john curl said:
I was only pointing out that the existance of even one 16/44.1 CD that sounds "good" means the format itself is not hopelessly flawed. I have many. I have also done the experiment and in non-headphone listening I can't hear an additional 17kHz brickwall filter.
Jan - I think the story about Beethoven's 9th is an urban legend fueled by the fact that Beta's 60min length gave VHS a nice target. The rate was based on 50/60Hz video compatibility, so they say.
Useless factoid - 44,100 can be factored as 2^2 * 3^2 * 5^2 * 7^2, i.e., the product of the squares of the first four prime numbers.
G.Kleinschmidt said:
Just use a bunch of 2N3055. Cheap, no oscillation problems there and really low base spreading resistance too 😀
Not joking now, I've seen this done once in a legit publication but I can't remember the ref.
Cheers,
Glen
There are several refs flaoting around using power transistors for MC, there are better than 2N3055 but probably not cheaper.
janneman said:
Mr. Jung and yourself could use some lessons about high performance/low noise regulators, isn't it? Joshua, with John's friendly support, beat you hands down with his outstanding design. You'd better burn your credentials and hide in the woods after this!
And FWIW what is the load regulation? 😀
scott wurcer said:
Not a coincidence, of course, but by design. The choice of 44,100 (as a product of the squares of the first four prime numers) for CD was based on the same requirements of video recorders which were used for the first CD recordings. It gave the greatest flexibility by giving enough factors to ensure divisibility by 50 or 60.
scott wurcer said:
Tom Holman also mentioned this solution: http://www.davidreaton.com/PDFs/Holman_AES_paper.pdf
syn08 said:
Yes indeed, very embarassing! I am now anxiously awaiting John's or Joshua's editing of the Wikipedia entry for 'voltage regulator' to straighten those guys out as well.
This design is also brilliant in that it has NO load regulation, so no discussion on that point either! As I said, brilliant!
Jan Didden
It seems like the value of R1 determines the output voltage, but I could be mistaken of course.
Jan, ook de beste wensen voor 2009!
lessons about high performance/low noise regulators
Mr Jung should be clever enough not to need any lessons on the subject. But someone may nudge him ever so gently and ask that he should also look into shunt regulators, as a follow-on project ... 😉
Mr Jung should be clever enough not to need any lessons on the subject. But someone may nudge him ever so gently and ask that he should also look into shunt regulators, as a follow-on project ... 😉
Re: lessons about high performance/low noise regulators
I think W. Jung has already published that.MRupp said:
courage said:
Yes of course, R1 and the current through it, which is set by Q1 no? Assuming that the treshold value of those followers is known with sufficient accuracy.
Jan Didden
Actually, many of these ideas are 'half baked' and obsolete, as to using power transistors, etc.
We have spent decades on trying different things, and a few points stand out:
1. Power devices MIGHT be quiet in the midrange frequencies, but usually they are not built for low noise, and impurities that might not be important for normal operation will show up as 1/f or popcorn noise. It is possible to find SOME power devices that are quiet, however. I once found a Fairchild comp pair that was designed for 2A out in a small package, initially designed for the discrete output stage of ' Sony Walkman' type amps. They were processed in parallel with smaller area devices and therefore were subject to the same quality controls as the 2N4403, for example. These were used in the Sota headamp, where 2 devices replaced 8 devices as used in the original Levinson JC-1 with essentially the same noise as we are discussing here with paralleled fets. I also recommended them to Dr. R. G. Meyer for a special low noise project back in 1980 or so. I was surprised that I had to supply samples, because Fairchild did not comp. professors of design engineering. Go figure!
2.Bipolar transistors require a large electrolytic cap on the output, and this could be easily detected in listening tests. That is why I now use fets for followers, rather than bipolar transistors, for my best designs.
