Beta Test: Multitone Loopback Analyzer software

[pkane]

Ok,bear with me here:

Here's a sweep of Khadas alone, just did it (don't mind level, is through a voltage divider I use to measure a strong pre) :

View attachment 433037
Started and ended just fine,perfectly ok.

..then I just got the cable and passed it through a pre and a x-over (other than the one measured in previous posts), changing absolutely nothing else (the low part of the x-over, so imagine 20-250Hz before levels fall) :


View attachment 433038

measurement stops right there and goes no further.
Now,at the high part of it measurement does not start at all.

Any explanation other than magic?

Magic is the only possible explanation

Can you figure out what last frequency it was trying to measure, and then just do the same level/frequency/settings, but without a sweep, while looking a the Spectrum plot? Does that get stuck like that also?

 

[Rantapossu]

I tested it with play gain signal level of -160dB, and it worked fine, so I didn't change things.

View attachment 433010

I cant get level sweeps working reliably if I set the start level to -60 dBFS or lower.


-59 dBFS works ok (The Y-axis scale is now -20 dBFS to -120 dBFS):

With -60 dBFS start level the Y-axis scale changes to+40 dBFS to -120 dBFS and if you look the level meter text on the bottom of the graph, it says -125.1 dBFS and the X-axis scale starts from -130 dBFS:

If I stop the sweep and change to the spectrum view, there is no fundamental at all:

If I make a single -140 dBFS measurement for the sanity check, the spectrum view works ok:

 

[Sokel]

Magic is the only possible explanation

Can you figure out what last frequency it was trying to measure, and then just do the same level/frequency/settings, but without a sweep, while looking a the Spectrum plot? Does that get stuck like that also?

It's at 1200Hz, so I did three, before, at it, and after:

Didn't stick and the results are as expected for the level (not the same with the sweep though)

And another, when I switched tabs from the sweep to spectrum, I got a very long error message about nulls and stuff.

(please don't tease me about the mains noise,cabling is terrible, 6 meters just for the test )

 

[Rantapossu]

I cant get level sweeps working reliably if I set the start level to -60 dBFS or lower.


-59 dBFS works ok (The Y-axis scale is now -20 dBFS to -120 dBFS):

View attachment 433042


With -60 dBFS start level the Y-axis scale changes to+40 dBFS to -120 dBFS and if you look the level meter text on the bottom of the graph, it says -125.1 dBFS and the X-axis scale starts from -130 dBFS:

View attachment 433057


If I stop the sweep and change to the spectrum view, there is no fundamental at all:

View attachment 433058


If I make a single -140 dBFS measurement for the sanity check, the spectrum view works ok:

View attachment 433059

Version 1.2.6 seeem to work with -60 dBFS start (And lower too, I tried -80 dBFS) level when sweeping.

Versions 1.2.7 and 1.2.8 don't like those lower start levels.

Actually it seems that -58 dBFS is the lowest starting point on my system, -59 dBFS stucks sometimes to this on the first round:

I tried -58 dBFS five times and it starts reliably every single time.

Something has happened to the level sweep between version 1.2.6 and 1.2.7, I'm afraid...

 

[Rantapossu]

Could the level sweep graph be autoscaling like the frequency sweep and the interval sweep are?

When for example the noise is below -120 dBFS, MT draws the line outside the graph:

Pressing the "Fit all data" -button helps for a second, but it reverts back to -20 dBFS to -120 dBFS scale for the next round.

 

[pkane]

Version 1.2.6 seeem to work with -60 dBFS start (And lower too, I tried -80 dBFS) level when sweeping.

Versions 1.2.7 and 1.2.8 don't like those lower start levels.

Actually it seems that -58 dBFS is the lowest starting point on my system, -59 dBFS stucks sometimes to this on the first round:

View attachment 433071

I tried -58 dBFS five times and it starts reliably every single time.

Something has happened to the level sweep between version 1.2.6 and 1.2.7, I'm afraid...

Sounds like I may need to go with my initial instinct and revert the handling of lower signal levels to what it was...

 

[pkane]

Hopefully this will solve the issue with low-level signal sweeps: https://app.box.com/s/ue7ll9xmvwogst817x2l1xg09opvgy47
(please uninstall the previous version before installing, as the version number is still 1.2.8)

This also changes the zoom behavior of the sweep charts: if at any point you change the zoom level or drag the chart around, it will no longer reset to default zoom on each subsequent point. If you want to get back to the original mode that includes most of the normal range, just click the Fit all Data button.

