From hifisonix.com we have:
"Firstly, make the amplifier linear in the open loop condition."
https://hifisonix.com/technical/the-case-for-feedback/ ...unfortunately there are some broken links in the article.
"Firstly, make the amplifier linear in the open loop condition."
https://hifisonix.com/technical/the-case-for-feedback/ ...unfortunately there are some broken links in the article.
Bart Locanthi worked on military servo systems and analog computers in the 1940's and 1950's so he understood a lot about feedback and what happens if you abuse it. The link below is to the JBL SA600 which he designed and it was launched in late 1966. If you look at the circuit (from over 50 yrs ago), you can see that many of the things we consider accepted practice now, he was already doing. Unfortunately, a few years later, Matti Otala came with a story that feedback caused problems by drawing some wrong conclusions. Bruno Putzeys discussed this in his 'F-Word' article
https://hifisonix.com/technical/the-sa-600-amplifier/
https://hifisonix.com/technical/the-sa-600-amplifier/
I'll have to fix those links - apologies for that. I wrote that article a few yrs ago after reading an anti-feedback article on Audio Note's website, and then Martin Collums 'infamous' feedback goes 'round and round' article in Stereophile.
🙂
This one and some old RCA app notes were the blue print for my designs in the early 70s. Nice to see it again.
I was actually taught the exact opposite, that overall feedback is in principle to be preferred over local feedback because it is in principle more effective, but that local feedback can be useful as a frequency compensation technique. Linearizing an amplifier with local feedback is therefore suboptimal unless it is needed for stability.
Still, from that point of view, an amplifier with degenerated input differential pair, common-emitter stage with Miller compensation and an emitter follower still makes sense. The degeneration, Miller compensation and inherent local feedback in the emitter follower all help to keep the thing stable, and the Miller compensation only provides local feedback at frequencies where it is needed for stability. Besides, the local feedback inherent in the emitter follower has essentially no local loop gain when it is driven from a quite high impedance, like it is at frequencies where the Miller compensation has not kicked in yet.
The only thing that is missing is an inductor across the emitter degeneration resistors of the input stage to kill the local feedback at low frequencies, where it is not needed for frequency compensation. In fact you could combine a decent slew rate with low low-frequency voltage noise by adding such an inductor, provided the hum it picks up is low enough.
Still, from that point of view, an amplifier with degenerated input differential pair, common-emitter stage with Miller compensation and an emitter follower still makes sense. The degeneration, Miller compensation and inherent local feedback in the emitter follower all help to keep the thing stable, and the Miller compensation only provides local feedback at frequencies where it is needed for stability. Besides, the local feedback inherent in the emitter follower has essentially no local loop gain when it is driven from a quite high impedance, like it is at frequencies where the Miller compensation has not kicked in yet.
The only thing that is missing is an inductor across the emitter degeneration resistors of the input stage to kill the local feedback at low frequencies, where it is not needed for frequency compensation. In fact you could combine a decent slew rate with low low-frequency voltage noise by adding such an inductor, provided the hum it picks up is low enough.
I once had a VHF oscillation in a relay control circuit cause low-frequency hum in a preamplifier. The relay control circuit worked on an unregulated supply with a large ripple, which modulated the oscillation, which coupled to the amplifier signal path and got demodulated there.
The frequency was way too high for the 20 MHz oscilloscope I had back then. It took years before I managed to figure out what was going on. (I had found a workaround, but hadn't a clue why it worked.)
I think it's a wonderful story, 🙂 however it assumes that nobody else did none of this as well?
I often get a sense that people think that products and circuits are always developed in a vacuum clean-room idea.
That is very rarely the case.
You pick up on things, reuse what others did, add your own twist and ideas and eventually come up with some unique tricks or approaches.
However, that is still after all that other information.
And before people already start drawing the wrong conclusions (like they often do), no that does absolutely NOT mean that people aren't smart and clever.
So my first question would be, if he was already doing many of these things we consider accepted practice, where there other people doing that as well?
