Just for fun, have made Victor,s 1kHz oscillator shielded with Cu-foil and measured again.
Drift is still present, but the whole histogram become more symmetrically.
Sure, change is quite small and probably not relevant for audio use, but is definite measurable,
and interesting to compare with unshielded measurement.
https://www.diyaudio.com/forums/attachment.php?attachmentid=773551&stc=1&d=1565287909
Drift is still present, but the whole histogram become more symmetrically.
Sure, change is quite small and probably not relevant for audio use, but is definite measurable,
and interesting to compare with unshielded measurement.
https://www.diyaudio.com/forums/attachment.php?attachmentid=773551&stc=1&d=1565287909
Attachments
Not speaking for Mr. Cordell...
If you consider "phase noise" to be frequency modulation of a tone (phase modulation would be similar), then the bandwidth of the harmonics would be that of the fundamental times the order of the harmonic. (Second harmonic being 2X, etc.)
And, if you go to the next thought, phase modulation of the conversion clock in a mixed signal system - like an audio DAC - effectively phase modulates all the tones at the output of the device. After all, a DAC is essentially a frequency multiplier aka mixer (in the RF sense, not the pro audio engineer sense).
Taking the next step with regard to an audio system, that implies that the musical information is phase modulated. Isn't this like moving a loudspeaker back and forth relative to the listener?
How much of this might audible and in what way is a really good question. After all, people are constantly moving their heads when listening to music, whether they want to or not. From what I've read, subtle head movements are part of the brain's overall system to localize sound.
Getting back to testing and audio oscillators, I think this is why using averaging to test a device can be misleading. Yeah, it allows you to dig deeper into the noise floor because while the signal itself is assumed to be constant over a lot of sweeps that allow for the averaging, the noise is assumed not to be. Thus, the noise largely gets averaged down.
However, if the noise is really chaotic with a small number of events that are really large and the rest very small, the large events get averaged down to almost nothing, too. But, they could be really significant in terms of listening. (This kind of thing is a consideration in measuring communications systems with a spectrum analyzer and internal demodulator. One symbol, which represents a number of bits, might be awful, but when averaged in with several hundred others, the overall effect is barely measurable in terms of dB. Those bits are still lost.) This may be a real consideration in terms of your observation about the size of FFT bins and the effect this has on measurements. Dunno, personally.
Taking the next step with regard to an audio system, that implies that the musical information is phase modulated. Isn't this like moving a loudspeaker back and forth relative to the listener?
How much of this might audible and in what way is a really good question. After all, people are constantly moving their heads when listening to music, whether they want to or not. From what I've read, subtle head movements are part of the brain's overall system to localize sound.
Getting back to testing and audio oscillators, I think this is why using averaging to test a device can be misleading. Yeah, it allows you to dig deeper into the noise floor because while the signal itself is assumed to be constant over a lot of sweeps that allow for the averaging, the noise is assumed not to be. Thus, the noise largely gets averaged down.
However, if the noise is really chaotic with a small number of events that are really large and the rest very small, the large events get averaged down to almost nothing, too. But, they could be really significant in terms of listening. (This kind of thing is a consideration in measuring communications systems with a spectrum analyzer and internal demodulator. One symbol, which represents a number of bits, might be awful, but when averaged in with several hundred others, the overall effect is barely measurable in terms of dB. Those bits are still lost.) This may be a real consideration in terms of your observation about the size of FFT bins and the effect this has on measurements. Dunno, personally.
Very interesting!
Any idea how much of this was thermal in nature? By thermal, I mean small changes caused by air movement. It might not be electrical shielding making the difference, but rather control of the air currents across the components. Or, it really could be electrical shielding. Or, both. Or, keeping the leprechauns out.
Indicates HUM pickup do not shielding...😀
Hmm, looks like hum modulation, but side freq. are not hum freq. related..
I learned from the victor's LDO:
1. has to be boxed completely 😱
2. the used LME op-amp may have some pop noise, do the substrate defect,
as my first op-map had this issue
Look at my pictures... 😀
BTW: connected the LDO as balanced
Hp
Attachments
may the new TO op-amp OPA1656 as a replacement or better choice ??
OPA1656: High-Performance CMOS Audio Op Amp
OPA1656: High-Performance CMOS Audio Op Amp
This brings up (the admittedly off topic topic of) PCB layout and design, and I'm convinced (from my informal studies of human factors) that one should make EVERY effort to make component placement and orientation consistent. Have all BJTs facing the same way, have pin 1 of all chips in the same direction, etc. The only exceptions would involve microwave frequencies, where an extra half inch of trace length really makes a difference.
#off-topic
Nobody is perfect. I know what can happen in a "hand picked" factory, despite a silkscreen.
(In a factory of Text Lite Ltd, Limerick, Ireland. Women with a grade in literature or so were working there)
Nobody is perfect. I know what can happen in a "hand picked" factory, despite a silkscreen.
(In a factory of Text Lite Ltd, Limerick, Ireland. Women with a grade in literature or so were working there)
Last edited:
Thanks for comments
I haven't done any further investigation, but change is clearly visible and repeatable, no doubt about that.
Just es i said before, the whole change is on extremely low scale and probably of no real meaning for audio purpose.
I am more concerned for oscillator stability and accuracy - every time when is it switched on, about 20min. is needed till mean frequency is achieved, but this frequency is always different - a few Hz more or less than 1kHz.
The question is: if that few Hz have impact et all on THD measurements or not?
All that what i have said, by any mean does not imply that oscillator have some strange behavior or error.
I have only look at it from "high frequency" perspective, once more-is that important for audio- i simply don't know?
The oscillator was already in metal box together with 4x9V battery, i have made only extra tight shielding of circuit board itself.
I haven't done any further investigation, but change is clearly visible and repeatable, no doubt about that.
Just es i said before, the whole change is on extremely low scale and probably of no real meaning for audio purpose.
I am more concerned for oscillator stability and accuracy - every time when is it switched on, about 20min. is needed till mean frequency is achieved, but this frequency is always different - a few Hz more or less than 1kHz.
The question is: if that few Hz have impact et all on THD measurements or not?
All that what i have said, by any mean does not imply that oscillator have some strange behavior or error.
I have only look at it from "high frequency" perspective, once more-is that important for audio- i simply don't know?
I know a friend who is an audio designer/manufacturer, he does that too. I can't accept that because it inevitably will compromise performance.
Jan
#off-topc
(BTW, I was talking about non-audio applications like the PX-1000)
Cheers, E.
There are always exceptions, but, as you know, they also prove the rule. 😉
(BTW, I was talking about non-audio applications like the PX-1000)
Cheers, E.
Yes. The oscillator frequency drifts for app +0.3% within this time. Heat from the LME49720 and the onboard shunt regulators warms up the polypropylene caps. These caps can be replaced with the polystyrene caps. Then the drift will be app. two times lower. Also the distortions may drop in this case, but this can be seemed only in lower than -150dB region.
Thanks for explanation