 

[Rantapossu]

Hopefully this will solve the issue with low-level signal sweeps: https://app.box.com/s/ue7ll9xmvwogst817x2l1xg09opvgy47
(please uninstall the previous version before installing, as the version number is still 1.2.8)

This also changes the zoom behavior of the sweep charts: if at any point you change the zoom level or drag the chart around, it will no longer reset to default zoom on each subsequent point. If you want to get back to the original mode that includes most of the normal range, just click the Fit all Data button.

Thanks!

Much, much better! Works with the low levels perfectly and doesn't get stuck.

But it's not perfect yet however. This is my first -100 dBFS sweep, it doesn't show any graphs until it reaches -90 dBFS level (The lowest I have tested before). Now I'm on round 4:

Round 14 (4 steps after the previous lowest start of the sweep) and now I have all the lines too, but not the 10 or so first points:

If I now restart the measurement at -100 dBFS, it scales the X-axis correctly and shows everything straight from the start:

If I now start a new sweep from for example -92 dBFS, it shows all, but it won't scale the X-axis right after the previous sweep and the lines don't start from the left edge of the graph:

 

[Rantapossu]

The realtime scaling works fine too!

 

[Sokel]

Yep, that fixed it, thanks!

 

[pkane]

Thanks!

Much, much better! Works with the low levels perfectly and doesn't get stuck.

But it's not perfect yet however. This is my first -100 dBFS sweep, it doesn't show any graphs until it reaches -90 dBFS level (The lowest I have tested before). Now I'm on round 4:

View attachment 433116


Round 14 (4 steps after the previous lowest start of the sweep) and now I have all the lines too, but not the 10 or so first points:

View attachment 433118


If I now restart the measurement at -100 dBFS, it scales the X-axis correctly and shows everything straight from the start:

View attachment 433121

If I now start a new sweep from for example -92 dBFS, it shows all, but it won't scale the X-axis right after the previous sweep and the lines don't start from the left edge of the graph:

View attachment 433123

If I understand what you're reporting, the X-axis scale was cutting off some of the early points in the sweep because they fell below -96dB cutoff there. The initial zoom level is set to include as much of the sweep as is normally useful for audio measurements. This is by design, so -100dB is usually outside the range, as is +100dB But, with this new version you can change your zoom level to anything you want, and it will continue to stay at that level until you hit Fit all data button. I think this is more reasonable than doing auto-scaling that will zoom-in on the first few data points, possibly showing a range from -100dB to -99 dB across the whole screen, and then continue to rescale with each added point.

 

[Rantapossu]

If I understand what you're reporting, the X-axis scale was cutting off some of the early points in the sweep because they fell below -96dB cutoff there. The initial zoom level is set to include as much of the sweep as is normally useful for audio measurements. This is by design, so -100dB is usually outside the range, as is +100dB But, with this new version you can change your zoom level to anything you want, and it will continue to stay at that level until you hit Fit all data button. I think this is more reasonable than doing auto-scaling that will zoom-in on the first few data points, possibly showing a range from -100dB to -99 dB across the whole screen, and then continue to rescale with each added point.

Actually the bug doesn't care about super low levels, it doesn't have to be -90 dBFS or -100 dBFS. 70 dBFS is low enough.

All you have to do is to take a couple of level sweeps and lower the sweep start point every time.

Don't skip any of these steps, if you want to reproduce the bug:

- Set the sweep start to -60 dBFS (Like it is on the default settings for the new user)
- Restart Multitone (This is important, don't skip this)
- Start the sweep (-60 dBFS)
- Everything works ok. Wait for the sweep to finish.
- Set the sweep start to -70 dBFS and start the sweep

And there is the bug!

Multitone draws the line outside the graph and you can't get the lines visible (Or at least I haven't been able to do it).

Using the mouse wheel over the X-axis doesn't scale enough to be able to see -70 dBFS and "Fit all Data" -doesn't scale it right anymore.

You'll have to wait until the level rises enough for the line to became visible and you miss the first measurement points.

So if you wan't to measure with the sweep start of -90 dBFS, you have to do it with one increment.

Changing straight the start point from -60 dBFS to -90 dBFS works, but the series of -60 dBFS -> -70 dBFS -> -80 dBFS -> 90 dBFS doesn't. The scaling breaks between -60 dBFS and -70 dBFS.

