A/B test Black Lion modded 24io with unmodded simulaneously

[EMRR]

Excellent to see this data. Interesting they 'add' so much noise (like to know more about this) that standard input loading versus open input makes no difference. That's downright abnormal; one would expect to see some amount of change. I have antique tube gear with better noise performance than that.

[zandurian]

BL modded 24 i/0

16 bit 44.1 -90

24 bit 44.1 - 89

[zandurian]

BL modded 24 i/0

16 bit 44.1 -90

24 bit 44.1 - 89

The unmodded shows - 90 for both sample rates

[Don T]

Hello,
Interesting, 50% less noise than my units. That's a substantial improvement. It makes me wonder if it could be further improved by the use of grain oriented steel shielding of the analog components and possibly the clock and PS. I guess I'm going to have to dig into my 24i to see what can be done. Wish I could get a schematic for it somewhere.

[willheim]

can someone clarify for a non tech user with bla mods - do they have less noise floor than the unmodded units?

[EMRR]

inconclusive; we have conflicting results in different systems.

[willheim]

thanks

[Don T]

Hello,
Might depend on how we are reading the meters too. I let it play for 15 seconds and read the peak noise. It is a random noise I suspect.

My units are not modified, yet have 50% more noise than his unmodified units. On his system, both the modded and unmodified units have essentially the same noise. So my guess would be that BL does not modify the dither / masking noise.

[zandurian]

24 I/O = -86db
24i = -89db
2408 mk1 = -86db @ 20 bit

My units are not modified, yet have 50% more noise than his unmodified units. On his system, both the modded and unmodified units have essentially the same noise.

50% more noise? I'm posting around -89db to -90db noise floor with both units, 0 to 1 db off of your 24i mark and 3 to 4 db off of your 24i/o mark. I'm not sure how you are getting the 50% figure.
- Byron

[EMRR]

he's saying 3db = 50

Unplug source from the input of the 24 I/O and record silence. Then play it back with either a couple of trim plugs with maxed out gain OR open the meter bridge and adjust the resolution of the meter to show the noise.

Do we know (1) there is no noise contribution from a trim plug (2) that the meter bridge resolution is to be trusted in that range? In other words, has anyone used any external methods to confirm accuracy?

[EMRR]

Veering more off topic, I ran my own A/D converter noise tests today after thinking about the numbers posted. I looked at the 2408mk3, 24I/O, and Alesis HD24XR, all at 88.2/24 resolution. Also checked the old Adat XT20 machine against the 2408mk3 at 44.1/24. Here's the data:

I found DP's ability to reveal meaningful noise stats iffy at best. With a 30 second simultaneous DP recording of the 2408mk3, 24I/O and HD24XR I got a consistent -90 dBFS noise metering that I could only see after applying a +40 dB boost with trim plugs. I could not see any action on the meter bridge with the resolution scaled correctly, so -90 dBFS comes from a meter bridge reading of -50 dBFS minus the +40 dB gain added by the trim plugs.

I then abandoned DP and had a look at the inputs in Spectrafoo. I had the metering hold time set at 20 seconds, and after letting each input settle in for about a minute I took notes. Spectrafoo metering gives input numbers for:

peak power / peak RMS / VU (avg. RMS)

So with those three in mind, here's what I got, in the same order, for each converter:

HD24XR -89 / -101 / -103
24I/O -87.5 / -100.75 / -102
2408mk3 -86 / -99 / -100 (best input)

Those numbers represent 'open' inputs with no source impedance present. I checked all three with a 560 ohm load resistor on the input and noted no change with the HD24XR and 24I/O. The 2408mk3 initially had terrible numbers with lots of power supply harmonics that varied widely channel to channel (as bad as -75 / -81 / -83), but improved significantly once I connected an input. Strange thing is, once I disconnected the source impedance the improvement remained. Some sort of input sensing circuit, or a poor ground scheme? The 2408mk3 inputs were all over the place when compared; note I did not look around at various 24I/O inputs to see if results changed. I did check two HD24XR inputs and got same results.

Switching to 44.1/24:

2408mk3 -86 / -99 / -100 (best input)
ADAT XT20 -90.25 / -103.5 / -104

As raw numbers, the old ADAT machine wins! Of course it's the worst sounding of the lot.

A check of adat dig input with no AD inputs enabled showed - inf. dBFS for reference.

Finally, noise weighting observations, which I think reveal the most; 88.2 kHz spectragrams showed the following:

HD24XR shows noise higher than approx. -117 dBFS only at 32 kHz and higher.
24I/O shows noise higher than approx. -117 dBFS only at 8 kHz and higher.
2408mk3 shows noise higher than approx. -117 dBFS only in the 4-8 kHz range and higher.

44.1 kHz spectragram for the 2408mk3 shows the noise level moving down about an octave (above 14-15 kHz) for the stuff above -117 dBFS in level and the -117ish stuff staying about the same in frequency range. The XT20 looked better at 44.1 kHz than the worst four inputs of the 2408mk3 at 88.2 kHz.

AC power related noise only appeared with the 2408mk3, and disappeared once I had something connected.

I took screen captures of various spectragrams, which I need to edit and then post for consideration.

[Don T]

Mr. Williams,
Very interesting indeed! Thanks for the in depth test, I know it's a lot of futzing around. I'll have to test the Digidesign 96 I/O's noise spectrum to see what it shows. I hadn't thought of testing my old gear (panasonic 3900 dat machine) to see how it compares. Oh wait a minute.......all my gear is old!

[EMRR]

Here's thumbnails with links to larger screen captures.

88.2 kHz 24 bit spectragram of A to D converter input noise. Left to right, Alesis HD24XR, MOTU 24I/O, and three MOTU 2408mk3 readings. 2408mk3 readings are best noise, worst noise, then effect of plugging a 560 ohm input load resistor into input. Input load negated the power supply bars seen in the first two. dB/color scale can be seen at top left.

http://farm3.static.flickr.com/2321/249 ... 1d53_o.png



88.2 kHz 24 bit spectragram of MOTU 2408mk3 A to D converter input noise. All 8 inputs shown side by side. Note this result was taken after connecting and disconnecting an input loading resistor, which cleared up a lot of power supply noise.

http://farm4.static.flickr.com/3217/249 ... 0575_o.png



44.1 kHz 24 bit spectragram of A to D converter input noise. Alesis XT20 Adat on the left, and MOTU 2408mk3 on the right.

http://farm4.static.flickr.com/3218/249 ... d9ce_o.png

[zandurian]

Thanks for posting info Doug.
- Byron

[EMRR]

sure thing. I think it's probably most important to note the noise shaping aspects of the readings, and consider the extreme difference of the peak 'any and all frequencies' reading versus the actual average levels at commonly audible frequencies. The peak meter is very misleading in a test like this.