Using the SPDIF output.....48k?

[kitar77]

Hi,

Don't post here often, but enjoy browsing all the great info at least once a day....
Quick question, Do I have to have my recording gear set to 48k in order to use the digital output on my m3? If so, why wouldn't they have used 44.1k?

Any help would be very appreciated....Thanks, Kirk

[Christian Baum]

Yep, 48 kHz only, no other way, no switch, no downsampling, no nothing.

Why? Probably because the whole internal sample system runs on 48 kHz (I guess the ROM samples themselves are 48 kHz). Permanenty downsampling it for digital output would probably fragg the whole thing up (maybe too much CPU usage or something).

I decided from the very beginning to rather use the analog outputs. At least I have six of those.

Best regards,

Christian

[sani]

I can only agree.
Having 48 kHz on the digital output is completely unnecessary and wrong.
I highly doubt that there is any audible difference between samples taken on 44.1 and 48 kHz. This is just a waste of memory space in the internal ROM.
On the other side, most records in studios are still done on 44.1 kHz. So, it makes no sense to record the digital outs on 48 and make a project based on 48 kHz because at the end of the work, the mixdown will most probably stil has to be at 44.1 for the CD audio.

[X-Trade]

I can only agree.
Having 48 kHz on the digital output is completely unnecessary and wrong.
I highly doubt that there is any audible difference between samples taken on 44.1 and 48 kHz. This is just a waste of memory space in the internal ROM.
On the other side, most records in studios are still done on 44.1 kHz. So, it makes no sense to record the digital outs on 48 and make a project based on 48 kHz because at the end of the work, the mixdown will most probably stil has to be at 44.1 for the CD audio.

However, standard for DVD quality 5.1 is 48k 24bit, and the M3's EDS 'enhanced definition' sound is ideal for film scoring and other such high fidelity audio work.

[Christian Baum]

Indeed, but mixing inside the M3 to a level that can compete with Logic and other DAWs and THEN put it out in stereo, is just not very sensible IMHO. If you get only single sounds or combis from the M3, that seems to be okayish, but as soon as you're up for more, it becomes tiresome. If at least the FW expansion worked like the Virus TI's USB connection... But it doesn't. So I think, even for film scoring (which I do a lot these days), analog outs are more flexible (if you have the inputs to take them on).

Best regards,

Christian

[Rob Sherratt]

The reason for a 48 KHz sample rate in all Korg keyboards is because of the audio codec chip that has been used for the sample-plus-synthesis sound generator. They are probably using a G.719 codec with AAC-SSR compression which has a fixed sample rate of 48 KHz. It also allows Korg to compress their samples up to 100% more than would be the case if a different CODEC was used. It is a very efficient codec.

Having made the decision to use a codec with fixed 48 KHz sample rate, the internal digital bus on the M3 has also to be synchronized to 48 KHz and the S/PDIF and Firewire hardware is also locked to this. It would cost extra to fit a real-time sample rate converter that would not cause loss of fidelity. You can not just convert A to D and then resample D to A at the new rate, otherwise the SNR and sound quality suffers. It has to be done by digital processing but it's complicated to do it in real-time. e.g. here's a sample rate converter that works in real time, you can see that it's not cheap.

http://www.fullcompass.com/product/233763.html

It is cheaper and more flexible to connect from the M3 using analog audio cables to a unit that has a variable sample rate A-D converter front end. I now use a Yamaha N8 with which you can set the internal digital mixing bus to whatever sample rate and bit depth you want.

Alternatively you can work your audio project at a 48 KHz sample rate throughout, and then use non-real-time digital processing to do a final mix-down export at say 44.1 KHz. This should not lose fidelity, it just takes a while to do the processing and the software can not keep up in real time. But most DAW software packages eg Cubase, Sonar, Pro Tools etc provide the capability.

Regards,
Rob

[jg::]

Korg made a far-sighted decision back in the 90s with the Trinity to adopt 48kHz as the internal sample rate. It's one of the reasons why Korg sounds good. Those extra few k may not seem like much, but it does add to the openness and transparency in the top-end, imho.

While some studios do still adopt 44.1kHz, I'd suggest that more pro studios are running at 48, 96, or even higher rates these days. The best engineers do this because they can hear a difference.

Fortunately, Korg use decent D/A converters, so using the analog outputs is actually a reasonable choice if you must use 44.

As Rob suggested, I prefer to work at 48kHz to take advantage of digital recording, and then convert the finished tracks to 44 in my DAW if I need to burn a CD, or for some other purpose.

jg::

[sani]

It is hard to say if downsampling from 48 to 44.1 is or should be losless. The program/converter has to decide which samples it will pull out. So, as an endresult there will be "holes" in the digital wave print at 44.1 kHz because the converter has to cut out about 4000 wavecycles out from the wave.
Sorry, but I don't see any reason why should anyone record at 48 kHz if at the and the master will end as a 44.1 kHz wave?!
It would make a sense if one would make a 48 kHz digital master and then transfer it to an analog media, but as far as we all finish our masters at 44.1 to be compatible with an CD audio media, I don't see any reason to record at 48. Even worse: often I receive some files which are done in other studios as part of a project I'm working on. All this files are always 44.1/24 file formats. So, no need and reason to use 48 kHz. This is a relict when masters were done on data recorders, those small data cassetes.
Everybody who uses a computer for recording or sequencing can make a simple test.
Record something at 94 kHz and then downsample it to 24 kHz for example or even less. What will happen? The sound will get muddy and dark. The same will happen if you record at 24 kHz or less from the beginning.
The only difference in recording at 48 is that you have one step more to do at the end of the mastering process.
Recording at 48 kHz makes only sense if the endresult will stay at 48 kHz.

