I'm reasonably new to using to taking measurements and I was wondering if someone could help me out with some aspects of interpreting the results, particularly the spectrum graphs and their peaks.
First of all, why do a lot of graphs have a large peak in the middle? Arta does this but RMAA graphs don't. Eg:
Some of my noise graphs in RMAA have a peak at 50hz, even when the mains isn't connected to my device. Any idea why?
Sometimes I get large peaks at 50hz and then at 150/250/350/450 etc. What would cause this?
Any advice appreciated!
First of all, why do a lot of graphs have a large peak in the middle? Arta does this but RMAA graphs don't. Eg:
Some of my noise graphs in RMAA have a peak at 50hz, even when the mains isn't connected to my device. Any idea why?
Sometimes I get large peaks at 50hz and then at 150/250/350/450 etc. What would cause this?
Any advice appreciated!
Sometimes I get large peaks at 50hz and then at 150/250/350/450 etc. What would cause this?
This could be from a couple of things. Firstly it could be mains ripple from inadequate smoothing of the rectified supply. Secondly it could be leaked flux from a mains transformer coupling into signal wires.
I'm reasonably new to using to taking measurements and I was wondering if someone could help me out with some aspects of interpreting the results, particularly the spectrum graphs and their peaks.
First of all, why do a lot of graphs have a large peak in the middle? Arta does this but RMAA graphs don't. Eg:
Some of my noise graphs in RMAA have a peak at 50hz, even when the mains isn't connected to my device. Any idea why?
Sometimes I get large peaks at 50hz and then at 150/250/350/450 etc. What would cause this?
Any advice appreciated!
In the first graph the large peak in the middle in the source fundamental. If you use sat a 1kHz tone then the fft will display this as a peak at 1kHz. Because the fundamental has the greatest power it will stand much taller than the other peaks. The other peaks are mostly harmonics of the fundamental tone at 2kHz, 3kHz, 4kHz and so on. If a very pure tone is used, one having few harmonics or very low in harmonic power, then beyond what the harmonics the tone generates is the harmonic contribution of whatever DUT, device under test, is contributing. This is a very overly simplistic explanation but it should give you an idea of what your seeing.
What I see in the third graph is the mains fundamental and it's harmonics. I think you've noticed the peaks are multiples of the 50Hz which mean they are harmonics of the 50hz.
The peaks at the upper end of the graph are something else.
From what I have read on this site and the advice that I have received through PM from very respected members ( you know who you are), RMAA is a bit, well, sloppy. And, it's quite limited in its accuracy that is compounded by the audio interface, which is why lab grade equipment costs tens of thousands of dollars compared to hobby usb analyzers.
Seeing that the noise is harmonically related to 50Hz AC, there is a chance that the dongle is likely picking up some power supply noise. With low-cost audio analyzers, both the test bandwidth and upper limits are restricted to preserve some integrity. The noise seen in the upper limit may be related to the carrier tome that the close modulates at, or may be something simpler. I have a few ideas about it.
Seeing that the noise is harmonically related to 50Hz AC, there is a chance that the dongle is likely picking up some power supply noise. With low-cost audio analyzers, both the test bandwidth and upper limits are restricted to preserve some integrity. The noise seen in the upper limit may be related to the carrier tome that the close modulates at, or may be something simpler. I have a few ideas about it.
Thanks for the reply everyone.
In the third graph there is mains connected so that 50hz peak is probably that. But in the second graph there is also a 50hz peak when there is no mains connected to the DUT?
In the third graph there is mains connected so that 50hz peak is probably that. But in the second graph there is also a 50hz peak when there is no mains connected to the DUT?
Does that mean I can safely say that the those 2/3/4Khz peaks are not caused by the DUT? And ignore them?If you use sat a 1kHz tone then the fft will display this as a peak at 1kHz. Because the fundamental has the greatest power it will stand much taller than the other peaks. The other peaks are mostly harmonics of the fundamental tone at 2kHz, 3kHz, 4kHz and so on
Yes. If I set it to external then I need another device to generate the test signal, right?Do you have ARTA's internal sine generator enabled ? Check that the 'Gen' dropdown box (just below left of the toolbar) is set to 'External'.
