Print

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

[JustinM]

Hi Guys,

I'm building new 3-way speakers that will be dependent on JRiver built in x-overs.. I was curious on the technical details of JRiver x-overs..   Are they FIR, IIR?
any info on the subject would be great..

[Matt]

Parametric Equalizer use 64-bit IIR filters for low and high pass filters.

Good luck with the build

[JustinM]

Thanks Matt for quick reply 

Your info will help me learn more about them.

Do You Think they will out perform quality passive x/o's ?       (64-bit is alot:)

Has JRiver/anyone ever tested performance??

I (with my little knowledge) that FIR have a post ringing (-130 or so db)  , I assume IIR are the same.... 

[Matt]

Digital filters should outperform anything you could achieve in the analog domain.

You might enjoy DSP Studio > Analyzer to watch the filters in action.  Calibration clips (Tools > Advanced Tools > Audio Calibration...) like pink noise are especially useful for visualizing the filters.

[JustinM]

Thanks Matt

Digital filters should outperform anything you could achieve in the analog domain.

I thought so ,  just fishing for confirmation/info before I placed an order another amp:)  (bryston 2.5b)   

I found this :
"The best way to reduce the quantization noise is improve the arithmetic accuracy. For the practical audio applications, I would recommend to have at least 32 bit data precision in IIR feedback path. That will keep the IIR quantization noise level well below -100dB in most cases"

thats at 32bit ,,,   64 bit is much, much better               64 bit + "my bad math" = -200db .... maybe      

[mwillems]

Hi Guys,

I'm building new 3-way speakers that will be dependent on JRiver built in x-overs.. I was curious on the technical details of JRiver x-overs..   Are they FIR, IIR?
any info on the subject would be great..

One thing that wasn't clear to me about JRiver's crossovers when I started using them (but is super important) is that they are even-order butterworth filters (12db, 24db, 36db, 48db).  You can cascade two to create a 4th order, 8th order, etc. Linkwitz Riley filter, but (to my knowledge) odd order butterworths, and other crossover filters (Bessel, etc.) are not currently supported.  

This turned out great in my bi-amping application, but I only really needed a plain vanilla 4th order Linkwitz Riley, and JRiver happily does that.  Matt, if there is support for other cross-over topographies and I'm confused, please correct me  

I can also confirm that JRiver's filters perform as expected when measured.  I have a hardware crossover that I measured separately with each channel of my bi-amped setup, and then measured JRiver crossovers on the same speakers.  The JRiver filters measure as expected (there's some variance for noise, but the crossover topography and behavior was correct).

[natehansen66]

One thing that wasn't clear to me about JRiver's crossovers when I started using them (but is super important) is that they are even-order butterworth filters (12db, 24db, 36db, 48db).  You can cascade two to create a 4th order, 8th order, etc. Linkwitz Riley filter, but (to my knowledge) odd order butterworths, and other crossover filters (Bessel, etc.) are not currently supported.  

Other target tansfer functions can be created using the PEQ as well, for example using a shelving filter to change the slope and peak/dip eq to change the Q of the filter corner.......it's pretty rare to hit an LR4 acoustic target with an LR4 electrical filter so this kind of work is necessary anyways IMO.

I've been using MC as my digital active xo for almost a year now with no issues that haven't been sorted by J River along the way, and I don't have any intention to try anything else at this point. It just works.

[mwillems]

Other target tansfer functions can be created using the PEQ as well, for example using a shelving filter to change the slope and peak/dip eq to change the Q of the filter corner.......it's pretty rare to hit an LR4 acoustic target with an LR4 electrical filter so this kind of work is necessary anyways IMO.

I've been using MC as my digital active xo for almost a year now with no issues that haven't been sorted by J River along the way, and I don't have any intention to try anything else at this point. It just works.

You can definitely do additional shaping with the shelf function (and I've added some minor "refinements" to my LR4 using PEQ, for sure), but do I understand you to be saying that you could emulate any other filter type using those methods?  

I guess what I wasn't sure about was whether such a filter would have the same phase/group delay consequences as the filter that was being emulated.  For example, if I use a series of PEQ filters to create a slope that (in terms of frequency response) looks like a Bessel curve, would it have the same phase/time domain characteristics as a Bessel?  Or does the frequency response of a filter completely dictate the phase (i.e. am I asking a stupid question)?