 

[pkane]

Actually the bug doesn't care about super low levels, it doesn't have to be -90 dBFS or -100 dBFS. 70 dBFS is low enough.

All you have to do is to take a couple of level sweeps and lower the sweep start point every time.

Don't skip any of these steps, if you want to reproduce the bug:

- Set the sweep start to -60 dBFS (Like it is on the default settings for the new user)
- Restart Multitone (This is important, don't skip this)
- Start the sweep (-60 dBFS)
- Everything works ok. Wait for the sweep to finish.
- Set the sweep start to -70 dBFS and start the sweep

And there is the bug!

Multitone draws the line outside the graph and you can't get the lines visible (Or haven't been able to do it).

Using the mouse wheel over the X-axis doesn't scale enough to be able to see -70 dBFS and "Fit all Data" -doesn't scale it right anymore.

You'll have to wait until the level rises enough for the line to became visible and you miss the first measurement points.

So if you wan't to measure with the sweep start of -90 dBFS, you have to do it with one increment.

Changing straight the start point from -60 dBFS to -90 dBFS works, but the series of -60 dBFS -> -70 dBFS -> -80 dBFS -> 90 dBFS doesn't. The scaling breaks between -60 dBFS and -70 dBFS.

I see, thanks for the detailed instructions on how to reproduce. Can you please test v1.2.9? The only change is to allow the level sweep plot to be scaled beyond the initial zoom settings: https://app.box.com/s/ue7ll9xmvwogst817x2l1xg09opvgy47

 

[Rantapossu]

I see, thanks for the detailed instructions on how to reproduce. Can you please test v1.2.9? The only change is to allow the level sweep plot to be scaled beyond the initial zoom settings: https://app.box.com/s/ue7ll9xmvwogst817x2l1xg09opvgy47

Thanks, seems to work correctly!

Is this now ready for release as the next official version?

This and many (If not all?) recent test builds have required you to uninstall the previous version before installing the new one, even though the version number has increased. Please make sure this installs over the previous official one if you decide to release this version.

 

[pkane]

Thanks, seems to work correctly!

Is this now ready for release as the next official version?

This and many (If not all?) recent test builds have required you to uninstall the previous version before installing the new one, even though the version number has increased. Please make sure this installs over the previous official one if you decide to release this version.

I believe it's ready for a release. I usually don't update the installer and version number if I release it for testing within 24 hours of a previous test release. All official versions automatically update any older versions already installed.

 

[pkane]

Multitone v1.2.9 is now official, thanks for all the help with testing and troubleshooting!

Changes in 1.2.9​

• Change: remove spikes and artifacts at the end of a capture caused by device or audio driver
• Change: improve zooming behavior on level and frequency sweep charts during capture
• Change: improve calibration file handling by removing auto-scaling
• Fix: level sweep plot could occasionally lose one of the measured lines when returning from another plot
• Fix: spectrum plot distortion panel not shown when captured from another tab
• Change: improved multitone notch filter design to handle wide-skirted fundamentals

 

[Rantapossu]

I had some free time to test my new Cosmos LPF.

It seems to be doing its job...

So close to -130 dB!

Equipment used in the measurement:

- Audient iD24 playing a 0 dBFS signal, optical output, L + R
- Topping E50 -DAC, optical input, balanced output, playing L+R, but only the left channel connected forward (To compensate the H3)
- Cosmos Scaler, 0 dB gain, as a buffer for the E50
- Cosmos LPF, 1 kHz low pass, 7 dB gain
- Cosmos APU, 1 kHz notch, 0 dB gain and the calibration file inserted
- Cosmos ADCiso, grade 0, THD +

 

[Rantapossu]

Hi @pkane !

You did a great job fixing the frequency and level sweeps. However, the interval sweep seems to be broken as well..

If you choose for example these measurements...

... after the first round (0 seconds) only two of them are visible as a text...

... and none of them are visible on the chart. No circles, triangles or crosses:

After the second round (60 seconds in my case) the text updates for the 3rd measurement...

... and now the lines appear too (But still no circles, triangles or crosses at the 0 second position, however):

 

[Rantapossu]

There is the same old color changing bug with the interval sweep too:

If you take a interval sweep...

... and go to some other tab and then go back to the interval sweep tab, the white line changes to green and the previous green line disappears:

 

[Rantapossu]

The scaling you set yourself also seems to disappear between rounds:

Before the next measurement:

After the next measurement it is back to default settings: