ProSoundWeb Community
Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => Audio Measurement and Testing => Topic started by: Mark McFarlane on July 11, 2013, 07:15:12 PM
-
I was noticing tonight using my Motu 896mk3 interface that I got a fair amount of noise in my equipment measurements (hardware tests, not speakers, unfiltered, unaveraged) if I let the reference or measurement inputs get into the yellow range. If I kept the meters right below yellow the traces were stable, if I got 1/2 way into yellow I got a lot of noise on the traces.
Is this expected?
I didn't notice this much sensitivity to noise when I tried two other interfaces over the past few weeks, but maybe my memory is bad.
-
Generally you only need enough gain to ensure you have a good signal to noise ratio. Typically we recommend shooting for -12dBFS for acoustical measurements. This should have good S/N and leave plenty of headroom.
For an electronic measurement you can get away with much less, I usually shoot for -30dBFS or more. Though if it's too low you may be tickling the magnitude threshold. So I generally run with some some averaging. I want it to be responsive so I use 2-16 FIFO or Fast exponential.
-
I was noticing tonight using my Motu 896mk3 interface that I got a fair amount of noise in my equipment measurements (hardware tests, not speakers, unfiltered, unaveraged) if I let the reference or measurement inputs get into the yellow range. If I kept the meters right below yellow the traces were stable, if I got 1/2 way into yellow I got a lot of noise on the traces.
Is this expected?
I didn't notice this much sensitivity to noise when I tried two other interfaces over the past few weeks, but maybe my memory is bad.
Mark, are you running the interface 16 bit or 24 bit? Quite frankly, given the limited dynamic range of many sound reinforcement systems and the noise floor of the rooms we find ourselves in, 16 bit is perfectly fine for sound reinforcement system measurements, even with the higher electronic noise floor of the interface. As Adam has pointed out, get a good signal to noise ratio.
Let me illustrate: So if a room has a background noise floor of 50 dB SPL, and the system outputs a max of 110 dB SPL, the difference is only 60 dB. However unless you want your ears to turn to mush, you don't measure a system with pink noise at 110 dB SPL unless you want to annoy everyone else. So usually I do my system alignments at a more comfortable level so that I'm at least 20 to 30 dB above the background noise of the room, so say 80 dB SPL in a 50 dB noise floor room, assuming the system behaves the same whether at 80 dB or 110 dB (it's supposed to be linear). Effectively only about 30 dB of difference. 16 bit is around 96 dB of dynamic range, several orders of magnitude greater than what you're trying to measure, so no need to sweat the small stuff here. Just watch your Coherence curve to ensure that background noise does not encroach on your measurement.
If you're measuring the electronic response of an audio device, or a optimizing a system in a recording studio, or doing some laboratory work, or making NC measurements of a room's mechanical system, that is a different issue altogether, and electronic noise floor of your interface can be rather important there. The lower the better here. Put away the high noise floor 1/4" measurement mics, and get out the lower noise floor 1/2" and 1" diaphragm test mics, and set your expensive A/D interface to 24 bit!! :)
-
Thanks Adam and Arthur.
I usually run at 24 bit. I don't think that will 'hurt' anything. I appreciate the +30db over noise floor target for acoustic measurements.
In my particular situation that prompted this post I was measuring hardware responses, no mic involved. Just reference before and measure after an EQ device to see what the EQ device was doing.
If I had the ref and measure meters in the yellow I was getting a lot of spurious noise in the measurement. Lowering the pink noise level so both meters stayed in the green gave more stable results.
I've been evaluating Smaart with 3 different interfaces, Roland UA-25, ART USB Dual Pre, and Motu 896mk3. It seems like I got stable hardware-only traces with the 2 cheaper interfaces when measuring in the yellow zone but the Motu gave me noise, but it could be bad memory, I'm over 50. I'm currently using a demo of Smaart and can't save the curves so I'll have to retest.
Below is a sample graph of the phenomena, the 'clean' curve is with the pink generator at -44dB, the noisy one at -8dB.