RME ADI-2/4 Pro SE - Auto-measured with Multitone

[pkane]

It looks like the signal is wrong for me.

I ran it by removing the other tests from the test plan at the beginning of this thread.

You can always run a single test from the plan by selecting it in the list and pressing the step button:

That said, perhaps the regional settings are again the problem with DIM signal specification -- I suspect the decimal point (.) is interpreted differently with your settings. [EDIT: just confirmed, regional settings is the issue. I'll post a fix soon]

 

[pkane]

Please test the updated version (v1.0.78) before I make it the official release (@Rja4000, @Sokel, @JohnPM):

MultitoneSetup.zip | Powered by Box

app.box.com

This version fixes noise/SNR calculation and also fixes DIM 30/100 (and other test signals with fractions in the spec) to always interpret the period as the decimal point separator, regardless of regional settings.

I also updated the test plan attached to the OP to rename the incorrectly labeled IMD SMPTE vs Frequency measurement step to IMD SMPTE vs Level. Thanks @Rja4000 !

 

[MC_RME]

@Rja4000 and anyone interested: most of the tests in this test plan are loopbacks. Loopbacks have an un-avoidable problem that makes it necessary to have a thinking engineer sitting beside the test (always good to have ), and either manually change settings at the unit or adjust the test for lower gen levels - which is not optimal as you don't get the highest values (SNR, THD+N...). Two examples/explanations:

- the loopback would typically have same ref levels on AD and DA. So a 1 kHz sine is output at +19 dBu, and received on AD at +19 dBu, giving 0 dBFS. This will easily go wrong and end in an overloaded input, as input and output levels are never 100% accurate. If the output has +19.05 dBu it is already enough to ruin all your results. But even -0.05 dBFS in loopback is not safe at all, as it might be frequency dependent. Not to mention that most ADC typically have a lot more THD near 0 dBFS (the ES9822Pro used here being one exception). So for such a loopback test simply dial output volume to -0.5 dB.

Even the used filter on DA and AD side can influence the ref level a bit, even more so with more complex test tones than sine.

- The intersample test: PK stumbled on the same error. You use loopback with same ref levels, but give out a signal that is 3 dB above full scale at the analog output - look at the unit's input level meters, OF COURSE the input is now overloaded! Clearly shown in your measurement. Easy solution for this step: change input to +24 dBu. That gives the headrom to measure the output correctly (and distortion will be very low as we have enough headroom for this test on the DA side).

Hope this helps!

And thanks again to Paul for his tremendous work!

Added: The Crosstalk test misses about 30 dB, most probably M/S mode can not be used with that.

 

[Music1969]

Wow this is super cool @pkane

This unit using AKM or ESS DAC chip?

I see a bit of an IMD hump where IMD level is higher than THD+N level

 

[Rja4000]

The Crosstalk test misses about 30 dB, most probably M/S mode can not be used with that.

Of course !
Stupid me. Sorry.

 

[Rja4000]

The intersample test: PK stumbled on the same error. You use loopback with same ref levels, but give out a signal that is 3 dB above full scale at the analog output - look at the unit's input level meters, OF COURSE the input is now overloaded!

This is obvious.
Intersample is a test for DAC side of the loopback, here, so we need the ADC to allow it.
And therefore another range.
I'll re-run this test accordingly.

 

[Rja4000]

@Rja4000 and anyone interested: most of the tests in this test plan are loopbacks. Loopbacks have an un-avoidable problem that makes it necessary to have a thinking engineer sitting beside the test (always good to have ), and either manually change settings at the unit or adjust the test for lower gen levels - which is not optimal as you don't get the highest values (SNR, THD+N...). Two examples/explanations:

- the loopback would typically have same ref levels on AD and DA. So a 1 kHz sine is output at +19 dBu, and received on AD at +19 dBu, giving 0 dBFS. This will easily go wrong and end in an overloaded input, as input and output levels are never 100% accurate. If the output has +19.05 dBu it is already enough to ruin all your results. But even -0.05 dBFS in loopback is not safe at all, as it might be frequency dependent. Not to mention that most ADC typically have a lot more THD near 0 dBFS (the ES9822Pro used here being one exception). So for such a loopback test simply dial output volume to -0.5 dB.

Even the used filter on DA and AD side can influence the ref level a bit, even more so with more complex test tones than sine.

- The intersample test: PK stumbled on the same error. You use loopback with same ref levels, but give out a signal that is 3 dB above full scale at the analog output - look at the unit's input level meters, OF COURSE the input is now overloaded! Clearly shown in your measurement. Easy solution for this step: change input to +24 dBu. That gives the headrom to measure the output correctly (and distortion will be very low as we have enough headroom for this test on the DA side).

Hope this helps!

And thanks again to Paul for his tremendous work!

Added: The Crosstalk test misses about 30 dB, most probably M/S mode can not be used with that.

It sure helps !

