small inaccuracy of riaa compensation

[morillon]

chrip1k 21000 riaa.PNG (46.66 KiB) Viewed 748 times

sorry for my "auto" english
after a lot of checking with different machines and software approaches, cal files or the precise use of the mathematical function, it turns out that there are accidents that appear with both files systematically ..... between 3khz and 10khz ... would it be possible to check or improve these cal files?
here a tested inverse riia and a normal riia card

small ps could you specify which riaa approach you use? there are so many "riaa" iec din etc ... (at home always a slight drop under 1k with your approach, but maybe due to my machines)

[morillon]

riaa inverse virtins.PNG (48.83 KiB) Viewed 747 times

.....

another machine...." inverse riaa"



(https://forums.melaudia.net/showthread. ... pid=202661 no accident with that or fonction math in MT)

[VirtinsTech]

Not sure how your reverse RIAA curve (i.e. reproducing RIAA) was obtained. It looks like that you used frequency sweep + peak hold in the Spectrum Analyzer. The abnormal undulation between 3kHz and 10kHz in the curve is likely due to improper settings, such as Record Length, FFT Size, or Sweep Speed.

Here is what the playback RIAA curve looks like by using the ReproducingRIAA.fcf as a frequency compensation file in Multi-Instrument. The test signal is a 10Hz~48kHz bandlimited "Periodic White Noise" digitally generated by the Multi-Tone function of the Signal Generator of Multi-Instrument.

ReproducingRIAA.png (110.96 KiB) Viewed 741 times

ReproducingRIAA.fcf is a CSV text file which can be viewed and edited using Windows Notepad as follows.

ReproducingRIAAFCF.png (52.23 KiB) Viewed 741 times

[morillon]

I have carried out numerous tests by chirp, long sweep, by white... a lot... with all types of fft etc and crossed this behavior between 3k and 10k with your riaa correction (and therefore also on an inverse riaa...) but not under rew or mt etc


here two different machines in very long sweep mode in duration..a normal riaa phono and an inverse ria using the inverse file...

(here in advanced and very detailed resolutions and (the resolution of your cal file is significantly lower than those I pointed out to you... maybe that's simply where it's at...?)

[VirtinsTech]

This test here is similar to a frequency response test. The testing parameters must be set properly in order to get meaningful results. For example,

(1) If you use the peak hold method, then the frequency resolution of FFT and signal generator sweep speed must be set correctly to produce a smooth final curve...

(2) If you use a non-repetitive normal white noise, you have to average the result to get a thin curve, unlike the case of using a "periodic white noise".

(3) If you use frequency sweep, you have to capture the entire sweep in one single oscilloscope frame from start to end, and the FFT Size must be greater or equal to the Record Length, and the window function must be rectangle ....

Here is an example of using a normal white noise.

ReproducingRIAA-WhiteNoise.png (143.12 KiB) Viewed 736 times

Below is an example of using a log sweep and 1/96 octave analysis.

ReproducingRIAA-LogFreqSweep.png (152.74 KiB) Viewed 736 times

(here in advanced and very detailed resolutions and (the resolution of your cal file is significantly lower than those I pointed out to you... maybe that's simply where it's at...?)

Definitely not. Multi-Instrument uses interpolation in between so the curve will be very smooth as shown in all the examples above (see the frequency resolution displayed at the bottom left in each screenshot).

[morillon]

playback.PNG (47.98 KiB) Viewed 732 times

another machine
just one a protocol but one with the "playback" cal, the other cal from your virtins, (the third Multitones but by the math function)
https://www.avnirvana.com/threads/using ... tage.5623/

[morillon]

.....

virtins.PNG (46.9 KiB) Viewed 732 times

[morillon]

.....

Frequency Response.png (86.3 KiB) Viewed 732 times

[morillon]

it seems to me that it is not the method but your cal file which has this particularity...

[VirtinsTech]

Here is the reverse RIAA curve measured using the peak hold method, with all testing parameters displayed. Generally, the longer the frequency sweep time per round, the smoother the curve. To achieve an even better curve at the low frequency end, the Record Length (currently 200ms) needs to be even longer.

ReproducingRIAA-PeakHold.png (101.03 KiB) Viewed 724 times

Not all of your testing parameters are visible. You can upload your peak-hold panel setting file for diagnostics.