Exactly, that is one of the other important points imho.
The next one would be to ensure that the measurement itself does not influence the signal processing chain.
That seems to be often overlooked- listeners are listening to a complete system and as an additional factor their perception is part of the system too.
For good reasons we still have to rely on results of listening tests.
And everyone who is preparing tests, participating in tests, or analysing test results should be aware of his own bias mechanisms.
Because of this there exist "triple blind" or even "quadruple blind" approaches, the last one beeing not an official phrase afaik.
@ qusp,
the "jitter test" linked is a good example of what is happening if people don´t think enough about the test goals (hypothesis to test) before setting up something.
Or don´t communicate what the objective of a test is nor what is needed to do the test in a sensible way.
Last edited:
The first mention of it I found in the literature is here, in the letter Bob Adams wrote to JAES to critique the work of Malcolm Hawksford on noise-shaped DACs. You'll find he begins the discussion of clock jitter sensitivity in the second-to-last paragraph in the left hand column of the first page : http://www.essex.ac.uk/csee/research/audio_lab/malcolmspubdocs/J16 Letter re J15 JAES paper.pdf
Jitter has become a major limiting factor in the quality of D/A conversion. The extensive use of serial communications protocols forces the use of PLL-based clock recovery circuits which often produce unacceptable amounts of jitter. The sensitivity of various D/A conversion techniques to jitter varies widely and is often signal dependent. A new VLSI chip has been designed to solve the problem of asynchronous rate conversion between arbitrary frequencies. A servo-control loop with very long time constants has the effect of heavily low-pass filtering the jitter spectrum, which largely eliminates the effects of jitter on signal quality.
Author:Adams, Robert
Affiliation:Analog Devices Semiconductor, Wilmington, MA
AES Convention:95 (October 1993)Paper Number:3712
Author:Adams, Robert
Affiliation:Analog Devices Semiconductor, Wilmington, MA
AES Convention:95 (October 1993)Paper Number:3712
From the EE times article (linked by fas42 in post 41978):
What can be shown from simple circuit theory is that bi-wiring will introduce a slight dip in response at the crossover frequency, unless the cable resistance is negligibly low. This is because the two currents (to HF and LF drivers, via crossover) will partially cancel in a single cable as they are not fully in phase. In two separate cables they cannot cancel so the voltage drop is greater. It is just about possible that this dip is audible. Any change is interpreted as an improvement when an improvement is expected. It would only actually be an improvement if the speaker had a small peak at the crossover frequency.
Note that 'widely accepted' does not necessarily mean 'true' - especially in audio.
What can be shown from simple circuit theory is that bi-wiring will introduce a slight dip in response at the crossover frequency, unless the cable resistance is negligibly low. This is because the two currents (to HF and LF drivers, via crossover) will partially cancel in a single cable as they are not fully in phase. In two separate cables they cannot cancel so the voltage drop is greater. It is just about possible that this dip is audible. Any change is interpreted as an improvement when an improvement is expected. It would only actually be an improvement if the speaker had a small peak at the crossover frequency.
http://www.wolfsonmicro.com/documents/uploads/misc/en/Specifying_Jitter_Performance.pdf
Eventually things went better, dual PLLs pretty much addressed the issue of jitter. Today, we have well under 50pS jitter PLLs supporting clocks up in the GHz range, both as ICs and design macrocells. A low cost example.
A very good point. In a transformer input preamp the impedance presented to the next stage will not be constant.
As Scott pointed out the PSSR of a complete preamp is also seriuosly affected by other issues such as CMRR.
Now privately "C" asked if the diode bypass capacitors allow noise to pass. They do to an extent, but if you also look at it they are in series across the secondary. My measurements as to diodes' capacitors here or not are quite inconclusive.
His other comment was that using additional resistors in the ground side of the filter chain may provide advantages. I will have to try that.
Now as to the regulator design. It is a late 60's Popular Electronics staple slightly modified. Before IC regulators (7805) The two transistor, zener and voltage divider was a popular and quite good design. this version has added two capacitors. One that seems to be a zener noise filter but also doubles as a gyrated inductor to the output, the other increasing the feedback gain of the regulator to output noise.
One issue that is now popping up is that most regulators have reasonably low noise and source impedance, but are not flat impedance with frequency. Using a frequency swept load shows many interesting issues with many regulator designs. I am still playing with that, but I did mention my technique for doing the tests a while back on this thread.
So where Jan's LA did a bit of measurements on power supplies and listening tests, the open issue is were they measuring the most critical parameters. A follow on did look at resonances.
Now when the first IC regulator chips came out one of the issues was to boost current. A late 70's approach was to use a PNP output transistor to increase current in the common collector configuration. This had the problem that as the PNP pass transistor heated up, the FT increased and many of them oscillated. Now Demian has pointed me to a thread where they look at the noise advantage of not using a follower. So I may play a bit to compare stability vs noise. Right now I think the issue is resonances in an audio power supply.
