At his time, he made the mixing desks the most appreciated by the majority of sound engineer that i know or knew. And on his desks were produced so many master pieces !
I remember those fights with the Automated process mixing desks, trying to get something "living", while just pluggin the same mike in a NEVE, and .... Bingo.
Both were expensive, both where automated, both had close performances on paper. But Ruppert did not agreed the sound of VCAs, so, if I remember well, he invented flying faders. As it was impressive for the customer, to see those faders moving by their own, Automated Process adopted the principle, but with... VCAs.
Scott, please tell me whether this is a fair representation of the output "distortion cancellation" scheme of the AD797? I find it easier to examine by equations. I've set Cn negative instead of showing the current mirror for simplicity. I also assume that Va=Vb (in the data sheet Fig 33) so I only need to consider Vb.
I have deliberately cut the feedback path to calculate the open loop equation Vout/Vin.
The result is an infinite open loop gain when Cn=Cc. This is the hypothetical condition for so-called "error cancellation". This is what reminds me of HEC which seems to me to propose the same hypothetical scheme. Naturally, infinite gain is not possible in practice and approaching it will result in unavoidable instability. Additional compensation is always required to keep the thing stable and this necessarily makes the scheme a standard NFB loop, as Hawksford later conceded.
My challenge to the data sheet is where it states:
Which I do not agree with because the practical implementation has to be a standard NFB loop in which distortion is ratiometrically reduced (assuming it's stable) and not "canceled" by simple subtraction or otherwise. Unless I have made a mistake, the maths does not support this IMO.
Attachments
Last edited:
Well, of course, distortion, bandwidth and linearity, slew rate, damping, every technical aspect has to be good enough.
I was talking about subtle sound details. You know, the difference between film caps and electrolytics, good and oversized resistances in the feedback path at low enough impedance , oversized power supplies and critical rails/speaker cables and printed circuit paths. Non magnetic chassis, non resonant radiators etc...
Not to talk about topologies, FET VS bipolar, CFA VS VFA, Class A VS class B VS class D etc ...
All those details and a lot more, that, placed end to end, makes a difference. At least to me.
(And, please, I'm not talking about outpriced esoteric components, a good quality price ratio matters more than subtle differences.;-)
Last edited:
It paints a horrible, bloody, picture if you think about an F1 driver trying to race a coarse they haven't driven a couple times.
It's probably more important to have track recall, than to have reflexes for them. Where as some other sports people do have better reflexes (cat like? I dunno, what does that even mean).
While most designing has been successful without it, if you knew it could be done in a manner where the results were not null, would you use it ever? Maybe do a serious of tests on things you aren't always sure about?
Hi Tournesol,
Every single product I have seen designed solely by ear has been a disaster. You've got to design and measure and measure and measure, then you listen and alternate measuring and listening. You can't get a good product without the instruments and ears. By doing both, you ensure that your design truly does reproduce music accurately. Depending on the source material, you can put the soul of the piece back out into the room.
-Chris
Every single product I have seen designed solely by ear has been a disaster. You've got to design and measure and measure and measure, then you listen and alternate measuring and listening. You can't get a good product without the instruments and ears. By doing both, you ensure that your design truly does reproduce music accurately. Depending on the source material, you can put the soul of the piece back out into the room.
-Chris
No it doesn't review neutralization of input capacitance on RF amplifiers, A can be 1 and you simply have C and -C, but A has a BW limitation even though it only has a magnitude of 1 at the limit you have 90 degrees of phase and -3dB and lose the cancellation.
Look at it this way, at the unity gain BW frequency the distortion in both cases is the same but with the error correction the distortion falls as 1/f squared while without it falls as 1/f. Take most popular amplifier's distortion vs frequency plots and you will see the rising THD at high frequencies, now superimpose a rising THD that starts later but intercepts the first at the UGBW.
You = me? I'm sorry I'm bending over backwards to be polite and respectful. Listening, waste of time nothing here to hear. Better bench measurements go for it, good luck.
My point was much simpler: I'm sure parts of the design were done without any careful listening test (and quite possibly the whole thing), and it's more likely than not that many things about his mixing desks are far from perfect. After all, such a desk would still be under development. Perfection is the enemy of good enough, but testing is necessary to push the art forward.
That's not a slight against these mixing consoles at all!
My understanding of neutralization is that it is used to reduce the effects of input capacitance of an amplifier by positive feedback of the output to the input via another capacitor. Like a sort of bootstrap. A drawback is that the positive feedback may increase distortion.
Where I may be making a mistake is that it looks to me like the AD797 makes Cn add negative feedback, the opposite of neutralization. Hence -Cn in my diagram. This time the negative feedback should reduce distortion. Otherwise wouldn't Cn have to connect to point B in Fig 33?
Sorry, by "you" I meant anybody. It wasn't a reply to you...by you I mean Scott Wurcer. Yoga is hard at my age.
Sorry, by "you" I meant anybody. It wasn't a reply to you...by you I mean Scott Wurcer. Yoga is hard at my age.
OK, if I have time I might have a plot to clarify. One more time though, the output buffer has a voltage gain of nearly 1, the extra capacitor is connected into the opposite side of the current mirror so it is -C (more exactly the current in C), the crossover distortion is a voltage error across the output buffer and the derivative of this voltage is an error current which is exactly (at low f) the same and opposite in sign from the error current in the compensation capacitor, they add to 0 (at DC and less so as f is increased).
Do the geometric approximation of the magnitude of the difference of two vectors with a small phase difference. You get the 1/f squared relationship.
- Status
- Not open for further replies.