Sample-rate mismatch woes

[BetaMark]

I've got my Kronos talking digitally to Logic via RME Babyface and have no trouble recording audio tracks from the Kronos. The problem starts when I try to add said track to an existing project, the hiccup being that the Kronos sample rate is fixed at 48 khz, and my existing projects are at 44.1 khz.

If I leave the project at 44.1 khz, I get constant reminders from Logic of the sample-rate mismatch, and of course, I can't input digital audio from the Kronos. If I change the project to 48 khz, any existing audio tracks in the project play back at faster than the original tempo and at a higher than original pitch.

Is there any solution to this conflict?

[popmann]

Other than recording it via analog connections?

I really thing that the "digital revolution" is being held back by people's perception that digital connections are necessary. It means limits everybong in a recording system to the lowest common denominator.

So, there ARE solutions...I mean, my digital recorder (Akai dps24) actually resamples the SPDIF input in real time so you can connect things of different rates. There are other digital converters that do sample rate conversion. But, you're going to be spending money for something not worth it. A serial digital transfer isn't perfect-add real time SRC--just get two really nice balanced cables (or 4) and record the analog audio out.

[popmann]

Side note...This is a real issue with software sample based instruments. They, are designed to live inside another clocked system (daw)...yet being sample based, those samples ARE sampled at a certain rate. Move it from that native rate, and their engines do real time SRC-some on the plug output, some note by note. Most you pay a sonic penalty...and where you don't pay too high a sonic price, it increases the CPU load exponentially to do the real time SRC.

If I'm doing string arrangments, my LASS is 48k and my VSL 44.1. Which do I use? Wait...the live tracks I'm sequencing to were tracked at 96k!

Anyway...my point is...digital instruments are designed to run at their optimum rate. connect analog and everything CAN run as designed. Digital...you're introducing side effects of SRC and jitter of externalclocking...frankly, becoming cumulatively not as "pure and better" as...different. Color of the converter trip. Color of all the digital processing to make it interconnect.

If you don't like the sound of the rme's AD, it would be better to spend the money on a nice AD than SRC gizmos to maintain digital connection. IMO.

[danatkorg]

Side note...This is a real issue with software sample based instruments. They, are designed to live inside another clocked system (daw)...yet being sample based, those samples ARE sampled at a certain rate. Move it from that native rate, and their engines do real time SRC-some on the plug output, some note by note. Most you pay a sonic penalty...and where you don't pay too high a sonic price, it increases the CPU load exponentially to do the real time SRC.

If I'm doing string arrangments, my LASS is 48k and my VSL 44.1. Which do I use? Wait...the live tracks I'm sequencing to were tracked at 96k!

The rate at which the samples were captured and the rate of the engine are two completely separate issues. Changing the pitch of a sample requires interpolation, which is essentially sample rate conversion. Most samplers should be able to play back samples recorded at essentially arbitrary sample rates; the KRONOS certainly can and does.

Now, the sample rate of the engine is a different story. It's certainly possible that a plug-in would always run at one rate and simply use SRC to communicate with the host. The KRONOS USB audio driver does this, for example.

You mention VSL, and from the benchmarks at (http://csdev.vsl.co.at/vienna_symphonic ... 19725.aspx) it appears that the per-voice CPU cost scales more or less proportionally with sample rate. This implies that the engine itself is running at the selected rate. If it was simply using SRC at the outputs, the CPU cost would not scale in this way. I expect that most plug-ins are similar to this.

Anyway...my point is...digital instruments are designed to run at their optimum rate. connect analog and everything CAN run as designed. Digital...you're introducing side effects of SRC and jitter of externalclocking...frankly, becoming cumulatively not as "pure and better" as...different. Color of the converter trip. Color of all the digital processing to make it interconnect.

Re SRC, see above. Re jitter, note that jitter is only heard in the analog domain (provided that it remains low enough that the clock can be recovered successfully), and cannot be captured when recording from synchronized digital signals. In other words, if you record something via S/PDIF (such as the KRONOS, for instance) using a standard, non-SRC digital input with proper clock connections and settings, jitter will not be recorded. For a little more detail, see Myth #5 here:

http://emusician.com/mag/emusic_debunki ... dio_myths/

If you don't like the sound of the rme's AD, it would be better to spend the money on a nice AD than SRC gizmos to maintain digital connection. IMO.

To me, this is a separate point from everything above, and in general I agree.

[SanderXpander]

Thanks Dan, that was very informative.

As a somewhat separate point, while there are certainly better ADs to be had, I think RME's are generally quite well regarded. Unless you are prepared to spend a lot or to audition a ton a develop a feel for different ADs I would advise against getting different ones purely based on someone's recommendation.

[popmann]

The rate at which the samples were captured and the rate of the engine are two completely separate issues. Changing the pitch of a sample requires interpolation, which is essentially sample rate conversion. Most samplers should be able to play back samples recorded at essentially arbitrary sample rates; the KRONOS certainly can and does.

Cool...I guess I'm projecting based on half knowledge... ...but, REAL every day knowledge of what happens when you take a VI out of it's native sample rate. I've had several that I could HEAR pitch being raised. Granted...this was always going from mostly 44.1 in a 96k project--not really 44 to 48...and it could be that the similar unrepeatable weirdness I've had with LASS (native 48k) running in 44.1 projects is unrelated.

But, that's a big stretch for me to think all are completely unrelated. I do get what you're saying about a sampler being able to stretch pitch up and down in real time...it only makes sense that they would be able to do a resample on the fly...

Re: the myths...see point 6 right after the debunking of jitter (which was definately not the right term on my part)...really-that's my point. A digital transfer is NOT a perfect copy. Nor, in this case, is it one you've ever HEARD--ie, you're listening and playing the Kronos through it's A/D, but then you want to record it via digital connection--how do you even know that sounds like you wanted it to?

I do appreciate you calling me on some of the tech details. My overall point being--digital interconnection is really not key to good sound. File level transfers from recording system to another is best. But, when you need to interface with gear--just hook it up analog, particularly if for some reason it's a PIA to go digital. Not a big deal.

[danatkorg]

Thanks Dan, that was very informative.

As a somewhat separate point, while there are certainly better ADs to be had, I think RME's are generally quite well regarded. Unless you are prepared to spend a lot or to audition a ton a develop a feel for different ADs I would advise against getting different ones purely based on someone's recommendation.

Yes, sorry, I didn't mean to imply anything re RME.

[danatkorg]

Cool...I guess I'm projecting based on half knowledge... ...but, REAL every day knowledge of what happens when you take a VI out of it's native sample rate. I've had several that I could HEAR pitch being raised.

That can certainly happen when things get confused about what the sample rate is supposed to be. Usually I've seen this when syncing to an external clock in systems which do not auto-detect the sample rate. For instance, you tell the system that the clock is at 44.1, but it's really at 48, and so everything's transposed up by a semitone or so.

But, that's a big stretch for me to think all are completely unrelated. I do get what you're saying about a sampler being able to stretch pitch up and down in real time...it only makes sense that they would be able to do a resample on the fly...

Yes - changing the pitch of the sample and changing the sample rate are two sides of the same coin. Both use the same basic process: you derive a continuous curve from the discrete sample data, and then choose new sample values from different locations on that curve. Whenever you transpose or re-tune a sample at all, even by a single cent, you need to do this (unless you're using very old-fashioned systems with individual D/A for each voice, which would transpose by changing the sample rate of the D/A).

Re: the myths...see point 6 right after the debunking of jitter (which was definately not the right term on my part)...really-that's my point. A digital transfer is NOT a perfect copy.

The myth that section addresses is the idea that digital transfers are *always* perfect. This doesn't mean that they're never perfect - just that they're not perfect in the specific cases described. The first part is about error correction on DATs and CDs; the second part is mostly about clock errors, with brief mentions of bad cables and insufficient buffer sizes.

My overall point being--digital interconnection is really not key to good sound.

Sure. It can solve problems sometimes, but it can also come with its own problems.

[Redrain]

a simple question :

I noticed that audio file added in song_A folders have a 48k added at the end, but not on all track.

I have in the same folders file, under the same song, with no 48k added at the end of the name, why ?

thank you

[apex]

I haven't had a chance to "re-read" all this information. But I was wondering about something...

I am trying a few 'new' things in my setup one being that I'm trying the ol aggregate device again. Combing my Onyx, My Kronos, and my XF.

Well when I create the aggregate device the Kronos option shows an error telling me that there is a sample rate mismatch (when I Combine the Kronos with the XF or the Onyx (the onyx and the xf are the same 44.1). Anyway, I can go in and change the sample rate of the onyx and the xf to 48 to match the Kronos and now the error is gone.

Is this a good or bad thing? are there real disadvantages or advantages to either of the options?

Thanks

[SeedyLee]

I say, let your ears be the judge. Almost any type of sample rate conversion results in a degradation in signal quality*, whether it's performed in the analogue or digital domain. Which works better will depend on the individual equipment involved. Some equipment is better at sample rate conversion than it is at ADC/DAC and vice versa.

For me, when I first started out I tried to keep everything in the digital domain believing it would help minimise noise. And whilst that's often true, the additional complexity meant that it was often a hindrance to making music. Which is worse than any barely audible signal degradation.

[ed_f]

I haven't had a chance to "re-read" all this information. But I was wondering about something...

I am trying a few 'new' things in my setup one being that I'm trying the ol aggregate device again. Combing my Onyx, My Kronos, and my XF.

Well when I create the aggregate device the Kronos option shows an error telling me that there is a sample rate mismatch (when I Combine the Kronos with the XF or the Onyx (the onyx and the xf are the same 44.1). Anyway, I can go in and change the sample rate of the onyx and the xf to 48 to match the Kronos and now the error is gone.

Is this a good or bad thing? are there real disadvantages or advantages to either of the options?

Thanks

I am doing the aggregate device thing too and have seen exactly what your talking about.
This is not an area where I am an expert but in my opinion, it is a good thing. If all of your devices have the ability to run at 48 then I think you are better off in all cases keeping them the set up to match. But in an aggregate situation it just kind of makes sense that "one" device should have the same rate no matter how you route in or out of it. Personally if a piece of gear could not be at the same rate, I would not try and make it part of the aggregate, because I think it attempts to pull everything down to the lowest common rate.

[apex]

But are these "rates" only applied when the devices are "aggregated" or does it change them PERIOD.

and as far as sound goes... what's the difference?

[ed_f]

But are these "rates" only applied when the devices are "aggregated" or does it change them PERIOD.

and as far as sound goes... what's the difference?

Here is what support says: You can set the sample rate and bit depth in the Format pop-up menus. Make sure they match the appropriate settings for your audio device, and that the input and output sample rates are set to the same value.

I guess it might be possible to have different rates going but I keep everything locked in the same so nothing is converting.

As far as sound. Well, that is a huge topic and you can find all kinds of information out there. I think it is safe to say the basic premise is the higher the bit rate the more information you capture and the more accurate the sound. The Kronos is locked in at 48K. I have set everything to that and had great results.

[apex]

So what would happen if say I put the Kronos in at 48, (via USB), but then analog the XF in with L/R cables.