Hey guys,

I recently purchased an emu 0404 and have just set it up. I've run some RMAA tests on it with two balanced TRS cables (Hosa brand, if it matters). However, when comparing them to the RMAA tests I found on google, it seems my card is underperforming.

Here's the website's measurements:

http://www.amb.org/rmaa/e-mu/E-MU_0404_USB...bal_16b_44k.htm

and mine:

http://unchained.nu/logomania/0404.htm

There seems to be something very fishy, especially with respect to the noise levels. I'm comparing the PCI (mine) to the USB (website) version, but that shouldn't make a difference should it? The graphs seem to indicate better performance than my card puts out. What could account for this difference? FWIW I've checked the "Normalize amplitude of test signals before analysis" box, but I still get "Normalize amplitude OFF" on the results page. I'm not sure how to change that, and I'm equally unsure if that change will effect the results noticeably.

Tips/Suggestions/Things I'm doing wrong?

Thanks.

Hi,

The pci and usb versions are very different animals. The things in common are the number of inputs (04) and then number of inputs (04). The usb version is generally perceived as being the better one.

Adriaan.

I don't see a significant difference. A couple dB more/less noise at that low level - no biggie, nothing you're likely to hear.

I'm not so much worried about the average levels, but moreso the levels at various intervals. I may be wrong, but I think those frequencies are at harmonics, and they're in a predictable pattern across the various measurements, which leads me to believe it's something that can be remedied.

Route a song you're familiar with with good dynamics and some quiet passages 5 times through the loop, then ABX it. If you can hear a difference, it may mean nothing (because it's "5x" the noise) - but if you cannot hear a difference, you have black on white that there ist no reason to worry. A one time loopback could be more practical, but the 5x can teach your ears what's actually happening inside the unit.

PS: Just don't mess up levels or it wont work. And I forgot (hello Arnold): you must time sync the record with the original before ABXing.

Here's the big picture:
unit A unit B What matters
Frequency response (from 40 Hz to 15 kHz), dB +0.07, -0.04 +0.02, -0.05 > +/- 0.3
Noise level, dB (A) -93.5 -96.1 < 80
Dynamic range, dB (A) 93.6 -96.3 < 80
THD, % 0.0013 0.0011 > 0.05
THD + Noise, dB (A) - 86.5 89.7 < 80
IMD + Noise, % 0.011 0.0045 > 0.05
Stereo crosstalk, dB -96.3 -96.3 < 20
IMD at 10 kHz, % 0.0057 0.0057 > 0.05

Now some of the numbers differ by as much as 3:1, but the point is that they are so small in either case that they don't matter. The FR curve has a vastly expanded vertical scale, which makes bumps that are tiny in the big picture look large.

Kinda strange... so I noticed an "RMAA" setting on the 0404's patchmix program. I picked it, and ran some tests, and got better numbers but IMD at 10khz was outrageous, something like 7%. So I switched back to the default settings and ran the test again, and got:

http://unchained.nu/logomania/0404p2.htm

The graphs still look worse than the website I'm using as a reference, but the measures are all better. I think I'm just going to chalk this up to me not understanding how to use patchmix properly.