Loopback measurement of RME ADI-2 PRO FS (Black Edition)

[polmuaddib]

It is original Capri. Not s2. I will first measure RME loopback and then return to Capri. But I will somehow measure output with voltmeter to find what values correspond to volts.
thanks for guidance.

 

[Rja4000]

Some example of measurements using REW with RME ADI-2 Pro fs (non R)

I managed to build a "preamp" with an attenuator and one of my Mic preamps (to get the +14dB gain, overall)

Idea: Test the preamp for 0dB gain, with a signal level of +4Vrms (approx 14.3dBu)
All measurements using Balanced In and Out

First, I setup REW for ASIO I/O, Analog 1 (1)
RME is set for Analog Input and Main Output @+19dBu range (All EQ disabled, "Sharp" filter)

Then I start the generator with around -4.7dBFS and measure the output (pin 2-3 on XLR) with a True RMS meter (I use a Keithley DMM6500, so I don't need to change the frequency. If you use a cheap RMS meter, you may want to use 80Hz instead of 1kHz)

I click on "Calibrate" and enter the actual measured Voltage.

I set the Preamp to get the same voltage at the Preamp output
I show the distortion measures (in DB) and also the Windowing parameters.
Default is Blackman-Harris 7, which is perfect for the job

THD+N (-SINAD) is -102.4dB here.
Not too bad ;-)

Does it improve if I change the generator level ?

Not significantly. I reach 103dB SINAD for -2.1dBFS, which is the best performance here.

If I go lower, SINAD will decrease due to noise.

If I go higher, the distortion will raise and limit the SINAD as of approx +22dBu (9.8Vrms).
(I have to change range and re-calibrate for +24dBU range In and Out on RME)

We now see the THD being the limiting factor.
So that's the best we can get.

What about varying the gain ?

Setting back the RME and calibration for +19dBu, I pushed the gain by 12dB

Of course, I have to lower the output signal to keep the same Preamp output level
-4.8dBFS - 12dBFS = -16.8dBFS (1Vrms)

Noise is hitting the performance here, which was expected: We have a 6dB SINAD penalty for 12dB more gain.


Then, let's see what happens if we lower the gain by 12dB

 

[3125b]

Here is my first loopback measurement of the ADI-2 Pro FS headphone output (single-ended):

111,7dB SINAD with a 3m long, unshielded cable. Not too bad, balanced would be a little bit better I assume.

But I have a few questions seeing all the measurements here:
- How do you display the distortion panel in dB?
- I have found test tone generation to be quite volatile across systems with most tone generators producing wildly varying results. What method do you use, is there one reliable method?
-What cables do you use? I'd be measuring mostly unbalanced consumer electronics.
- The sampling rate is selected via the driver, wich works fine. However I can then only select 16 or 24 bits depth in the Windows sound menu - is that normal, am I missing something, is it capable of 32 bit via USB?

Other numbers I've come up with so far:
- iPad Pro 9.7 (2016): 95,9dB SINAD (at max. 1Vrms out)
- Asus Crosshair VI Hero (front panel out, so that might be holding it back significantly): 76dB SINAD
The iPad seems about rightseems about right at least.

 

[polmuaddib]

Thank you @Rja4000 for a great guide. Could I use spare XLR (other channel) output to make a cable which would be connected to Voltmeter all the time as I am measuring or just use voltmeter once to find which preamp volume level (number) corresponds to 4V output?
Will try to make new measurements later in the week if i gather all necessary equipment.

 

[Rja4000]

Could I use spare XLR (other channel) output to make a cable which would be connected to Voltmeter all the time

Hi
Yes, you can.
Another way (the way I do it) is to purchase a few Y XLR cables and use them to allow easy plugging of voltmeter (and to build an xlr to banana short cable for convenience)

The benefit is that voltage may vary with load, and here you're able to measure the actual voltage under load. Remove the voltmeter cable for THD measurement.

I use Cordial CFY cables

 

[KSTR]

But I have a few questions seeing all the measurements here:
- How do you display the distortion panel in dB?
- I have found test tone generation to be quite volatile across systems with most tone generators producing wildly varying results. What method do you use, is there one reliable method?
-What cables do you use? I'd be measuring mostly unbalanced consumer electronics.
- The sampling rate is selected via the driver, wich works fine. However I can then only select 16 or 24 bits depth in the Windows sound menu - is that normal, am I missing something, is it capable of 32 bit via USB?

