Watts?

[kg]

No.

thank you.

I'm saying that power supply's limits make tube amplifiers clip, just like those very same limits make solid-state amps clip. And the clipping is not in all cases nice, soft and smooth, on a contrary. As is demonstarted by the oscilloscope capture that shows a Trainwreck clone overdriven to power amp distortion. I could post several more of such scope captures, from Fenders, Marshalls, Mesa/Boogies etc. all portraying about similar hard clipping characteristics.

all of which run out of g1 drive voltage as they pull grid current, since none of them use direct coupled drivers/inductive loads/interstage transformers. so where is the "power amp distortion" actually occurring?

Yes, generic transformer coupled tube amps similarly hit a "brick wall" when clipping. They can't magically exceed the limits dictated by their power supply.

inductive load Q begs to differ with you.

[loudthud]

Yes, generic transformer coupled tube amps similarly hit a "brick wall" when clipping. They can't magically exceed the limits dictated by their power supply.

Ah, but they appear to! Observe the clipping level with a resistive load, then do the same test with a speaker connected. There are certain amps that produce more voltage across a speaker load than they do with a resistive load. We call them Tube amps.

The assertion that more voltage applied to a speaker makes it louder is left as an exercise for the student

[teemuk]

Ah, but they appear to! Observe the clipping level with a resistive load, then do the same test with a speaker connected. There are certain amps that produce more voltage across a speaker load than they do with a resistive load. We call them Tube amps.

You can replicate that phenomenon with solid-state amps too and you don't need an output transformer for it. In either case (SS or tube) both can generate voltage peaks exceeding levels they'd had when clipping loaded by a resistive load. BUT, thanks to speaker inductance, the output power remains constant and only the phase of voltage and current change, thus output current also relatively drops during those peaks. In the end, loudspeaker is a current driven device, so I'm not entirely convinced if those voltage peaks have any considerable effect on loudness. Just because the voltage response shows "there's something more" doesn't mean there is.

[Steve Conner]

I think it's easier for the tube amp to generate those peaks. Tubes are so horribly inefficient in normal use, that it's easy for them to magically find some more headroom when the load impedance gets high. (accompanied by dangerously glowing screens-but that is another story )

But a good SS amp pulls almost all the way to the rail on its rated load to start with, so it doesn't have much more voltage to give. And, the rails don't sag as much.

To mimic the tube amp, the efficiency at rated load of the SS amp would have to be deliberately degraded, and that would lead to extra heatsinking and transformer iron, making it almost as big and heavy as a tube amp of the same wattage. Not quite as heavy, since the transistor one has no OPT weighing it down.

So, to sum up, according to my theory, watts is the wrong unit of loudness, it should be pounds. A 50lb transistor amp should be as loud or louder than a 50lb tube one.

Switchmode power supplies and Class-D power stages should break that trend, if they can stay in one piece long enough.

[loudthud]

BUT, thanks to speaker inductance, the output power remains constant and only the phase of voltage and current change, ...

When the phase between voltage and current shifts from what it is in a pure resistance, the power dissipated by the load goes down. Durring part of the cycle, current is sunk by the amplifier. There are times when the amplifier is putting out a positive voltage but the current is flowing the wrong d#%ned way, it's negative.

In the end, loudspeaker is a current driven device, so I'm not entirely convinced if those voltage peaks have any considerable effect on loudness.

So you would agree that if the current to the speaker load is increased, the loudness would increase?

The assertion that more voltage applied to a speaker makes it louder is left as an exercise for the student

For this to not be true, the volume would have to remain the same or decrease when the voltage is increased. That sounds like magic if you can make that happen.

[Steve Conner]

Loudthud is right. IMO at least. Say a tube amp can apply 3dB more voltage to the speaker at some frequency, such as its resonant frequency, then it will make 3dB more noise at that frequency.

A solid-state amp with current feedback will do the same for small signals. But a tube amp keeps doing it, although to a lesser extent, even when clipped as hard as possible. That isn't easy to emulate: the SS amp just wants to smash into its rails and make a plain square wave. The question is whether it's relevant. The small-signal behaviour certainly is, it gives several dB boost at resonance and in the upper midrange. But I'm not sure how you would even measure the effective boost under hard clipping, except by grinding out power chords in an anechoic chamber and measuring the SPL.

[teemuk]

To mimic the tube amp, the efficiency at rated load of the SS amp would have to be deliberately degraded, and that would lead to extra heatsinking and transformer iron, making it almost as big and heavy as a tube amp of the same wattage. Not quite as heavy, since the transistor one has no OPT weighing it down.

You actually just need to introduce clipping below the voltage limits of the supply rail.

