Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mattkhan]

What is this thread for?
This thread is intended to capture features that would improve jriver's ability to implement an active audio system for both stereo and multichannel use. JRiver has a v rich feature set for DSP that caters to some advanced use cases but also is missing some basic, and not so basic, features that standalone devices using off the shelf DSP chips have implemented. Filling these gaps would round out the offering nicely.

Active in this context is used somewhat loosely to refer to the use of digital filters to implement active crossovers and speaker/room correction.

Why do we need a thread for this?
The subject has come up in at least 3 threads and various people (e.g. natehansen66, mwillems) have mentioned similar features before. Collecting them into 1 place makes it more likely that a cohesive design can be constructed and individual features prioritised accordingly. Obviously all this is predicated on jriver deciding to satisfy demand here.

"DSP" is also a particularly tricky string to search for without getting tons of noise in the results

What features do we want?
I have put these into broad categories driven by how users would benefit.

PEQ/Crossover Flexibility

1) add support for different HP/LP filter types (bessel, linkwitz-riley)
what is it?
- let the user select the HP/LP filter type

any workarounds?
- cascade 2 filters to implement an LR
- no workaround for bessel

2) add a new filter type "all pass filter" to PEQ
what is it?
- allow user to select order (1st or 2nd), Fc (in Hertz) and Q of the filter
- the following definitions are used

1st order APF = 180 degree phase rotation
2nd order APF = 360 degree phase rotation
Fc = point at which 1/2 of the total phase rotation has been accomplished

- the reference implementation is the APF filter in http://www.musicdsp.org/files/Audio-EQ-Cookbook.txt

any workarounds?
- no

3) add a new filter type "advanced biquad" to PEQ
what is it?
- allow user to enter biquad parameters as per https://yabb.jriver.com/interact/index.php?topic=98863.0
(NB: this would be sufficient to implement the previous 3 requests)

any workarounds?
- no

4) allow shelf slopes >1
what is it?
- the existing PEQ configuration for low/high shelf filters does not allow Q (shelf slope) > 1, this can be useful and is not precisely replicated by stacking shelf filters

any workarounds?
- effect can be approximated by manually stacking peaking filters on top
- the "advanced biquad" would suffice in place of this

NB: also flagged in the too easy thread -> https://yabb.jriver.com/interact/index.php?topic=98734.msg699035#msg699035

5) allow channel count and mixing target to vary independently in Output Format
what is it?
- separate dropdowns for channel count (1,2,5,...,n) and mixing target (2.0, 2.1, 5.1, 7.1 etc)

any workarounds?
- apply downmixing manually using PEQ
- NB: this workaround would be much easier to apply if (Ease of Use)/1 were implemented as then we should add something to wiki which provides the set of filters required to downmix correctly

Ease of Use & Interoperability
1) provide the ability to export/import PEQ filter banks from/to a defined text format
what is it?
- a way to make life easier when you have a lot of filters to enter
- a way to interoperate with auto eq tools like Room EQ Wizard
- implementation could be similar to the convolution implementation or could be more like a DSP load/save
- should also be exposed as a MCWS/MCC command

any workarounds?
- no

2) provide greater granularity in Room Correction for distance setting
what is it?
- allow choice of units in both distance (ft, mm) and time (ms)

any workarounds?
- use PEQ to specify delay

3) allow user to rename channels
what is it?
- a way to make it easier to remember which channel is which in an active setup, e.g. replace "channel 10" with "SL-1" or Left with "L-2" (or whatever arbitrary string the user comes up with)

any workarounds?
- no

4) provide a graphical representation of the filters applied to each channel
what is it?
- a sanity check on what you have configured

any workarounds?
- play a sweep through jriver and measure the electrical result via loopback in your measurement software (e.g. REW)

5) allow user to select which channels are displayed in the analyser
what is it?
- put checkboxes next to each channel

any workarounds?
- install Voxengo Span which is a free 8 channel VST Plugin
- does not work for higher channel counts

6) change the name of the text field on the HS/LS entry screen from Q to S
what is it?
- pedantry
- see http://yabb.jriver.com/interact/index.php?topic=99096.msg686082#msg686082 for details

any workarounds?
- no

[natehansen66]

Matt - I requested the all-pass filter. It's well defined what they are and what they do.....I'm not sure what you're asking?

Thanks for starting the thread.

[DarkSpace]

From how I understand your post, it's meant to collect all DSP Feature Requests. Your title suggests that it's only about EQ though, so if I misunderstood, please tell me and I'll remove my post.

Here, I proposed an additional Output Format tab in the DSP, hoping that before that 'final' Output Format, the channel count etc. would be variable (i.e. VSTs could change it).
I also had an idea about allowing the first Format tab (after which channel count etc. would still be variable) to output an Ambisonic audio format, but I haven't done nearly enough research to even guess whether this might just be a Really Bad Idea. I may eventually make a thread about this (or not), depending on how useful I consider the idea. Currently, my main motivation would be to be able to (again, theoretically) do a 'perfect' binaural headphone mix without having to worry about the channel layout of the source audio.

[mwillems]

Thanks for getting this rolling Matt.

Two suggestions for additions:

1) Short of the whole monty of what Darkspace mentioned, I think it would be very helpful to at least be able to specify the channel count and the mixing target separately in output format. That seems more incremental and has been requested several times by close to a dozen people.  I think it would help a lot of people's use cases.
2) I think some sort of graphical representation of what PEQ/EQ is doing (in the abstract, not an RTA) would be extremely helpful to everyone.

Allpass filters have a solid mathematical definition, they're effectively a method of altering phase relationships: https://en.wikipedia.org/wiki/All-pass_filter

Additionally, I'm not sure I understand what is meant about orders for HS or LS filters?  Currently we can specify Q which is the same as order.  A .5 Q shelf filter is a 1st order shelf.  A 1 Q shelf is a 2nd order shelf.  Do you mean more or different orders?  The answer from Matt previously was that shelves with a slope greater than 1 Q were not stable(?). I think that's the word he used.

[mattkhan]

Matt - I requested the all-pass filter. It's well defined what they are and what they do.....I'm not sure what you're asking?

I know what an all pass filter OS but what I meant is what controls do you want exposed? At the moment, it is like saying "we want a filter" but exactly how do you want to define that filter? This wasn't obvious to me from the posts I saw and an all pass filter is a very broad term.

[mwillems]

I know what an all pass filter OS but what I meant is what controls do you want exposed? At the moment, it is like saying "we want a filter" but exactly how do you want to define that filter? This wasn't obvious to me from the posts I saw and an all pass filter is a very broad term.

That makes sense, we could well wind up with what's effectively a flange filter if we aren't more specific.  Maybe Nate can talk more about his use case? 

I'm not entirely sure what I would even do with an allpass as all the kinds of phase correction that I need to do couldn't be done with an allpass.

[natehansen66]

As I understand it, higher order ap filters can start to get nasty in the time domain, so I'd like 1st and 2nd hi and low ap filter sets with an adjustable corner freq. My use is for directivity control. I could likely get it done with FIR filtering but I'd prefer to stick with IIR.

I think Matt might be referring to my request for adjustable slope and q xo filters rather than shelving filters in the other thread.

While I'm at it I'd like the ability to right click copy/cut/paste/delete single and multiple sets of filters in the peq.

[mattkhan]

As I understand it higher order ap filters can start to get nasty in the time domain, so I'd like 1st and 2nd hi and low ap filter sets with an adjustable corner freq.

OK so to be clear on definitions

1st order APF = 180 degree phase rotation
2nd order APF = 360 degree phase rotation
Fc = point at which 1/2 of the total phase rotation has been accomplished

If so, this implies a fixed Q factor (slope of the phase shift) for this filter but that is not defined so far. Do you have a tool you use at present for this? or a reference implementation that you've seen in use?

[mattkhan]

Additionally, I'm not sure I understand what is meant about orders for HS or LS filters?  Currently we can specify Q which is the same as order.  A .5 Q shelf filter is a 1st order shelf.  A 1 Q shelf is a
2nd order shelf.  Do you mean more or different orders?  The answer from Matt previously was that shelves with a slope greater than 1 Q were not stable(?). I think that's the word he used.

AIUI Q != order. For example, a high Q HPF will reach the (6*order)dB/octave rolloff v quickly and can get quite peaky whereas a low Q HPF would take a while to do so (c.f. butterworth Q=0.707, bessel Q = 0.577).

I am not entirely sure what order even means for a shelf filter but I don't believe it is the same thing as S (shelf slope) or Q. For example look at the section titled "Shelving Filters with a Variable Transition Bandwidth or Slope" in http://uk.mathworks.com/help/dsp/examples/parametric-equalizer-design.html which shows what happens when you raise S (or Q) for a shelf filter (you get increased ripple which shows as a peaked response at either end of the shelf). It then raises the order to damp the ripple. I don't know how this is implemented though.

[mwillems]

AIUI Q != order. For example, a high Q HPF will reach the (6*order)dB/octave rolloff v quickly and can get quite peaky whereas a low Q HPF would take a while to do so (c.f. butterworth Q=0.707, bessel Q = 0.577).

HP and LP are different than shelves.  Q does not mean the same thing in both contexts

I am not entirely sure what order even means for a shelf filter but I don't believe it is the same thing as S (shelf slope) or Q. For example look at the section titled "Shelving Filters with a Variable Transition Bandwidth or Slope" in http://uk.mathworks.com/help/dsp/examples/parametric-equalizer-design.html which shows what happens when you raise S (or Q) for a shelf filter (you get increased ripple which shows as a peaked response at either end of the shelf). It then raises the order to damp the ripple. I don't know how this is implemented though.

Check out this thread:

http://yabb.jriver.com/interact/index.php?topic=98998

You can see measurements there of a 1Q shelf from MC and a 2nd order (12dB/oct) shelf from a DCX 2496.  The slope is identical (although the frequency is selected differently).  There are also some good links in that thread as well.  The Q in MC is the slope, and .5 Q has an identical slope to a 1st order (6dB) shelf in REW, and a 1Q MC filter has the same slope as a 2nd order (12dB shelf) in REW.  

The Q in MC does not introduce ripple (at least not within the currently allowable range). If you look at the Q graph in your mathworks link, the shelves with Q's below one don't peak, but the one with the Q of 2 does peak.  MC doesn't allow shelves with a Q over 1, so that may be the difference.

See also, http://www.hometheatershack.com/forums/rew-forum/67464-generic-dcx2496-shelving-filters-center-frequency.html

[natehansen66]

OK so to be clear on definitions

1st order APF = 180 degree phase rotation
2nd order APF = 360 degree phase rotation
Fc = point at which 1/2 of the total phase rotation has been accomplished

If so, this implies a fixed Q factor (slope of the phase shift) for this filter but that is not defined so far. Do you have a tool you use at present for this? or a reference implementation that you've seen in use?

Afaik that's correct. I'm not using one right now but I'm working on something where it might be needed. No examples at the moment sorry.

[mattkhan]

From how I understand your post, it's meant to collect all DSP Feature Requests. Your title suggests that it's only about EQ though, so if I misunderstood, please tell me and I'll remove my post.

I think we should sharpen the focus of this thread as atm it is described in implementation terms (the jriver dsp engine has some (P)EQ features) rather than the actual use case. Let me take a stab at that and if people agree then I'll update the 1st post.

This thread is intended to capture features that would improve jriver's ability to implement an active audio system for both stereo and multichannel use. Active in this context is used somewhat loosely to refer to the use of digital filters to implement active crossovers and speaker/room correction.

How does that sound? Active feels like the wrong word to use but I can't think of a better one atm.

[mattkhan]

HP and LP are different than shelves.  Q does not mean the same thing in both contexts

Check out this thread:

http://yabb.jriver.com/interact/index.php?topic=98998

You can see measurements there of a 1Q shelf from MC and a 2nd order (12dB/oct) shelf from a DCX 2496.  The slope is identical (although the frequency is selected differently).  There are also some good links in that thread as well.  The Q in MC is the slope, and .5 Q has an identical slope to a 1st order (6dB) shelf in REW, and a 1Q MC filter has the same slope as a 2nd order (12dB shelf) in REW.  

The Q in MC does not introduce ripple (at least not within the currently allowable range). If you look at the Q graph in your mathworks link, the shelves with Q's below one don't peak, but the one with the Q of 2 does peak.  MC doesn't allow shelves with a Q over 1, so that may be the difference.

See also, http://www.hometheatershack.com/forums/rew-forum/67464-generic-dcx2496-shelving-filters-center-frequency.html

I think this indicates that jriver is using incorrect terminology in the shelf filter screen, it refers to Q but it means S (shelf factor).

http://www.musicdsp.org/files/Audio-EQ-Cookbook.txt tells us that you convert one to the other via

1/Q = sqrt((A + 1/A)*(1/S - 1) + 2)

where

A = sqrt( 10^(dBgain/20) )

This means S=1 converts to Q~=0.707 (because it reduces to 1/sqrt(2) irrespective of the gain) and the Q associated with S=0.5 varies with gain (so a +10dB S=0.5 is Q=0.432 whereas +3dB S=0.5 is Q=0.498). S>1 (and Q>0.707) results in increased ripple (try it in the minidsp spreadsheet to see what I mean) and this is exactly the sort of peaking you see in a high(er) Q [HL]PF filter so it seems to me that what Q means remains much the same in either case.

FWIW the minidsp biquad spreadsheet uses Q. A value of 0.707 results in a filter that looks a lot like the [HL]S12 filter at first glance.

I don't think this means that this is a 1st or 2nd order shelf filter, it just means that the slope associated with S=0.5 or S=1 is, when in the nearly constant slope phase, v close to that associated with the rolloff of a 1st or 2nd order (high/low) pass filter.

[natehansen66]

Another request.....the ability to rename channels as we see fit in the PEQ.

[Audionut11]

In Room Correction, left or right click toggle of the distance unit.  ft, M, cm, ms.

[jjazdk]

1) Short of the whole monty of what Darkspace mentioned, I think it would be very helpful to at least be able to specify the channel count and the mixing target separately in output format. That seems more incremental and has been requested several times by close to a dozen people.  I think it would help a lot of people's use cases.

I agree, which is sort of what I have requested here:
http://yabb.jriver.com/interact/index.php?topic=99401.0

The current limitation in the amount of output channels does not make much sense (f.x. with my 20 channel soundcard, I can only use 16 channels).

[njoak]

I have a few requests/suggestions:

1) A graphical representation of the filters applied to each channel. I am thinking of something similar to what minidsp has, for those familiar with those devices. Something for channel mixing/routing would also be nice. I feel this would help to get an overview of what you are doing and also help double-check your work and make "sanity checks".

2) A quicker and simpler way to enter and/or edit many filters/entries. For instance an option to write scripts/text files with the filters. When you have more than a few filters/operations to enter, I feel the current method can be a bit laborious.

3) Ability to import filters produced by REW (Room Equalization Wizard).

[mattkhan]

I updated the 1st post with everything collected so far. Let me know if I missed anything (or any corrections required)

[mwillems]

PEQ 4) and Ease of Use 2) both have workarounds.

Independently specifying the mixing target and channel count can mostly be worked around by manually downmixing from a higher channel count in PEQ.

"Distance" in Room correction is just delay by another name, so lack of granularity can be worked around by manually specifying delay in PEQ which has almost arbitrary precision.

[mojave]

I started typing the same thing as mwillems.

5) allow user to select which channels are displayed in the analyser
what is it?
- put checkboxes next to each channel

any workarounds?
- no

Voxengo Span is a free 8 channel VST Plugin that I use all the time. You can see what is going on in each channel or even group channels. Of course it won't work for higher channel counts.

[mattkhan]

Ta. I updated the 1st post. I think the channel count/mixing issue would be ameliorated if we had the import/export PEQ feature.

I also think we should try and prioritise these from a user perspective. Obviously this is a bit tricky but if we can criteria by which we prioritise then at least some sort of MoSCoW categorisation should fall out.

[Audionut11]

Those requests that have no workaround should receive greater priority IMO, these add functionality.  The others just make life easier (for instance my distance request).

[natehansen66]

Thought I'd bump this thread to the top, and add that the more I think about it an advanced biquad implementation would take care of all of my dsp requests in one fell swoop....

At least for now 

[Fabith]

Better downmix for Headphones

[JimH]

Better downmix for Headphones

Have you tried the settings in DSP Studio?  What's missing?

[Matt]

Next build will have this:
NEW: Added a Linkwitz Riley filter to Parametric Equalizer.

Testing appreciated once the build is available.

[Hilton]

Next build will have this:
NEW: Added a Linkwitz Riley filter to Parametric Equalizer.

Testing appreciated once the build is available.

Cool!

Will it be available in room correction DSP too?

[mattkhan]

Thought I would bump this one based on a few recent posts in http://data-bass.ipbhost.com/index.php?/topic/285-the-bass-eq-for-movies-thread/page-10#entry10251

That thread shows one use case for the "import peq definition from a text file" feature (and also a shelf with S>1 though that can be handled in other ways so is not as important imv though would be nice for flexibility)

[mattkhan]

That thread shows one use case for the "import peq definition from a text file" feature (and also a shelf with S>1 though that can be handled in other ways so is not as important imv though would be nice for flexibility)

for reference here... an example to demonstrate jriver using S for Q,

50Hz +5dB low shelves with S=1 (or Q = 1/sqrt(2) ~= 0.707)
 
green: jriver Q=0.5
red (hidden behind blue): jriver Q=1
blue: acourate Q=0.707
brown: acourate Q=1
 
blue is almost exactly the same as red, the minor variation is because I used 0.707 rather than 1/sqrt(2)

jriver output captured using a loopback into REW

[jjazdk]

*facepalm*

What you are saying is that JRiver does not even use the conventional Q in "S", hopefully that is a mistake and not intentional.

Somebody from JRiver, go fix, immediately.. Together with the still existing convolution-channel+10 issue!!!

[bluescale]

Thought I would bump this one based on a few recent posts in http://data-bass.ipbhost.com/index.php?/topic/285-the-bass-eq-for-movies-thread/page-10#entry10251

That thread shows one use case for the "import peq definition from a text file" feature (and also a shelf with S>1 though that can be handled in other ways so is not as important imv though would be nice for flexibility)

I would love to see the S>1 limitation lifted.  I'd have to assume that this is low hanging fruit.  I've stopped doing BEQ corrections right now because it's a pain in the ass to both convert Q to S, and then figure out the correct stacked filter to make it work.  I should probably build an excel spreadsheet to do this for me, but I'd love to see JRiver get this corrected.

[v_erich]

MS Encoder (Mid Side) would be a good feature to output a Stereo signal not in L+R but in M+S.

Thanks,
Erich

[mattkhan]

MS Encoder (Mid Side) would be a good feature to output a Stereo signal not in L+R but in M+S.

Thanks,
Erich

The ultimate point of this thread is/was to try to provide a strong case for further enhancements to DSP studio by explaining what we want, how widely such a feature is applicable and show how such features don't have workaround.

To that end it would be useful if you can give the use case for that feature, i.e. what would MS encoding enable you to do that you can't do today, and also comment on whether there is a known workaround (e.g. apply some external VST plugin?).

I have only come across 1 use for MS encoding myself, which I don't think is applicable here, so would need some guidance from people who want that feature to be able to promote it into the 1st post.

[v_erich]

I want to use an acourate cleaner or a passive solution with lundahl transformers for better audio quality, see the link with explanations.
http://www.audiovero.de/en/acouratecleaner.php

I used an rme soundcard before which has built-in feature of ms encoding.
Since a few month I convolve and convert all my music (all wav, lots of HD music) with the convert function of JRiver (convolving, level, upsampling,...) and store it seperately on my nas.
I can use an VST plugin for MS encoding, but then I can only convert 1 file, not a few in parallel.
Last time I tried that the conversion of my complete music colletion lasted on a Xeon Server 1 week!
So it is not possible for me to use VST tools for changing something in my music collection because of the time consuption to convert everything.
On the fly conversion is not a solution because the quality degreed to much compared to only playback a converted file (please no discussion, I can hear that here easily).

So an internal MS encoding function would be very helpful for me and my setup.

Thanks and BR
Erich

[AudioMunchies]

Just want to show my strong support for the feature requests in the first post of this thread.

We may not look like a lot of people who are keen on having these features implemented but all of them are important and solving these issues will make JRiver interesting to a broader market.

Just look at the success of miniDSP, and all of the people who use that are already using a computer to control their DSP devices. Big potential market.

[bluescale]

I would love to see the S>1 limitation lifted.  I'd have to assume that this is low hanging fruit.  I've stopped doing BEQ corrections right now because it's a pain in the ass to both convert Q to S, and then figure out the correct stacked filter to make it work.  I should probably build an excel spreadsheet to do this for me, but I'd love to see JRiver get this corrected.

Bump...I'd like to know if this is being considered at all.

[Matt]

Next build we'll try this:
Changed: Tried lifting the limit on Q for a high or low shelf to 5.0 from 1.0.  Testing appreciated.

I'm not familiar enough with what it's supposed to do to know if the change works well or not.  So I'd really appreciate a little testing.

Thanks.

[bluescale]

Thanks, Matt!

There are no new builds of JRiver 21, right?  If I want to test this out, I'll need to upgrade to JRiver 22?

[JimH]

Yes.