I always check the input levels.
As an example, when the +24dBu signal was saturating, I'm 100% sure the level was not the issue. Not even on one of the 2 inputs.
(Well, obviously, I didnt check when I ran the intersample one with the automatic test plan.)

The goal here is to provide automation, which is a great addition to the available tools.
But, as you say, we shouldn't chase the last 0.1dB of measured performance by pushing the limits too far.
-0.5dB sounds like a good target to standardize ADC level for such tests.

 

[Rja4000]

@pkane
A few suggestions:

Maybe you could add a warning to change range or attenuate level for the intersample test ?

And would it be possible to allow running just one single step of the test plan ?
And / or to restart it at a given step ?
Or, even to run a set of steps in loop ?

Maybe also, add the timestamp in the output file names, so we never overwrite one run result by the next one ?

And also, could you add the ADC level in the measurements ? With using dBr we have no clue of the level. (Unless I miss something).

Added: also, in the measurements figures, I like to see the input signal crest factor. This helps to quickly identify saturation. (And also to check that the test signal matches what you expect. Particularly important for multitone signal).

 

[VintageFlanker]

This unit using AKM or ESS DAC chip?

ESS ES9038Q2M

 

[morillon]

@pkane
A few suggestions:

Maybe you could add a warning to change range or attenuate level for the intersample test ?

And would it be possible to allow running just one single step of the test plan ?
And / or to restart it at a given step ?
Or, even to run a set of steps in loop ?

Maybe also, add the timestamp in the output file names, so we never overwrite one run result by the next one ?

And also, could you add the ADC level in the measurements ? With using dBr we have no clue of the level. (Unless I miss something).

the plan makes it possible to multiply a given test, with the rapids, not sweep, it is practical and comfortable, then have chosen...
simple, but effective...

 

[Rja4000]

Please test the updated version (v1.0.78) before I make it the official release (@Rja4000, @Sokel, @JohnPM):

MultitoneSetup.zip | Powered by Box

app.box.com

This version fixes noise/SNR calculation and also fixes DIM 30/100 (and other test signals with fractions in the spec) to always interpret the period as the decimal point separator, regardless of regional settings.

I also updated the test plan attached to the OP to rename the incorrectly labeled IMD SMPTE vs Frequency measurement step to IMD SMPTE vs Level. Thanks @Rja4000 !

Thanks !

I re-ran the DIM30 test
It looks better, indeed.

I still can't zoom in the Waveform, though

 

[morillon]

I still can't zoom in the Waveform, though

(with windows 10) it works on my side via scrolling.. even if a little clumsy ((the full screen of the measurement would be very useful in this specific case)


(accessory but...
can be reflected on the fact that during the "automatic plan mode".. the upper information banner is not normally of too much interest..
maybe the display of the measurements could be full screen...
?)

this "test plan" approach corresponds so much to the somewhat particular approach (with its richness of tests) of m-t which is not the "realtime" way of other software, that we can only welcome this "obvious" option
;-))

 

[VintageFlanker]

@pkane
Any plan to add some kind of file generator feature for each specific Multitone test, the same way does REW (I mean: saving a WAV file of the tone(s) with custom parameters)?

I would need that when measuring DUT with no regular digital inputs.

 

[Sokel]

@pkane
Any plan to add some kind of file generator feature for each specific Multitone test, the same way does REW (I mean: saving a WAV file of the tone(s) with custom parameters)?

I would need that when measuring DUT with no regular digital inputs.

You can already do that,just set whatever you want to generate and then press the highlighted on the left:

 

[VintageFlanker]

You can already do that,just set whatever you want to generate and then press the highlighted on the left:

Thanks @Sokel !

It was there all along... Stupid me.

 

[morillon]

is it possible to carry out this mode of plan... but or each measurement is stored in mode data being able then reprocessed a posteriori? like in for example just visual formatting?
("dats" ;-) )

would also seem quite logical with the particularity of the approach of M-T
(ps
nobody is fooled.. it is the practice of the race for figures, to display, which led to ""running so close to 0dbfs""(and ""asr"" has a lot to do with it at 0.1db of sinad ;-) )and not a reasonable and very useful systematic headroom... ;-) )

 

[Music1969]

ESS ES9038Q2M

I wonder if the IMD hump comes from the ADC

Because I've seen many DACs using 9038Q2M without the ESS hump

Cheap Pro-ject Pre Box S2 for example (designed by John Westlake years ago)

 

[VintageFlanker]

Because I've seen many DACs using 9038Q2M without the ESS hump

And there're many with it.

I wonder if the IMD hump comes from the ADC

It does not. Here is the PO100 AK (AK4493) measured with the ADI-2/4 PRO SE:

(Cosmos does have the hump, tho)

 

[morillon]

ps there are still some measurement titles that do not match ...
basic d imd becoming log sweep etc.
little things to "clean" as you go
;-)

 

[pkane]

Wow this is super cool @pkane

This unit using AKM or ESS DAC chip?

I see a bit of an IMD hump where IMD level is higher than THD+N level

ESS