The only time you'll see a tiny bit of hum from the SYS-2722 analyzer is when measuring the noise floor of very low noise sources using high-res FFTs. I don't recall exact figures, but I think for the unit I'm using the spikes are well below -140 dBu (more like -160/-180 dBu). Unlikely of relevance for THD/THD+N measurements.
Samuel
Samuel
The 60Hz spikes on mine are below -140dBu, but the noise floor seems only a little lower than -155 dBu
Just checked, 150 Hz is at -146 dBu for both channels. The left channel (closer to the power transformer) shows additional weak signals probably related to mains, so for critical work it might pay off using the right channel.
The -160/-180 dBu I mentioned before is clearly not what it is, I probably remembered these numbers from THD ratio measurements--for a +20 dBu signal -146 dBu is at -166 dB. Yet, hum from the analyzer should not be an issue unless you're working with input signals well below 0 dBu.
This depends on the FFT settings (sampling rate, resolution, window), as the FFT is not scaled for noise. For the measurements I've just done it was around -163 dBu.
Samuel
The -160/-180 dBu I mentioned before is clearly not what it is, I probably remembered these numbers from THD ratio measurements--for a +20 dBu signal -146 dBu is at -166 dB. Yet, hum from the analyzer should not be an issue unless you're working with input signals well below 0 dBu.
This depends on the FFT settings (sampling rate, resolution, window), as the FFT is not scaled for noise. For the measurements I've just done it was around -163 dBu.
Samuel
The issue is becoming weird for me, atm. I tried to see what hum component look like on my 2322 and only found that mentioned same -- and irrelevant -- little 150Hz spike, on the left channel, right channel is clean**), just like Samuel says. But I definitely measured much higher -- but still inside spec -- components (with inputs shorted) on both channels a year ago when I got it, and I personally witnessed the noise levels and the improvement when my colleague did the transformer transplant about 2 1/2 years back. Our lab is in an electrically very noise industrial zone, railroad tracks running along right behind the building, power plant nearby, transformers of various equipment starting/stopping to hum at will (dynamic DC-offsets, probably)... so mabye we were looking at the effect of highly disturbed mains at levels the toroid couldn't handle... will check what the 2722 looks like these days, next...
**) We always see a (non-mains) component at around 80Hz, level depending on time of day (much better in the evenings) which I have not been able to identify or remove yet.
Whish we had a pro mains synthesizer to see if and how much susceptible the AP power supply really is.
**) We always see a (non-mains) component at around 80Hz, level depending on time of day (much better in the evenings) which I have not been able to identify or remove yet.
Whish we had a pro mains synthesizer to see if and how much susceptible the AP power supply really is.
Last edited:
The ac power mains could very well be where the noise is coming in.... the spike.... I routinely apply ac line filtering and ground isolation transformer to test gear (Topaz Ultra-Isolator). Sometimes a quality line regulator as well depending on test/load. Its a habit just from such T&M issues showing up with data in the past.
Thx-RNMarsh
Thx-RNMarsh
FINALLY! I installed both IM boards yesterday afternoon. Don'task. I am sooo far behind.... Multitasking like crazy to keep up.
Thx-RNMarsh
Presumably this will fix the #1 priority interrupt that Dolby circuits (a $4-5K option) would try and never let me do what I wanted to do... now they are removed. Part of installing the new IM boards -- remove Dolby.
-RNM
Thanks to Samuel, I have the A-P running on my WIN8 system
See line # 3253 at Low Distortion Audio Range Oscillator thread.
In a self-test I see -145dB at 2H and 3H on the A-P 2722. This was checked against the ShibaSoku AD725D and they are within 1dB of each other down to at least -130dB... I haven't checked at lower harmonic levels, yet. Expecting close to -150dB, I expect.
Thx-Richard Marsh
See line # 3253 at Low Distortion Audio Range Oscillator thread.
In a self-test I see -145dB at 2H and 3H on the A-P 2722. This was checked against the ShibaSoku AD725D and they are within 1dB of each other down to at least -130dB... I haven't checked at lower harmonic levels, yet. Expecting close to -150dB, I expect.
Thx-Richard Marsh
correct configuration needed...
Old thread, but I was wondering if anyone can point to more information on this. What is the correct configuration, but also if indeed - as RNMarsh points out - the DSP option adds noise (?):
Not sure I understand the whole thing.
Thanks much in advance!
Old thread, but I was wondering if anyone can point to more information on this. What is the correct configuration, but also if indeed - as RNMarsh points out - the DSP option adds noise (?):
Not sure I understand the whole thing.
Thanks much in advance!
You don't need a dual domain system, as you are not using the digital i/o. An analog+DSP system is sufficient.
Groner is right on the dot here. I don't know how the various settings are done in an S1 (I have a SYS21722) but there's two area's where you need to set it up for max ADC performance. On the analog side, make sure that you don't overload the ADC input. Normally the internal autoranger takes care of that, but you may be able to hit the sweet spot with a manual range setting*.
Secondly, you need to set up the parameters for the FFT. Make sure that the sample rate, FFT length and # of samples are matched. If you are using the AP generator, you can set the Window to None or Equiripple (not sure if AP's equiripple window is available in the S1).
Groner has an online article about measuring a whole series of opamps and iirc he discusses the AP setup he used. Anyway, read anything from Groner you can find, he knows his stuff!
Jan
*using an passive notch at the AP input means that the dynamic range between the levels of the fundamental and the harmonics are decreased by the notch dept. That eases the job of the ADC considerably; it allows me to routinely measure to below -150dB.
