Valve sound?

[Radio Wrangler]

Originally Posted by Synchrodyne

One could say that very good amplifiers, whether valve or transistor, don’t have a “sound” and are essentially indistinguishable. However, in the descent from the “very good” plateau, valve amplifiers tend to move along a vector that would be described as “valve sound”, whereas solid-state amplifiers move along a vector that would be described as “transistor” sound.

So to return to the original question, it is probably fair to say that “valve sound” does exist, and in a generalized way represents the sound typically produced by valve amplifiers when they depart from the “very good” level. But it is not an immutable property of valve amplifiers; a competent designer whose objective was a true hi-fi amplifier would make sure that it was not there.

Be that as it may, it is clear that no amount rationality and good science will supervene the pseudo-science that is invoked in support of individual beliefs and preferences, not just in the audio field, but generally. “Flat earthing” is alive and well in the “information age”. Or, in colloquial terms, bovine 3-methylindole baffles brains.

That, I think, is the perfect synopsis.

I'd add that it applies to 'very good' amplifiers when not driven into clipping. If that happens, differences can become apparent. Musicians sometimes wish to do this as part of their performance and their amplifier becomes an extension of their effects pedals. It's worth noting that only a single instrument's output is passing through this treatment. Things would be very different if all the band's instruments were distorted together. This is the big divide between guitar (and Jon Lord's organ) amps and hifi ones.

David

[Argus25]

I have attached a pdf (as images of text don't do too well) It should be readable with a bit of zooming. It was Authored by Sidney Corderman Vice President of R&D at McIntosh Labs.

It appears to be as they were transitioning into transistor amps. It looks at differences in output transformer-less amps (OTL's) and transformer amps in transistor power amplifiers. A lot about getting max output power while still in the safe operating limits of the transistors (see fig 11 & 12). Also though some interesting points about McIntosh's low leakage reactance output transformers too. I'm sorry I do not know the publication origin or date, it is a photocopy from my archives.

[Nuvistor]

There is a footnote re the FTC Nov 1974 regulations so it must be later than that. According to Wikipedia the MC2500 was introduced in 1980.

[Radio Wrangler]

1974 would be very late for even considering transformer coupled transistor amplifiers.

It's easy to think that as the amplifier output is going to wind up in magnetic form in the speaker anyway, so what is so evil in having an added trip to magnetism and back along the way?

There are at least two issues with this consideration: The speaker's magnetic bit isn't within the feedback loop, so any accumulation of phase shift doesn't put a stability-imposed limitation on the feedback. Secondly, the chunky iron polepieces in the speaker are carrying the massive flux of the speaker's magnet, not that of the signal. (This is why the pole pieces don't need to be laminated to keep losses down)

David

[Nuvistor]

I understand the article but I don’t have the knowledge to question it, I just noticed the date and found what appears to be the release date of one of the amplifiers noted in the article.

[GrimJosef]

Originally Posted by Radio Wrangler

...

You can drive 8 ohm speakers with direct-coupled valve amplification - no output transformer. It takes a stupidly large number of parallelled bottles, doubled up for a totem-pole push-pull circuit. I know, I was stupid enough to build one! The amount of heater power in a vast array of line output valves is enough to make it pointless.

That does depend on how much power you're happy with. I spent some time measuring a commercial OTL amplifier where the 'large number of paralleled output bottles' was, er, two (PL519s). They were indeed then doubled up into a totem-pole (the so-called 'Futterman' configuration) making four per channel i.e. eight for stereo. The amp measured quite respectably: 0.25% THD+N at 1W into 8ohms, 0.65% THD+N at 10W, passing through the 1% threshold at 16W in one channel and 20W in the other. The harmonics were predominantly the second at 1W and the second and third at higher powers, so not the nastier sounding ones. The output impedance was 0.7ohms, so not quite the '10 times lower than the speaker resistance' point at which its effects become negligible, but not far off.

OK, this amp didn't compete with a Radford STA25, which measures hugely better in every respect and whose output stages only use four EL34s in total (interestingly both amps had near-identical small-signal and drive stages - the OTL designer had clearly learned from Bailey !). The Radford would definitely pass your 'sounds just like the transistor one' double-blind test.

Cheers,

GJ

[G8HQP Dave]

Originally Posted by Argus25

I don't know about the first remark, most of the transistor amps I have tested are definitely biased out of cross over distortion even for very low volume signals unless they are faulty or out of adjustment. So at normal listening levels I cannot see how that could be too significant, unless the amp was faulty or incorrectly biased.

In some designs there is only a narrow bias window between crossover distortion and gm-doubling distortion. Correct bias may thus be unlikely, quite apart from thermal drift.

[Radio Wrangler]

It's very difficult to avoid distortion of one sort or another in the crossover region. I've always considered both a gap in between the areas covered by the source and sink transistors and the case of them overlapping as being crossover distortion. On the whole the overlap condition is nicer because at least you always have some Gm to provide influence over the output.

(16 PL519s in the output stage of a mono amp, similar to Futterman. I was out to equal what I'd done with transistors, peak current, bandwidth, distortion and all. The driver stages were cascoded and quite low impedance to get the bandwidth. Stage gains were somewhat restrained. Yes I did realise that if I'd decided to put it into production there would be a few idiots who'd have considered buying one, but I didn't want to deal with them. It was made to do a job. To prove to myself that I could engineer the valve sound out of an all valve [except rectifiers!] amplifier and to find out if anyone could hear the difference.)

David

[Refugee]

Originally Posted by Radio Wrangler

There are at least two issues with this consideration: The speaker's magnetic bit isn't within the feedback loop, so any accumulation of phase shift doesn't put a stability-imposed limitation on the feedback.
David

Did Phillips not do an amp with a feedback sensor on the speaker cone itself a few years ago?

[Argus25]

Originally Posted by G8HQP Dave

In some designs there is only a narrow bias window between crossover distortion and gm-doubling distortion. Correct bias may thus be unlikely, quite apart from thermal drift.

"Unlikely" correct bias in cheap low range audio gear and lower power range output amps as often the bias isn't adjustable to allow for different transistor specimens and on top of that emitter resistors sometimes are not used so there are more issues with thermal stability if the bias is too high. So the base emitter voltage of the output transistors tend to have little or then a sudden effect on the quiescent current. It is all less than ideal I agree.

But in most reasonable grade Hi Fi amps (with BJT's) there is adjustable bias with a pot and a transistor Vbe multiplier arrangement and emitter resistors too, and the designers not shy to waste some quiescent power in the transistors and the resistors. Then the cross over distortion is not a major issue, that is, if its adjusted properly. The global negative feedback cleans up the gain related square law non linearity in the region of cross over. This is completely unlike peak clipping though, once that occurs, the feedback is effectively useless and cannot correct it.

[Radio Wrangler]

They did. Philips Motional Feedback loudspeaker.

It had an accelerometer on the cone of the bass driver. The bandwidth of this loop was limited to just equalising the bass output at the bottom end.

David

[julie_m]

The OTL valve amplifier certainly sounds like something that would be good for those long, cold winter nights ..... Under consumer law, you'd probably have to describe it as a space heater you can listen to music with, if you wanted to sell such a beast!

Gain is gain, irrespective whether it's being achieved by turning back some of the charge carriers in a stream, narrowing the pipe so only some of them can get through or attracting them towards yourself but they overshoot. Any "characteristic sound" associated with a particular technology really is just a consequence of its inherent deviations from the ideal; and once those imperfections are understood and dealt with, all you are left with is a faithful reproduction of the original sound.