Lots of valuable insights emerging out of the smoking remains of this initial derailed thread. Thank you all.
Marcel, an exceedingly important contribution of LTP emitter degeneration is to open up the linear operating region so that with fast transients, the LTP cannot switch. With typical 0.5V degen (from say 100 Ohms in each emitter and 10 mA tail current), you get about 0.5V linear operating region, whereas without degen and say 2 mA tail current (a typical operating regime in early amplifiers) the linear operating region of the LTP is a few mV and with a fast transient, the LTP switches and the amp slews because it’s now operating open loop.
You don’t see this in opamps where emitter degen is often omitted for reasons of noise, but in those cases, Cdom is sized accordingly and the onset of slewing is 200kHz (modern opamp).
Because he came to audio from another branch of engineering where this stuff was already well understood. And let me ask you this, if we know feedback is actually a good way to improve amplifier performance, why are there ‘gurus’ like Colloms and Qvortrup pushing the ‘feedback is bad’ schtick in 2023?
In short, no one in audio was doing what Bart Locanthi was doing because they were stuck in a vacuum tube mindset and the designers did not understand feedback. Early 1970’s Japanese amps with low distortion but flawed sonics are another example. You only have to open a transistor amplifier from that era, or look at the circuits, to understand this was the case.
Last edited:
Sure. But what do you do with people who think low distortion automatically implies impeccable sonics (not anyone here of course; let's say, hypothetically speaking, maybe at some other forum)? Wouldn't that be in a way much like Colloms, just off in some other direction of misunderstanding?
Last edited:
As someone that remembers those days, I can vouch for the "vacuum tube" mindset. Original transistor circuits were basically the same as tube circuits, and they worked (and sounded) worse. Even output stages were still transformer coupled.
I would like to point out that even today, competency with linear circuits is not a universal given. In fact, it seems to me that fewer people are learning about linear circuits; when I tell people I design linear circuits, they roll their eyes like I'm a dinosaur. And I'm still learning because the topics of linearity and stability are deep rabbit holes.
Feedback was used in early solid-state amplifiers to try to hide horrific non-linearities. No amount of feedback can straighten a transfer characteristic whose slope becomes zero at the origin. Feedback got blamed when the real problem was poor open-loop linearity.
Ed
Ed
Why wouldn't that show up in a distortion measurement?
Really, I'm more interested how much distortion measurements tell us today. Do standard measurements tell the whole story of sonics? Are there any remaining loopholes of things that may get missed?
I will even go so far to say that I personally think some things can get missed. What I wonder about is what other people think? Are there enough people here who think there are remaining issues with how some people typically measure (say, if they mainly rely on push-button AP-type measurements; IIRC, something Scott Wurcer subtly warned about), to make an informative article for Jan?
To be clear, I'm talking not about crazy audiophile claims. I'm talking about what are perceived as respectable engineering companies.
Last edited:
Apologize for so many edits in the last post while trying to figure what's bothering me here. I guess its this: If we have articles about how feedback is good, and how low-distortion and low-noise are good, does that sort of thing indirectly lead to readers towards becoming more easily convinced that believing in such things as SINAD rankings, as the scientific way to know what is good audio equipment? Then its only a matter of finding the best SINAD numbers per dollar to know the value of a potential purchase.
Sorry, but such thinking exists. IMHO it is replacing one type of audiophoolery with another.
I understanding trying to teach audiophiles something about technology, which is laudable. However, we also have a bunch of semi-recovered audiophiles who have swung in the opposite direction so far they are in some other not-quite-real world. Sure, its less expensive than high-end, but its still not scientific enough. What about informative articles that cut it down the middle, which is probably closer to were the truth of the matter really is? Might that be the best way for consumers to be informed?
Measurement as a part of design is one thing, and it can be very extensive in that context. Measurement as part of standard published specifications is another, and it is often significantly more a marketing tool rather than tell-it-all engineering data. Who will educate people about that?
Last edited: