INTERACT FORUM

Please login or register.

Login with username, password and session length
Advanced search  
Pages: [1]   Go Down

Author Topic: DSP regression?  (Read 2174 times)

mattkhan

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 4269
DSP regression?
« on: July 13, 2024, 02:33:43 pm »

has there been any undocumented change to the DSP engine in MC33?

I ask because the same DSP configuration in MC33 produces totally different output vs MC32 in that multiple channels come out much quieter. The behaviour is almost like PEQ2 isn't applied.

I need to dig into it further to work out exactly where the problem is but thought I'd post now in case there is some undocumented change that may trigger such behaviour.
Logged

mattkhan

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 4269
Re: DSP regression?
« Reply #1 on: July 14, 2024, 04:22:07 am »

OK it's actually really simple and it is just completely broken (for >8 channel use cases)

I made a new zone and added DSP to simply copy L to every other channel, played full bandwidth pink noise into that channel and output to disk

every channel above 8 is silent

repeat in 32, every channel is identical

dsp file attached
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10968
Re: DSP regression?
« Reply #2 on: July 14, 2024, 04:31:33 am »

Matt enabled the new extra channel system by default in MC33, sounds like it still needs work. I have let him know to look at it.
Logged
~ nevcairiel
~ Author of LAV Filters

Matt

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 42441
  • Shoes gone again!
Re: DSP regression?
« Reply #3 on: July 14, 2024, 09:38:59 am »

There's a little change in extra channel routing now.  You used to pick something like "#9" but now you pick something like "Extra 1".  If you pick the extra channels discretely does it work properly?
Logged
Matt Ashland, JRiver Media Center

mattkhan

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 4269
Re: DSP regression?
« Reply #4 on: July 14, 2024, 09:47:36 am »

I can try

It doesn't make any sense to me that it shows the extra channels and the numbered channels though. I don't get the point of this change at all btw, what is the benefit?
Logged

Matt

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 42441
  • Shoes gone again!
Re: DSP regression?
« Reply #5 on: July 14, 2024, 10:19:48 am »

We're trying to clean up the extra channels a bit.  In the past picking them would confuse JRSS because they just looked like channels with data.

Now they're discrete and marked as not having data.  This way JRSS leaves them alone.

I could make Analyzer draw channels even if they aren't marked as having data if people think that would be nice.
Logged
Matt Ashland, JRiver Media Center

mattkhan

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 4269
Re: DSP regression?
« Reply #6 on: July 14, 2024, 10:34:08 am »

Ok right, I will test later to see if that's the issue

Can you confirm what the internal indexes are for those channels? I will also have to update my UI in beqdesigner to cope with this change.
Logged

Matt

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 42441
  • Shoes gone again!
Re: DSP regression?
« Reply #7 on: July 14, 2024, 10:45:00 am »

Can you confirm what the internal indexes are for those channels? I will also have to update my UI in beqdesigner to cope with this change.

I don't think the index should change.  It's just how they're labeled in our internal audio format object.  They're now marked as extra channels.

Let's spend a few weeks working on this together and see if we can all be happy with the new approach.  If not I'm fine adding an option back.

Thanks.
Logged
Matt Ashland, JRiver Media Center

Matt

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 42441
  • Shoes gone again!
Re: DSP regression?
« Reply #8 on: July 14, 2024, 10:46:11 am »

I'll add this to build 3:
Changed: Analyzer in DSP Studio includes channels not marked as having data (like extra channels) because they could get filled by other DSP.
Logged
Matt Ashland, JRiver Media Center

mattkhan

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 4269
Re: DSP regression?
« Reply #9 on: July 14, 2024, 12:05:16 pm »

change makes sense to me so I don't think you'll need an option, I think the current impl is confusing & inconsistent though in a few ways

1) currently both new and old variants of the channels are shown in the DSP studio dialog, e.g. given 7.1 + 6 extra then I get the option to extra 1-8 and channels 9-14. The latter (explicitly named channels) has the right no of channels but they're actually totally ignored and shouldn't be there. The former (extra channels) has the wrong number listed.

