The No.1 Website for Pro Audio
 All  This Thread  Reviews  Gear Database  Gear for sale     Latest  Trending
Sample Rate Conversion
Old 4th March 2019
  #31
Lives for gear
 
thermos's Avatar
Also something to be said for the charm of converter artifacts. Sometimes jitter/certain op amps etc can make things sound brighter or more alive. But it certainly isn't more accurate.

I remember reading that Dave Collins and the Metallica guys spent a ton of time trying to figure out the best way to get from 96k to 44.1. In the end, they dropped the record by .1 db and used Izotope RX.

Also, the math of 88.2 to 44.1 is a total myth from what I understand. Most SRCs upsample first before downsampling.
Old 4th March 2019
  #32
Lives for gear
Quote:
Originally Posted by thermos View Post
Also, the math of 88.2 to 44.1 is a total myth from what I understand. Most SRCs upsample first before downsampling.
Somewhere in these pages, Fabien TDR has said that in the case of 88.2>44.1 the SRC in fact DOES just drop every other sample and there's no upsampling necessary.
Old 4th March 2019
  #33
Lives for gear
 
thermos's Avatar
Quote:
Originally Posted by scraggs View Post
Somewhere in these pages, Fabien TDR has said that in the case of 88.2>44.1 the SRC in fact DOES just drop every other sample and there's no upsampling necessary.
Interesting, I doubt that's the case with every SRC but I could definitely be wrong.
Old 4th March 2019
  #34
Lives for gear
I could very well be wrong here too!

But when I used to use R8Brain Pro, 88.2>44.1 was one step and took two seconds, 96>44.1 was 3 steps and took a minute.
Old 5th March 2019
  #35
Lives for gear
 
Trakworx's Avatar
Quote:
Originally Posted by scraggs View Post
But when I used to use R8Brain Pro, 88.2>44.1 was one step and took two seconds, 96>44.1 was 3 steps and took a minute.
You made me curious so I tested with Pro Tools Tweak Head SRC and 88 to 44 took the same amount of time as 96 to 44. FWIW.
Old 5th March 2019
  #36
The SRC first searches the lowest common multiple between both rates, and then:

1. Add samples as needed (zero-filling) to reach the lowest common multiple.
2. Nyquist filtering as required by the target rate.
3. Remove samples as needed to reach the target rate.

In cases where the lowest common multiple already equals one of both samplerates, depending on the direction, either 1) or 3) can be ignored.

A single lowpass filter suffices. But weird ratios obviously ask for more memory and cpu cycles (or possibly a quality compromise).

For example, the Nyquist filter in a 44.1 to 88.2 kHz conversion will have to operate at 88.2 kHz. But the Nyquist filter of a 44.1 kHz to 48 kHz conversion has to operate at 7,056 kHz. That's a little more than 7 MHz! A huge difference of course (80 times more data to process!).

Last edited by FabienTDR; 7th March 2019 at 01:46 AM..
Old 5th March 2019
  #37
Gear Nut
 
12th & Vine's Avatar
 

Weiss Saracon
Old 5th March 2019
  #38
Gear Addict
 

Verified Member
Quote:
Originally Posted by Disease Factory View Post
Converters? if so which ones?

r8brain?
rx7?
Sox?
Weiss?
CDfile?

What else is there?

What settings? When and how?

From my experience, out ond dac and into a 44.1khz adc is good. r8brain at 32 bit float is ok. but not perfect.. Is anyone getting a good detailed non 2d sounding conversion from 48khz? I hear its better to convert from 88khz. is this right?

DF
The higher you convert from the better in theory. Possibly exception that I would expect 88 might convert to 44 better than 96 would. But I would prefer 192 to 44 over 176 to 44.
Old 5th March 2019
  #39
Lives for gear
 

Verified Member
It's also worth noting RX SRC isn't one fixed algo, it's about as adaptable as I could expect most filters to be.
Old 10th March 2019
  #40
Lives for gear
 

I just found a newer one where you can choose between SoX or SSRC. It has graphic representations of LPF Frequency Response and LPF Impulse Response, and a number of other options. It's called Resampler-V DSP plugin for Foobar. If you download the zip, there's an installer in there that you have to use. Then when you reopen foobar2000 it will be there.

Old 10th March 2019
  #41
Lives for gear
 

Quote:
Originally Posted by jdunn View Post
I just found a newer one where you can choose between SoX or SSRC. It has graphic representations of LPF Frequency Response and LPF Impulse Response, and a number of other options. It's called Resampler-V DSP plugin for Foobar. If you download the zip, there's an installer in there that you have to use. Then when you reopen foobar2000 it will be there.
Yes, that foobar plugin is very nice, especially for batch processing or real-time resampling for playback (like when someone feels, it can be beneficial to hit his DAC with higher rates and with different filter), as it has the table, where you might set different parameters for each source and target frequency.
It's bit PITA to setup all possible variants at its UI (if you don't set up the profile for particular combination, it will just pass samples without touching), but it's only tool, I know, which allows you to do that.
Compared to other soxr based plugin by lvqcl A new resampler DSP for foobar2000
It allows you to set bit different parameters for filter.