I certainly wasn't trying to complain, just trying to add a little extra info on the "off the shelf" crossover settings (no pun intended). I've been very happy with JRiver's versatility when it comes to DSP, and I couldn't run my system without JRiver at this point. It does definitely "just work".  

[JustinM]

You can definitely do additional shaping with the shelf function (and I've added some minor "refinements" to my LR4 using PEQ, for sure), but do I understand you to be saying that you could emulate any other filter type using those methods?  

I guess what I wasn't sure about was whether such a filter would have the same phase/group delay consequences as the filter that was being emulated.  For example, if I use a series of PEQ filters to create a slope that (in terms of frequency response) looks like a Bessel curve, would it have the same phase/time domain characteristics as a Bessel?  Or does the frequency response of a filter completely dictate the phase (i.e. am I asking a stupid question)?

I certainly wasn't trying to complain, just trying to add a little extra info on the "off the shelf" crossover settings (no pun intended). I've been very happy with JRiver's versatility when it comes to DSP, and I couldn't run my system without JRiver at this point. It does definitely "just work".  

Good question mwillems,

something I've also wondered, and been meaning to find out... If I remember correctly :   Typically a passive  2nd order high-pass filter has a +90 degree phase shift at the crossover point, and a 2nd order low-pass filter has a -90 degree phase shift at the crossover point...
Does this hold true for JRiver digital (iir) x-overs??

[natehansen66]

 

I guess what I wasn't sure about was whether such a filter would have the same phase/group delay consequences as the filter that was being emulated.  

I've wondered that myself..... My thinking is that they would match up as freq and phase go hand in hand.

[mwillems]

I've wondered that myself..... My thinking is that they would match up as freq and phase go hand in hand.

Phase shifts that result from cross-over filters definitely vary with frequency in a predictable way, but I'm not sure that phase is dictated by frequency response in a 1 to 1 fashion in all cases.  For example, FIR filters (which are a different animal than most analog crossovers, and the IIR filters that JRiver uses) can (I'm told) create linear phase crossovers with flat frequency response.   So phase and frequency response have to be decoupled to some extent (or perhaps only in some circumstances).

None of which gets us any closer to answering my previous question i.e. would a filter (made out of various JRiver PEQ filters) that emulates the frequency response of a Bessel filter have the same phase characteristics as a Bessel filter?  My guess would be not, but I don't actually know and haven't tested it (and may not have the know-how/ability to do a real test of it myself).

This is all academic, anyway, as I don't need any additional crossover functionality from JRiver at the moment, but I like understanding how things work  

[JustinM]

Just wanted to follow up:
IIR digital filters do have phase shift....  (non-linear)

So far the only info I've found on IIR has been very technical. Good for designing/ implementing, not so good for basic understanding.

I  believe: The digital signal is mathmatically conververted to represent an analog waveform, (using Bilinear transform) Where they Simulate a Cross-overs transfer fuction (eg :2nd order butterworth,) Then Take the result and convert it back to Digital (again using Bilinear transform)...   I believe the "simulation" of the cross-over transfer function is recursive.. which is where the phase shift comes from..  This makes sense why JRiver felt the need to go 64-bit for this process..  (thankyou again JRiver)

Please feel free to correct my understanding if something is wrong..

I'll be time aligning my new speakers,, can't wait to start lissening and take measurements without the passive xo's

[mwillems]

Just wanted to follow up:
IIR digital filters do have phase shift....  (non-linear)

So far the only info I've found on IIR has been very technical. Good for designing/ implementing, not so good for basic understanding.

I  believe: The digital signal is mathmatically conververted to represent an analog waveform, (using Bilinear transform) Where they Simulate a Cross-overs transfer fuction (eg :2nd order butterworth,) Then Take the result and convert it back to Digital (again using Bilinear transform)...   I believe the "simulation" of the cross-over transfer function is recursive.. which is where the phase shift comes from..  This makes sense why JRiver felt the need to go 64-bit for this process..  (thankyou again JRiver)

Please feel free to correct my understanding if something is wrong..

I'll be time aligning my new speakers,, can't wait to start lissening and take measurements without the passive xo's

Well that makes sense (as to where the phase shift comes from).  Thanks for doing some digging. 

