I don't have your program for i use Nuendo, but it would be basically file -> export -> mixdown ... or Audio -> Bounce to disk ... or something of that sort, and then when the program asks you what to save as, chose..

Wav
16 bit/44.1 KHz
PCM uncompresses
Stereo

or something of that sort... its as easy as that, and the program will convert all the files into a 44.1 KHz mixdown 8-)

easy enough, but there is no selection for the clock, only bit rate. (8, 16, 24) am i missing something?

and also, when i do convert to 16 bit and apply dithering, i actually come out with 65kHz. can you explain this?

not sure about logic express, but you might try starting a new session at 44.1 and then importing the files. it should convert them to the 44.1 format, maybe.

logic, my platinum 5.5 anyway, can't sample convert when bouncing..
guess you'll need a seperate program..
the dithering is only for bitrate reducing.

while we're on the subject:
how do you mastering guys downsample?..
i've heard of mastering engineers preferring just going to analog and back again using hiquality converters..
thanx
huub

It depends on the material.
Sometimes it can be great to downsample analog through the converters.
But I've had a project where this changed the sound too much in the wrong direction.
So I did the SRC in the DAW witch was much better in this circumstance.

Best Regards

'kay...thanx..
and what program/algorithm do you use for that?
huub.

Try Voxengo r8brain. Freeware, stand-alone, quite decent.

First remember to back up this album before going into the abyss.
I'm not familiar with your platform. But you apply dithering when going from 24 to 16.
You apply anti aliasing when down sampling from 96 to 48/44.1, no dithering there.
Pro Tools lets you do these things in BTD (Bounce To Disk), it opens up a dialog where you can apply dithering if you want to.
Try and find a BTD or Process buttons in the File dropdown list
Your App should have the manual in a help file or whatever. Try that and tell us what you came up with.

not trying to be an a$$ about this, but if all your files are to be used for is audio then next time simply begin the session at 44.1 - when you downsample from 96k to 44.1 you lose a lot of what you captured - you are basically telling you computer to throw out RANDOM spots of information...or if you absolutely for some reason want to record at a higher sample rate choose 88.2 The computer can much more accurately and easily downsample from 88.2 to 44.1 since it only has to throw out or discard "every other sample" as opposed to "a sample here and a sample there" as with 96k to 44.1 does. Just a thought. :)

sorry should have clarified this - if all your files are to be used on CD media format, not DVD or any other format where a higher sample rate is required. :)

Well you aren't throwing out random information, it's very predictable. Some processors are said to sound better operating at higher sampling rates. The quality of the conversion is completely dependent on the converter. Some are very good and some suck. The benifit of processing in HD is that the filters are at a much higher freq and as a result the high end is less effected by these filters in the audible range. If you can retain this until the end, then you will reap the benifits.

I would still say to go 88.2 rather than 96k. I have found that downsampling from 96k to 44.1 as opposed to 88.2 to 44.1 just doesn't sound as good. I run SONAR 3.1 PE with MOTU HD192 and it has been my experience that there really is no logical reason to record at 96k unless you plan on using your files for DVD Audio. As far as retaining the high end and reaping the benefits...I still have yet to meet a project that needed me to use 96k to have a good high end. It just makes no sense considering our audible range is really only about 30hz-16khz on average and not 20-20. Even at 20-20 full range it still doesnt make any sense to me. Just my two cents.

Oh I agree that we can't hear that high. I'm talking about processing at higher sampling rates. Everytime you eq something in the digital domain, there are filters applied to the process. These filters have a ripple effect that you can hear in the audible range. So if you do all of your processing at higher sampling rates, the filters are at a much higher freq range, beyond the audible range. the results are a much smoother high end in the audible range. If say you process at 44.1k, filters are lying in the audible range and you can hear it.

RAIN0707 wrote: when you downsample from 96k to 44.1 you lose a lot of what you captured - you are basically telling you computer to throw out RANDOM spots of information...or if you absolutely for some reason want to record at a higher sample rate choose 88.2 The computer can much more accurately and easily downsample from 88.2 to 44.1 since it only has to throw out or discard "every other sample" as opposed to "a sample here and a sample there" as with 96k to 44.1 does. Just a thought. :)

The SRC process to 44.1 from 96 is definitely NOT random. The signal needs to be interpolated (inserting samples) up to a very high integer multiple, then decimated (removing samples) to 44.1 at a different integer multiple. Therefore, multiple digital filters are required to pull this off which opens the door for artifacts and aliases to appear in the final (converted) spectrum.

