I did some experimental IM test beside of the traditions IM frequencies root while:
1. Do not cover listening freq. range (the sensitive range)
2. To use frequency in a such range/combination that no cancellation or addition will happen
May others contribute with their experiences or combinations 😀
My own test combinations:
1. 1.0 / 1.5 / 2.0 / 2.5 kHz space starting at 10kHz
2. 0.5 / 1.0 / 1.5 / 2.0 / 2.5 kHz space starting at 10kHz
what I have seen so far:
1. Setting the level near max level, shows more IM products
2. getting much over (as on -40 dB level) the IM products are more or less gone
see pictures done on 96kHz & 192khz using RME ADI-2 Pro in a loop-back configuration using 0.5 / 1.0 / 1.5 / 2.0 / 2.5 kHz space starting at 10kHz
Hp
1. Do not cover listening freq. range (the sensitive range)
2. To use frequency in a such range/combination that no cancellation or addition will happen
May others contribute with their experiences or combinations 😀
My own test combinations:
1. 1.0 / 1.5 / 2.0 / 2.5 kHz space starting at 10kHz
2. 0.5 / 1.0 / 1.5 / 2.0 / 2.5 kHz space starting at 10kHz
what I have seen so far:
1. Setting the level near max level, shows more IM products
2. getting much over (as on -40 dB level) the IM products are more or less gone
see pictures done on 96kHz & 192khz using RME ADI-2 Pro in a loop-back configuration using 0.5 / 1.0 / 1.5 / 2.0 / 2.5 kHz space starting at 10kHz
Hp
Attachments
Did you see the spreassheet for spectral contamination I posted some time ago? Someone (I forget who) calculated a set of frequencies where neither the harmonics or the IM products overlap so every nonlinear product shows. In any of these tests you do need to be sure of the crest factor (I'm sure you are) and not overload. They are used more by transducer engineers to address problems in cones and domes.
In your plots it looks really pretty good. The IM products are all at least -110 below the peak. You can also see the rising noise floor of the AKM 5385 ADC in that box. I find it interesting that people seem to prefer the RME with that ADC to the Lynx with its higher performance AK5394A ADC. However it may be the stuff around the ADC that matters.
In your plots it looks really pretty good. The IM products are all at least -110 below the peak. You can also see the rising noise floor of the AKM 5385 ADC in that box. I find it interesting that people seem to prefer the RME with that ADC to the Lynx with its higher performance AK5394A ADC. However it may be the stuff around the ADC that matters.
I am not aware of this.. .any pointer to the content?
In this RME ADI-2 pro the ADC is an AK5574 (quad as dual configuration) and the DAC is the AK449, also 768kHz SR support as up to DSD256.
hp
The AK5574 is a newer generation and possibly better than the 5385. For measurements the AK5394A is still the best performing. For various reasons AKM chose to go a different direction on subsequent designs. Most likely chip area since the 5394 is in a very large package for a 2 channel part.
The spectral contamination spreadsheet is attached. The attribution is included. You can find the original article in the AES library. AES E-Library » Spectral Contamination Measurement
The spectral contamination spreadsheet is attached. The attribution is included. You can find the original article in the AES library. AES E-Library » Spectral Contamination Measurement
Yes, currently it is the only company who produces ADC and even supporting DSD ADC for audio. My personal interest was always in a ADC for measurement (no audio only). This means for me SR larger than 768kHz and low SFDR, low THD & low noise.
Some PDF link .... by google 😀
http://www.tmr-audio.com/pdf/jon_risch_biwiring.pdf
hp
Czerwinski et al. published a two part Article on multitone testing in the JAES in 2001, which is quite interesting (although excessive in a way 🙂 ):
Czerwinski, E., Voishvillo, A., Alexandrov, A., Terekhov, A., Multitone Testing of Sound System Components " Some Results and Conclusions, Part 1: History and Theory, Journal of the Audio Engineering Society, Vol 49, No 11, pp 1011 - 1048 (Nov 2001)
Czerwinski, E., Voishvillo, A., Alexandrov, A., Terekhov, A., Multitone Testing of Sound System Components " Some Results and Conclusions, Part 2: Modeling and Application, Journal of the Audio Engineering Society, Vol 49, No 12, pp 1181 - 1192 (Dec 2001)
Czerwinski, E., Voishvillo, A., Alexandrov, A., Terekhov, A., Multitone Testing of Sound System Components " Some Results and Conclusions, Part 1: History and Theory, Journal of the Audio Engineering Society, Vol 49, No 11, pp 1011 - 1048 (Nov 2001)
Czerwinski, E., Voishvillo, A., Alexandrov, A., Terekhov, A., Multitone Testing of Sound System Components " Some Results and Conclusions, Part 2: Modeling and Application, Journal of the Audio Engineering Society, Vol 49, No 12, pp 1181 - 1192 (Dec 2001)
The SC file I used here John Curl's Blowtorch preamplifier part II and here John Curl's Blowtorch preamplifier part II is available for download here: John Curl's Blowtorch preamplifier part II
there was an initial problem with that page, ie the lack of the linked files. corrected now.
Alan
And what I get from those
1 & 2: IMHO for digital analysis not clean enough 😀
3: Crest factor 4.251, requires some larger FFT & a related FFT Window to view the lower freq. cluster
Cheers
Hp
Attachments
Thanks for validating the signals. The first two were for verifying lower level distortion both in loopback and on analog analyzers. Anything below -160 really is ot meaningful in that domain. The multitone looks clean enough for analog systems as well.
Keep in mind, as given, the file name with -3dB is really -10dB. Additional phase shifts would help to get a lower crest factor.
Professionals purchased the SW to deal faster with various & additional settings (as required dithering) done as a snap, as moving xxx MBytes around the sky 😀
PS: Could not resist 😀
Hp
I'm playing with your eval software. I used it to generate the same -70 dB test file but its different in some way I have not figured out yet. I'm getting odd results from the Shibasoku's. Does the generator move the tone to bin centers automatically? The Shibasoku needs the harmonics to be exact multiples or it rejects them.
There is a separated "Sine & Harmonic" / "Multi-Wave" generator function implemented.
First deals with 100% exact harmonic multiplies given from fundamental.
The Multi Wave places/calculates the entered freq. dependent of the coherent none coherent switch. This means exact harmonics is possible but could a bit more tricky.
May use the preset to define your individual sets and use the Tool tab to alter values in quicker manner.
Let's use the support channel to get this done 😀
Hp
The 3/04 issue of AudioXpress had a construction article for a 1PPM IM analyzer;
that used frequencies of 19K (actually 19,2) / 20K; 9K (actually 9,6) / 20K and 9 /10 /20Khz.
Author was Dick Crawford. He made PC boards available. Built one and it works quite nicely.
Charles
that used frequencies of 19K (actually 19,2) / 20K; 9K (actually 9,6) / 20K and 9 /10 /20Khz.
Author was Dick Crawford. He made PC boards available. Built one and it works quite nicely.
Charles
I think I have a set of those PCB's. I also picked up from eBay a "proto" of a similar analyzer built by Chuck Hansen that would an opamp upgrade gets to the same -.0001% threshold. Needless to say once I had a built unit assembling PCB's was not interesting.
- Status
- Not open for further replies.