Groner is right on the dot here. I don't know how the various settings are done in an S1 (I have a SYS21722) but there's two area's where you need to set it up for max ADC performance. On the analog side, make sure that you don't overload the ADC input. Normally the internal autoranger takes care of that, but you may be able to hit the sweet spot with a manual range setting*.
Secondly, you need to set up the parameters for the FFT. Make sure that the sample rate, FFT length and # of samples are matched. If you are using the AP generator, you can set the Window to None or Equiripple (not sure if AP's equiripple window is available in the S1).
Groner has an online article about measuring a whole series of opamps and iirc he discusses the AP setup he used. Anyway, read anything from Groner you can find, he knows his stuff!
Jan
*using an passive notch at the AP input means that the dynamic range between the levels of the fundamental and the harmonics are decreased by the notch dept. That eases the job of the ADC considerably; it allows me to routinely measure to below -150dB.
If you concern yourself with a single number THD+N measurement, then yes, adding an A/D converter to the analyzer residual must increase the noise, and so the THD+N number will be worse.
The good part is that there is no reason to use THD+N as a distortion metric when you have a DSP analyzer that can show you the actual level of each harmonic in isolation. Even with amplifiers that I feel have too much distortion to be considered "clean", their THD+N results will be dominated by noise, so a single THD+N used to rank amplifiers will sort amplifiers only based upon their noise levels, and will have nothing to do with their relative nonlinearity. Yes, in that undesirable situation, adding DSP to the residual will increase analyzer's effective noise and reduce the sensitivity, but again, you're performing a noise test here, and actual nonlinearities are completely obscured by noise, even in some audibly flawed amplifiers.
How is this FFT 'noise reduction' possible? Like the ear, an FFT separates the energy of the analyzer residual into many separate frequency bins and calculates the level in each bin. A pure tone, such as a distortion harmonic, has a definite frequency, so its energy will always fall into the same FFT bin based on its frequency, so its level will be essentially the same after an FFT. Noise however, which is best thought of as a density over frequency, will have its energy spread among many frequency bins, so the noise level in each bin will be less than the total noise energy as seen in a THD+N number.
Since a distortion harmonic's energy will land in the same FFT bin regardless of how many FFT bins are available, by using more FFT bins (a larger number of samples for the FFT), you can spread the noise among more bins, while retaining an accurate level for each harmonic. You get noise reduction of sorts, since noise that is not right on top of a harmonic will be spread away from the one bin that houses the energy from that distortion harmonic.
So, as Jan mentioned, having an analog analyzer with DSP will indeed allow you to have far higher sensitivity to distortion, since the broadband noise will be peeled away from the nonlinearities by the FFT process.
I recently made this point graphically about noise v. signals in this thread: https://www.diyaudio.com/forums/software-tools/346959-noise-fft-condundrum.htm
AP has a macro "FFT_Scaling.apb" which uses the windowing filter and sample size to accurately chart noise. The paper which goes with it explains in detail.
The analog analyzer in the SYS2722 still seems to have a lower noise figure than the digital analyzer, when the bandpass filter is engaged.
Not a note on the analyzer (so a bit sideways from Jack's point), but on the generator - one of my takeaways from this fall's AP seminar in Seattle is how they keenly focused on an improved analog generator with the latest generation ("B series"). They seem to consider it something like a "foreseeable future" conclusion that they can provide lower source distortion from this (I guess they call it "EAG" - ADC Test Mode and Enhanced Performance Arrive with B Series APx555 - Audio Precision) than any synthetic/digital solution.
Radu.
Monte - thank you very much for your thoughtful and elaborate response. I think you do a very good job at providing clarity with the background of this.
I think I probably completely misunderstood RNMarsh's point. I thought the mere presence of the module inside the AP box has detrimental effects on the unit's overall noise figure. I was thinking of that as a design flaw of sorts.
In effect, I am currently using my working AP S1 with an EMU 0404 (modified to lower its noise) hooked up to the monitoring outputs. I am using the AP to do all interfacing with the DUT (auto-ranging, providing a very low noise, calibrated front end, etc.) while getting out most various measurements it can do; then, the FFT is done with the EMU - which is exceptionally quick, low noise, etc. Doing the FFT through the DSP seem to take quite a few seconds to perform (snapshot style), and overall be lower resolution, etc. So all and all, it seems to me the FFT through the EMU is a far better deal.
Radu.
Last edited by a moderator:
Hi I got S1 and it worked fine with APwin but doesn't with S1.exe at first. After some trial and error I found a solution.
My PC is intel core2 generation laptop and APIB is PCM-WIN card.
The S1.exe did not work under win98's command prompt or native DOS. It crushes screen or, with some options, left error message: runtime error divide by zero.
The same error is reported here.
Audio Precision System One
From divide by zero, I reminded win95 K6 patch, which fixes win95 could not boot with fast CPU. I did not understand well but the bug is related to divide by zero.
So I tried to slow down the CPU using CPUSPD.
CpuSpd - A Hardware Based CPU Speed Control Utility for DOS/Win9X Retro Gaming \
VOGONS
Reducing clock multiplier and disabling cache, yeah it worked
My PC is intel core2 generation laptop and APIB is PCM-WIN card.
The S1.exe did not work under win98's command prompt or native DOS. It crushes screen or, with some options, left error message: runtime error divide by zero.
The same error is reported here.
Audio Precision System One
From divide by zero, I reminded win95 K6 patch, which fixes win95 could not boot with fast CPU. I did not understand well but the bug is related to divide by zero.
So I tried to slow down the CPU using CPUSPD.
CpuSpd - A Hardware Based CPU Speed Control Utility for DOS/Win9X Retro Gaming \
VOGONS
Reducing clock multiplier and disabling cache, yeah it worked