As you stated, SRC to 44.1 from 88.2 is just a simple 2:1 decimation to get to the final spectrum. Inherently a cleaner processing path, far simpler filters, lower clock rates required (less potential for jitter).

Michael Fossenkemper wrote: Oh I agree that we can't hear that high. I'm talking about processing at higher sampling rates. Everytime you eq something in the digital domain, there are filters applied to the process. These filters have a ripple effect that you can hear in the audible range. So if you do all of your processing at higher sampling rates, the filters are at a much higher freq range, beyond the audible range. the results are a much smoother high end in the audible range. If say you process at 44.1k, filters are lying in the audible range and you can hear it.

I thought all you dudes use fancy outboard analog EQ's. :wink:
If you know of a good digital EQ to use over a mix; do tell. I have the UAD Cambridge (sucks except as track EQ), and PultecPro (better, but not flexible enough) as well as the TC EQSat (decent for slight changes, fairly sterile without the Saturation). I am probably going to get the UA Precision EQ and give that a whirl. The Pultec and Precision both upsample to 192 i believe? Does that mean it is not necessary to actually upsample the audio file itself?

BTW, anyone wanna sign my petition for a new standard sample rate (66.15 kHz) for the next standard audio media format to be produced? Just enough above hearing range to get everything that matters through unadulterated, without overburdening our crappy little silicon chips with a processing load they will screw up. :)

Disclaimer: I am not a mastering engineer.

The weiss eq is great. The algorythmix plug sounds good too.

That's all we need is another sample rate.

Reggie wrote:
BTW, anyone wanna sign my petition for a new standard sample rate (66.15 kHz) for the next standard audio media format to be produced? Just enough above hearing range to get everything that matters through unadulterated, without overburdening our crappy little silicon chips with a processing load they will screw up. :)

This is Lavry's take on this, correct? But I thought his idea was to just not try and promote 192khz and leave 96khz as a standard. I could be wrong.

That is his view.

I'm not sure, but I would be happy with anything around 60k or so probably. I'm still not convinced 96k is necessary for my ears.

That algorythmix EQ plug looks nice; hadn't heard about it. But DANG that's an expensive plugin. Might have to stick to Sonalksis or something for now...... :cry:

dpd wrote: As you stated, SRC to 44.1 from 88.2 is just a simple 2:1 decimation to get to the final spectrum. Inherently a cleaner processing path, far simpler filters, lower clock rates required (less potential for jitter).

That's not how it works. SRCs go up to a common sample rate regardless of the source or destination sample rates. So 88.2, 96, 192 all go up to something like 35MS/s and then are downsampled at integer rates to 44.1kS/s. I don't know of any SRCs that do a simple 2:1 decimation. The programming from doing that would be unnecessarily brutal.

For this reason, the claim that 88.2 conversion to 44.1 sounds better than other source rates is dubious.

Nika

Nika wrote: [quote=dpd]As you stated, SRC to 44.1 from 88.2 is just a simple 2:1 decimation to get to the final spectrum. Inherently a cleaner processing path, far simpler filters, lower clock rates required (less potential for jitter).

That's not how it works. SRCs go up to a common sample rate regardless of the source or destination sample rates. So 88.2, 96, 192 all go up to something like 35MS/s and then are downsampled at integer rates to 44.1kS/s. I don't know of any SRCs that do a simple 2:1 decimation. The programming from doing that would be unnecessarily brutal.

For this reason, the claim that 88.2 conversion to 44.1 sounds better than other source rates is dubious.

Nika

Nika - sorry for the delay on this. I wanted to talk to a couple of my signal processing experts at work about this whole SRC issue. On one of my products we do a simple 2:1 decimation to get to a final telemetry rate. For this application we don't have to achieve 24 bit aliasing reduction (kinda hard to do in a 16 bit fixed point DSP) but the filters can be designed with no horrible difficulty to achieve better alias reduction than what this particular application requires.

