Well, I did some testing tonight on my 990 and here's what I've found...

Testing with pink noise, even using the same 30 second snippet of pink noise designed to have very regular distribution results in some anomalies that suggest that testing with test tones may be easier. However, the data appears significant enough to draw some conclusions.

I ran three tests. I used the IASCA 1995 competition CD, track 27 which is pink noise, and a Sound Devices 744t recorder set to analog input at 16-bit, 44.1kHz with levels ~ -1dB peaks. I recorded the first 30 seconds of the pink noise track while playing back from my CD player:

First: CD Player RCA outputs > Sound Devices 744t
Second: CD Player > 990 set to Large speakers, bypass mode > RCA L/R line outputs > 744t
Third: Same as Second but with Small speakers.

I opened up the recordings and did a spectral analysis of the same section of audio (Blackmann-Harris with Fast Fourier Transfer size of 8192 samples.) The Second recording (990 Large) looks very much like the First in the bottom 2.5 octaves. But the Third run shows some dramatic dropoff in the lower registers. It's not so even and regular looking that I can absolutely identify the crossover type or slope, but extrapolating through some data, I would guess that it's a 12dB/Octave slope with a -3dB (crossover point) of 70Hz.

Some interesting data...
- down ~ 26dB at 20Hz
- down ~ 17dB at 40Hz
- down ~ 10dB at 50Hz
- down ~ 3dB at 70Hz

Hopefully we can all agree this is a VERY audible filter. I can't say if it's digital or analog, but it's there... and it's not supposed to be.

More tests if feedback warrants.

- Jason