Daniel,
You seem to be coming over to the dark side. Remember that when one measures no change to a circuit under test there cannot be any perceived change, even when one is looking in the wrong place, for the wrong signal or my favorite when the data presented can be easily misinterpreted.
The best example is the higher frequency power supply harmonics, the ones between 3,000 to 6,000 Hz. If one just uses an oscilloscope these will probably not synchronize with the sweep generator and will be lost as "noise." But with a spectrum analyzer one may note the harmonics raise the noise floor in several critical bands and result in a perceived loss of midrange clarity.
Now when you just use a spectrum analyzer, you might have an oscillation way above the audio band that it misses. Now the oscillation may seem to increase inband noise which the FFT may catch depending on windowing, span etc. Of course a wideband oscilloscope will catch that.
I will only briefly mention things that common tests might miss like thermal modulated distortion, vibration induced voltages or even long term or periodic power line issues.
So as a potential darksider, you might want to listen and then when you hear a difference use the test equipment to try and find out what is actually different.
Mark,
Seems you might be a candidate also. Spending three dollars on an issue that is easily measured but some believe you can't perceive! I do recall when you mentioned six layer PC cards were the way to go.. didn't want to burst your bubble and suggest odd numbers of layers sound better.
BTY many years ago Digikey had a new vendor of 1n400x rectifiers. It seems these diodes had enough switch off noise just about everybody noticed even digital designers. The manufacturer pointed out that turn off was not specified for that model. They did however fix it to be about the same as other folks. I used to have a bunch to demonstrate the bad example. Unfortunately the last time I looked for them they were missing. Hopefully they didn't get used!!
JN,
My magic cable just came in, should test it tomorrow.
ES
Last edited:
And if you are playing music at normal listening levels at the same time you won't see anything in either case. Not saying there can't be PS problems but no need to totally ignore masking. Did you ever do the "Sousa" test?
When I had a super-cost-challenged powered speaker to design for Harman, I was given the idea by a seasoned professional, when I mentioned my concern about switching noise and other lower-frequency noise entering the signal chain due to the proximity of the a.c. input switching on the potentiometer to the pot elements, to switch the ground side of the line between the main bulk cap and the rectifier bridge. As they say, seemed like a good idea at the time (the originator has since asked me not to reveal his identity).
Now I was concerned that a long loop would still occasion the detection of reverse-recovery spikes, which could be readily demodulated by the bipolar-input op amp used for some minimal gain and EQ. So I placed a 10uF 'lytic and a 100nF ceramic local to the rectifier output, to partially snub things a bit.
To spend a penny on a ceramic cap and another penny on the electrolytic was not insignificant in the multi-million volumes anticipated, but as I knew 1N400X diodes were not all created equal, I felt it justifiable insurance were we to encounter a particularly messy batch of them.
The disaster that ensued required two factors, and in one terrible case three: (1) the manufacturer of the boards substituted a higher-voltage electrolytic than I had specified, (2) the ceramic cap vendor experienced a yield problem from shoddy dielectric material and turned off their tester, allowing some very-low-voltage parts to be installed, and (3) in one case the thermal fusing of the wall-wart adapter also failed. In addition the vendor substituted a leaded 100nF part (the design used an SMD part) and located the leaded part almost touching the sidewall of the plastic enclosure.
The perfect storms occurred when a user would turn the speaker power off with the potentiometer switch, rather than at a plugstrip. The overall circuit behavior became mains voltage doubling (!) coupled across to the 100nF and 10uF parts, via the interwinding capacitance of the adapter. Given sufficient time, and with no other discharge path other than capacitor leakage current, the voltage rose indefinitely. What might have been held down by 'lytic leakage was now occurring at a considerably higher voltage, enough to turn some of the defective 100nF caps into little ceramic heaters, and in one case (with the defective adapter) enough to produce a flame which one girl in a dormitory observed emerging from the side of her speaker. Had there been flammable material in the vicinity a fire could well have resulted, with possible injuries and deaths.
The additional irony was that the remedy for the batch of speakers thus afflicted was to snip out the 100nF. My boss had to explain to upper management how I could have included a part that was dispensable. I was not a fly on the wall but I would have liked to have heard the explanation, which properly would have included how diodes could generate noise.
Scott,
As usual...
Some of us use radio tuners, microphone or even phonograph preamps.