One thing that filter designers take into account is what the true spectral content is that is being filtered. For 88.2 Khz sampling, we are still dealing with spectra that should only occupy ~20 Khz of bandwidth, with the transition band rolling off due to the combination of the natural audio data spectral balance and whatever filtering is/has been applied. In other words, to get effective aliasing performance, the filter stop band performance can be reduced based on what is being filtered. The spectra being filtered probably isn't white, in other words.

Also, for a situation where the decimation from 88.2 to 44.1 is being directly followed by dithering for bit reduction, I would think that this would allow the filter to not require full 24 bit aliasing performance, and could back off to better than 16 bit (maybe 18?).

Additionally, the concept of interpolating up just adds complexity and additional aliasing and artifact issues that must be dealt with prior to the decimation back to the final sample rate.

My guys believe that the simpler path is better and a straight 2:1 decimation should be achievable. Alas, we are in a different industry, but the concepts and applications are very, very similar.

Thanks for the post. Not sure I understand why the audio industry goes up, then down - but I appreciate the correction.

dpd wrote: My guys believe that the simpler path is better and a straight 2:1 decimation should be achievable....

Thanks for the post. Not sure I understand why the audio industry goes up, then down - but I appreciate the correction.

David,

Of course "your guys" are correct. It is a simpler path and it is indeed achievable. The issue is that we have many potential source and destination sample rates to write code for. We have to get from, say, any combination of 32, 44.1, 48, 88.2, 96, 176.4, and 192 to any other of those. That means a bevvy of possible banks of coefficients. The simpler approach is to use a single set of coefficients for each sample rate and then use a common multiple. This way, going from 44.1 to 48 means first applying the 44.1kS/s coefficients (to get TO the common multiple) and then applying the 48kS/s coefficients (to get FROM the common multiple). This makes programming very easy, requires less calculation, and if it is all done with reaonsable competence, can be done as transparently. If the THD and other byproducts are kept outside of human audibility for each pass then there should (and won't be) any disadvantage to upsampling first.

Of course, if we only have to talk about one conversion - say 88.2 to 44.1 and the SRC doesn't need to handle any additional tasks it would be easier just to write the simple, single set of coefficients for that, but if the SRC is designed to handle multiple rates, the upsampling method is a less cumbersome overall solution.

Does this make sense?

Nika

Nika wrote: [quote=dpd]My guys believe that the simpler path is better and a straight 2:1 decimation should be achievable....

Thanks for the post. Not sure I understand why the audio industry goes up, then down - but I appreciate the correction.

David,

Of course "your guys" are correct. It is a simpler path and it is indeed achievable. The issue is that we have many potential source and destination sample rates to write code for. We have to get from, say, any combination of 32, 44.1, 48, 88.2, 96, 176.4, and 192 to any other of those. That means a bevvy of possible banks of coefficients. The simpler approach is to use a single set of coefficients for each sample rate and then use a common multiple. This way, going from 44.1 to 48 means first applying the 44.1kS/s coefficients (to get TO the common multiple) and then applying the 48kS/s coefficients (to get FROM the common multiple). This makes programming very easy, requires less calculation, and if it is all done with reaonsable competence, can be done as transparently. If the THD and other byproducts are kept outside of human audibility for each pass then there should (and won't be) any disadvantage to upsampling first.

Of course, if we only have to talk about one conversion - say 88.2 to 44.1 and the SRC doesn't need to handle any additional tasks it would be easier just to write the simple, single set of coefficients for that, but if the SRC is designed to handle multiple rates, the upsampling method is a less cumbersome overall solution.

Does this make sense?

Nika

Nika - yup, makes sense, I'm with you. The compromise the audio industry has to make is in having multiple 'standard rates'. Where they screwed up, IMO, is setting 44.1 as a standard for CD delivery and that's a story we've all heard. Had they settled on 48 and integer multiples, or even 44.1 and integer multiples (but just one base sampling rate) we probably wouldn't be discussing this. :D

Oh, btw, name is Paul - I'm a middle-name guy.

Nice to meet you, Paul. My name is Nika (I live on the second floor...)

Thanks a lot, Nika - Now I've got THAT running through my head all night... :?