I've seen that done many times. It really isn't a scheme that is almost without hesitation abandoned. An important point is that current draw to sustain the swing to such voltages that are high enough to reproduce the "inductive" peaks is usually required for only brief transients, therefore even a lighter power supply can cope with that as the entire process just taxes the loading of filter cap a bit more than usually. Naturally, the supply will sag down and clip such peaks if overdriving becomes sustained for long periods. Usually it doesn't.

Anyway, these things fall down to hundreds and hundreds of design details and their effects. Kinda like negative feedback loop in a tube amp removing most interaction with speaker inductactance, including the habit to generate those peaks when clipping, or at least massively surpressing their magnitude. Design is everything.

So, to sum up, according to my theory, watts is the wrong unit of loudness, it should be pounds.

Ain't that the truth. :-)

[kg]

i agree that a few things are going on:

• the tube power stage output impedance is typically an order of magnitude higher (sometimes even more than that) than a corresponding SS design
• that higher impedance manifests itself as an output signal that is not so much like a voltage source but more like a current source
• due to this higher output z, the amp generates MORE output terminal voltage as output impedance rises
• tubes are high impedance devices thus need a output transformer
• that output transformer represents an inductive load which can generate voltages far in excess of power supply voltages

all of these factors relate to and are dependent on one another.

Anyway, these things fall down to hundreds and hundreds of design details and their effects. Kinda like negative feedback loop in a tube amp removing most interaction with speaker inductactance, including the habit to generate those peaks when clipping, or at least massively surpressing their magnitude. Design is everything.

during hard clipping, what is the magnitude of the "error signal" introduced by that loop NFB?

[teemuk]

during hard clipping, what is the magnitude of the "error signal" introduced by that loop NFB?

That depends on at what threshold the amp clips. What is the magnitude of difference between input signal and clipped input signal to feedback node. The clipping does not competely eliminate feedback, it just limits its maximum magnitude to a certain level.

[diagrammatiks]

you made it a ss/tube issue...

at any rate...valve loads are continuously variable and all parts of a tube don't breakdown all at the same time, in the same order every time or anytime.

at any rate if you controlled all the other variables, same speaker efficiency, clamped down the tube amp so that it was for sure operating in the same class of operation as the comparative solid state amp 100 watts would be a 100 watts.

100 watts isn't really that loud.

1000 watts is something like maybe twice as loud as a 100 depending on which day of the week it is, who you're asking and the direction of the wind.

A small cinema has what on average...5000-6000 watts for its surround system?

[MWJB]

"1000 watts is something like maybe twice as loud as a 100 depending on which day of the week it is,"

Statements like this ignore the practicalities of amp building. They rely on "other factors being equal", which when considering 2 amps with a x10 power difference, they never are. Let's just look at speaker array, a 100W amp can get away with 1x100W 100db speaker....the 1000W, if using the same speaker, would need 10 of them...127dB SPL.

In the same way a 100W P-P tube amp, running 500vdc B+, in fixed bias, with a pair of 102dB speakers might only be twice as loud as a 10W amp...a 10W P-P, 500vdc, well filtered, fixed bias amp with 2x102dB speakers...& you don't see many of those around here! ;-)

[diagrammatiks]

yeah. it's really silly to compare solid state to tube amps since even comparing them against other tube amps is impossible.

that's what I said right?

[kg]

That depends on at what threshold the amp clips. What is the magnitude of difference between input signal and clipped input signal to feedback node. The clipping does not competely eliminate feedback, it just limits its maximum magnitude to a certain level.

actually, the error signal becomes enormous when any part contained within the feedback loop becomes nonlinear.. iow, clips.

[diagrammatiks]

doesn't the feedback loop fall apart at clipping.

[teemuk]

actually, the error signal becomes enormous when any part contained within the feedback loop becomes nonlinear.. iow, clips.

True, but how "enormous" it gets depends on the difference of the input signal and the feedback signal. Even when the amp is clipping the feedback signal more or less tries to follow the input signal, resulting into less difference between those signals than in a condition in which feedback is removed competely.

So, all in all, there is still little feedback going on and usually even this tiny bit can make a considerable difference to amp's damping characteristics. Just try it: Remove the feedback loop from, say Marshall amp, and drive the amp in open loop hard to its limits (basically to square wave clipping territory) and then do the same thing with closed loop amp. The latter will still pose some of those damping characteristics even when clipping and this will result into lesser magnitude of "inductive" peaks. You know, it might even save your OT from getting arced through.

doesn't the feedback loop fall apart at clipping.

Yes and no. Depends on how you define "fall apart". The amp will try to use some or all of its open loop gain to correct the difference it sees between input and negative feedback input. In other words, the feedback loop is still doing its exact job but you usually run into problem of combination of high open loop gain and requirement for high magnitude error signal pushing things into clipping as well.