Which, as I've alluded to previously, turns out to be a lot less interesting than it sounds as though it should be. So if the destination is boring, then you might as well at least try to have an interesting journey -- hence some people opt to mountain-bike, pony-trek, hang-glide or run dressed as chickens. And there's a certain undeniable, sadistic pleasure to be taken in watching a grown adult run 26 miles through the streets in an ill-fitting costume, not because they're actually having fun -- which they quite evidently aren't -- but because they think that's what they are supposed to do.

[GrimJosef]

Originally Posted by julie_m

... there's a certain undeniable, sadistic pleasure to be taken in watching [someone], not because they're actually having fun -- which they quite evidently aren't -- but because they think that's what they are supposed to do.

Some people claim to be able to hear a valve sound. Some people claim to know when others are having fun. I can check the first claim with a double-blind test (although in practice that's a great deal harder than many imagine). But I'd hesitate to be so judgemental about others' experience of fun . (The use of the word sadistic was a very nice touch .)

Cheers,

GJ

[Synchrodyne]

Regarding the use of output transformers in transistor amplifiers, my understanding is that McIntosh did this with its first solid-state model, the MC2505 of 1967, and then continued the practice for quite a few years afterwards. But as far as I know, the transformer was used simply as an impedance matching device, and was outside of the main feedback loop. The attached block schematic for the later MC2205, attached, confirms this arrangement.


Cheers,

[Argus25]

the transformer was used simply as an impedance matching device, and was outside of the main feedback loop. The attached block schematic for the later MC2205, attached, confirms this arrangement.

Well I guess they were confident that their transformer was not adding significant distortion that required it to be inside the loop. Also, because it is possibly an auto transformer ( figure 5 for the MC2255) or trifilar wound type (figure 9), like in the article I posted, there are negligible issues with leakage inductance which is a major gremlin in many audio output impedance matching transformers.Also as noted in the article better impedance matching isn't just simply that on its own, it allows more power output while keeping the transistors inside their safety limits.

[GrimJosef]

Quad's first valve amplifier, the 50D (later revised to the 50E) also used an output transformer http://www.geocities.ws/ResearchTria...22/q50ecir.jpg. This allowed relatively easy choice of load matching and, in this case, was inside the feedback loop.

Cheers,

GJ

[Radio Wrangler]

First non-valve amplifier ?

Just trying to prove I'm awake and paying attention...

David

[GrimJosef]

Yes (I clearly wasn't at 07:55 today ).

Cheers,

GJ

[mark_in_manc]

Originally Posted by Radio Wrangler

They did. Philips Motional Feedback loudspeaker.

It had an accelerometer on the cone of the bass driver. The bandwidth of this loop was limited to just equalising the bass output at the bottom end.

David

That's interesting. I was once involved for quite a while in projects looking at loudspeakers which could measure their own acoustic radiation impedance, in attempts to control the way they excited the acoustic room mode soup through which the LF end of the program material has to be perceived. Some of these attempts used accelerometers on the cone (Knowles offered a good, cheap, linear, miniature device which we were later somewhat disturbed to learn had been developed for service in anti-personnel mines), but at least one approach (B&O) used a two-microphone technique to get the velocity component. In fact it might have been one motorised microphone (no relative cal issues), which must have looked very groovy indeed as it did its robot thing.

Oh - and your point above about putting a whole band through the same non-linear transfer function made me smile and think how it is that Deep Purple songs don't contain so many jazz chords - Fm7b5 etc - the root and fifth being about all you can get away with with 'fuzz' before sum and difference inter-modulation products get seriously dissonant, and not in a good way

[Radio Wrangler]

There's a good minute of a Hammond C being distorted to hell and back at the beginning of the Machine Head version of 'Lazy' It's quite obvious that the fuzz is well pedalled back when he wants to play a chord. There's a whole lotta expression going on!

For jazz chords in rock, it has to be Rick Wright. There are some gorgeous key change sequences in Dark side of the moon, and as for Celestial Voices, Ah Now I've got to go listen to it. There's a Farfisa Compact Duo recreated version on youtube by nickhirst999 which is superb!

David