Generally, both ssrc algorithm (extracted from Shibata's tool) and soxr (standalone resampling library from SoX) has quite extensive control about filter response, only at most of tools or DAWs, where it is employed, full controls aren't exposed to users.
For example soxr:
The SoX Resampler library / Code /
[945b59]
/src/soxr.h


So it's like with iZotope SRC implementation in their RX, where you have full control about filter parameters, it's not really possible to speak about the one inherent sound of particular algorithm, because you can significantly alter its response and pretty much mimic other resamplers with fixed parameters.

One can go from very steep filter with short transition band and more significant ringing (like say FinalCD with its sharp filter) up to gentler filters with wider transition band, which deliberately allows some aliasing (akin to Saracon, apodizing resampler in Pyramix or some DACs with "fast" filters for example) and everything in-between. Similarly one can adjust or switch phase response (and thus also where is energy of ringing - eg. symmetrical, post).

Michal

P.S. to mimic that Saracon/Pyramix/RX default response in Resampler-V plugin with SoX algorithm, link pass-band with stop-band frequency and set it to roughly 91%. Select lin-phase response and adjust it for maximum available stop band attenuation.
That's usually what I like as default settings for most of things.

Last edited by msmucr; 10th March 2019 at 10:56 AM..
Old 10th March 2019
  #42
Deleted 691ca21
Guest
What settings are people liking in RX for up and downsampling, or does it change depending on sample rate etc? I have never strayed very far from the defaults.
Old 12th June 2019
  #43
Gear Nut
 

I would love it if someone smarter than me, maybe someone from Prague ;-)) could assist with RX SRC settings to approximate FinalCD goldilocks filter. I'm working on a classical release recorded at 96 with deliverables at 44,48,96. Still testing & listening but in a first cut, in order of least to most stressful to my ears (on this one) I'm rating it thus:
- FCD goldilocks
- FCD sharp
- Sox Resampler-V set to mimic that Saracon/Pyramix/RX default response (as posted above).

Thanks for any insights.
S.
Old 12th June 2019
  #44
Lives for gear
 

Verified Member
Quote:
Originally Posted by Deleted 691ca21 View Post
What settings are people liking in RX for up and downsampling, or does it change depending on sample rate etc? I have never strayed very far from the defaults.
Somewhat crude but I like to think what I would be using in Equilibrium in response to phase response settings in Free mode.

i.e. this sounds like this, if I was to cut out high frequency where would that slider be.

As RX has a similar slider between linear and minimum (of some description).

That gives me a pretty bang on result to start with if it's something critical where SRC is going to do something which can make a difference further down the chain.

I've never intentionally let it alias mind, I can't think of many situations where you would let the filter over shoot like that, but I would be interested to hear if anyone has a positive experience.
Old 12th June 2019
  #45
Lives for gear
 

Hi S.,
"Goldilocks" in FinalCD is pretty similar to what RX uses in default (32, 1, 1). In the sense, it it's not "brickwall" with maximum attenuation happening right bellow Nyquist. So there is some intentional aliasing allowed to reduce overall amount of ringing.
Only "Goldilocks" is even bit gentler with wider transition band. So to approximate that in RX resampler, I would further reduce steepness parameter to say 24-25 and leaving others at default (eg. no shift, linear phase).

I don't have later version of RX here, using RX4, but I don't think, those parameters were changed and resampler is essentially the same (perfect ).

Michal
Old 12th June 2019
  #46
Lives for gear
 

Verified Member
Quote:
Originally Posted by msmucr View Post
Hi S.,
"Goldilocks" in FinalCD is pretty similar to what RX uses in default (32, 1, 1). In the sense, it it's not "brickwall" with maximum attenuation happening right bellow Nyquist. So there is some intentional aliasing allowed to reduce overall amount of ringing.
Only "Goldilocks" is even bit gentler with wider transition band. So to approximate that in RX resampler, I would further reduce steepness parameter to say 24-25 and leaving others at default (eg. no shift, linear phase).

I don't have later version of RX here, using RX4, but I don't think, those parameters were changed and resampler is essentially the same (perfect ).

Michal
It appears to be the same, the batch processing has a pretty crappy workflow though, wish they would make that a main feature, must be a whooole lot of audio being batch processed these days.
Old 12th June 2019
  #47
Gear Nut
How can you hear ringing at 22 kHz? I can't hear ringing even at 13-15 kHz, but hear sound up to about 17-18 kHz. But sharp brickwall filter is 22k....it is not possible
Old 13th June 2019
  #48
Gear Nut
 

Hey Michal - thank-you, I really appreciate that insight. I am always interested to read what you have to say on a topic.

Stereo Flux - no I do not claim to hear that at all. Those were subjective impressions, but I do plan to attempt a blind AB test. I am just sensitive to "digititus" and have not come up with a better idea than trying to advance everything incrementally without becoming OCD. I will say two further things. Surely if all SRC's were equal, it wouldn't matter which one you choose? Also I once gave a blind demo to a collegue who's ears I greatly respect, of 1 Saracon and 2 different RX SRC's and gave away nothing. Her descriptions of the 3 precisely matched my own that were no doubt heavily polluted with confirmation bias. So I guess I don't know what I don't know.

Last edited by potscrubber; 13th June 2019 at 10:55 AM.. Reason: forgot the important "not"
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump