You are right, I was thinking of Radiospares and not RadioShack/Tandy.
A brit here on the board.
Wrinkle
Yup, he said hi a few weeks ago, still thanks me for the 797. I hope to send my stuff Monday, I've got 4 football game's to work over this weekend.
I don't think the ST was designed with any eye to stuff below -110dB. Certainly its internal oscillator was -106dB at best. With passive filtering of the oscillator and an external twin-tee I was able to measure down to -150dB with a spectrum analyser (my most missed piece of test equipment ever, the 3Hz-40MHz HP, all analog, 7nV noise floor). If anyone thinks the AP removes/misses the artifacts, I don't see any other way to rationalize Bruno's and JC's results, they are dreaming.
/OT
I had the oportunity to say Hi to Bruce at the last AES convention, courtesy of his buddy Bob Cordell. Nice guy and a very, very smart designer.
/OT off
I presume you mean not watch them, or do you moonlight? It is a bit of a downer here to setup for a playoff series, go all the way and then watch the team lose. The joke used to be after we did a stadium (ballpark, arena) the team would make the playoffs and then lose. Held up for four projects in a row! Number 5 didn't pay the bill and lost more than won. The sixth team got their rings!
I though the ST was brilliant for the mid 70's! But if you could get -150 out of it that would indicate to me the notch filter/analyzer had to be far superior to the signal generator section.
But I am pretty sure you are aware of what it takes to make a clean sine wave!
Yes an interesting issue is if we are so reliant on FFT use, is there some area of errors that it misses?
in principle a fft's complex points exactly represent the input digital data's "information content" as they are mathematical duals
you have to meet the assumptions for digitizing in the 1st place
bandwidth limitation to < Fs/2
sufficient ADC bit depth, S/N, linearity to capture the signal information you are looking for
windowing and record length affect resolution too, you often can just increase record length a few times over the "theoretical minimum" to get the required resolution
there are also finite word length, truncation issues so you may need to keep the fft results in huge bit depth to be able do a "bit perfect" inversion to recover the original time series data
but subject to well understood limitations the fft does contain "the same" information - but in a format that allows us to pick up on certain features much more readily than the time domain response would
a particular point I keep returning to is that the Otala' "TIM" IMD products are perfectly "visible" in a fft - the 2 tone a*F1 +/- b*F2 (a,b integers) IMD frequency products can in principle be resolved by the phase information in the complex valued fft to distinguish "AM" vs "FM" products - so this distortion is Not "missed" in fft - even without the phase information you still see the IMD frequency components in the |fft|, just summed with any other other distortion mechanism's distortion product at those frequencies - no |fft| components above the noise floor at the IMD frequencies means No TIM
Last edited:
To my mind its daft to test cables with an AP in isolation if the aim is to try to uncover audible differences. That's because the AP is careful not to corrupt its measurement with common-mode effects (transformer isolated source), and audible cable differences in real-world systems do seem to be the result of these same CM effects that the AP takes pains to avoid.
As jcx points out, the time domain and freq domain are duals, provided we accept that the FFT's output is complex. However the FFT is normally presented as magnitude only, so some information is missing. The windowing function employed trades off between dynamic range and frequency resolution, so there's also the potential to miss something there.
The trend towards greater and greater frequency resolution in FFTs (and more and more time averaging) means time resolution suffers. So noise could be coming and going within the time window of interest (audible noise modulation) and this would be totally missed.
The aim isn't to uncover audible differences from known and well established effects. It's to look for evidence of the non-established effects, such as those claimed in the marketing literature of cable manufacturers.
se
Not quite sure what this means, computing power has advanced so far beyond the audio problem that this is sort of meaningless. You can pose any problem and constrains that you want. You could take the FFT of an entire CD easily now (lots of frequency resolution) not sure of the point though
Last edited:
I am not sure about this. My measurements were done without transformer coupling, and I have not measured any 'cable distortion' for 150dB resolution.
To me, this ('cable harmonic distortion')is completely wrong track. Why do not we rather speak about EMI/RFI pickup into various cables, loop induced interference voltage - dependent on audio instruments and audio system configuration. It is easy to measure these issues, and any system gives different results. This is not discussed, but strange 'distortions', non-existent or being somewhere at -300dB, are discussed.
Well I'm not one to get hung up over syntactics. So what would you like to call the effects which manifest as a result of common-mode loops if not 'cable effects' ? They are generally (though not exclusively) an effect plaguing single ended (aka unbalanced) connections, so how about 'unbalanced interconnection effects' ?
PMA said:
That was the general direction I was heading, yes. Agree - they're an artifact of the whole system (emergent?) rather than something pertaining to the cable per se.
scott wurcer said:
Perhaps if I put it into another field. Does the plethora of computing power now available render security issues in computing moot? After all practically any CPU worth its salt now won't flinch in doing a Diffie-Hellman 2048 bit public key exchange.
Last edited:
It isn't clear to me what these graphs are supposed to show. If this suggests the cable's response including spurious generated noise, it is obviously flawed data. I can't speak for the other cables but insofar as the RS cable is concerned, the apparent falloff of output starting at around 19 khz and falling at around 10db per decade can't be right. The proof was the insertion of the cable in the video signal path between a VCR output and a high quality NTSC TV input. As I stated, I compared the TV set's own tuner on the same channel as the VCR's tuner using a cable feed. This test was similar to a shunt test. Not only would there have been serious degradation of the video signal quality by the RS cable, the color burst signal wouldn't even have shown up, it would have been too degraded and weak to have worked at all. You can easily verify this yourself by repeating my test.
Also, the 15db spike at 16 khz and the noise between 1 and 2 Khz would be clearly audible if the cable were used between a turntable output and the high gain magnetic phono input when a record wasn't being played. In addition to the general signal boost of the magnetic phono preamp to bring it up to line level, there is a 15 to 20 db additional boost at 16 khz as the result of RIAA equalization.
The similarity of the three test results in noise between 1 and 2 khz, the 16 khz spike, and the high end falloff strongly suggests that the equipment or test design is defective.
Last edited:
- Status
- Not open for further replies.