Regarding the crispness and good sounding of the Korg keyboards due to the 48 kHz sample rate, I disagree here completely.
A good sound for me is a well sampled sound: more samples per octave, longer recording times, longer loops and other factors. Not the sample rate. A Motif or Kurzweil certainly doesn't sound worse because they use 44.1 or even less sample rates (as Kurzweil does).

Remember the famous original Triton Classic Piano patch? It is reported to be 3 Mb in size and it is sampled at 48 kHz. Does it sound good? I don't know anybody who ever liked that piano. On the other side, Kurzweil made a piano for the K2600 which is 4 Mb in size but stereo. So it would be about 2 Mb if it was a mono sample. And the vast majority likes the sound. It is the same sample that Kurzweil uses even now in its latest models (PC3) and most people stil like that piano. And it is definitely not sampled at 48 kHz. Sounding crisp and sounding good are different terms.

[X-Trade]

It is hard to say if downsampling from 48 to 44.1 is or should be losless. The program/converter has to decide which samples it will pull out. So, as an endresult there will be "holes" in the digital wave print at 44.1 kHz because the converter has to cut out about 4000 wavecycles out from the wave.

A bad downsampler would do this.
A good quality downsampler will re-scale the actual sample rate and interpolate new values for each sample that falls in between the current samples based on its difference between the old ones.

Much cleaner, pretty losless. But obviously you do lose that advantage of having 48k in the first place when you downsample to 44.1 - if you can hear the difference.

[sani]

Just to make one thing clear:
downsampling and losless aren't going together.
Downsampling means always a loss.
Downsample 48 to 12 kHz with the best algorythm available and see if its losless or not.
It's another question if you hear the difference between 44.1 and 48 but it goes worse. There is no interpolation of any values because you are taking something away. You are not upsampling. I'm not an expert in all this things, but all I know is that there is no audible advance in having a keyboard with 48 kHz samples over all those on 44.1. It is practically inaudible and it is a hassle to use in studios if the endresult has to be something at 44.1/16 bit.

[X-Trade]

There is no interpolation of any values because you are taking something away. You are not upsampling

I think you somehow just completely negated my point. I'm not saying its totally lossless, but it doesn't just take a bunch of samples out.


If you have two different rates you get an effect like this:

A....B....C.....D....E....F.....G....H....I.....J.....K
|.....|.....|.....|.....|.....|.....|.....|.....|.....|.....| 48kHz
|......|......|......|......|......|......|......|......|.... 44.1kHz
a.....b......c......d.....e......f......g.....h......i.....

| = 1 sample


So instead of having a situation where a = A, b = B, c = C, d = D, e = E, f = F, then drop G, h = H, etc
what it does is it generates completely new values for those that fall between the old ones, so for example D = average between values for D and E (because it falls exactly halfway between). Other better algorithms may take even more points into account and use a spline point vector to calculate the curve of the waveform and produce an even smoother result.

I'm not an expert in all this things

Well, I do know what I'm talking about. I'm taking a degree which involves computer audio processing and programming.


Not saying that its any easier to downsample to 44.1, or that there isn't a disadvantage, but a good converter should produce the least distortion of the audio. Main point is most converters won't just drop a few samples.

And 44.1 is commonly used for CDs, but there are plenty of other formats now that use 48, so I don't have a problem with going for the higher standard and downsampling later. You shouldn't end up with anything worse going down from 48 than if you'd recorded in 44.1 in the first place, unless you're converting with something from the 80s.


edit:
forum condensed my spaces, used dots instead.

[Rob Sherratt]

Thought for the day

The audible frequency range for humans is from approx 20 Hz to approx 22 KHz. The audible frequency range for dogs is from approx 15 Hz to 24 KHz.

If you wish to transmit this spectrum digitally, then according to the (proven) Nyquist-Shannon theorum, you need to sample the signal at a rate of 2f, where f is the highest frequency component you will be able to transmit.
http://en.wikipedia.org/wiki/Nyquist%E2 ... ng_theorem

A 44.1 KHz sample rate captures 100% of the human audible frequency range with no perceivable losses. A 48 KHz sample rate captures 100% of the audible frequency range for most dogs, with no losses.

Conclusion? - the sample rate of 48 KHz was chosen by Korg because they thought most of their customers would be ... dogs

And now being serious - when you down-convert from a 48 KHz sample rate to a 44.1 KHz sample rate using a good down-converter operating exactly as described by X-trade, there is a loss of fidelity because the high frequency signals in the range from 22 KHz to 24 Khz will be removed. But only the "dogs" will notice!

Regards,
Rob