gone as usual. BUT, now the "parallel diode" hogs most of the current, which significantly decreases how much current the actual tube grid draws (Hint: it would usually skyrocket at such stage). Effects of such grid current draw would in turn reflect to screen current (and henceforth plate current) and although they can be designed to become important parts in dynamic characteristics of the power amp (e.g. Mesa "DynaWatt") they generally tend to "push" the power tube(s) even into further "ovedrive". The tubes operate under greater stress for starters, and if "audible issues" are an issue the "deeper" overdrive (prevented by preventing output tube grids from drawing current) may introduce harmonic distortion, which is not as "musically pleasing" or "contextually fitting" as generic generic crossover distortion "fizz" and blocking distortion "sputter" many of us are accustomed to.

[Mike Sulzer]

The Paul Ruby mod allows grid current to flow in the power tube on the positive half of each cycle rather than just when the .022 microf cap is charging. But it assures that the current can only flow when the drive is large enough to charge the cap to the a bit more than the zener breakdown voltage.

In fact, the PR mod can have no effect unless grid current does flow. A cap allows no dc current, and therefore if there is average current through the zener in one direction, that must be equal to the average grid current in the other direction. That is, the PR mod allows an ac load to exist, and thus an ac current through the capacitor.

The other mod prevents significant grid current from flowing (assuming that it is otherwise possible) and so stops the tube from entering the region associated with grid current flow.

[Malcolm Irving]

The Paul Ruby mod allows grid current to flow in the power tube on the positive half of each cycle rather than just when the .022 microf cap is charging. But it assures that the current can only flow when the drive is large enough to charge the cap to the a bit more than the zener breakdown voltage.
....

Mike, not sure what you mean here? The PR zener and its series diode can only conduct when the power tube side of the coupling cap is 15.6 V or so negative relative to ground. While grid current is flowing the power tube side of that coupling cap must be positive relative to the cathode voltage.

[Mike Sulzer]

Mike, not sure what you mean here? The PR zener and its series diode can only conduct when the power tube side of the coupling cap is 15.6 V or so negative relative to ground. While grid current is flowing the power tube side of that coupling cap must be positive relative to the cathode voltage.

Exactly. The zener diode with the diode in series prevents the capacitor from charging too far so that on the other half of the cycle grid current can flow. The simple version of this circuit is just two oppositely facing diodes connected in parallel fed through a capacitor. Remove one diode and the cap just charges. Use both diodes and a large enough signal and you have current through the capacitor and each diode.

[Malcolm Irving]

Exactly. The zener diode with the diode in series prevents the capacitor from charging too far so that on the other half of the cycle grid current can flow. ....

OK. Understood.

[Mike Sulzer]

OK. Understood.

The significance of allowing grid current has not been discussed, I believe. With the 13 volt zener in the cathode, the plate current is not limited by the cathode resistor under clipping conditions. Therefore, the plate current can increase somewhat over what might be expected. It is not a huge amount since the PI cannot provide much current, but none the less you can expect the power in clipping to rise somewhat.

[SoulFetish]

Great thread!

I love the sound you can get out of EL84 output stage. For my money, I like to drive them hard into clipping. But they're sensitive output tubes and blocking distortion and heavy crossover distortion can be an issue. I'll never forget the first time I heard it, I thought someone dropped some hardware on my speaker cone. The nice thing about Ruby's (and other's) circuits is that they offer a pretty affordable and easy way of retro-fitting existing amps. I think Nickb actually timmed the fat on this circuit and came up with a simpler version which was a functional equivolent a few threads back.
Personally, I was faced with redesigning my amp anyway, so I decided on a DC coupled output stage. Plus, now I have a bipolar supply that I can experiment with (I've been inspired by Sulzer's "high gain/low noise" design).
Chuck, any thought to solving your problem by getting rid of your coupling caps all together?

[Chuck H]

Not as yet. The coupling caps DO serve another function WRT guitar amps. Because "guitar amp speakers" (and almost always the cabinets) are so poor at frequencies below about 70Hz, why have the power tube squander watts down there? Coupling caps can be used to trim unneeded frequencies from reaching the output tube grids. If the tube is going to make watts, it may as well make useful watts instead of LF that is only causing extra wear on the tubes and is inaudible to the user. I suppose you could trim those frequencies from the stage preceding the PI instead. I haven't tried it. I think there was an old Gibson model that used a triode/pentode tube for the output stage that was directly coupled (EH150?). I'm not looking that up right now though.

EDIT Okay, I DID look it up Some EH150's were actually transformer coupled. There are many different designs for that model and it looks like some early ones did use a dual tube power tube and some later ones used a cap coupled tube PI. So many different designs that you couldn't even know what someone meant if they said EH150 and didn't attach an era.

[trobbins]

Other gremlins could easily come out of the woods if output stage coupling caps were removed. Microphonics would be one - especially in combo's.