(I assume this issue is the root cause of my problem, those channels shouldn't be even listed in the 1st place)

I think you should list the exact no of channels available along with User 1 & 2, in this case L-RR and Extra 1-6

2) I think this upgrade is going to be a bit painful for people who do use >8 channels, currently I don't see how anyone could do anything except redo it manually (which, given a high channel count, is really painful). Automated migration would be nice & might be quite straightforward (just replace references to channel x with channel y) but not sure if you have an easy way to know that the config came from <33 and hence migrate it

3) I still don't understand the channel indexes, currently L = 2 and this runs upto RR = 9 then we have User 1 / 2 at 11 and 12 and then channel 9 = 13 and then it just increments by 1 til we get to channel 32 = 36 . It would make sense to me, given this scheme, if Extra 1 = 37 (or 38 if you like to leave a gap) and continue from there. However you said "they won't change" which implies that a given index could now be either one thing or another? if so, that would be quite unpleasant and I would revisit that design.

Logged

Matt

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 42441
  • Shoes gone again!
Re: DSP regression?
« Reply #10 on: July 14, 2024, 12:19:06 pm »

1) currently both new and old variants of the channels are shown in the DSP studio dialog, e.g. given 7.1 + 6 extra then I get the option to extra 1-8 and channels 9-14. The latter (explicitly named channels) has the right no of channels but they're actually totally ignored and shouldn't be there. The former (extra channels) has the wrong number listed.

Good find.

Next build:
Changed: Extra channels could appear two ways in Parametric Equalizer.
Logged
Matt Ashland, JRiver Media Center

mattkhan

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 4269
Re: DSP regression?
« Reply #11 on: July 14, 2024, 04:21:42 pm »

using X1-6 instead of C9-14 works fine

I also checked the internal indexes and those are 37-42 so that seems fine as well

it means it's totally possible to provide a migration from old to new but I guess you've never done it before so probably not going to start now for (what I assume is) a niche use case.

if you can confirm the above then I will add such a migration capability to beqdesigner (because I can't be bothered to do it manually so might as well make it theoretically reuseable)
Logged

whoareyou

  • Galactic Citizen
  • ****
  • Posts: 439
Re: DSP regression?
« Reply #12 on: July 26, 2024, 09:31:48 pm »

I have configuration of 5.1 with 2 extra channels, and this configuration uses a VST3 (HLConvolver) for convolution expecting 7 channels (5.2 system). 
With old mapping system 5.1/2extra created a 7.1 output container and I'd use order channels to match my DAC's channels.
Now, I can't seem to properly change channel order. 

Not sure how to troubleshoot further, but perhaps parametric order channel is not working correctly with new system?  Does adding extra channels still change the output container layout? 

Having said all that, 7.1 does work for my situation but something seems amiss, or I'm confused as usual :).
Logged

Matt

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 42441
  • Shoes gone again!
Re: DSP regression?
« Reply #13 on: July 27, 2024, 12:11:30 pm »

Coming soon the new extra channel system will be optional.

I just tested and 5.1 with two extra output channels still tells the VST it's opening in 8 channel mode.

When you order channels, you'll need to refer to the extra channels by themselves instead of using channel 7 and 8.
Logged
Matt Ashland, JRiver Media Center

whoareyou

  • Galactic Citizen
  • ****
  • Posts: 439
Re: DSP regression?
« Reply #14 on: July 27, 2024, 02:59:52 pm »

Ok.  I think I got it but was confused becase I forgot that JRiver treats the 5.1 mapping as FL, FR, C, SW, SL, SR, but after adding 2 extra channels the mapping changes to 7.1 container FL,FR,SW,E1,E2,SL,SR.  Does it make sense to maintain standard 5.1 mapping and add the extra channels after the SL & SR?

One other question about the order channels config option.  That  only handles up to 7.1 channels, so could the extra channels be added to the list of available channels to reorder?
Logged

Matt

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 42441
  • Shoes gone again!
Re: DSP regression?
« Reply #15 on: July 27, 2024, 04:48:13 pm »

Thanks.  You're right that the order channels needs to get updated for the new method as well.  I'll take care of that soon.

I think we'll need to make it so that you have to add a new order channel effect to load the channel configuration.  So if you change to more or less extra channels, you'll need to add a new channel order filter.

Thanks again.
Logged
Matt Ashland, JRiver Media Center
Pages: [1]   Go Up