- dB scale is selected in the Distortion Settings panel of the RTA.
- use REW as test tone generator. It's impeccable, 100%.
- for local loopback, cable choice is irrelevant, "ground loop" issues can simply not happen.
- RME's ASIO driver is 32bit, but only few software feed ASIO with a 32bit stream, but this is irrelevant in the end. For measurements with REW under windows you must use ASIO (Java only supports 16bit in windows).

 

[Rja4000]

- dB scale is selected in the Distortion Settings panel of the RTA.
- use REW as test tone generator. It's impeccable, 100%.
- for local loopback, cable choice is irrelevant, "ground loop" issues can simply not happen.
- RME's ASIO driver is 32bit, but only few software feed ASIO with a 32bit stream, but this is irrelevant in the end. For measurements with REW under windows you must use ASIO (Java only supports 16bit in windows).

Agreed 100%
About cable: If you see mains harmonics in your FFT, change cable.

Here is where you find the setting "Show distortion in DB"

 

[JohnPM]

- How do you display the distortion panel in dB?

V5.19 doesn't have that option, you will need the latest V5.20 release candidate.

 

[3125b]

Thanks for the answer.

- for local loopback, cable choice is irrelevant, "ground loop" issues can simply not happen.

I know, the issue is picking up mains noise, that cable is really prone to that. Not an audible issue and not even really concerning for measuring lower end devices, but to get that loopback measurement, I had to fiddle around with it quite a bit.

The 24 bit limitation dosen't bother me too much, seems to have no impact, but it's weird since they advertise 32 bit. The higher sample rates via their ASIO driver work but are pointless for the most part. Anything above 384kHz and Windows doesn't recognize the device anymore, but since many programs won't work with anything higher than 192kHz that's not really a limitation anyway. REW seems to default to 48kHz anyway, and I haven't noticed the slightest difference in measurements anyway.

you will need the latest V5.20 release candidate.

Oh, I've been meaning to get that for a while now ...
Thanks for that great software btw.

 

[KSTR]

The 24 bit limitation dosen't bother me too much, seems to have no impact, but it's weird since they advertise 32 bit.

There is no limitation from RME's driver, the limit comes from the application feeding it. Reaper for example passes 32bit bit test files, other players don't (as they often use 32bit floats which have only ~24bit precision when converted to integer).

 

[3125b]

Whatever the reason may be, this is the selection in the Windows sound options (with 192kHz selected via the driver):

 

[KSTR]

Use ASIO

 

[3125b]

I think I do?


If I'm missing something here, is there an explanation anywhere?

 

[KSTR]

Use ASIO in the application, like in REW. REW plays 768kHz just fine. The driver is multiclient, that means several application have shared access to the driver which then adds their inputs ... but all must be requesting the same sample rate. If a SW uses a different sample rate than what the driver is currently working on, all application with different samplerates are stopped.

 

[3125b]

You're right, REW works completely fine with ASIO (though it does set itself to 24 bit also).
Most programs however simply don't have an option to select the sound output, not even foobar without an additional plug-in.
Selecting 192kHz in the RME driver and Windows works fine with all applications.

 

[polmuaddib]

Forgive my ignorance, but isn't 32 bit pointless for pure playback? Even RME can't surpass 24 bit dynamic range...
I was led to believe that 32 bit is important for DSP, volume control and all that... Something that DAC does internally... And also a bigger number for advertising.

 

[Rja4000]

Forgive my ignorance, but isn't 32 bit pointless for pure playback? Even RME can't surpass 24 bit dynamic range...
I was led to believe that 32 bit is important for DSP, volume control and all that... Something that DAC does internally... And also a bigger number for advertising.

32 bits (float) is important for sound processing (mixing, adding effects, mastering,...).
Not for any input or output.
24 bits (integer) is way above what current commercial ADC and DAC can provide.
And anyway way above what human might ever need.

 

[Rja4000]

If you measure at different levels, you'll be able to plot some interesting graph like this

It's manual with REW (you'll have to measure for each level, then copy value to Excel and plot from there)
That's why I use Virtins MultiInstrument instead, which I'm able to automate from .Net.

 

[JohnPM]

You can make stepped level tests in REW.

 

[restorer-john]

Focusrite 2i2v2 sweet spot (loopback) for reference.