The Sousa test has rather attenuated high frequencies and always maintains a 30 dB margin. Not the case for much audio gear.
JN
Attached is the TDR using the same technique as before. The data sheet does not give either value but if I measure it I'll PM you as this is intended to give me a market advantage.
Attachments
Yes, it should be assumed the test in question be done carefully, and at high level. Trick is, a near max-power imd test should show that incursion and fold back into the audio spectrum. And that's before sweating whether it's truly perceptual.
Whichever side I'm on, it's the one interested in the truth, as best we can discern. Yes, there's a good amount to learn about how we listen, but a first run of a distortion stress test to sniff out spurs before we go into further diagnostics.
Derfy,
There is no truth only perception. Those without test equipment do their best with ears only. With test equipment you can chase better numbers. Now harmonic distortion may not be perceptable at .01% 2nd harmonic. But a unit that measures a hundred times better may also have better performance on parameters that do affect perception.
The best example is record vs CD. CDs were developed with precise measurements. Records had their standards evolve over time by listening preferences. Yes there are measurements involved but the development was perception based. Now even a mediocre CD measures one hundred times better than a vinyl record. But if you compare the two accurately perception is not that far apart.
Then there is the staring assumption issue. Why look at high power results when it could be a low power issue.
https://www.fairchildsemi.com/techn...covery-and-Its-Effect-on-Switching-Losses.pdf
High power = better stress test to see PSU-related (err... most anything related) defects. And, yes, I realize this isn't exactly 50/60 Hz.
My point was that we should not lose track of the ostensible method of "fix first what is broken or inferior". I.e. check for supply incursion, if none, go worry about something else.
Distortion tests/FFTs are quite useful diagnostics, not exclusively for marketing material.
Last edited:
Ed maybe you misunderstood I meant Bill Waslow's Sousa at 60dB below the program material. It was inaudible to at least two of our esteemed members and I find it hard to believe line harmonics at -60dB are. The threshold of audibility of a 3kHz tone in the presence of nothing but a 30Hz tone to me at least is meaningless. Oh sorry I forgot, this line of discussion is pointless.
Scott,
I think it was closer to 50dB down, being the amount it was attenuated before mixing in. As far as I know no one has been able to hear it -- there is a set of 7 file pairs, each having a "with marching band" and a "without marching band" version, people who thought they could hear it were asked to listen to all pairs and say which had the Sousa band. They could verify it was indeed there by using my DiffMaker program to subtract one from the other -- the intent of the test was originally to counter claims by people that their hearing was "much more sensitive" to hearing differences than any computer math might be able to find. I got only a couple of responses from listeners, and of those neither did better than 4 out of 7 (the listeners who voted didn't claim to even hear the band btw). The test is still up on my website at
Audio DiffMaker example files,
but don't vote now (I don't have the list of which was which and have no way to know anymore). The Diffmaker program is at
Audio DiffMaker
I don't know whether it still works on more recent versions of Windows, it's been a long time now.
I think it was closer to 50dB down, being the amount it was attenuated before mixing in. As far as I know no one has been able to hear it -- there is a set of 7 file pairs, each having a "with marching band" and a "without marching band" version, people who thought they could hear it were asked to listen to all pairs and say which had the Sousa band. They could verify it was indeed there by using my DiffMaker program to subtract one from the other -- the intent of the test was originally to counter claims by people that their hearing was "much more sensitive" to hearing differences than any computer math might be able to find. I got only a couple of responses from listeners, and of those neither did better than 4 out of 7 (the listeners who voted didn't claim to even hear the band btw). The test is still up on my website at
Audio DiffMaker example files,
but don't vote now (I don't have the list of which was which and have no way to know anymore). The Diffmaker program is at
Audio DiffMaker
I don't know whether it still works on more recent versions of Windows, it's been a long time now.
I thought is was 30. At -60 it will take quite a bit to pop up. But with selected frequencies it is possible. I also recall the Sousa piece had attenuated HF.
Yes often our attempts to communicate are unsuccessful, but still interesting.
Did you hear the one about Santa Claus, the totally honest lawyer and honest politician?
The cheapest SA is a PC with soundcard and software RealTimeAnalyzer (ARTA or similar).
The Rigol SA costs about 1500€ and is worth any cent of it.
It covers 1.5GHz - fine for em-pre-compliance measurements.
I love it.
But it is not suitable for measurements in the audio band.
- Status
- Not open for further replies.