All times are UTC+01:00




Post new topic  Reply to topic  [ 166 posts ]  Go to page Previous 113 14 15 16 17
Author Message
PostPosted: Wed May 01, 2019 10:22 pm 

Joined: Sun Feb 03, 2013 2:39 pm
Posts: 309
Nah, we use proper math. Maybe we should sell you our tools. :lol:

RMS of the output differences just for the 1000 samples of the 2nd part:

  1. 9.30 vs 9.31: -inf dB (exact match of every digit)
  2. 9.23 vs 9.31: -56 dB
  3. 9.23 vs 9.00 (16 bit int): -96 dB
  4. 9.23 (16 bit int) vs 9.00 (16 bit int): -93 dB
  5. 9.23 vs 8.00 (16 bit int): -73 dB
  6. 9.23 (16 bit int) vs 8.00 (16 bit int): -73 dB

Fun fact: We have a tool that can match the settings of 2 audio processors to have their output as close as possible. Some years ago we matched the ST peak detection to one of the RMS modes. I hope I remember correctly and don't tell lies, but I think the match was something around -70 dB.

(Just for safety: Our tools aren't publicly available. Any requests will be disregarded.)


Top
   
PostPosted: Thu May 02, 2019 10:39 am 
User avatar

Joined: Tue Mar 17, 2009 2:56 pm
Posts: 3480
Really? Can you match ST to Orban and make zilion people happy.. and we can finish that story?

_________________
control point


Top
   
PostPosted: Thu May 02, 2019 11:56 am 
Site Admin
User avatar

Joined: Mon Mar 17, 2008 1:40 am
Posts: 9893
Quote:
  1. 9.23 vs 9.31: -56 dB
Ok, I found what's causing it - it's the bug fix to avoid overreacting to certain tones in RMS mode. Now, if you hear clicks, those clicks must have been present in the old version, not the new one. So unless this change is actually causing issues (and I don't think it is - with slower (more realistic) attack/release the difference is also much smaller), the new version should actually be better.


Top
   
PostPosted: Thu May 02, 2019 4:48 pm 

Joined: Sun Feb 03, 2013 2:39 pm
Posts: 309
Quote:
Really? Can you match ST to Orban and make zilion people happy.. and we can finish that story?
Haha. Matching only a few parameters can take from several minutes to several days. Matching ST and Orban would probably take decades or centuries but it should give the closest match eventually.
Quote:
Ok, I found what's causing it - it's the bug fix to avoid overreacting to certain tones in RMS mode. Now, if you hear clicks, those clicks must have been present in the old version, not the new one. So unless this change is actually causing issues (and I don't think it is - with slower (more realistic) attack/release the difference is also much smaller), the new version should actually be better.
The clicks aren't in ST. You asked me to process the same test signal with the same preset in 9.23 and 9.30 and listen to the difference. The difference signal is not silent. Because it's short I loop it. That makes it sound like clicks but I guess if I use a long test signal instead it would give an ongoing tone.

Besides that I don't get why the peak compressor needs or is affected by a RMS mode fix, the problem seems to be you consider it a minor change because you didn't meant to introduce a new type of compressor. However the difference between 2. and 3. in my above post is 40 dB. If that doesn't qualify as a new type then I really don't get it.

Anyhow if you like the new peak mode better you unintentionally introduced fine. But we need a legacy mode with the old attack curve behaviour or we're stuck at 9.23 (which might not be the end of the world).


Top
   
PostPosted: Thu May 02, 2019 4:59 pm 
User avatar

Joined: Tue Sep 15, 2015 12:22 pm
Posts: 257
Does the change audibly alter the way your presets sound? I would not imagine a difference of -56dB makes that much of a difference. The 40dB you mention is a difference of differences which does not mean that much (-160dB and -200dB also differ 40dB but it's not like you're going to hear that).

_________________
Software Engineer @ Thimeo Audio Technology B.V.


Top
   
PostPosted: Thu May 02, 2019 5:42 pm 

Joined: Sun Feb 03, 2013 2:39 pm
Posts: 309
Audible, yes. That much, well a bit. The RMS gives just an average. The error is bigger in the critical section where the level curve makes a bow.

EDIT: Just had a look at the timing constants. Release needs to be a tiny bit slower now, attack quite a bit faster. Match to 9.23 is still bad by measurement but I will do a listening test when I have a fully compensated profile and report back.

EDIT 2: I finished a speed compensated profile (mainly faster attack). Now the left part of the level curve before the bow is the same as 9.23. However the right part after the bow takes ages to reach the exact target.
This means that the sound is a little bit louder, less compressed, less controlled, less sharp, more airy, open and analog. Does that sound better? Definitely. Does that help with our problem considering it a new type of compressor? Not at all. We stay at 9.23 for the time beeing and have a 3rd party compressor which is much closer to 9.23 than 9.3x is, as 2nd best option. ST 9.3x is now banned here.
Btw can the RMS fix be done without a change to the attack curve?


Top
   
Display posts from previous:  Sort by  
Post new topic  Reply to topic  [ 166 posts ]  Go to page Previous 113 14 15 16 17

All times are UTC+01:00


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Limited