Protools really is bad at SRC, this is what you're hearing. Your best bet would be a hardware SRC or you can try a program like Barbabatch. If these are out of your budget, then maybe you know someone with another program that you can try it on and see what it sounds like. In none of these are an option, you can do a D/A/D capturing at 16/44.1 if you have good converters.

Not that I am an authority, but I didn't hear any mention of him dithering to get from 24 bits to 16. That should help some, right?

that would help too.

What ProTools system are you using?

First of all, thanks to everyone for their good info. I am using Pro Tools LE through a Digidesign 002. I understand that dithering is important, so I am assuming the "Tweak Head"
conversion in protools does dither. Is this correct? Also, I checked out Barbabatch. Looks like a great program, but I am running a Windows system and they don't have a windows version. Does anyone have any other suggestions for good windows software to convert down to 44.1/16bit, or any other good information to do the best conversion possible? Thanks

I'm with Michael - I haven't heard to many programs that do as nice of a job as a loop through some quality converters.

Voxengo's R8Brain is a fairly nice stand-alone if you want to try it. Free, also. There's a new "Pro" version as well if I'm not mistaken.

Thanks guys, I truly appreciate the help. So if I'm not mistaken, I should be able to get the best results by playing the songs back at 24/96, then recapturing from analog at 44.1/16? Is there any way to do this using pro-tools on a digidesign 002 (i.e. playback two 24/96 tracks while recording two 44.1/16 tracks?) Part of my problem is that I only have one piece of hardware that can record digitally, the 002. Is there any hardware you would recommend around $500 or less?

Thanks again!

unfortunately you can't do it on protools. You can look at other options like a dat machine, they go for real cheap nowadays. you can also look at something like the masterlink. maybe a stand alone CD recorder. You can find these things used for under $500. The key to doing it this way is the converters. The better the converters, the better result. Welcome to the world of never ending gear purchases. I like the masterlink as a tool to record and playback from, but I don't like the converters, some on the otherhand do.

Here are my ideas for you to get your stuff sounding better...each one will probably get you incrementally "better", but "better" is a pretty subjective thing.

First off, you should definitely going through some sort of dither, if nothing else use digidesign's Power Dither as your LAST LAST LAST plug-in. Set it to 16 bit. That will help some.

I'm not sure about the 002's clock, but back when I had a home system on an Audiomedia III card, it would help ALOT if I clocked off an external converter without going out then "back in". The 888's used to suck, too. The 96 and 192 I/Os are much improved, but like I said, I'm not sure about the 002.

But try the dithering plug-in and let us know how much it helps...if any!

Maybe we should have a conversion "shoot-out" one of these days...take a 24bit/96KHz file and convert it down to 16/44.1 using different systems, then do a blind test. Would be interesting.

a discussion on dither and pro tools is on page 480 of the latest manual I have...look for "dither and bounce to disk". Hope this helps. 8-)

mixandmaster wrote: Here are my ideas for you to get your stuff sounding better...each one will probably get you incrementally "better", but "better" is a pretty subjective thing.

First off, you should definitely going through some sort of dither, if nothing else use digidesign's Power Dither as your LAST LAST LAST plug-in. Set it to 16 bit. That will help some.

I'm not sure about the 002's clock, but back when I had a home system on an Audiomedia III card, it would help ALOT if I clocked off an external converter without going out then "back in". The 888's used to suck, too. The 96 and 192 I/Os are much improved, but like I said, I'm not sure about the 002.

But try the dithering plug-in and let us know how much it helps...if any!

Maybe we should have a conversion "shoot-out" one of these days...take a 24bit/96KHz file and convert it down to 16/44.1 using different systems, then do a blind test. Would be interesting.

This would be interesting to do.

I can easily do a ProTools HD conversion...maybe we should do two styles of music, maybe a heavier, rhythmic thing, and a dynamic acoustic thing...Anyone else?

I find this to be very interesting too. 8-)

I´ll offer to do Samplitude 7. Not much web space to put it up on though.

Gunnar

ahhh, sound like another project is brewing.

Let's see
so far:
samplitude 7
protools HD
apogee rosetta 200

anyone else?