Does that mean I can safely say that the those 2/3/4Khz peaks are not caused by the DUT? And ignore them?
No and no. You have to do some work baselining your test, then get an idea of repeatability. Once you've got that nailed, you can separate DUT distortion from test distortion.
Still mains related pickup from somewhere. It can be difficult to get rid of. Try to keep DUT and connection leads away from powered items, or turn them off. You can get it down to a certain level but then you find it varies by time of day, the neighbour's washing machine, aliens tunneling, etc. Actually -120dB for a quick test hookup is not so bad, I've had worse.... in the second graph there is also a 50hz peak when there is no mains connected to the DUT?
ARTA has a high pass filter you can set to exclude most of the mains related junk from the THD measurements.
Yes.If I set it to external then I need another device to generate the test signal, right?
Last edited:
Thanks! I've adjusted a few wires and got it down to -138dB now.Try to keep DUT and connection leads away from powered items, or turn them off. You can get it down to a certain level but then you find it varies by time of day, the neighbour's washing machine, aliens tunneling, etc. Actually -120dB for a quick test hookup is not so bad
Yes, I've done loopback tests as a baseline and these peaks are there tooNo and no. You have to do some work baselining your test, then get an idea of repeatability. Once you've got that nailed, you can separate DUT distortion from test distortion.
I think you can benefit from reading this thread.
http://www.diyaudio.com/forums/equipment-tools/205304-low-distortion-audio-range-oscillator.html
The thread has been running for years and it's hugely long but well worth the read. Every means of measuring TH distortion have been discussed and a few enhancements for SC - FFT measurements evaluated. It's all there. It would be redundant to discuss it here.
http://www.diyaudio.com/forums/equipment-tools/205304-low-distortion-audio-range-oscillator.html
The thread has been running for years and it's hugely long but well worth the read. Every means of measuring TH distortion have been discussed and a few enhancements for SC - FFT measurements evaluated. It's all there. It would be redundant to discuss it here.
because its not audio?
in fact not physically realizable at all - you have practical bandwidth limits in all systems - even air
and of course ADC?DAC, digital signal processing - our more highly resolving and now often cheapest tools are very strict about bandwidth limitation
your 1960s Sine-Tri-Sq benchtop signal generator is fine for debugging "is it working" questions at very crude levels
but todays PC sound chipsets often 24/192k are quite an advance - and can give much more information for actual audio frequency signals
in fact not physically realizable at all - you have practical bandwidth limits in all systems - even air
and of course ADC?DAC, digital signal processing - our more highly resolving and now often cheapest tools are very strict about bandwidth limitation
your 1960s Sine-Tri-Sq benchtop signal generator is fine for debugging "is it working" questions at very crude levels
but todays PC sound chipsets often 24/192k are quite an advance - and can give much more information for actual audio frequency signals
Last edited:
I have a question, why little to no square wave testing being done anymore..?
Who says?
umm - does the thread title have any relevance?
in spectrum plots square wave all odd harmonics, all at differing level aren't great for eyeball interpretation
certainly compared to frequency sweeps to determine high/low frequency corner peaking/flatness rather than interpret square wave harmonic lines relative levels, for square wave the time domain is more natural, looking for "tilt" or overshoot
we can do ADC at >100 MHz for amp stability evaluation - Gsample/s digital 'scopes are cheep too now
in spectrum plots square wave all odd harmonics, all at differing level aren't great for eyeball interpretation
certainly compared to frequency sweeps to determine high/low frequency corner peaking/flatness rather than interpret square wave harmonic lines relative levels, for square wave the time domain is more natural, looking for "tilt" or overshoot
we can do ADC at >100 MHz for amp stability evaluation - Gsample/s digital 'scopes are cheep too now
Last edited:
I very rarely see them around here...
Look no further than my articles.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
- Home
- Design & Build
- Equipment & Tools
- How to interpret measurement spectrum graphs and peaks