Had some fun playing with cable and put the results up here- Interconnects
Curious what people think. Misguided bunk or a clue?
Curious what people think. Misguided bunk or a clue?
Interesting stuff... i did not thought mutch about dielectric absorption before.
As i understand you have only connected one end of the cable to one of the two differential inputs? The second end was floating?
The value of your mica cap must be 2x greater, as the difference voltage is too high for 270p?
Thx - Udo
As i understand you have only connected one end of the cable to one of the two differential inputs? The second end was floating?
The value of your mica cap must be 2x greater, as the difference voltage is too high for 270p?
Thx - Udo
As I expected, really good questions!
udok- There are various things I didn't include, like the lengths of cable and standard measured parameters. I have a spreadsheet for those. The amplitude of the square wave is proportional to capacitance, or at least close to it, so a 270 pF cap and 255 pF cable will produce similar amplitudes. Unless I inadvertently changed a scope setting, which I hope I didn't! Most of the cables I tested (10) were about 40" long, but capacitance was all over the map. Yes, one end was unterminated, which is similar to the usual hifi practice of low drive impedance and high driven impedance.
SY- I didn't measure inductance but the closely spaced conductors of a cable won't have much. I came to dielectric absorbtion by elimination of everything else, dissipation factor mostly, and the time constant of the droop can't be accounted for (AFAICT) by the values of any RLC associated with the cables. I think the timebase was 0.1 mS/div, so the droop is a pretty slow effect. Any "simple" DA measurement I would do would be near DC charge, short and measure affair, using an electrometer, and that's not the frequencies of interest. Is there another way? Finally, inductance is magnetic and not influenced by dielectrics, so cables of near identical geometry wouldn't have wildly different waveforms if that were the cause.
AndrewT- Yes, the triangle to square is a bit of a handy trick. The best way to see it is to draw two triangle waves, one on top of the other, but offset in phase slightly. You'll see the rising portions are parallel, as are the falling portions. That difference is what's amplified and being constant for the half cycle, gives the flat top and bottom of the square wave.
This test is very close to my first love, bridge measurements. I suspect the exact same results could be had simply by making a standard bridge measurements, but exciting it with the triangle and using a wideband detector rather than a tuned detector. The use of tuned detectors with bridges eliminates depth-of-null problems caused by capacitor distortion and dielectric absorption! No reason to trouble people with those pesky imperfections. In all my GR literature, and it's a lot, I don't remember ever seeing DA mentioned once. My guess is they're interrelated and even attached to dissipation in some way, but that's physics way beyond my pay grade.
The one thing that bothers me is the error is numerically on the border of where I would have guessed it to be audible. It does, however, seem to get far worse as the frequency rises, so maybe that's why the sonic differences are mostly transients and things like cymbals and the noise-like portions of Dire Straits Industrial Disease.
udok- There are various things I didn't include, like the lengths of cable and standard measured parameters. I have a spreadsheet for those. The amplitude of the square wave is proportional to capacitance, or at least close to it, so a 270 pF cap and 255 pF cable will produce similar amplitudes. Unless I inadvertently changed a scope setting, which I hope I didn't! Most of the cables I tested (10) were about 40" long, but capacitance was all over the map. Yes, one end was unterminated, which is similar to the usual hifi practice of low drive impedance and high driven impedance.
SY- I didn't measure inductance but the closely spaced conductors of a cable won't have much. I came to dielectric absorbtion by elimination of everything else, dissipation factor mostly, and the time constant of the droop can't be accounted for (AFAICT) by the values of any RLC associated with the cables. I think the timebase was 0.1 mS/div, so the droop is a pretty slow effect. Any "simple" DA measurement I would do would be near DC charge, short and measure affair, using an electrometer, and that's not the frequencies of interest. Is there another way? Finally, inductance is magnetic and not influenced by dielectrics, so cables of near identical geometry wouldn't have wildly different waveforms if that were the cause.
AndrewT- Yes, the triangle to square is a bit of a handy trick. The best way to see it is to draw two triangle waves, one on top of the other, but offset in phase slightly. You'll see the rising portions are parallel, as are the falling portions. That difference is what's amplified and being constant for the half cycle, gives the flat top and bottom of the square wave.
This test is very close to my first love, bridge measurements. I suspect the exact same results could be had simply by making a standard bridge measurements, but exciting it with the triangle and using a wideband detector rather than a tuned detector. The use of tuned detectors with bridges eliminates depth-of-null problems caused by capacitor distortion and dielectric absorption! No reason to trouble people with those pesky imperfections. In all my GR literature, and it's a lot, I don't remember ever seeing DA mentioned once. My guess is they're interrelated and even attached to dissipation in some way, but that's physics way beyond my pay grade.
The one thing that bothers me is the error is numerically on the border of where I would have guessed it to be audible. It does, however, seem to get far worse as the frequency rises, so maybe that's why the sonic differences are mostly transients and things like cymbals and the noise-like portions of Dire Straits Industrial Disease.
Last edited:
Amazing. THANK YOU!
I was suprised to see Belden 8420 to be so bad. I was going to try that out if I went with XLR and a Balanced setup.
Its great to see RG-62 perform so well. Something which I've used myself is high quality RG6 Quad Shield with high quality crimp-on BNC connectors. I've settled on that for all of my stuff.
And I'm going to be using twisted pair inside of my equipment!! 2 for 2!
http://www.diyaudio.com/forums/anal...ople-using-bnc-out-there-rg6-quad-shield.html
&
http://www.diyaudio.com/forums/lounge/258647-about-quality-connectors-audio-equipment-6.html
I was suprised to see Belden 8420 to be so bad. I was going to try that out if I went with XLR and a Balanced setup.
Its great to see RG-62 perform so well. Something which I've used myself is high quality RG6 Quad Shield with high quality crimp-on BNC connectors. I've settled on that for all of my stuff.
And I'm going to be using twisted pair inside of my equipment!! 2 for 2!
http://www.diyaudio.com/forums/anal...ople-using-bnc-out-there-rg6-quad-shield.html
&
http://www.diyaudio.com/forums/lounge/258647-about-quality-connectors-audio-equipment-6.html
Last edited:
Thanks, Conrad. My suspicions are aroused because of your silver mica results- that dielectric has very high DA, higher than nearly any polymer, so if that's what you're actually measuring, it ought to show up here. Any ideas?
I think I mentioned that, as mica caps are the example always used for very low DF with surprisingly high DA. It's true, but that "high" DA is in comparison with better dielectrics and my guess is it isn't even on the map compared to the stuff they make cables out of, with the exception of Teflon. Or, maybe it only shows up in a different frequency range. I don't know if DA is entirely predictable once you have a single point measured. When I get back to the scope I'm going to add an air variable cap to the other input. Then I can null the square wave, increase the gain, and see much more subtle differences. My guess is I'll be able to see the DA of the mica cap then. I need to find a similar value NP0 to compare- I have about 8 billion components here, but never the right values! Have you ever seen a 68000 pF silver-mica? I have one. 270 pF or similar NP0? Not a one!
The reason I like differential measurements so much is they show errors regardless of cause, then the fun is explaining them.
The reason I like differential measurements so much is they show errors regardless of cause, then the fun is explaining them.
Last edited:
I agree with you, that the effect is very small. About 5 mV droop to 24 Vpp = 0.02%.
Most people can not hear such low level differences.
But perhaps the spectrum does change more dramatically...
The effect is mostly in the fast rising edge of the signalgenerator.
May it be that the droop is even induced by the fast rising edge?
Most people can not hear such low level differences.
But perhaps the spectrum does change more dramatically...
The effect is mostly in the fast rising edge of the signalgenerator.
May it be that the droop is even induced by the fast rising edge?
Remember, no edge here! The input signal is a 2 kHz triangle wave, not square, and the edges you see on the scope are the result of the triangle switching direction. The droop is a relatively slow thing in comparison.
I know for certain my ears are not golden and I don't know about hearing .08% of anything (there's no 24Vpp, the input is 12Vpp), so it either gets much worse at higher frequencies or I'm not looking at the math right. I may not be choosing the cutoff point of the droop correctly on the relatively fuzzy scope image.
If memory serves (I don't have my tables with me), the worse DA among polymers is from PVC, and it's still 5 times lower than mica. That is going off memory, though, and I'm old.
I tend to believe that whatever changes to the signal you cannot hear are multiplied by whatever other chances you cannot hear in every other component in your system to the point where you hear a definite change from what you /should/ be hearing.
Everyone knows what a Trumpet sounds like. now put that through a hundred components and then try to make it sound the same.
https://www.youtube.com/watch?v=d70fiI2Mn_A
Last edited:
I've fooled with that in the past and don't remember much, but it's a very good idea. This is stuff that really needs to be crosschecked using different methods.
BTW, anybody with a scope can, and should, duplicate this test. Almost all two channel scopes will let you invert one channel and then sum them (A+B mode), giving you a differential amplifier. Find a couple matched resistors, a function generator and Bob's your uncle. You won't have the gain available on my diff plug-in, but the signals here should be quite visible with normal 5 mV/div settings.
yes, use any polymer film dielectric before Mica https://web.archive.org/web/20051226052157/http://www.national.com/rap/Application/0,1570,28,00.html
and some rambling What's all this soakage stuff, anyhow? | Analog content from Electronic Design
and some rambling What's all this soakage stuff, anyhow? | Analog content from Electronic Design
Yes, but not in the usual AC sine wave sense. Think more in terms of charging the cap with constant dV/dT (triangle) through a resistor, rather than a sine function.
Kastor- yes, I tested some Teflon aircraft coax. Fantastic stuff on the waveform, also has a dissipation factor of 0.00002, though moderately high capacitance.
I have to reread Pease, but actually don't agree with never using mica. I've used it in RIAA sections with no sonic problems I could detect. Mylar is a far worse performer for distortion and I avoid it when possible.
An interesting thing about mica caps is that I've never heard of them being a problem in any AC circuit, only with slow speed stuff like integrators. Interesting that they also look good in my test. Does anybody know anything about dielectric absorption frequency ranges? Does performance in a 1 second test predict high frequency behavior? This seems like stuff we need to know to predict audio performance. Later today I'll try to run some traditional DA tests on the same cables.
Now, I think it's Rod Elliott that makes a good argument for dielectric absorption not being audible or a factor in audio. I respect him greatly and he makes a good argument, but I also think circuit locations and impedances are everything when it comes to what's important.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.