• Welcome to the General Discussion forum for UAD users!

    Please note that this forum is user-run, although we're thrilled to have so much contribution from Drew, Will, and other UA folks!

    Feel free to discuss both UAD and non-UAD related subjects!

    1) Please do not post technical issues here. Please use our UAD Support Forums instead.

    2) Please do not post complaints here. Use the Unrest Forum instead. They have no place in the the General Discussion forum.

    Threads posted in the wrong forum will be moved, so if you don't see your thread here anymore, please look in the correct forum.

    Lastly, please be respectful.

Sample Rate Conversion Help

BENBLEASE

New Member
Just wondering if any more experienced people could advise me on this;

I'm quite up on my stuff when it comes to digital audio, but I can't seem to find any answers on this topic. (not even in my prized bob katz book!)
.
The question is this: if i record, sample, and process at 32float/96khz, I know I need to dither down to 16bits, but is there a more graceful way of coming down to 44.1khz?
I use Cubase sx2.2, and so far at mixdown, am just simply adding dither noise to the masterbus, and exporting the audio at 44.1! does cubase do the SRC for you, or should I actually convert it down? are there different methods/alogorythms? Does it matter?

Thanks in advance.

P4, 3GHZ, 2GB RAM, RME FIREFACE, UAD1, POWERCORE.
 

akisd28

Member
I think it'd be better if you left your project @ 96/32 and leave the mastering engineer do the rest for you. If, however, you have to do this, the process you're following sounds right to me. If you want a better sample rate converter, you may want to try Voxengo r8brain pro. From what I've read, it's supposed to be better than Steinberg's SRC algorithm.
 

Von

New Member
I think it'd be better if you left your project @ 96/32 and leave the mastering engineer do the rest for you.
I agree totally...Most mastering engineers prefer to have as much of the source material as possible. There's nothing worse than getting files that have been poorly dithered and converted (on top of terrible normalizing/limiting used by many newbie engineers/producers with DAW's).

Von
 

.mr chris

Active Member
r8brain *PRO*, dither(ApogeeUV22HR or Pow-r Type 1), fades.

as always: IMHO, you know...
HTH,
.c
 

fgp303

Member
BENBLEASE said:
The question is this: if i record, sample, and process at 32float/96khz, I know I need to dither down to 16bits, but is there a more graceful way of coming down to 44.1khz?
.
Why dont record at 88.2khz at 24bit if u want produce 44.1kHz final material ? 88.2 are whole number multiple of 44.1 (2*) which so easy to convert down to 44.1... it will much better than convert from 96khz
 

MASSIVE Mastering

Active Member
That's not how it works... Audio is UPsampled to the lowest common multiple and then DOWNsampled to the target. It's not just dropping every other sample...
 

TheEastGateMS

Active Member
maybe i am missing something in the math, but if the numbers we are dealing with are really 88.2 and 44.1, then the lowest common multiple is in fact 88.2. besides, this also would make sense with r8brain pro's behaviour. there is one step when converting 88.2 to 44.1 (and is about 1/5th the processing time of 96 to 44.1) and 2 steps when converting 96 to 44.1 (obviously following the logic/process MM is talking about, but takes forever at highest quality settings).

also, the guys from lynx put it in layman's terms in just that way... that 88.2 is a cleaner conversion to 44.1 due to it being a simple halving instead of 96 to 44.1 which leaves remaining samples to be \"smoothed out.\"
 

neil wilkes

Venerated Member
TheEastGateMS said:
maybe i am missing something in the math, but if the numbers we are dealing with are really 88.2 and 44.1, then the lowest common multiple is in fact 88.2. besides, this also would make sense with r8brain pro's behaviour. there is one step when converting 88.2 to 44.1 (and is about 1/5th the processing time of 96 to 44.1) and 2 steps when converting 96 to 44.1 (obviously following the logic/process MM is talking about, but takes forever at highest quality settings).

also, the guys from lynx put it in layman's terms in just that way... that 88.2 is a cleaner conversion to 44.1 due to it being a simple halving instead of 96 to 44.1 which leaves remaining samples to be "smoothed out."
Not right, I'm afraid.
LCD (Lowest Common Denominator) works by getting the ratio between the actual sample rate & the desired sample rate.
With R8Brain Pro, going from 96 to 44.1 is done by upsampling by a factor of 147, then downsampling by a factor of 320.
96000 * 174 / 320 = 44100.
You use an LCD of 300, as this is the LCD - try it.
44100/300 = 147
48000/300 = 160
etc.
You can arrive at the exact math by always dividing the starting sample rate by a factor of 300.
Then divide the required rate by 300.
Reverse the numbers, and you can prove to yourself it works.

Remember - and this is what a lot of people forget - you are not dealing with 88.2, but 88200 - same goes for 96, it's not really 96, but 96000.
So there is absolutely no advantage to be gained by using 88.2 unless you are encoding for CD, and wish to use DTS-Pro, which can be encoded using 88200 and will play back at 44100 (due to clever bits in the encoder) when played through a DTS decoder that does not usually allow 88200 sample rates.
Nearly all SRC these days use LCD, very few use anything else.
Simply halving does not work well, as you will introduce artifacts because of the inevitable rounding that has to occur.

Have you seen the real figures for this stuff on http://src.infinitewave.ca/
where you will clearly see only the Saracen Weiss is better than R8Brain Pro, and that is a seriously expensive SRC.
And with R8Brain Pro, all noise is below -138dB - in short, inaudible.
Plus - and this is the biggest plus going - R8Brain Pro has a mode called Linear Phase - emulates going through serious, reference quality DA/AD to do the SRC. Sounds a lot better to me than Lin Phase does.
 

.mr chris

Active Member
neil wilkes said:
where you will clearly see only the Saracen Weiss is better than R8Brain Pro, and that is a seriously expensive SRC.

Plus - and this is the biggest plus going - R8Brain Pro has a mode called Linear Phase - emulates going through serious, reference quality DA/AD to do the SRC. Sounds a lot better to me than Lin Phase does.
Saracon Weiss = *great* AND(!) it includes POW-r dither, yes! 8)

Neil meant, the *MINIMAL* phase mode. I personally use combinations of both types of SRC...

Cheers,
.c
 

neil wilkes

Venerated Member
Eek.
Spot the typo!
Thanks for that - I appreciate the correction.
 
UAD Bundle Month
Top