Drop in level when looping audio interface out to in - is it normal?

[rob_gould]

I'm just starting to test a project and I've found something during set up for the test which I did not expect.

I've just got a new audio interface - Motu 16A, and when I generate a 1kHz test tone in Cubase and then loop it out of the soundcard and directly back into inputs on the 16A, there's a drop in level of about 1.5dB.

I'm using 60cm balanced patch cables to connect the interface outs to ins directly.
I'm generating the tone using Cubase's internal tone generator and a free meter plugin to see the levels more easily than Cubase's built in mixer meters. 
The faders on Cubase's mixer are flat. 
The meter plugin is set to post fader. 
No processing, eq or additional gain settings or anything else are applied either on Cubase or in the Motu mixer GUI. 
Basically, nothing should be boosting or attenuating the signals going out or coming in.

Two questions...

- Is this normal?  Should I expect this level drop when just looping audio interface output to input?

- If this is normal, what's the best way to compensate when doing things like setting unity gain on equipment I'm testing?  Presumably I increase the gain on the input channels by 1.5dB to compensate for the drop so that it doesn't appear that my equipment under test is dropping the signal by 1.5dB.  Is this the correct way to compensate?

Attached is a screenshot.  Right hand side is tone out; left hand side is tone back in.

Cheers,

Rob

 

[Humner]

What happens if you do the same test, but make sure to use a mono channel in Cubase instead of stereo?

Also, do the meters read the same in Motu's Cuemix software?

 

[rob_gould]

What happens if you do the same test, but make sure to use a mono channel in Cubase instead of stereo?

I get exactly the same result.  See attachment.

Also, do the meters read the same in Motu's Cuemix software?

The 16A doesn't use Cuemix software.  It's a new mixer / GUI in a web browser.  But all faders are down and all channels are set to unity gain.  It shouldn't matter anyway because for Cubase to work correctly, all the channels have to be muted on the mixer of the audio interface itself.  If I unmute the channels, I'll get feedback loops.

 

[emrr]

It's normal.  Look at the specs. Differing in/out values.  You can get closer with the trims, but not exact. 

 

[rob_gould]

It's normal.  Look at the specs. Differing in/out values.  You can get closer with the trims, but not exact.

Thanks, good to know.  But looking here

http://www.motu.com/products/avb/16a/specs.html

Which in / out values are you referring to?  The dynamic range?

And does this mean that if there's 1.5dB loss when going from directly output to input, when I'm trying to test a device using the same setup I should compensate by increasing the gain by 1.5dB at the input?  In other words, I should set unity gain by connecting output directly to input by adjusting the gain at the input, and only then should I insert my device under test into the audio path?

 

[Andy Peters]

It's normal.  Look at the specs. Differing in/out values.  You can get closer with the trims, but not exact.

Thanks, good to know.  But looking here

http://www.motu.com/products/avb/16a/specs.html

Which in / out values are you referring to?  The dynamic range?

The specs actually don't indicate the conversion levels (dBFS to voltage). Usually, you'll see something like "+4 dBu referenced to -12 dBFS" or whatever, but for some reason MOTU doesn't bother with that.

And does this mean that if there's 1.5dB loss when going from directly output to input, when I'm trying to test a device using the same setup I should compensate by increasing the gain by 1.5dB at the input?  In other words, I should set unity gain by connecting output directly to input by adjusting the gain at the input, and only then should I insert my device under test into the audio path?

Look carefully at the specs: the maximum line input level is +24 dBu but the maximum line output level is +20 dBu. This implies some amount of attenuation on the inputs to allow it to accept that input level (remember that the inputs attenuate that +24 dBu level to 0 dBFS on the input, which might be something like 2 volts).

On the output side, my guess is that 0 dBFS maps to Full Tilt Boogie, or whatever the op-amps can swing to.

So if you are concerned, trim the input up by 1.5 dB, because you know that the output cannot clip the input under any condition. That is, unless you patch in something that has gain, but even then you shouldn't worry because clipping is well above nominal.
-a

 

[rob_gould]

Many thanks Andy. This explanation makes it clear.

 

[emrr]

MOTU 16A:

The input trims are analog before the ADC.
The output trims are digital and use the DAC's internal trim.
Gain steps on either are 1dB.

Spectrafoo balanced loop test, output has to be set 0 and input has to be set +2 to get +0.26dBr WHEN COMPARING INTERNAL SIGNAL GENERATOR TO RETURNING LINE.
FS output will clip input at +2 input setting.

600 ohm output signal gen sending +24dBu reads 23.7dBm at outputs with multimeter, -0.25 dBFS incoming to 16A set input trim 0. 
16A output set 0 (max), output reads +21.7dBm with multimeter. 
Impossible to get same analog output level as analog input level without DSP gain.

I have a Monitor 8 as well and it's level difference is not the same as the 16A, though on paper it looks like they'd be the same. 

The DAC side is described as 32 bit input and filter architecture, selectable roll-off freq, and volume control.  In my Mac under 'audio MIDI setup' the 26A is described as a 32 bit device.  MAYBE if you have a DAW set to run 32 bit depth it is transmitting all info to/from the converter.
Since it's stepping gain digitally, does it dither it's output?  Don't know.  It does not define an output bit rate for the analog side. 

 

[emrr]

I sent MOTU a suggestion that they implement a unity loop setting on these interfaces.  I don't know if it's possible to toggle the gain steps in fine enough increments.  You can hit them here:

http://www.motu.com/other/feedback/suggestions

 

[abbey road d enfer]

- Is this normal?  Should I expect this level drop when just looping audio interface output to input?

It is not normal, but some calibration error is to be expected from a piece of equipment that is not designed to metrologic standards. Practically, for music applications, a few dB variation is not a major issue.

- If this is normal, what's the best way to compensate when doing things like setting unity gain on equipment I'm testing? 

You need to have a solid reference, such as a meter, and check the actual output level. Then you can compensate.
You may well find that both the input AND output calibrations are out of whack! Any measurement done internally (plug-in) is not reliable because it depends on the actual calibration.
It should not matter much, since most relevant measurements are made at moderate level, generally about 10-20 dB below clipping. Except, of course for evaluating clipping levels.
Remember that loading may introduce a level drop, but ususally it is limited to a fraction of dB.