Relationship between output valve resistance, OT and speaker load

Arkless_Electronics

I disagree that a feedback amp always has a load! The point is that the feedback itself causes a reduction in drive to the output valves as the no load output voltage rises and hence the NFB signal rises.

ICBM

jpfamps said:

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.

With a few oddities, yes. eg the Deluxe Reverb's wrong OT ratio, and the 6W Champ's over-hot bias - both of which I suspect were caused by changes after the initial circuit design - in the case of the Champ, the wrong (lower, from the old 4W PT version) voltages are actually shown on the schematic. Fender employed proper electronics engineers who knew what they were doing - it was still fairly new technology back then.

Unlike Marshall, who seemed to copy other maker's circuits and make accidental changes without knowing what effect they would have… eg the JTM45's much higher negative feedback compared to the 5F6-A Bassman, and the 18W having much higher voltages than the WEM Dominator. The first time they tried to design an amp themselves the result was the 200W 'Pig', which could probably best be described as a bit of a lash-up!

jpfamps

Their seems to be two potential mechanisms here.

1) A phase shift in the feedback loop that causing positive feedback and hence oscillation.

Almost certainly due to capacitance in the reactive load. One of the problems designing a reactive load is stray capacitance in the inductive components; whilst resistors and capacitors approximate ideal characteristics, it's impossible to wind an inductor without significant stray capacitance.

Furthermore it's expensive to wind inductors consistently; what you need to do is overwind the part, measure the inductance and remove winds until you get the correct inductance.

If this mechanism is the reason you are getting oscillation then you will have phase shifts within the feedback loop that combine to cause a 180deg phase shift at a frequency where the loop has greater than unity gain.

There are quite a few places where this can occur, including the output transformer, however as the effect is noticeable with 6L6s and not EL34s it's possible that the greater input capacitance of EL34s is causing enough roll off to bring the loop gain below unity.

2) Large voltages caused by very high loading.

This is akin to the open load situation, and the feedback path is due to coupling via stray capacitance to the input.

The output transformer can have a significant role in both the mechanisms.

If the transformer has a relatively low high frequency role if it's likely to exacerbate phase shifts in the audio band. If it has a large leakage inductance (which implies a relatively poor HF response) then ringing when driven into clipping will be worse and may exacerbate mechanism 2.

Interesting we are constantly told that guitar amps don't require much HF response as a) there is not much HF coming from the guitar, and b) guitar speakers can't reproduce much over 6KHz anyway, and thus an output transformer with extended HF is not needed. 

Of course it goes without saying that an OT with extended HF response it more expensive that one without.....

However as we can see from this thread extended HF response (and by implication low leakage inductance) could be beneficial.

jpfamps

Arkless_Electronics said:

I disagree that a feedback amp always has a load! The point is that the feedback itself causes a reduction in drive to the output valves as the no load output voltage rises and hence the NFB signal rises.

You're wrong; a negative feedback amplifier always has a load, even if it's very light.

In the usual feedback arrange in guitar amps the feedback resistor is in parallel with the load (the input end approximates a virtual earth) and needs to be driven by the amplifier, so a feedback amplifier always has some load, even if it's very high with the typical 100k resistor found in amps with a presence control.

In feedback amps the value of the feedback resistor limits the maximum load seen by the amplifier, and as I said above, is I think the reason that Fender reduced the value of the feedback resistors in their BF amps.

Driving the feedback network is not a problem in valve amps as the feedback resistor is relatively large, but can be an issue in op-amp designs where low resistances are used to reduce noise.

Arkless_Electronics

jpfamps said:

Arkless_Electronics said:

I disagree that a feedback amp always has a load! The point is that the feedback itself causes a reduction in drive to the output valves as the no load output voltage rises and hence the NFB signal rises.

You're wrong; a negative feedback amplifier always has a load, even if it's very light.

In the usual feedback arrange in guitar amps the feedback resistor is in parallel with the load (the input end approximates a virtual earth) and needs to be driven by the amplifier, so a feedback amplifier always has some load, even if it's very high with the typical 100k resistor found in amps with a presence control.

In feedback amps the value of the feedback resistor limits the maximum load seen by the amplifier, and as I said above, is I think the reason that Fender reduced the value of the feedback resistors in their BF amps.

Driving the feedback network is not a problem in valve amps as the feedback resistor is relatively large, but can be an issue in op-amp designs where low resistances are used to reduce noise.

I'm right. A load of 100k is no load worth worrying about and counts as open circuit to the output of the amp. What you are saying is like saying a lorry is carrying a load because there is a feather in the back!

Your last sentence is something that does need an eye kept on it but generally if signal levels are so low as to make noise an issue then the output swing will be quite low anyway. Op amps such as a TL071 which don't like to drive anything below 2k or so should be avoided in such circumstances! (but it's noisy anyway!)

SunDevil

Gents - I'm thinking of knocking together a simple restive load box to test the premise that it's the reactive load that's the issue here - plan would be to take a line out from that into the line level in on the Power Station and bypass it's load.

I'm reading that a 25 Ohm, 225W wire wound resistor should be the load (actually struggling to find one, but I can get 2x 50 Ohm / 225W to run in parallel)

The rest of the circuit calls for an 82k 1W resistor and a 5k linear pot - assume these are 'in line' so I have

Input jack

2 wire wound resistors in parallel

82k resistor

5k pot

Output Jack

Earth just runs from input sleeve to output sleeve via the ground lug on the pot

Just that sounds about right?

ICBM

Yes, although 2x 225W resistors is somewhat overkill even by my standards .

Have you tried running the amp set one step higher than the load yet?

SunDevil

Thanks @ICBM - I'd go for lower rated, but the 100W are about 30p cheaper than the 225s, so might as well go large

I've not tried the amp on the mis-match, sorry - I get that the impedance is all over the shop, but I'm a little terrified of smoking the OT

Also, having tried the amps un-attenuated, I think I'm getting a much less invasive, but still present phase issue on all of them (that's been there all along, but didn't both me until now) so i'd like to see if a resistive load will solve that problem (and potentially prevent OT issues down the line)

ICBM

SunDevil said:

I've not tried the amp on the mis-match, sorry - I get that the impedance is all over the shop, but I'm a little terrified of smoking the OT

Don't be. Valve amps are safer with too low a load than too high - if there's any risk, it's to the valves and even that is fairly unlikely.

One of the possible reasons for the oscillation is because the reactive load is currently presenting too *high* an impedance, which is far more of a risk to it.

Just try it. You'll know immediately if it's even part of the cause, hence you'll have a much better idea of which way to go for the solution before you start spending money on things that might not help...

SunDevil

Fair enough - talking sense as ever!

I'll give it a whirl this evening ..if I get home from work before midnight

SunDevil

Alas, same problem, but worth a try

The amp sounds great with EL34s (it's a Plexi style amp, so that's absolutely its core thing)

I guess it's just an odd cocktail to be avoided.

ecc83

Baz, I wouldn't faff about for resistors. Just get one or more of these,

http://uk.farnell.com/te-connectivity-cgs/hsc10010rj/resistor-100w-5-10r/dp/1174288

Virtually indestructible.

Dave.

ICBM

In that case jpfamps is right - it's likely to be the NFB loop going into positive feedback due to a phase shift caused by the reactance of the loads, but suppressed by the input capacitance of the EL34s.

If so you either need to decrease the NFB slightly, or add some caps to the grids of the power valves - as Fender did in the late 60s on the early Silverface amps when they also had stability problems.

SunDevil

Thanks gents, I have a schematic for the amp and I'm assuming there's a NFB resistor going from the output?

What's the easiest way to non-destructively test the NFB option? - put a larger resistor in parallel?

ICBM

SunDevil said:

Thanks gents, I have a schematic for the amp and I'm assuming there's a NFB resistor going from the output?

What's the easiest way to non-destructively test the NFB option? - put a larger resistor in parallel?

No, that will increase it - more voltage being sent back. You need to either increase the resistance in the NFB loop or decrease the resistance at the ground end - which in a 5F6-A/JTM45 circuit is the presence pot, so you could put a resistor in parallel with that.

SunDevil

..sorry, yet more questions!

Re the DIY load box - the 24 ohm load was based on an assumed 16 ohm output.

I typically run at 8 ohms, so it is as simple at bringing the load down to 12?

Looking at a few more schematics for this, most seem to show the load going straight to ground with smaller resistor and pot feeding from the 'front' of the load with a larger (750k in one case rather than 82k) resistor going the pot (now 100k rather than 5k) and then the line out jack

So is my original understanding of the load, 82k and 5k pot all in series a different flavour of this or did I just plain mis-understand the concept and these are one in the same with different resistor and pot values?

ICBM

For a resistive load 12 ohms is fine. In fact anything between about 8 and 24 won't cause any trouble for an 8-ohm amp. Even 24 would probably be OK.

The load resistor goes between the input and ground. The pad resistor and the pot are in series with each other, and the pair are in parallel with the load - not all in series, that would give you an 87K load on the amp, effectively an open circuit.

SunDevil

Thanks again ICBM - appreciate all the help 

Glad I checked re the loadbox!

Re the presence pot approach to reducing NFB - assume I can just put a resistor across the positive and wiper of the presence pot?

Any suggestions on what size of resistance I should try? ..I've a decent selection of 1/4 watt at home if they're suitable?

ICBM

SunDevil said:

Re the presence pot approach to reducing NFB - assume I can just put a resistor across the positive and wiper of the presence pot?

Any suggestions on what size of resistance I should try? ..I've a decent selection of 1/4 watt at home if they're suitable?

Across the two outer terminals, not the wiper.

I would try roughly a 4K7 first, ie the same resistance as the pot. That will halve the NFB ratio.

If you want to try caps on the power valve grids, use 2200pF (100V minimum, preferably higher) caps from the grid inputs to ground, before the grid stopper resistors if it has them.

SunDevil

Much obliged - will report back