Relationship between output valve resistance, OT and speaker load

SunDevil

I've been chasing an oscillation for a few weeks now and finally have an answer

I have in effect 2 identical amps and both of them will present with this oscillation if fitted with 6L6s and  they see a reactive load (again, I have 2 types available)

Remove the reactive load and plug direct to the cab = no oscillation

Alternatively, fit EL34s in either amp and run via the reactive load = no oscillation

Having spoken to the builder, he tells me that the OT was specced to sit in between the ideal primary impedance for both valve types and that the reactive load is weighing down the 6L6s to the point where it's causing a problem

I've read around this and I still don't quite get the relationship between the power valve resistance, the OT primary and how this is 'mis-loaded' (not sure if it's too high/low?) with the reactive load?

Could anyone explain the relationship and why one is bad and the other fine, in idiot speak and without the need to resort to algebra?

Finally, what value am I looking for on the valve data sheets? ..I'm ultimately trying to understand if KT66s, 6550s or KT77/88 might not cause the problem I have with the 6L6s?

Assuming it is resistance, then I see confusing data from different manufacturers

Eg

SED

EL34 plate resistance 2k

6550 plate resistance 3k

6L6 plate resistance 5.6k

JJ

EL34 - R(tiny k/tiny f) of 20K ..WTF?

Cheers

..a very confused Baz

ICBM

I don't know, but my first observation is that if there is no problem with a real speaker connected and only with a reactive load, then the problem is the load, even if two different ones do it.

SunDevil

Blimey - your knowledge isn't actually infinite then? ..just ridiculously capacious! :-)

The two reactive loads are a Fryette Power Station and a Two Notes Torpedo Reload - very different re-amp, but I suspect both may use the same basic reactive load circuit in the Aiken Amps site

And the valve type definately makes the difference..

ecc83

I do not pretend any greater technical than ICBM but would like to make a few points?

The "Anode slope resistance" of a valve is RELATED to the "optimum load" but is not the same as it. The load specified (usually a-a for ppull) is one which give the best balance between output power and distortion. Remember ALWAYS, valves were designed and the ratings given for QUALITY reproduction!

Then, this OL will depend upon the particular circuit conditions: HT V, G2 V anode current. AFAICT guitar amp designers, sucked it and saw and if it sounded ok and did not eat valves (too badly!) Rock n Roll!

I strongly suspect those amplifiers have negative feed back ? If so I am not surprised that under SOME conditions they oscillate. It is likely that the happy-go-lucky, catch all traff loading gives slightly more overall gain with the 6L6 than the EL34.

Usual design fixes for a screamer? A pole in the PI, that is 100puff or so across one or both PI anode loads. Or a cap between the two anodes (needs to be 1kV rating) or you could try a shunt cap across the series feedback resistor.

But at the end of the day amplifiers are designed to drive "usual" loads. Had this been one of "ours" I would just say "tough"!

Dave.

SunDevil

Thanks Dave - And correct, both have NFB with a presence control

In this case, it's pretty much unplayable at the point of breakup with the 6L6s, but sounds great with the exact same settings with the EL34s

Take the reactive load out of the loop and no issue with either

I guess I'm working on the mis-guided assumption that an attenuator with a 4-8-16 Ohm selectable reactive load should be getting pretty close to mimicking a cab - more marketing BS and reality by the sounds of it?

Is there any easy way for me to measure the input impedance of the reactive load?

Baz

ICBM

SunDevil said:

Is there any easy way for me to measure the input impedance of the reactive load?

No - you need to be able to measure AC voltage and current at the same time across the audio frequency range, and preferably look at the phase relationship between them, to calculate true impedance.

My guess would be that the loads don't accurately follow the impedance of a real speaker at some frequencies, and possibly may be causing a resonance which is enough to trip the 6L6s into instability.

Have you tried deliberately using the wrong impedance on the load? The THD Hotplate has a problem with causing screen overload on Marshall-type amps when used at the "correct" impedance, but works fine when used with the amp set to the next one down - eg 8 ohms when using a 16-ohm Hotplate - this is because the Hotplate has too low an impedance at high frequencies.

SunDevil

ICBM said:

SunDevil said:

Is there any easy way for me to measure the input impedance of the reactive load?

No - you need to be able to measure AC voltage and current at the same time across the audio frequency range, and preferably look at the phase relationship between them, to calculate true impedance.

My guess would be that the loads don't accurately follow the impedance of a real speaker at some frequencies, and possibly may be causing a resonance which is enough to trip the 6L6s into instability.

Have you tried deliberately using the wrong impedance on the load? The THD Hotplate has a problem with causing screen overload on Marshall-type amps when used at the "correct" impedance, but works fine when used with the amp set to the next one down - eg 8 ohms when using a 16-ohm Hotplate - this is because the Hotplate has too low an impedance at high frequencies.

I've not, but happy to give it a go - so if I set the Power Station to expect a 16 Ohm load and the amp to output 8 Ohm?

SunDevil

Scratch the last question - just tried it - no better unfortunately (might even have been a touch worse)

Good shout though - apprecaite the idea!

ICBM

SunDevil said:

Scratch the last question - just tried it - no better unfortunately (might even have been a touch worse)

Good shout though - apprecaite the idea!

Try it the other way round if you haven't already. Which way might work will dependent on the direction the impedance curve diverges from the speaker, and I don't know that either .

ecc83

If you really want to get into this Baz you could try a "Zobel" network across the input of the load.

Start with a resistor, chunky one, equal to the working Z and a capacitor of 100nF then slap on more 100nFs and you Might find a value that turns the phase shift back to goodliness. I am assuming the "reactivity" is inductive?

IC I take you point about G2 overload. The fact is that if an amp is putting out say 20V rms (50W >8R) into a big reactance no power is dissipated IN that reactance but it has to go somewhere and is lost in the active OP devices.

The Quad electrostatic speaker claimed the lives of many power transistors because it was almost a pure capacitance. To be fair many of the amps blew because they were not stable!

Note! The Zobel C and R are in series!

Dave.

SunDevil

I'm pretty sure the resistance is inductive but you lost me with most of the rest, I'm afraid - I'll do some background reading!

So I'm ok to run the amp at 8 ohms into a 4 ohm load ..thought that was bad for the OT?

I'm picking up an old scope next week, so I may have pictures (..bring popcorn)

ICBM

SunDevil said:

So I'm ok to run the amp at 8 ohms into a 4 ohm load ..thought that was bad for the OT?

No, it's not bad for the OT (although it can sometimes be a bit harder on the valves) - and anyway, the point is to check whether the impedance of the load is wrong, you're trying to create a *better* match to the impedance curve of the real speaker. If the load has too high an impedance then by deliberately lowering it relative to the amp, you're making it a better match.

jpfamps

ecc83 said:

If you really want to get into this Baz you could try a "Zobel" network across the input of the load.

Start with a resistor, chunky one, equal to the working Z and a capacitor of 100nF then slap on more 100nFs and you Might find a value that turns the phase shift back to goodliness. I am assuming the "reactivity" is inductive?

IC I take you point about G2 overload. The fact is that if an amp is putting out say 20V rms (50W >8R) into a big reactance no power is dissipated IN that reactance but it has to go somewhere and is lost in the active OP devices.

The Quad electrostatic speaker claimed the lives of many power transistors because it was almost a pure capacitance. To be fair many of the amps blew because they were not stable!

Note! The Zobel C and R are in series!

Dave.

It's the capacitance in the load that will be causing the problem due to phase shift, so if the load was purely inductive there would be no issues with stability (there may of course be other issues.....).

If the load is as ICBM says (how do you get the @ICBM to work?) and like the Aiken reactive dummy load (and there is something very similar in RDH4 and published by Douglas Self), then as you see it's a mixture of inductance and capacitance:

http://www.aikenamps.com/index.php/designing-a-reactive-speaker-load-emulator

Now there are some hefty inductors required for even moderate power levels.

As I'm sure you are aware in the real world it's impossible to make anything like a pure inductor; there is always stray capacitance and series resistance to contend with, and I would hazard a guess that it's the stray capacitance that's causing the stability problems.

I can imagine that the unaware may build such a device assuming that a 1mH/3A inductor is a pure inductance.

ICBM

jpfamps said:

(how do you get the @ICBM to work?)

That works .

I don't know for sure because I'm not familiar with these two loads, but my experience of attenuators which are claimed to be 'reactive' and 'mimic a speaker' is that they don't, not closely enough to be sure they won't cause trouble. This is despite what their manufacturers usually say too - the Hotplate being a case in point, it very much does put more stress on the amp and the valves than a real speaker. Deliberately using the 'wrong' impedance is just a get-around based on what I know of how its impedance curve differs from a real speaker (much lower at high frequencies).

The Marshall Powerbrake goes the other way though - not as badly, but if you had a problem with that I would try the opposite and have the amp set to higher impedance than the attenuator.

Arkless_Electronics

Even a short circuit is fine on the output of a valve amp but will blow a solid state amp (if there is no protection against shorts). Working into no load is fine for a solid state amp but can damage a valve amp, especially if it is open loop and has no negative feedback.
The oscillation issue is a strange one. Competently designed amps are "unconditionally stable" and most valve guitar amps use fairly low feedback anyway.... The idea above from "ecc83" to try a Zobel network is a good one but should not really be necessary...
Jez.

ICBM

There are a very few valve amps I can think of which are unstable with no load and no input signal - the Vox AC50 is one, and the Mesa Dual Rectifier another, but only in 'Modern High Gain' mode where the NFB loop is turned off.

It sounds like this one is marginal enough that if there's no other obvious solution, increasing the NFB slightly might fix it.

Arkless_Electronics

I've known a few that do that but it's due to output back to input coupling usually. Amps which use tone controls in the main NFB loop can sometimes be marginal...

SunDevil

This from Randall Aiken on TGP seems very pertinent as to why a reactive attenuators load could cause issues with oscillation?

"Resistive loads are indeed safer than inductive loads for higher impedance loads, because inductive loads can rise much higher than the nominal impedance at the resonant point or at the upper frequencies. A resistive load at a 2:1 mismatch will remain mismatched at 2:1 over all frequencies of interest, while a 2:1 nominal mismatch with a reactive load will vary from 2:1 up to possibly 12:1 or even 24:1 at the resonant point. Higher than normal impedances are usually responsible for transformer, tube, or tube socket arcing failures.

In addition, resistive loads are much less likely to cause instability problems, like parasitic oscillations, because they don't have the phase shift associated with reactive loads. A purely resistive load will not have the same level of overshoot/ringing on square wave edges, either, so it won't have the higher-voltage "spikes" on these edges that can cause arc-over.

Typically, it is not the just the higher resistance load that causes problems. A 2x, 3x, or even 10x resistive load is not really a problem for the amp, because the tube voltage is limited by the plate voltage - the output can only swing to a max of twice the plate voltage on either side of the output transformer. For example, if you had a 500V supply, the max you could ever get is 1000v on each plate, or 2000V p-p across the entire output transformer primary. The transformer is usually designed to handle this. In reality, the output tube cannot drive to 0V at full saturation, especially in class A1/AB1/B1, and the transformer winding resistance also limits the upper range, so you wouldn't even get the full 1000V swing on each plate, more like 950V or so. If you drive the amp to full clip, you'd never get more than around 1900Vp-p across the transformer primary, no matter whether you had it connected to the nominal 8 ohm load or a 32 ohm load.

The problem comes in because transformers are not ideal, and they have leakage inductances and things that can cause the output waveform to not be a pure, flat, square wave. It will have "ringing" or overshoots at the edges, because of the fast transition of the edge, and this overshoot can be very high, depending on the transformer design. Here is where the load impedance characteristics come into play. A reactive load, and especially a mismatched one, can cause this overshoot to be very large, and the spikes can exceed the voltage breakdown ratings of the output transformer or tube socket insulation. The same will happen with a resistive load, but not to as great an extent, but there will still be higher voltage spikes if you run into an impedance that is 2x or 3x the nominal value than you would see if the amp was loaded with the correct impedance."

Arkless_Electronics

A good explanation that.... and also the reason for valve amps being dodgy with no load, as I mentioned earlier. With no negative feedback it gets much worse as there is nothing reigning in the drive to the primary of the OPTX!
It's not though likely to cause continuous oscillation on its own unless there is a "perfect storm" of other factors....
As jpfamps says above, a capacitive load would be much more likely to cause a problem than a pure inductance... but a combination of various reactances (+ & - j) can get pretty nasty!
This is one of those interesting ones that if I had it on the bench here I could probably quickly solve, but is virtually impossible to analyse beyond speculation at a distance. I would wager a few beers it ends up being a decoupling or grounding issue though! :-)

ICBM

Randall Aitken's explanation is why it's worth trying the amp deliberately set higher than the (claimed) load impedance.

jpfamps

Arkless_Electronics said:

A good explanation that.... and also the reason for valve amps being dodgy with no load, as I mentioned earlier. With no negative feedback it gets much worse as there is nothing reigning in the drive to the primary of the OPTX!
It's not though likely to cause continuous oscillation on its own unless there is a "perfect storm" of other factors....
As jpfamps says above, a capacitive load would be much more likely to cause a problem than a pure inductance... but a combination of various reactances (+ & - j) can get pretty nasty!
This is one of those interesting ones that if I had it on the bench here I could probably quickly solve, but is virtually impossible to analyse beyond speculation at a distance. I would wager a few beers it ends up being a decoupling or grounding issue though! :-)

Actually a negative feedback amp always has a load: the feedback network.

On amps with a presence control this is likely to be high enough to effectively be open circuit, eg 100k.

It's interesting that when Fender ditched the presence control they reduced the value of the feedback network. To more you look into the pre-CBS Fender amps you realise that they really did know what they were doing.