The software that I have is Sonar 4 PE and Wavelab 4. I really would not dare using either of them for SRC (fortunately, I have not ever had to SRC anything). Although, I am very interested in hearing the comparisons of the other gear. If anyone would like, I could do a conversion in both programs for the sake of comparison.

-Erik

Well since Rosetta and Samplitude is taken, I have a Masterlink here, so I could take it for this purpose.

Richard

Michael Fossenkemper wrote: ahhh, sound like another project is brewing.

Let's see
so far:
samplitude 7
protools HD
apogee rosetta 200

anyone else?

I fear I will be laughed at for this but I have Izotope's Ozone with the MBIT+ dither. :oops: Call me nuts but I like it better than Apogee.

p0llen_p0ny wrote: I fear I will be laughed at for this but I have Izotope's Ozone with the MBIT+ dither. :oops: Call me nuts but I like it better than Apogee.

2 votes for MBIT+.... I'd still be interested in the hearing the shootout!

Also on the MBIT+ team. I have the POW-r dither (in sonar) but I like the MBIT+ better. Dont know if it may be an issue with the way sonar implements it though.

I still think that Ozone is worth the price even if you JUST use the dither.

I don't know, I prefer the realtime hear what you get approach.

Michael Fossenkemper wrote: ahhh, sound like another project is brewing.

Let's see
so far:
samplitude 7
protools HD
apogee rosetta 200

anyone else?

If we're comparing software src I can throw in SAWStudio 3.9c & Sound Logical's ReSample & Voxengo's r8brain 1.6 to the mix. It'd be good just for reference if people clocked the processing time each one of these software src and logged the cpu/ram used on their box for the conversions.

If we're also comparing loopback I can offer the Mytek Stereo96. Michael - are you going to throw in the Weiss SRC-1 as a reference too?

People are mentioning various dithers also - so I could toss in Waves IDR, Sweetboy RDR, & ReSample and I also have Sony & SAW native dither options (though both of these last 2 I don't think are that good), + the "SuperShaper" option on the Mytek.
I also want to agree that Megabitmax is a really fantastic algorithm - really I hate the sound of Ozone's processors - but the dither included in it is truly a great one.

Things could get mighty complicated though if we're trying to deal with all the combinations of src + dither.
I think it might be more productive in these tests to just stick with straight src and not involve word length reduction in them - it will make the testing have a lot less variables and be easier to do some apples to apples comparisons.

Anyway - besides music it'd be also good to convert a series of test tones also so that we can do some FFT analysis a little easier.

Another possible interesting thing is to use each of these to downsample and then upsample and see whether the results null. Thing is that level matching loopback could be difficult.

Best regards,
Steve Berson

I can supply the web space. I have 6 gigs free on my server.

Sorry for newbie question, but as I understand, it's better to recrd at higher res for better plug in performance. But in the case of teh original poster, would'nt it be better to record at 44.1 16bit going thru the best converter available and avoid the bouce/dither process?

Set me straight, cause I'm soon to be moving up to the 24 bit world, recording simple rock and roll with mediorcre equipment. ( Tascam 1884 or RME fireface)

Thanks!

stonewall40 wrote: Sorry for newbie question, but as I understand, it's better to recrd at higher res for better plug in performance. But in the case of teh original poster, would'nt it be better to record at 44.1 16bit going thru the best converter available and avoid the bouce/dither process?

If you're mixing in digital I find that my tracks recorded in 44k/16 sound nicer and more natural than tracks I've recorded in 96k/24 and then dithered down to 44k/16. But this is all subjective.

If you're mixing on an analog console go for the 96k/24. Dither won't be an issue. There is a slight difference in recording at 96k compared to 44k. But personally I try to stay away from dither. Call me superstitious.

Ok, I call you superstitious. Everyone should know what I think by now. but you get a GOOD SRC and decent dither, the benifits outweigh the negatives. We shall soon see on the next project.

Excellent thread.

This comparison really needs to be done, and I am anxious to hear the results.

In our experience, the sample-rate conversion has not been as much of a problem as the bit-depth conversion.
I have yet to hear a dithering algorithm I was happy with.
Since we are going for the highest recording res possible, this immediately creates the companion problem of downsampling.

I am interested in all software/hardware solutions to this issue.

Michael Fossenkemper wrote: Protools really is bad at SRC, this is what you're hearing. Your best bet would be a hardware SRC or you can try a program like Barbabatch.

Not disagreeing, but what information can you cite for your ProTools comment? And, if the data stays in the digital environment, SRC is a software process; therefore, isn't it the algorithm being graded, vice software vs. hardware?

my ears. software and hardware SCR is not the same. Hardware incorporates chips specifically designed to do the job. reclocking the signal and filtering. software is just an algorythm and is an offline process. Not saying that some software can't do a good job, but all the best SRC's are hardware for a reason.

Massive Mastering wrote: I'm with Michael - I haven't heard to many programs that do as nice of a job as a loop through some quality converters.

I think this is especially dubious. It would be far easier to write a good SRC than to design a good A/D converter. An A/D converter essentially has an SRC in it to start with - it converts the high delta sigma rate of 2.8224MS/s or so down to 44.1kS/s. D/A converters do the opposite. Therefore, going through this path inherently has SRC happening anyway, but in addition to that you also get the analog distortion and noise, etc.

It would be far easier to just write that SRC algorithm onto a DSP chip or in code and do it within the computer. I think Digi's new tweakhead setting has THD or SNR of around -120dB FS? It would be very difficult to attain anything near this going through ADCs and DACs. While one may prefer this route because of the coloration added due to noise and distortion, it is highly doubtful that even top of the world converters could rival a modestly priced piece of software.

Nika

well I was referring to hardware SRC's that are out there now. up until a few years ago, I did prefer a loop through some d/a/d because I thought it just sounded better. I have the apogee rosetta 200 SRC now and I think it's really good and prefer that. Of coarse I think the weiss is the bee's knee's but I can't afford everything weiss. Still can't stand the way PT's SRC sounds no matter how slow they tweak the head. It has gotten better over the years but to me they are way behind the game in this regard. Barbabatch sounds pretty good and i've heard Sonic's is really nice. But these are all offline where as the hardware units allows you to hear it as it happens which also makes me feel more comfortable.

This is interesting. Coming from the electronics industry where we implment a LOT of sample rate conversion algorithms (in a variety of DSP hardware), the simpler path (not going through more analog/digital conversion) has always resulted in less artifacts and aliasing distortion.

It appears that real-time processing in hardware is far easier to implement with a higher degree of quality than doing so in generic PC-based software. Maybe what you all are hearing are the differences in these algorithms.

Thanks for the information.

dpd wrote: It appears that real-time processing in hardware is far easier to implement with a higher degree of quality than doing so in generic PC-based software.

David,

Nice to see you here. This is, in fact, not correct. It is not easier to implement on dedicated chips than it is on a computer's CPU. This is not to say that outboard processor manufacturers have or have not done a better job for any number of reasons.

Nika

if you are recording on a high res software that does 32 bit float calculations then your gonna have to dither anyway. if your recording into any software that is above 16 bit you should dither back down to 16 bit. if your not recording in high calculation software then i really dont think high resolution is your concern. in other words it's safe to record in higher bit rate becuase the calculations are more accurate. It' s like if you had 2 calculators one had 16 places and the other one had 24 places and you start doing the same calculations on both of them, which one will have the more accurate final outcome after multiple calculations? you guessed it the one with 24 spaces. dithering helps retain that xtra information when going back down to 16 bit.

Nika wrote: [quote=dpd]It appears that real-time processing in hardware is far easier to implement with a higher degree of quality than doing so in generic PC-based software.

David,

Nice to see you here. This is, in fact, not correct. It is not easier to implement on dedicated chips than it is on a computer's CPU. This is not to say that outboard processor manufacturers have or have not done a better job for any number of reasons.

Nika

Why do you say this? I would think that a chip dedicated to the function at hand would not only be easier to impliment, but also better at it than a processor that is designed to do many things. For instance, my daw has one clock and any process that requires 2 clocks can only be done offline without being clocked to anything. Now i'm not pretending to be a programmer or a digital wiz, but doesn't this seem limited. Now you take a dedicated hand coded dsp chip with multiple stable clocks and do the same thing, i'm betting it's going to sound better. And if this was so easy to impliment, then why would there be a demand for dedicated outboard SRC's that go for good money. I know that i've tried a handfull of software src's and hardware src's and the hardware wins. not only in quality but also in implimentation such as hearing it while it's being done.

perfectwave wrote: if you are recording on a high res software that does 32 bit float calculations then your gonna have to dither anyway. if your recording into any software that is above 16 bit you should dither back down to 16 bit. if your not recording in high calculation software then i really dont think high resolution is your concern. in other words it's safe to record in higher bit rate becuase the calculations are more accurate. It' s like if you had 2 calculators one had 16 places and the other one had 24 places and you start doing the same calculations on both of them, which one will have the more accurate final outcome after multiple calculations? you guessed it the one with 24 spaces. dithering helps retain that xtra information when going back down to 16 bit.

We aren't talking about bit reduction. We are talking about sample rate reduction and the best method to use. There is not argument that 24 bits is better than 16 bits or that dither is or not needed.

i was just responding to pollen pony and his quote:

If you're mixing in digital I find that my tracks recorded in 44k/16 sound nicer and more natural than tracks I've recorded in 96k/24 and then dithered down to 44k/16. But this is all subjective.

sorry for the confusion micahel im still trying to find out how to use the quote feature for this site.

Michael Fossenkemper wrote: Why do you say this? I would think that a chip dedicated to the function at hand would not only be easier to impliment, but also better at it than a processor that is designed to do many things. For instance, my daw has one clock and any process that requires 2 clocks can only be done offline without being clocked to anything. Now i'm not pretending to be a programmer or a digital wiz, but doesn't this seem limited. Now you take a dedicated hand coded dsp chip with multiple stable clocks and do the same thing, i'm betting it's going to sound better. And if this was so easy to impliment, then why would there be a demand for dedicated outboard SRC's that go for good money. I know that i've tried a handfull of software src's and hardware src's and the hardware wins. not only in quality but also in implimentation such as hearing it while it's being done.

Michael,

First we have to decouple the issues of doing the task from doing the task in real time.

As far as doing the task is concerned, this task is merely the upsampling of data and then the downsampling of data. Both the upsampling and the downsampling require filtering, and the filtering needs to be capable of a significant anti-alias/image rolloff which requires very precise (hard-coded) coefficients and then basic convolution. There is nothing about a dedicated chip that is any better at this than a computer CPU. On the contrary, the computer CPU has as many as 64 or 80 bits of floating point processing power available, which, when used properly, can yield very accurate coefficients which can be used to make very steep filters.

Of course any computer CPU could be plausibly put into an outboard box, but in general the outboard chips do not provide the same type of onboard RAM or processing capabilities for anywhere near the same price point as a computer CPU.

As far as doing the process in realtime is concerned, one need not have more than one clock to do this. Multiple clocks can be derived from a single clock source. This is not to say that it has been effectively done for this task in a computer - I'll grant that. If you need to listen in real time it seems you have no choice right now other than hardware SRC. This is not to say, however, that it can't be effectively done.

Nika

perfectwave wrote: sorry for the confusion micahel im still trying to find out how to use the quote feature for this site.

In all of these examples substitute [ and ] for { and }

To quote, put {quote} before and {/quote} after the section you want to quote.

To bold, put {b} before and {/b} after the section you want to bold.

Same, but i for italics.

To post an image, post {img}then the url, then{/img}

Nika

ahhhh ic, i didnt under stand what was going on when i pushed the quote button, now i kinda do, haha thanks nika. Im gonna try it out with one of my old quotes to see if i got it.

perfectwave wrote: i was just responding to pollen pony and his quote:

If you're mixing in digital I find that my tracks recorded in 44k/16 sound nicer and more natural than tracks I've recorded in 96k/24 and then dithered down to 44k/16. But this is all subjective.

sorry for the confusion micahel im still trying to find out how to use the quote feature for this site.