Time alignment can be kind of fiddly, I wish you good luck!  I tried physical distance measurements and calculating my delay that way, but it didn't work quite right (I didn't have flat frequency response at the crossover point, and it sounded odd). 

Ultimately I just had to get out a microphone and some RTA software and use the "greatest null" method to determine what the ideal delay setting was.

But tweaking is half the fun :-)

[natehansen66]

Anyone know of any (free) software that you can use to generate magnitude/phase data for the various transfer functions? It would be nice to have these as an impulse or text file. I'd like to see how an e.g. Bessel filter created in the PEQ matches up to the real deal.  I used to use a miniDSP to create filters to use as the target in REW's eq window to do eq/crossover work but recently sold the mini's. I'm an LR crossover type guy so I use the software I mention below to create the targets, though it creates FIR filters so I don't know if the phase is that of a textbook filter.  I don't see why it wouldn't be but it'd be nice to test.

For time alignment what works well for me is to add a 5ms or so delay to one of the drives, then do a limited bandwidth measurement just around the crossover (with the crossover filters in place). I like to add the delay to the tweeter, then it's pretty obvious where the tweeter peak in the impulse is in relation to the mid/whatever driver and you can go from there.

If anyone is curious about FIR/linear phase stuff here's a pretty cool free software that's ongoing development: http://www.diyaudio.com/forums/multi-way/221434-rephase-loudspeaker-phase-linearization-eq-fir-filtering-tool.html

[mwillems]

Anyone know of any (free) software that you can use to generate magnitude/phase data for the various transfer functions? It would be nice to have these as an impulse or text file. I'd like to see how an e.g. Bessel filter created in the PEQ matches up to the real deal.  I used to use a miniDSP to create filters to use as the target in REW's eq window to do eq/crossover work but recently sold the mini's. I'm an LR crossover type guy so I use the software I mention below to create the targets, though it creates FIR filters so I don't know if the phase is that of a textbook filter.  I don't see why it wouldn't be but it'd be nice to test.

For time alignment what works well for me is to add a 5ms or so delay to one of the drives, then do a limited bandwidth measurement just around the crossover (with the crossover filters in place). I like to add the delay to the tweeter, then it's pretty obvious where the tweeter peak in the impulse is in relation to the mid/whatever driver and you can go from there.

If anyone is curious about FIR/linear phase stuff here's a pretty cool free software that's ongoing development: http://www.diyaudio.com/forums/multi-way/221434-rephase-loudspeaker-phase-linearization-eq-fir-filtering-tool.html

Thanks so much for that software link.  That looks absolutely incredible.  I'm fooling with it right now and it's blowing my mind a little (I wasn't aware there was free software to do this with this kind of easy interface)

EDIT: I've been tinkering with the software.  It will make minimum phase crossovers (which have the normal phase characteristics), but it will also make linear phase crossovers (that definitely don't have the "usual" crossover phase characteristics).  Most importantly, it will "linearize" the phase response of a target external crossover (such as one applied in JRiver), by generating an inverse phase convolution filter.  So you can use it with JRiver to create linear phase crossovers for free. 

Which is huge (at least for me).  Really, really cool

[mwillems]

So I was talking to a friend of mine who is an EE (who does some audio design) about this conversation, and he indicated that if JRiver's filters are all minimum phase (which I believe they are), a given frequency response curve completely dictates the phase.  So to bring it back to the previous point, if you use a series of minimum phase filters to approximate the slope of a Bessel filter, it will have the same phase characteristics as a Bessel filter, and so will be, for all meaningful purposes, a Bessel filter.  Of course, all that goes out the window when you're discussing non-minimum phase filters (i.e. linear phase filters)

So if my friend is correct, JRiver (in the active crossover context) is even more flexible than I was giving it credit for.  You could (with adequate math background, visualization skills, or trial and error) use the tools in PEQ to make any conventional crossover topography (or even unusual ones like Horbach-Keele). 

That linear phase software you dropped in above is incredibly easy to use, and provides the ability to effectively "linearize" your existing crossovers (and box-based phase ripples too!).  I've spent the whole afternoon A/Bing the same set of EQ/crossovers, switching back and forth from linear to minimum phase.  I was surprised at how immediately audible the difference is (at least one of my guests could immediately identify which was which with her eyes closed).

I can't decide which one I prefer, but I really need to do some additional measurements before I get too far down that road.