AAAAgghr only 16bit

[FrankH]

I read this thread and thought: "Huh....really?"

I'm on a Mac/ OS10.6.8. I'm running a MOTU 2408mk3 as I/O between the D8B (v5.1) and the Mac via three OPT-8s. The D8B and the MOTU 2408 are externally clocked by a Symetrix Lucid GENx192. I have three DAWs. Primary is Cubase 6.5.1. Secondarys are LogicPro 9.1.7 and StudioOne2 v1.6.5. I own Ozone 4 and a little Bit analyzer by Schwa called "Bitter" It's AU only.

I ran some tests recording audio sourced from the D8B's Mic and Line preamps. Sent to the DAWs @ 44.1kHz/24bit.

Tried this on all three DAWs (Cubase can't run Bitter) and as far as I can see (with UV off across all 3 cards) I'm getting full 24bit recordings.

(1)With UV on and the Pre's at minimum gain, I can clearly see the last 8-9 bits (16-24) getting filled with dithering noise.

screenshot_01.png (Array KiB) Viewed 1680 times

(2) With UV off nothing at minimum preamp gain.

screenshot_02.png (Array KiB) Viewed 1680 times

(3) With UV off and a -2.5dbfs 1kHz tone on the way in, I get about 23bit registering.

screenshot_03.png (Array KiB) Viewed 1680 times

So I have to ask: What are the settings on the RME? Can you post a screenshot?
I also get about the same thing in Bitter (in Logic and StudioOne) but the Mac version seems to have an entirely different UI than what you've posted.

Maybe it is a v3 issue. I dunno. I haven't used v3 since I started installing v5 betas when we were developing it.

[alfspanners]

Hey Frank,

thanks so much for checking it out. Ok I'm not at the studio today, I'm at home so I can't do every test that I want to, but I have plenty of audio recorded at the the studio on my home drive that I've been able to test. I should have posted more screenshots, but for various technical reasons, it was more difficult than it should have been - I've sorted this now.

A couple of questions - when you have your trim to minimum - is that just your analogue trim on the D8b? To have no bits coming out suggests you have an amazingly quiet (or non existant) noise floor. When I did this experiment, the only ways I could get the 16th bit to switch off (in Bitter at least) was to put the digital trim to off or switch on the the channel filter - which killed the sound dead. Is your channel filter on?

My version of Bitter is recent and runs on VST so I can use it in cubase, though I've taken some readings on Ozone too if that's what more people are used to seeing.

Now, before i go any further I can rule out the RME card straight away. We got our Pc in 2008/9, but I had files on my drive going back to 2006 which were recorded when I was recording directly onto the Mackie HDR24/96. I have checked them out this morning and it confirms it - cubase still says they are 16bit when they were recorded as 24bit wordlength files.


Ok

Definite 24bit signal, created on my home system

proper 24bit.JPG (Array KiB) Viewed 1674 times

The above file is a screenshot of my version of Bitter showing a definite 24bit signal which i had recorded on my home system (Rme fireface 400, Synergy PC, Cubase 6). This is how bitter shows a mono 24bit signal - note how that maximum width is half the screen size.

Now the D8b

D8b "24bit" + cubase stat of file

Bitter and stats showing lower 8 bit anomaly.JPG (Array KiB) Viewed 1674 times

Note the extended bits in the 17- 24 region, the bitter manual states:

" A line the full width of the window (unusual) indicates a stuck-bit error".

The above file has a decent level on it, but the next pic is a snapshot of low level mic/preamp hiss at a quiet passage in a vocal take.

D8b low level

bitter low level signal.JPG (Array KiB) Viewed 1674 times

Note the full extension of the 16th bit and the mirrored/flat line of the 17th to 24th bit, these are moving left and right in unison - like they are all doing the same thing. Here's some examples in Ozone.

Ozone comparisons

Ozone low level comparison.jpg (Array KiB) Viewed 1674 times

The pic on the far left is a proper 24bit signal, but the middle and right had pix are time slices of what I'm seeing in the bottom 8bits ie with a low level signal they're just flashing on and off together with a signal which has a maximum rms at about -61db.

Interestingly when I increase a signal level the bottom bits are added to rather than altered/replaced and that's why the bottom 8 bits often extend to the edge of the page in Bitter.


Finally a mocked up comparison of minimum serttings

Photoshopped representation of our D8b at min settings vs my rme

Comparison at minimum.jpg (Array KiB) Viewed 1674 times

I had to photoshop the D8b one from memory as I'm not there to sreenshot the test this morning, but the 16bit never switched off even with analogue trim to 0 and digital trim to -66 (the next setting down on digital trim is the off position which cut everything)


Anyways, It's doing my head in seriously. I can rule out the RME and also the opt8 cards and cables, as files before the rme have the same problems and if I send a digital signal in and out of the desk it is unharmed.


????????
Mark

[alfspanners]

Finally, although i have not been able to use the same scource signal -this comparison should demonstrate the point.

Waveform comparison.jpg (Array KiB) Viewed 1672 times

Both files were processed with 40db of extra gain, then viewed in cubase's editor at maximum zoom. You can see clearly on the right hand file that the D8b is working at a lower resolution than the left hand 24bit file. A number of the D8b bits have the same value and larger steps between them.

Mark

[FrankH]

I'm beginning to suspect flawed methodology.
The image below displays info on a file generated in Cubase. The file is that of a 1k tone fed into the D8B's Line In. This was piped out via OPT-8 to a MOTU 2408mk3 feeding Cubase 6.5 Project set to 24 bit/44.1kHz.

Cubase's Statistics window estimates this is a 16 bit file. This bit of information made me think: Holy Shit! Could the D8B actually be generating 16 bit files instead of 24 bits? So I looked at the file info opened in Amadeus Pro (lower left) and Quicktime (upper right). The Cubase wave editor is at max resolution. I also opened the file in AudioFinder (but didn't get a screenshot). All three concur that they believe it to be a 24bit file. The lone dissenter being Cubase's estimated resolution.

SInGenLineInD8BRecInC6.JPG (Array KiB) Viewed 1637 times

So the next thing I did was to generate a 1kHz sine wave directly in Amadeus Pro at 44.1kHz/24 bit and save it out.

SinGenInAmadeus.JPG (Array KiB) Viewed 1637 times

As you'll notice, Amadeus, Quicktime (and please trust me, AudioFinder) see it as a 24 bit file. The waveform in Cubase appears no different than the D8B generated version (other than a slight frequency difference...my sine wave generator is not dead on). However, Cubase's estimated resolution puts this file at 12 bits. A 24 bit file generated entirely in the digital domain.

The only thing I can conclude is that Cubase's analyzing function is flawed. And that doesn't surprise me at all.

Other than both showing me that there are 24 bits present, both Bitter and Ozone simply don't show enough detailed analysis/definition/delineation in its GUI to make an accurate assessment of what's going on. They are barely "scientific tools" by my definition.

As far as I can tell, The D8B, running version 5.1 puts out a 24 bit signal from a source generated at the Mic/Line input.

[alfspanners]

Hi Frank,

many thanks for getting back and doing some more tests.

your results are very interesting, but I don't think using a 1khz sine is an effective test of bitdepth.

Using Cubase's own test generator if I render a 1k sinewave, 10th of a second file at full scale, Cubase will tell me it has a resolution of 12 bit. If I render that 1k tone to a 10 second file - I get 19bit. Likewise if I render a 100hz tone at a tenth of a second cubase stats say it is 17bit, 10 seconds and cubase calls it 24bit..

