Higher Sample rate?

[byrd62au]

I've been reading here and elsewhere lately about Nyquist Theorem and the idea that cheap converters can sound better at higher sample rates.

I have a Mackie Onyx 800R that I send via ADAT to my 002. I have been recording at 44.1 24 bit with quite good results. The pres and converters on the Mackie are generally ok for the level I'm at with my recording chops and better than the ones on the 002.

I'm doing a session with a singer songwriter next week involving acoustic guitar and vocals with a bit of harmonica and violin, no drums or electric instruments and small track counts so I thought perfect chance to try 88.2 or even 96K.

I had hoped to use the Mackie Pres and converters for 4 channels at that rate but I see the 002 won't do SMUX on the ADAT optical so I'm limited to 48K. The 002 apparently takes SPDIF at 88.2 or 96K but only 2 channels. I want to record at least 4 tracks simultaneously.

So it I can't get audio from the Mackie into the 002 at 88.2 shall I go to 48K and keep using the Mackie converters or give the 002 converters a spin at 88.2 or 96K? I can still use the mic pres on the Mackie and line into the 002.

Any thoughts on what will give the better sound? I suppose the best thing is just to get in and try them both.

[lowfreq33]

You should be fine at 48. The Nyquist frequency at 48 is 24khz, the roll off probably starts around 22.5khz, which is far enough above what most humans can hear.

A sample rate above 48 is probably the last thing you should be worries about. It really is a tiny difference, and some people can't even hear it (I can, but it's not enough for me to slow my system down over).

Conversely, the difference between 16 bit and 24 bit is HUGE. You should def. be tracking at 24 bit if you aren't already.

[theblue1]

Actually the roll off has to be complete by the so-called Nyquist point (22.05 kHz for 44.1, 24 for 48).

So, if you wanted flat response to 20kHz, your input filtering would have to do a relatively steep roll-off within a very small part of an octave (between 20kHz and 22.05kHz)

OTOH, using a higher sample rate allows the use of an easier-to-accomplish gentle roll off.

That's why a very good converter may produce virtually transparent AD for a full frequency bandwidth of 20 kHz but a "lesser" converter may have an intrusive filter roll off at the high end (or -- worse -- digital artifacts) -- while either one does relatively well at a higher sample rate.

That said, over 96 kHz, the computational overhead becomes so heavy that realtime conversion may become difficult enough to become problematic.


One proviso -- while it's technically possible to do transparent sample rate conversion from 48kHz, 88.2kHz, 96KHz, etc down to 44.1kHz -- not all sample rate conversion software (or dedicated hardware) is created equal.

Check out these SRC comparisons: SRC Comparisons

[byrd62au]

Yes I'm mostly interested in how the theory of cheap converters (002) sounding better at higher sample rates plays out. Coming back down to 44.1 is another story I suppose. I just got Logic 8 which includes compressor. I'm interested to try the SRC in that too.
Thanks for the replies.

[theblue1]

Well, you can check out the sample rate converter comparisons in my post above -- but they only have Logic 7.2.3 -- which did not do all that well in SRC transparency and bandwidth compared to other DAW's included SRC.

But, happily, you can get the free R8brain from Voxengo which does considerably better that Logic 7's SRC in this comparison. (Don't know how Logic 8's is.)

[PettyCash]

Originally Posted by theblue1

Actually the roll off has to be complete by the so-called Nyquist point (22.05 kHz for 44.1, 24 for 48).

So, if you wanted flat response to 20kHz, your input filtering would have to do a relatively steep roll-off within a very small part of an octave (between 20kHz and 22.05kHz)

OTOH, using a higher sample rate allows the use of an easier-to-accomplish gentle roll off.

That's why a very good converter may produce virtually transparent AD for a full frequency bandwidth of 20 kHz but a "lesser" converter may have an intrusive filter roll off at the high end (or -- worse -- digital artifacts) -- while either one does relatively well at a higher sample rate.

That said, over 96 kHz, the computational overhead becomes so heavy that realtime conversion may become difficult enough to become problematic.


One proviso -- while it's technically possible to do transparent sample rate conversion from 48kHz, 88.2kHz, 96KHz, etc down to 44.1kHz -- not all sample rate conversion software (or dedicated hardware) is created equal.

Check out these SRC comparisons: SRC Comparisons

Thanks for posting those SRC comparisons, that was very informative. Any ideal on when comparisons from the newer versions of DAW programs like PT, Nuendo, and Logic Pro 8 will be available?

[Firefox]

simply do your session at 44.1k, 24 bit

the conversion from 48k to 44.1k SUCKS when done by most methods ITB

fyi

keep it simple!!!

[theblue1]

Originally Posted by Firefox

simply do your session at 44.1k, 24 bit

the conversion from 48k to 44.1k SUCKS when done by most methods ITB

fyi

keep it simple!!!

Well... I think, as a general principle, you're probably more likely to get good results on your computer (with good software) than with a realtime conversion box. If there's more time to do the mathematical processing, as a rule, greater accuracy can be assured, if I understand the issues.

That said, a bad algorithm is a bad algorithm. If one looks at some of the SRC tested (and assuming their methodology is good which I have no reason to doubt), you can see some pretty bad SRC... while some of the "lesser" SRC results are still beneath the level of audibility on most systems -- there are a few that have anomalous results spiking well into the audible dynamic range -- and sometimes in a number of frequency bands.


I have to say, the first time I tried SRC in my computer (I think it was using CoolEdit 96) the results were painfully, obviously flawed. No testing required. No closing of the eyes and straining required. The results sounded like someone had taken a rat tailed rasp to them to texture them up. And I ended up doing analog dumps of all the material I had to SRC -- another layer of 16 bit DA and AD was still far superior. But that was then.

[Rick360]

From what I read on Voxengo's website it's only available for Windows OS? Any other good SRC software out there? I've been using Logic on a Mac Pro which is my primary DAW If there is something better out there I'dlike to try it...

[Marty J]

Give BIAS Peak Pro 5 a try. They have an LE version too, but I'm not sure if it uses the same SRC algorithms. I would imagine it does.

[Shadow_7]

My hda-intel (nVida MCP61) does get a better sounding capture at a higher sample rate. Not to imply a good capture.

From what I've heard, the higher sampling rates make the most difference in a mix. So if you're not planning on mixing what you capture, then there probably isn't much need for the higher capture rate.