We have spent decades on trying different things, and a few points stand out:
1. Power devices MIGHT be quiet in the midrange frequencies, but usually they are not built for low noise, and impurities that might not be important for normal operation will show up as 1/f or popcorn noise. It is possible to find SOME power devices that are quiet, however. I once found a Fairchild comp pair that was designed for 2A out in a small package, initially designed for the discrete output stage of ' Sony Walkman' type amps. They were processed in parallel with smaller area devices and therefore were subject to the same quality controls as the 2N4403, for example. These were used in the Sota headamp, where 2 devices replaced 8 devices as used in the original Levinson JC-1 with essentially the same noise as we are discussing here with paralleled fets. I also recommended them to Dr. R. G. Meyer for a special low noise project back in 1980 or so. I was surprised that I had to supply samples, because Fairchild did not comp. professors of design engineering. Go figure!
2.Bipolar transistors require a large electrolytic cap on the output, and this could be easily detected in listening tests. That is why I now use fets for followers, rather than bipolar transistors, for my best designs.
Re: lessons about high performance/low noise regulators
http://www.audioxpress.com/magsdirx/ax/addenda/media/jung2778.pdf
http://www.audioxpress.com/magsdirx/ax/addenda/media/jung2779.pdf
I'm not sure if his two recent articles on regulators cover shunt regulators, but he writes on more than just current regulators.MRupp said:
http://www.audioxpress.com/magsdirx/ax/addenda/media/jung2778.pdf
http://www.audioxpress.com/magsdirx/ax/addenda/media/jung2779.pdf
Yes, thank you very much. I have seen his recent articles on current sources, which would obviously be one half of a supershunt (CCS + actual shunt reg). So the other half is still missing, plus I think I read that he will continue his work on CCS. Maybe we have just to wait a bit longer.
that's pretty good, since J.C. already approves of the Audio quality.
48mA of output and just 410mV of drop across the pass transistor.
48mA of output and just 410mV of drop across the pass transistor.
Now, I hope to explain things a little better.
What we have here is a Norton equivalent of a Zener-resistor combination.
In this case, we have a SELECTED fet-SELECTED resistor combination with a cap bypass, in order to make a QUIET reference down to DC. I select EVERYTHING. That is why Vendetta's are so hard to make.
It is true that the output impedance is relatively high, BUT it is linear, quiet, and consistent in impedance over a wide frequency range. With a class A, low level preamp, this seems to work OK. It is true, that I worried about the 10-15 ohm output Z, but at least it doesn't overshoot and ring, or drive extra noise significantly above 10-15 ohms into the input stage.
It is best to use SEPARATE regulators for each gain block to prevent xtalk to another gain block.
It should be considered part of the input stage, itself, rather than a remote buffer.
What we have here is a Norton equivalent of a Zener-resistor combination.
In this case, we have a SELECTED fet-SELECTED resistor combination with a cap bypass, in order to make a QUIET reference down to DC. I select EVERYTHING. That is why Vendetta's are so hard to make.
It is true that the output impedance is relatively high, BUT it is linear, quiet, and consistent in impedance over a wide frequency range. With a class A, low level preamp, this seems to work OK. It is true, that I worried about the 10-15 ohm output Z, but at least it doesn't overshoot and ring, or drive extra noise significantly above 10-15 ohms into the input stage.
It is best to use SEPARATE regulators for each gain block to prevent xtalk to another gain block.
It should be considered part of the input stage, itself, rather than a remote buffer.
OT
The reasoning is based on pure psychoacoustics.
So it would lead to a design of filters with a lot less pre-ringing and a lot more post-ringing instead.
Cauhtemoc said:
The reasoning is based on pure psychoacoustics.
So it would lead to a design of filters with a lot less pre-ringing and a lot more post-ringing instead.
Cauhtemoc said:
Nonsense. What you see is a sinx/x - like response of anti-alias and reconstruction filter. You can make different design of filters, with different transfer function, without ringing. Then you change passband and stopband attenuation. To use different filter, you need sampling higher than 44.1
- Status
- Not open for further replies.