attenuation, voltage, unbalanced connection

Another probably dumb question, and I was not sure whether it should go here or to Hardware/Soundcards. Moderator, please move if appropriate.

I was playing with my new M-Audio Audiophile 192. I was introducing attenuation either with an external attenuator, or with the card's mixer application. I noticed that loudness differed between the two methods when the same attenuation values were used.

I measured the output (p-p of sine wave) voltage and after conversion to dB it turned out that the external attenuator worked fine, but the attenuation from the mixer application was a half of what had been set.

Now, the card has balanced TRS connectors, but my amplifier accepts unbalanced only, so I used TS jacks. Is this the reason for the attenuation mismatch?

If i understand your situation:
As the cause for a half db error in scaling across the range - No.  At least not because you use one end of a balanced output instead of both.  That cuts the signal in half but it doesn't change the attenuation scaling, unless you have also created an impedance mismatch of some kind.  

There would be this kind of discrepancy error if one meter was reading off power in db while the other amplitude.  Just a thought.

unless you have also created an impedance mismatch of some kind.

Thanks. Does plugging a TS-to-RCA cable into the sound card's balanced TRS socket on one side, and into amp's RCA socket on the other side constitute an impedance mismatch?

The impedance mismatch is irrelevant - I'm afraid you only got a textbook answer to the question, not a real-world one.

For a start, there's no guarantee that the scaling values you read on the mixer are correct - and it would only take one to be incorrectly scaled as power to cause this discrepancy. It's been the norm in audio ever since telephone engineering standards were dropped some time in the 1950's to source audio from a low impedance, and bridge it with a higher one as far as 'normal' signals go, and this is why the concept of an impedance matched connection doesn't arise. Normally, you'd be sourcing from somewhere in the region of 50-200 ohms, and bridging with anything from 10 kohms upwards. The reason for this is that it's then possible to feed several inputs from one output without upsetting the levels at all.

As to whether you got a reduction in the output by only using half of a balanced feed rather depends on how the balance was achieved. It's quite common for the output to be artificially floated, in which case its level won't alter when one side is grounded - you simply won't get an output from the cold pin at all, because it will simply connect to a resistor (that has the same value as the source impedance) that's grounded. This is quite a common way of producing a virtual balanced output these days - AKG and Neumann use it with their microphones, and Mackie use it too. You'd only get half of the output from a centre-tapped transformer output, or an electrically balanced output - and you don't get the former on soundcards at all. I'm not even sure how many electrically balanced outputs there are either these days - the single resistor trick is way cheaper!

unless you have also created an impedance mismatch of some kind.

Thanks. Does plugging a TS-to-RCA cable into the sound card's balanced TRS socket on one side, and into amp's RCA socket on the other side constitute an impedance mismatch?

Not in itself, if you have the wiring correct.  I was considering the possibility of a component that has been forced outside its range of compliance somehow, possibly by using miss-matched connectors, or partial component failure.  The db meters should either read the same or...  Does one meter read off at precicely half the other throughout the full range?  That will rule out component failures, in favor of a power reading db meter.

The db meters should read the same or...  Does one meter read off at precicely half the other throughout the full range?

Maybe I did not explain it correctly. It is not about dB meters anywhere in my computer.  It is not about the voltage output from M-Audio being generally wrong. Actually, the non-attenuated voltage was almost identical to the specifications (for non-balanced).

It is about the mismatch of the applied (with card's mixer/control panel applet) attenuation and the actual attenuation, measured extrenally.

Here is what I did exactly.

The setup was:

An audio player application  (foobar2000) playing a wave file with a sine wave (1 kHz, -3dBfs AFAIR) *

M-Audio Delta Control Panel application *
controlling:
M-Audio Delta Audiophile 192 sound card (with balanced TRS connectors)

Hosa TS to RCA cable

RCA to BNC connector

BNC cable

A programmable hardware attenuator (Tucker-Davis Technologies PA-4) *

BNC cable

An oscilloscope

and, parallel to the oscilloscope, Haffler SE 120 power amp which was then disconnected from the speaker

A known amount of attenuation (measured in dB) could be applied using any device marked with * (but I did not touch it in foobar)

I measured peak to peak voltage with the oscilloscope when no attenuation was introduced - both PA4 and the M-Audio control panel were set to 0 dB (V0).
Then, I measured it when attenuation of -10 dB was applied with PA4 only and the M-Audio control panel was left at 0 dB attenuation (VP).
Next, I set PA4 back to 0 dB and applied -10dB with M-Audio Control Panel, and measured again (VM)
Then I calculated actual attenuation levels using formulas 20*log(VP/V0) or 20*log(VM/V0)

The attenuation measured when PA4 was used was approx. -10 dB. When M-Audio Control Panel was used, it was approx. -5 dB.

I tried to set other values, like -20 or - 30 dB, and the results were the same: -20 and -30 for PA4 vs. -10  and -15 for M-Audio.

I asked M-Audio technical support, and their first response was "yes, unbalanced level is lower than balanced level, this is normal", which means that they did not understand the question. Maybe it was my fault so I re-asked the question trying to be more specific and they answered, yes, this is normal when you use unbalanced connections. Still, I do not understand why should this happen with unbalanced. And I have bad experience with the quality of their answers, so I am in doubt, and I'm not sure what's going on.

I asked M-Audio technical support, and their first response was "yes, unbalanced level is lower than balanced level, this is normal", which means that they did not understand the question. Maybe it was my fault so I re-asked the question trying to be more specific and they answered, yes, this is normal when you use unbalanced connections. Still, I do not understand why should this happen with unbalanced. And I have bad experience with the quality of their answers, so I am in doubt, and I'm not sure what's going on.

The answer that they should have given you was 'whoops, we got the scaling on the attenuator wrong'. They have scaled it for power, and you have used (and calculated) the values for level - which is correct. They cannot determine the power output from their card - it doesn't have one!