Now an issue that many seem to have missed is the noise in the power supply even when it is unplugged. There is not just the 60 Hz spike but all of the harmonics that occur in normal rectification. If you look at the power on graphs you see many more harmonics but these are not dropping with frequency as would be expected. that is because they are not just from the power supplies' rectifiers but also from other ones on the same AC line. The transformer response to high frequencies is a rising one. So much of that noise passes and mimics the internal rectifier noise but does not follow what would be expected for HF attenuation.
Now the circuit John showed is a CLC filter into a follower pass transistor that is fed from a filtered zener that is powered by a current source.
Now John has put a resistor across the L which will limit it a bit but much more importantly will damp any resonances inherent to inductor construction. He also uses two different value filter capacitors that should avoid building resonances. However the larger capacitors have fairly low resonances so they will pass some EMI muck. The small rel caps should get most of that. But he is counting on his regulator to fix most issues. His choice of semiconductor works to higher frequencies than my bipolar one. But not having measurements I can only speculate on how much line noise makes it past his circuit.
ES
Last edited:
Antelope gives us the worlds first 'atomic digital preamplifier'.
Looks like we are finally making progress
Antelope Audio to unveil World?s First Atomic digital preamplifier | Bornrich
Looks like we are finally making progress
Antelope Audio to unveil World?s First Atomic digital preamplifier | Bornrich
Indeed, Bob Adams' paper is 20 years old. One would hope that some progress has been made after that...
jan
Just think, had Mr Levin stuck with the Aardvark tradition, it would have been named anti-lope (or simply bok)
I found it odd that the entire biwiring paragraph was completely unsupported. Just opinion, no actual facts.
The analysis for biwiring is a tad more complex that the simple analysis you allude to, but that is not an excuse to provide opinion as fact in an EE times article..
jn
Of course that Mr. Moscode in the comments section is George Kaye. For a time, my benchmark local audiophile was using Moscode amplifiers. From time to time Kaye would contact him and persuade him to ship them back to him to be modified for even "better" sound. And of course to the audiophile they always sounded "better", as expected.
It may be a delicate matter to probe, but the last system incarnation I heard was using Bryston power amps. No word of what became of the "better-sounding" low-feedback designs.
ahh, but that is when it ceases to be an informed opinion worth listening to. how to reconcile that? with listening as measurement, what happens when taste or mood changes? does the quality of the amps change also?
Of course I'm being somewhat ironic there, as I have always heard good sound at this guy's house. However he is a strange mixture of schooling as an undergrad in the hard sciences, which gave way to grad school in music and a career as a professor and composer.
He also claims to hear the effects of Bybee Quantum Purifiers and that "sonic completion" box that Atkinson said appears from measurements to be a Blumlein shuffler. The latter he is enjoined to not open and look at, I guess the magic smoke could escape.
I don't know if the man would submit to double bacon lettuce and tomato or not. Probably not, because he'd be afraid his electrostats would arc over during some switching event.
To the guy's credit he does do a lot of hands-on work, and has among other things built additions to his house, always overbuilt way beyond code. He and wife and structures came through the 1994 Northridge quake rather well, and they weren't all that far from the epicenter. He's also written some decent music --- there's a particular piece for solo oboe that also plays well on alto saxophone, and plenty of other stuff.
Referring to the schematic in the above post (#41951) I'd sleep better at night if there were reverse-biased diodes across the error amp transistors' (Q2. Q4) base-emitter junctions, protecting them from reverse breakdown in the event of an output short circuit. The collapse of the input voltage could be a bad thing too, but that's very unlikely with all of the capacitance there. Of course once everything is in a box and has survived to that point, there won't be a problem.
Thanks for the explanation of what I put up, Ed. It is the CLC passive filter that is important to note. The rest is 'boiler-plate'. The Blowtorch has a more elaborate power supply with a shunt regulator as well, but the output of the JC-2 power supply still goes through another mosfet sourced cap multiplier, before it sees any audio ckts.
No peeking! Didn't Carver make a CD player with this in it years ago? We had a laugh with it at an AES meeting. Of course it's audible, it grossly processes the stereo signal.
Yes I don't question the audibility of what I suspect is in the sonic completion box! I suspect one of the reasons the maker made my friend agree not to open the box is what it likely has inside, a few eight-legged critters that he's not being bothered with the effects of, not knowing that they are in his signal path.
Additional irony: when my friend "upgraded" his system he used to cast off certain equipment, and sometimes I'd get them. Via this route I got a CD player (PS Audio) and before then, that infamous Carver "Digital Time Lens".
Eventually I bought one of the guy's old Sony players, at that time their top-of-the-line, which he had been using by then as a transport, having fitted into it a glass fiber optic link that drove his Wadia DAC (he was a beta site for Wadia for a time). The Sony had individual manual high-order-bit adjustments, which when out of adjustment also produced audible effects
It was quite satisfactory for years, until it got fussier about things and eventually failed altogether.- Status
- Not open for further replies.