You know exactly what I mean. You are just confusing folks. The inductance of the Bybee is higher than just a straight piece of wire. Which is a bit interesting. However not high enough to act as a filter at the 1/F low frequencies of claimed interest.
The resistance is decently accurate. So I can get a reasonable comparison between a resistor and a Bybee.
Now generating a good 1/F noise source is a bit more. I suspect I'll start with an avalanche diode and run that noise through an amplifier or two.
It is, but Ed just wants to verify that it is, since you keep claiming that it isn't.
jn is doing a fine job of pointing out some of the fine details of measurement which need to be attended to if you want to measure vanishingly small numbers at the edge of the resolution of one's test gear. Error bars may not be fun, but they're essential. The alternative is to do what I did and pick a merely ridiculously low threshold rather than a stunningly low threshold. 😀
Well, peoples, we might be in luck: people have been screaming out for real data about perceived differences in digital sound reproduction depending upon seemingly irrelevant aspects - like how it's played back ... and I've happened to chance upon one, purely by accident ...
My PC, decent commercial level HP, runs simple plastic Harmon Kardon monitors, and I was curious as to how high a treble they could reproduce, and my ears pick up. I knew that I could pick up an 18kHz sine wave at decent levels, so this is what I did:
Using Audacity, generated a 30 sec long, pure 18kHz tone, at 44.1k sampling rate, 32 bits, at 0.8 max level amplitude. To make it easier to pick up something happening I added a volume envelope so that the sound faded in from silence, to max at 15 secs along a linear slope, then, faded away to silence in the same fashion - in other words, the waveform envelope has a diamond shape.
The good news was that I could pick up the tone, but, distortion was coming in - could this be the dreaded S-D noise modulation?? It was, is, relatively low frequency, very similar to the background buzz of a bad line when using an old fashioned telephone. And, it was coming in at a certain point along the volume envelope, and disappearing again at precisely the equivalent point on the attenuating side.
So, was this the monitors distorting on the signal, or the PC electronics? Simple tests implied the monitors were not guilty, so to confirm this, I upsampled to 88.2, 176.4, 352.8 - the speakers were theoretically receiving the same signal, so should distort at the same point ...
And the results were? The precise point where the spurious distortion appeared and disappeared varied depending upon the sampling rate, this was consistent, easily measured by ear alone.
As a first example: for 44.1 rate, 11.5 secs; 176.4, 6.5 secs.
And, two other phenomenon are occurring: in Audacity, if the only difference between A and B playback is altering the project rate - absolutely nothing is done to alter any aspect of the waveform itself - the distortion onset time alters; and, the onset time varies with time - set to continuous repeat, the onset time 'improves', gets later and later over a significant time period.
My PC, decent commercial level HP, runs simple plastic Harmon Kardon monitors, and I was curious as to how high a treble they could reproduce, and my ears pick up. I knew that I could pick up an 18kHz sine wave at decent levels, so this is what I did:
Using Audacity, generated a 30 sec long, pure 18kHz tone, at 44.1k sampling rate, 32 bits, at 0.8 max level amplitude. To make it easier to pick up something happening I added a volume envelope so that the sound faded in from silence, to max at 15 secs along a linear slope, then, faded away to silence in the same fashion - in other words, the waveform envelope has a diamond shape.
The good news was that I could pick up the tone, but, distortion was coming in - could this be the dreaded S-D noise modulation?? It was, is, relatively low frequency, very similar to the background buzz of a bad line when using an old fashioned telephone. And, it was coming in at a certain point along the volume envelope, and disappearing again at precisely the equivalent point on the attenuating side.
So, was this the monitors distorting on the signal, or the PC electronics? Simple tests implied the monitors were not guilty, so to confirm this, I upsampled to 88.2, 176.4, 352.8 - the speakers were theoretically receiving the same signal, so should distort at the same point ...
And the results were? The precise point where the spurious distortion appeared and disappeared varied depending upon the sampling rate, this was consistent, easily measured by ear alone.
As a first example: for 44.1 rate, 11.5 secs; 176.4, 6.5 secs.
And, two other phenomenon are occurring: in Audacity, if the only difference between A and B playback is altering the project rate - absolutely nothing is done to alter any aspect of the waveform itself - the distortion onset time alters; and, the onset time varies with time - set to continuous repeat, the onset time 'improves', gets later and later over a significant time period.
Last edited:
Further, it seems to be sensitive to distant interference as well. Just now I switched on my laptop, from a hibernation state, quite some metres away, but running off a power block on the same power circuit; and this has very significantly increased the distortion period. Which is not what I was expecting ...
This has all the makings of being able to get some interesting, yes(!!), data ...
This has all the makings of being able to get some interesting, yes(!!), data ...
I'm positively sure the Checkered Demon's hammer was Pat-aphysical.
(Mr Bradley, Frank said 'simple plastic Harman Kardon', pat him with your pinot bottle)
(Mr Bradley, Frank said 'simple plastic Harman Kardon', pat him with your pinot bottle)
Yes noticed that. Actually the two-piece JBL Duet is a pretty decent monitor-flanker. That's what I do very casual near-field listening on. For once the industrial design married well with acoustical advantages (dispersion, structural rigidity). If the transducer had a little less IM distortion we'd all be happier, but the bill of materials was about 10 bucks iirc.
The h/k branded entry-levels for Dell, going back about 10 years now, were just atrocious, and nonetheless had (until one brief disaster involving ceramic caps turning into side-melters) the lowest customer complaint incidence of anything Dell bought. I suspect some of that was just low expectations, but they usually worked and didn't break.
I went to see about S. Clay Wilson and learned that he's in pretty sad shape after, apparently, a bad fall or attack. He stood alone among the rather baroque cartoonists I think, and also will be remembered for his statement: Beer is infinite.
The end of the line awaits all of us, a rare few have the visionary talent to predict it.
http://www.primal-urge-magazine.com/Oct - Nov/s clay/Wilson_Demon_DCrumb.jpg
This is the sort of thing that people hear and ascribe all sorts of meaning to. But the standard tests usually done dont show anything or dont explain it. I like this sort of thing.... its fun to figure out.
Thx-RNMarsh
"We fought indians, we fought Dicks! Why remember when Richard Dicks came to town and tried to take over?" (Blazing Saddles)
voltage regulator
After looking at Ed's mic preamp regulator I went back to some old designs to review. In the process I noticed a few effects I'd mostly neglected, particularly on line regulation.
Attached is something suitable for moderate currents and set for a 15V output. The clamping at the pass transistor Q5's gate is primarily for a rough current limit that should allow a fumble with a probe but not much more than that. Device threshold variations will also enter in, so this is not offered as a production-ready design.
Line rejection and output impedance are exemplary for a relatively simple design. The noise will be dominated by the reference, despite the heavy lowpass filtering of R13/C2. A better reference constructed of doubly-cascoded JFETs as a current source to develop 5V across a 665 ohm resistor shunted by 100uF makes the e sub n of Q1 and Q2 the limiting noise source, so lower rbb' parts could help there. But the output noise even with moderate input ripple should be quite low indeed, even with the TL431.
The cascoding with the 4392 parts helps line rejection substantially. Also aiding is the use of a fairly small PDMOS as the pass transistor, with relatively small capacitances. Aluminum polymer output caps are shown for an ESR of about 11mohm, which will become the output impedance at high frequencies if inductances are kept low with a tight layout. I am assuming lots of passive filtering and big bulk caps ahead, with C3 just something local.
After looking at Ed's mic preamp regulator I went back to some old designs to review. In the process I noticed a few effects I'd mostly neglected, particularly on line regulation.
Attached is something suitable for moderate currents and set for a 15V output. The clamping at the pass transistor Q5's gate is primarily for a rough current limit that should allow a fumble with a probe but not much more than that. Device threshold variations will also enter in, so this is not offered as a production-ready design.
Line rejection and output impedance are exemplary for a relatively simple design. The noise will be dominated by the reference, despite the heavy lowpass filtering of R13/C2. A better reference constructed of doubly-cascoded JFETs as a current source to develop 5V across a 665 ohm resistor shunted by 100uF makes the e sub n of Q1 and Q2 the limiting noise source, so lower rbb' parts could help there. But the output noise even with moderate input ripple should be quite low indeed, even with the TL431.
The cascoding with the 4392 parts helps line rejection substantially. Also aiding is the use of a fairly small PDMOS as the pass transistor, with relatively small capacitances. Aluminum polymer output caps are shown for an ESR of about 11mohm, which will become the output impedance at high frequencies if inductances are kept low with a tight layout. I am assuming lots of passive filtering and big bulk caps ahead, with C3 just something local.
Attachments
- Status
- Not open for further replies.