But that was one of Peter Walker's party tricks with QUAD ESLs. And you can see why people would be impressed by that!
Maybe you misunderstood me - in the eighties pictures of perfect 1kHz-squares where quite common in the hifi magizines of that time. So I did my private test with the A77 - just to relate these plots with audible quality.
I deduced my conclusion of limited slew from the well known fact, that saturation level of magnetic tape quickly drops with increasing frequency - an overdrive condition for higher frequencies. All in all, I am not nostalgic with magnetic tape machines now - the audible limitations were quite obvious at their time to me - and I do not claim to be fitted with "golden ears".
First, you have to understand English is not my native language. Not easy to explain nuances, or be as humoristic and funny that i am, by habit, in French.
Then about this protection circuit. It was an old idea. And i am out of the design business from decades. So, I thought you were closer and better than me to find the right 'today' components.
Remember this story of mechanical relay. I thought you had all the luck to know a better part than I. And, once you had chosen the one you like, to chose the transistor with enough current to switch it (2N2222 ?).
So, i only concentrated myself on RDSS to find the best solid state one.
An other thing. When we are not stuck by some commercial contract or source, and when you make an artwork, you can solve some difficulties, using two resistances in serial instead of one, to arrange some good looking, to chose an other component witch better fit in your board avoiding a lot of moves of done work for few millimeters... see what i mean ?
I was in this spirit. A community project.
In fact, i was so often obliged to obey to stupid orders that i am, now, allergic to give directives more than the ones witch can really help.
Alex was more than a board designer. he made a masterpiece.
That you took for laziness was confidence and respect.
Again, in my mind, a community project where everybody could take as much part in creativity he wanted to take.
I realized it was a little an utopia ;-)
Last edited:
To the first question, as far as a raw transducer, not often. Once you introduce crossover networks all bets are off. At least we have tools for analysis and of course for measurement.
As far as choice of impedance, from a cost perspective we used to think that going to somewhat higher Z ought to be slightly less costly---the silicon area was going to be larger for lower Z, hence wafer yields lower, compared to delivering the same power at higher voltage. But confounding this (which was David McCorkle's contention, and was plausible) was the tendency for Philips (now NXP) to get a premium for their very decent parts. In fact they seemed to be about the only integrated amplifier parts that really worked, although the National ones were not bad IF you never pushed them into localized thermal overload (their "Spike" protection) which mutilated the audio. Philips was hardly free from faults, but they worked out some of their problems, particularly with capacitively coupled single-ended output overload (which peak currents would elicit "audio holes"). Now this low-budget stuff really is misplaced here---who would put a big electrolytic in series with the loudspeaker to begin with---but since you asked.
One funny bit with Philips: about the time I stopped doing low-budg stuff, there was one pretty good budget part you could use as a four channel C-coupled or a two-channel d.c.-coupled bridge. But it had a persistent on-chip parasitic oscillation at a fairly low level, iirc about 1MHz. When I asked about it they told me So what?
As far as I know they decided it wasn't worth fixing. There were also some questions about management, and I am sure I didn't hear both sides of the story, but to let Bruno leave (whether a push or a pull) and not integrate the UcD approach, seemed quite odd to me.John, I'm curious. What were the problems with those layout people? Lack of skill? Lack of communication? Not understanding small signal stuff?
Concur. The (folded in this case) cascode is a nearly noiseless way to increase output impedance and eliminate most of the Miller capacitance multiplication, without entailing higher voltages at once from a standard cascode. There is a little noise contributed by the pullup/down resistors, but for the high-gm JFETs this will be small. One is still left with the substantial input capacitance which varies with voltage, but the complementary arrangement makes the variation cancel to some extent, and ideally so if we had truly complementary N and P devices.
It would be a nicer world if holes had the same mobility in silicon as electrons.
Who knows, maybe one day someone will work out how to use positrons without leaving a large smoking mess!
When I was observing with colleagues at Mt. Hamilton, I joked at dinner with the great astronomer George Herbig (of the eponymous Herbig-Haro objects, admittedly not quite a household word) that we were anticipating a quark-based photometer, with noise benefits due to the fractional charge.
Quad used square waves into a reference speaker, and a test production speaker connected out of phase facing it,
a couple of feet away. They checked for a null on a mic between them as quality control.
... this is what it's about - there is a process of learning sometimes needed to be able to steer one's hearing to where a problem is - and from then on it's "obvious"; the brain has learned the trick of focusing on the "bad bits". Which doesn't mean that people can't pick it, unconsciously - many wouldn't be able to say why the sound isn't right, doesn't appeal; they just know they're not happy with the overall presentation.
Thanks for posting the drawings and your comments John. That's what this thread is all about.
Now, there are downsides to this simple design:
One, is that it is sensitive to noise and interference from the power supply. So the power supply has to be QUIET!
Second, it will only run class A.
Third, it is inverting and adding feedback can be a real compromise to the S/N.
Finally, the optimum set-up unfortunately 1/2's the output voltage swing.
One, is that it is sensitive to noise and interference from the power supply. So the power supply has to be QUIET!
Second, it will only run class A.
Third, it is inverting and adding feedback can be a real compromise to the S/N.
Finally, the optimum set-up unfortunately 1/2's the output voltage swing.
Now, what about replacing the '300' ohm resistors with current sources. In fact '300' is a NOMINAL value that depends on the number of paralleled input devices, and the intrinsic Idss of each part. The ultimate part pair would be 1ea, 2SK389V, 2SK109V matched to each other to 1-2ma. High Idss is preferred, the more the better.
> Third, it is inverting and adding feedback can be a real compromise to the S/N.
That can be solved by using a complementary diff pair as Input instead.
> I could replace the 300 ohm resistors with 25ma current sources, BUT it is usually impractical, if not impossible.
7.5V should be enough headroom even for a cascoded CCS ?
Keeping the current constant and tracking for both top & bottom CCS might be tricky.
But then there is a servo to compensate for that ?
BTW they call it Pass LSK Pre over here :
http://www.diyaudio.com/forums/pass-labs/261442-lsk-line-stage-original-schematic-baf-2013-a.html
Patrick
That can be solved by using a complementary diff pair as Input instead.
> I could replace the 300 ohm resistors with 25ma current sources, BUT it is usually impractical, if not impossible.
7.5V should be enough headroom even for a cascoded CCS ?
Keeping the current constant and tracking for both top & bottom CCS might be tricky.
But then there is a servo to compensate for that ?
BTW they call it Pass LSK Pre over here :
http://www.diyaudio.com/forums/pass-labs/261442-lsk-line-stage-original-schematic-baf-2013-a.html
Patrick
Last edited:
- Status
- Not open for further replies.