Sample Rate Conversion - what's your plugin chain these days

WaveLab
1

I do a fair bit of vinyl restoration - inititally recorded in 32bit float @ 96khz and then I need to convert the ready files down to 16/44 for standard consumption purposes.

Within WL 8.5 (and using it’s standard supplied plugs) I current use the following chain in this order:

  1. Crystal Resampler
  2. PeakMaster
  3. Izotope MBit+ Dither.

However I recently acquired both iZotope RX4 Advanced and Ozone 6 - which both feature the full blown Izotope SRC resample module and of course MBit+ dither.

Given my current plugin chain - and these new tools - I feel like I can do better in the down-sampling department.

So - what’s everyone else using out there? And - if you have access to the above tools - have you taken advantage of the built in modules OR maybe used Ozone within your WL plugin chain for conversion?

Appreciate any updates.

Cheers!

VP

2

I favor the iZotope SRC in RX4.

Aside from how it sounds…I dislike using the global master section in general for montage based projects because plugin chains must be saved and reloaded separately. I do like the function of the Montage Master Section but of course the Crystal Resampler is not available in the Montage Master Section.

Because of this, I do any SRC with the RX4 batch converter before assembling the montage. This of course complicates having an easy way to render 96k files and 44.1k files but that shows you how much I dislike the Crystal Resampler, and the global master section.

I also don’t like the idea of the SRC running live on each render of the project. To me, it’s just one more thing to potentially go wrong, and also makes for a longer render time.

3

So how are you using SRC in RX4 - Batch Processor? Loading the files into RX? Or other?

VP

4

I use the RX4 batch processor and save the 44.1k versions to a dedicated 44.1k folder for each project.

If the client needs 96k versions of a master, I recreate the montage at 96k. It’s not easy but I hope Wavelab 9 has a way to easily change a montage to another sample rate, populate the correct sample rate files, and not have your markers all shift in time.

5

And before I had RX4 Advanced, I was using the iZotope SRC via Sample Manager by Audiofile Engineering.

I actually prefer the workflow of Sample Manager to the RX4 standalone app, but since Sample Manager has not been updated in some time now, I assume that RX4 advanced has the latest SRC algorithms.

RX4 standalone app is missing the critical “process in place” feature in the batch converter. Sometimes you have a folder of files and you just want to SRC them, not make additional copies anywhere else.

6

Any specific SRC or dither settings to share?

I tried a test run this morning in RX4 batch processor using the Resample module using the standard 44.1 (CD) template followed by the Dither module using the CD Dither - Safe template.

I then A/B’s the result against the originals - sounded excellent to me.

It’s too bad that RX4 ADV will not afford us standalone plugin versions of the Resample (SRC) and Dither modules so we could plug them ina DAW scenario.

Also wish they would allow their Batch Processor to run standalone instead of having to launch an empty RX4 shell all the time.

VP

7

Can’t you just set Copy Results = Original File’s Folder and make the file name the same. That should do the trick (overwrite with new SRC). Yes?

VP

8

I’ve been pretty happy with the presets within the batch converter and haven’t had much time to experiment with various settings.

I’m in agreement on the empty shell thing too. There are a lot of small workflow things that could make RX4 standalone be better.

I’ve contacted iZotope support about a few of them but haven’t seen any action.

9

No, when you do this, it just gives you a new file with the same name, but with the numeral 1 after it. So you are still left with two sets of files, and typically the files you want to use now have a 1 added to the name which is not convenient.

Sample Manager is great as far as workflow. You can save files to a new folder, or simply “process in place” which does exactly that. Changes the file directly and keeps the same name, no redundant files.

10

Annoying :slight_smile: Especially when the interface clearly states it’s going to keep the file name the same. But it’s not really keeping it the same at all - by add a number to the end.

Was also going to ask - have you ever recorded right into RX4? I have been testing this capability and while it works great - I am getting a nasty “Crash Reporter” window whenever I press Stop to stop a recording. Not sure what is going on there. That was the first time I saw this “reporter”. I was asked for some details and then it presumably sent a report directly to IzoTope.

VP

11

Nope, I had never tried recording into RX4 before.

I just did a quick 15 second test recording and didn’t get an error when pressing stop. I’m on OS X 10.10.2

12

Running Win 7 Pro here. Might have something to do with my C drive (which is the default path for recorded files). My system drive has had all sorts of bizarre UAC issues since I installed it several years back. Sometimes still requires me to “grant” admin access for certain things - even tho I am always on as uber-admin :slight_smile:

I changed the default record path and will try some more recording this weekend.

VP

13

I like Wavelab Crystal Resampler, but think there are others that are probably just as good. And Wavelab Internal dither, TPDF settings (type 1 no noise shaping). I have Izotope RX4, but find Crystal equivalent in SRC quality, and the convenience and workflow of a built-in SRC is definitely a major consideration here. I just wish Crystal Resampler was as fast as Izotope, or SoX (the fastest I’ve seen for the quality). But I think they’re all excellent quality.

Looking at filter slopes, Izotope RX4 SRC “CD” preset is not as steep as Crystal “Ultra”, so the Izotope CD preset might be more equivalent to Crystal Resampler “High” setting than “Ultra”. I would think Crystal “High” would have a gentler filter than “Ultra”, hopefully with the same “quality” (since “quality” seems to be a matter of opinion regarding SRC filter slopes and transitions). Would be interesting to see charts for Crystal “High” and “Standard”, but they weren’t done for infinitewave.

I’d like to be able to use SRC plugins too, but I believe PG said the VST spec doesn’t allow sample rate change in VST plugins. Maybe AU does (?), but still I’ve never seen an SRC plugin available anywhere, in AU, VST, RTAS, or TDM.

If I was going to use an external batch processor, I’d probably go with Foobar with the SoX component, or RX4, or Sample Manager (for Izotope or Apple afconvert bats).

14

I’m guessing the new version of Sample Manger which will be called Myriad will have the Goodhertz SRC rather than the iZotope.

I heard something about iZotope discontinuing the licensing of their SRC soon, or at least drastically changing it (maybe it’s more expensive now?) so looking to the future, apps that are using iZotope will have to change to something else.

In Wavelab, the need to put the SRC before the final limiter causes a problem for me because I really prefer the montage master vs. global master section so all the plugins save within the montage vs. saving and reloading.

I know some people might not care but to me it’s a big workflow preference.

15

I’m with you on that preference. I live with Crystal, but I too-often forget to save my global master section. Fortunately I keep it pretty sparse, and even with Crystal in use, only 2 or 3 settings might change, and those not much. But when that happens I can’t honestly get back to whatever I previously delivered, which just feels wrong and janky. I’d strongly prefer Crystal living in the montage master section.

Crystal sounds ok, but that too could be better. Izotope’s sounds better and is more flexible to boot. I’d love for WL to license the algorithm’s Izotope’s currently using (I thought they’re in-house, not licensed, but apparently I’m mistaken?).

16

I’d like to see SoX considered. I don’t know about its availability, but I believe it’s open source. It’s available in Foobar on Windows and XLD on Mac. Also in Audacity, but not with all available settings.

I ran a test of SoX against Izotope RX4 and r8Brain Pro, converting a 96KHz 24 Bit 3:45 song to 44.1 32 bit float in each of the programs, using settings closest to “ideal” as shown on infinitewave. I was shocked to find that all 3 results nulled exactly to each other down to about -110db. I made difference files of the nulls, and turned them up 60db in Wavelab to hear what was there, listening in headphones at a loud volume, but there was no discernable audio in the differences at all. I would call the 3 programs identical at these settings. I could be wrong, but I don’t think anyone could tell these 3 apart in blind testing, using these settings, with nulls down to -110.

These are the settings I used:

Izotope RX4
filter steepness 2000
cutoff shift 1.0
time to convert 0:38s

SoX in Foobar
passband 99%
phase 50% (linear)
time to convert 0:13s

r8Brain Pro
linear phase
ultra steep mode
time to convert 0:26s

Crystal Resampler Ultra
time to convert 0:55s

The closest in XLD is VHQ Linear Phase, not quite as close as Foobar for some reason.

With these settings, as well as being equivalent in quality to Izotope imo, SoX is also almost 3 times faster than Izotope, and more than 4 times faster than Crystal Resampler, which makes a big difference when converting 96k and especially 192k albums.

The version of SoX in Foobar has sliders for continuous adjustment of passband and phase, and a control for aliasing.

Other more euphonic, analog-like settings could probably be compared, since they have become more common, but that should be possible with continuously variable parameters.

I’d be ok with Izotope (although I much prefer the speed of SoX), but if Izotope is a problem, and SoX is less of a problem, I think SoX could be a great alternative.

17

Great test Bob, fair enough!

I wasn’t clear whether Crystal Ultra nulled with the 3 SRCs above it or not from your post. Even though it takes 4X longer, on a fast mac, delivering to multiple targets in WL makes sense to me if it does… I can always find something else to do during deliveries.

18

No, Crystal seems to be different from the other 3 in this regard. I even tried sub-sample alignment with it to try to get better null to the others, but could only get null down to about -40db. It’s not an indication of Crystal’s quality (don’t think), but does seem to indicate Crystal maybe has something different in filtering/decimation, while the others are using near identical filtering/decimation. I think? Open to correction.

19

I sample rate convert in Weiss SaraCon (stand alone) and then dither to 16 bit in WL (batch processor).

20

The Weiss SaraCon is a great sample rate converter. I also tend to vary between the RX4 Mbit+ dither and the PSP xDither, but find myself using the PSP xDither more often than not. Just a preference of sound to me, but they’re all very reliable software that tend to get the job done correctly. I haven’t done a shootout with the system dither in WL8 so that might be an interesting project to test out.

As for printing down from a 3296 file, I tend to have 5 generations. The original 3296, printed down to a 2496. From that file it’s printed down to a 2444 file which is checked for errors. Once that’s good, the dither is chosen and applied for the 1644 file, exported, and proofed before DDP creation (if necessary). The DDP is also proofed. It takes time but allows for a safe and smooth transition down the conversion chain that catches errors before they blow up into big problems. Serial proofing.