Ok so why would I get different figures....well it goes back to what i was saying before. Cubase's estimated resolution feature doesn't use wordlength to give it's result -IMO it's just looking at the number of discrete levels that there are in a piece of audio or the smallest stepsize to arrive at it's result. if you look at your pix above there are only 22 discrete levels between peak and trough, 44.1 samples per full cycle, this is a repeating waveform with very steep sides sliced into only 44.1 samples so 1 cycle could never represent 16bit (65,536 discrete levels) nevermind 24bit (16,777,216 discrete levels)> The 1k wave will gradually drift inside the sample grid, but it may take some time to arrive at 24bit for cubase's test. A more accurate test would be a lower frequency sine wave with a shallower incline (allowing itself to be desribed by more samples) such as the 100hz tone I did, which came out as 17bit at a tenth of a second (5bits more than the 1k tone) or achieving 24bit resolution at 10 seconds. Incidentally a 10hz wave (the shallowest slope of them all that is a full cycle at 1 tenth of a second) , at 1 tenth of a second, achieves 23bit.

You also show a comparison between the cubase sine and the amadeus waveform, but in my opinion you're niether at anywhere near the magnitude required, nor can the nature of the 1k sine wave show up any differences, because it itself cannot achieve 24bit resolution from just one cycle.

It's also important to note that different test tone generators will output slightly different waveforms, both in amplitude, time and consistancy.

Yes, over time, there have been many bugs in cubase and it was my first reaction to believe that this was one, but when you think about it why should I think there are any less bugs in the mackie when there have already been 2 16bit issues over the years. Youre other software could only be measuring bit depth by the wordlengths in the file ( a test would be to render a known 16bit wave at 24bit and see what the say about the new file).

had it just been cubase's resolution though, I would be less certain that's true, however I's also Bitter, it's also Ozone and it also shows up at the maximum magnification using realworld files. bear in my that I have tested the spdif output of my yamaha motif to the spdif of my rme and it's clearly 24bit in cubase resolution and on bitter - why is it only files that have been recorded with the mackie that are showing up as 16bit.

You also pointed out that you got 16bit resolution in cubase estimated resolution when you sent the sinewave to the line in on the D8b. The line in would have added hiss/noise which would increase resolution (discounting for a moment the actual file length - which was longer 9secs approx?) and help achieve the 16bit (assuming that that was all the desk could really output). In theory, if v5 has this problem too then no matter how long you record this line in for - it will always be 16bit in cubase's estimation.


anyway - I feel we are getting closer to the truth.

thanks again

mark

[anyhorizon]

It's also important to note that different test tone generators will output slightly different waveforms, both in amplitude, time and consistancy (sic).

Why is it important?

Peter

[alfspanners]

Hi peter ,

it's important because the test generators we are using are all different, so we wont get identical results. Certainly in cubase it may take some sine waves longer to reach 24bit resolution depending whether they are 1k or 999.99hz or whatever as it will take longer for the wave to cycle through the 44,100 samplerate using all of the resolution.

Mark

[Petersueco]

For those of you that have RME, download DigiCheck and run the Bit Scope.



Much better bit meter option IMHO.

Peter Holmquist.

[anyhorizon]

Hi peter ,

it's important because the test generators we are using are all different, so we wont get identical results. Certainly in cubase it may take some sine waves longer to reach 24bit resolution depending whether they are 1k or 999.99hz or whatever as it will take longer for the wave to cycle through the 44,100 samplerate using all of the resolution.

Hmmm. Cubase is a moving feast and always has been. With every fix, two other things get broken. I wouldn't be expending too many words using it as the measuring device. Don't get me wrong, I love its MO but it is a fragile little beast. Also, I don't think 0.01 Hz is gonna make all that much of a difference. YMMV.

Peter

[FrankH]

For those of you that have RME, download DigiCheck and run the Bit Scope.

Thanks for posting that, Peter. It only reinforces the last bunch of images I'm going to post. To sum up: what Cubase reports is not to be taken at face value. Bitter, at best is a toy.

This is whole thread started based on flawed methodology. What my last bunch of images clearly show is the inconsistencies trying to get a clear answer from inadequate and flawed tools. The results are empirical at best. I'll take the word of RME over whatever nonsense Steinberg has jammed into Cubase any day of the week.

Untitled-1.JPG (Array KiB) Viewed 1586 times

Untitled-2.JPG (Array KiB) Viewed 1586 times

Untitled-3.JPG (Array KiB) Viewed 1586 times

Untitled-4.JPG (Array KiB) Viewed 1586 times