I used to design valve amplifiers. I designed one that measured extremely well. Guess what it sounded like! I changed my measurement technique to include real world loads, and real world practical conditions of inputs and power sources, and my later amplifiers sounded much better. They still did not all sound exactly the same. and I would not call any one perfect, or even "the best". However even now I class amplifiers by sound as "good",(even if sounding different from one another), or "bad" regardless of claimed specification.
Hi Trevor,
Thanks for bringing that patent to my attention. I have not seen it and have not had a chance to look at it in depth yet. I've got a whole chapter on DC servos in my book, with numerous arrangements, but I don't discuss a case like this one appears to be. When I get a chance to read the patent, the key things I will be looking for are what problems it solves as compared to more conventional servos, and are those problems common to a large number of well-designed amplifier topologies.
Cheers,
Bob
Thanks for the reply, Jan. I should remember that our vernacular use of irony in Oz is not universal. My intention was to use to use a term contrasting with additive "niceness" but already used here on forum to describe blameless sound quality. I may not agree with the term or such impressions either and certainly "transparent" does make much kinder copy in print, but the term has the same strict meaning to me. Doubtless others will use it describe the presence of some specific distortions that make an amplifier sound awfully sterile, but that is a measureable quality, I believe.
I hope you find no real offense in my accommodating the negative views of others here too.
I agree. Yet some people claim that a bland sterile SS setup can be improved by adding a tube buffer, either as a separate item or incorporated into one of the components. They seem to believe that the valve is subtracting nastiness and adding niceness. My own view on this is that the original SS setup had little or no nastiness, and the valve is adding their favourite niceness i.e. they prefer distortion. Niceness and nastiness are actually different flavours of the same thing: distortion. It may be that a little niceness can mask the presence of a little nastiness, too. Perhaps the real problem with SS is not the presence of low levels of high-order distortion, but the absence of high levels of low-order distortion.
My working hypothesis is a little different. Since to me 'nastiness' consists of RF induced intermodulation products, I would suggest that a valve buffer might be relatively immune to such. If the valve buffer is, for example, following a passive I/V stage for a DAC, then no intermod gets created because the valve has no RF-demodulating junctions to flood. Then assuming its reasonably band-limited, it passes on no out-of-band hash either. Thus in my (tentative) audio view, niceness consists of absence of intermod and accompanying RF. The valve of course adds some of its own 'nastiness' (low order THD) but this is far more benign.
Not being at all an expert with valves, I believe that avoiding RF rectification can be achieved with SS, if sufficient care is taken. In this case the aim is no added nastiness from either low order harmonics or out of band intermod products.🙂
I would add here that I have no measurement evidence so far to support this view, but I am working on it, slowly...
Given that I have ignored the connotations of what "sterile" may mean applied to audio, I think I would agree with this and appreciate the comments.
A cathode follower won't be affected much by RF hash, but unless deliberately band-limited it will pass on most of it. A CD player should not be emitting RF anyway, and a decent amp will have input filtering. If a "tube buffer" genuinely improves the sound then it means the surrounding components are poorly designed, which is the opposite of what some people believe.
There is one stage where it may be genuinely worthwhile to use a valve instead of SS, in an otherwise fully SS item: the buffer immediately following a DAC. A valve can handle the HF hash from a DAC better than an op-amp, but at the expense of more low-order distortion. Some valve CD players/DAC do this. Others just cheat and add a valve after the op-amp so you get the worst of both worlds: op-amp handles hash badly, then valve adds distortion! They can get away with this because they know that most users and many audio journalists can't read a circuit diagram.
There is one stage where it may be genuinely worthwhile to use a valve instead of SS, in an otherwise fully SS item: the buffer immediately following a DAC. A valve can handle the HF hash from a DAC better than an op-amp, but at the expense of more low-order distortion. Some valve CD players/DAC do this. Others just cheat and add a valve after the op-amp so you get the worst of both worlds: op-amp handles hash badly, then valve adds distortion! They can get away with this because they know that most users and many audio journalists can't read a circuit diagram.
Well said. I, not being a native speaker, tend to use the original meaning of words where sterile means 'not contaminated'. That's how I want my sound to emanate from my amplifiers. Sterile 😉 .
jan didden
a buffer can only improve the connection, when the source is incapable of driving the interconnect and the receiver.
I can remember "sterile" being used as a put down when reading reviews of Audiolab equipment.
But, many bought them anyway for they liked the sounds that Audiolab gave.
But, many bought them anyway for they liked the sounds that Audiolab gave.
I think many people's designs have poor PSUs - an amp should never really be too dependent on external sources IMO. I suspect a simple Salas style current-source and voltage clamp shunt regulator would add a whole world of definition to many amps.
In addition I regard resistors as a futile and pointless load for an amplifier - a synthetic speaker may be OK perhaps - as if you've ever done an A/B switch between a resistor and a speaker you'll realise they make the amp behave very differently.
Hopefully, there's a considerable degree of bandlimiting going on if its preceded by a DAC, doubly so if that DAC is a noise-shaped version.
But lots of them do - not necessarily enough to fail EMI regs of course. EMI regs measure quasi-peak levels so one way of getting through them is deliberately to jitter the clock and spread out the energy. Note I'm not saying this actually happens in the case of audio equipment, (its an option in computer BIOSes) - just that its a limitation of the standards that QP levels are used at particular frequencies rather than setting limits on total energy. DVD players which have nowadays replaced CD players are potentially even worse offenders than CD players because of additional (high speed) video circuitry and Blu-rays are beyond DVD.
The standard RC single pole input filtering is often compromised by the grounding. Even in Doug Self's book, the input ground from the signal input is shown routed direct to the long-tailed pair. Thus RF is actually assisted into the LTP by having a shunt cap to that contaminated local ground.
Ha, has me rolling around laughing my *** off😀
I bet it must be very tempting for CD player designers aiming for the section of the market where a jitter test is unlikely to occur!
I regard deliberate clock jitter as cheating, but I know it is widely done. The solution is to reduce the test threshold by, say, 20dB for jittered devices. The current limits are based on the assumption that a little energy here is offset by much less there, so the risk of interference to other equipment is small. This assumption breaks down when jittering is used.
Yes, that's what I thought when I was introduced to imaginary (complex) numbers.
Which is a bit off topic, but the topic isn't going anywhere. There are those of us who wash our hair (what remains of it in my case) with plain old soap (boys), and there are those who like shampoo with every known variety of fruit or vegetable included (girls).
w
Must make a bigger effort to get in touch with my feminine side.
Hello Hugh
Yes quite true, but sometime...
One of my best friend are a jazz pianist who was, for years and years, listening musics on very crappy sound systems. He live quite far from me so we was mostly talking by phone. But wen he come to pick me up to invited me to see his new home, I decide to bring my best cd and my cd player and also my last diy amp, he borrowed good speakers and I installed all that system...
For him and his wife it was a revelation, they did, for many hours until late night, listening at the music and the day after he throw out his crappy sound system. So I've found him a good used cd player and made him an amp and he did buy the speakers he borrowed... now he is happy.
Bye
Gaetan
Last edited:
Hello
A good amp need a good PSU filtering, I alway use C-R-C or C-L-C pi filter.
And for testing my amps prototypes after simulation, I use my loudspeakers as a load to do 10khz squarewave stability tests and FFT to see the thd harmonics. After that I do listening tests.
Bye
Gaetan
Last edited:
I have CLC filtering on my SE amp at the moment. I feel the L is important as it's a nice way of filtering without energy loss, but when I can I'll build a Salas style shunt regulator for an SE power amp. I would upgrade the current source to cascade mosfet but basically I expect it to sound much better: experimentation will tell!!.
The current source effectively eliminates ripple/hum and all mains plug/cable/hash stuff, then the voltage clamp means the music is not being played through some serial transistor in the PSU or some electro capacitor but via the far superior shunt.
CLC cannot compete with that IMO, although they can sound very good. The next step up would be batteries - which is an idea. A few 6S lipos - for 440V odd volts I'd need 20, at $35 each that would be a $700 PSU with need for careful charging regime (to prevent fire) and careful discharge limiting (to prevent killing them). Expensive but probably the best ever - if you went with smaller cells it would be cheaper.
I don't think you'll get much disagreement there, so long as the PSU is not considered as the sole source of noise and distortion. Class AB amps, for example, inject considerable switching sawtooth currents on their own rails that can make diode switching artifacts look tame. At the peak currents of say, a 100W/8 ohm OP stage, the power required from the PSU with a shunt reg. would be substantially greater than a conventional supply.🙁
Assuming the supply is stiff and does not sag below nominal rating at full rated power, it would probably have near constant dissipatiion of the voltage/current supply overhead. If, however, you are only thinking of the supply to the first and second stages, it would seen a reasonable way to ensure higher PSRR. than simple RC filtering yet without excessive waste.
"If, however, you are only thinking of the supply to the first and second stages, it would seen a reasonable way to ensure higher PSRR. than simple RC filtering yet without excessive waste."
That is one aspect of some of Doug's designs that I feel could be improved.
I believe that the use of 10R and 1,00uF in the -Ve rail supply to the front end is capable of considerable improvement as suggested.
SandyK
That is one aspect of some of Doug's designs that I feel could be improved.
I believe that the use of 10R and 1,00uF in the -Ve rail supply to the front end is capable of considerable improvement as suggested.
SandyK
- Status
- Not open for further replies.