IIR vs. FIR filters in Jriver

dallasjustice

Member Sponsor
Apr 12, 2011
2,067
8
0
Dallas, Texas
Nyal and I have been working with REW. I am stunned at how powerful REW is compared to what I knew before using XTZ room analyzer. Nyal generated some excellent filters using several measurements we took using REW Filter Adjust. I attached a screenshot of my jriver PEQ tab. I added a couple of small low shelf filters just for fun. :cool: All of these filters are below 200hz. I think they sound better than any other filters I've ever tried.

My understanding is that Jriver uses IIR filters and dithers them to 64 bits. I have no problem with the sound. I think it sounds great and I can't hear any degradation using these filters. However, I am interested to know what the same filters would sound like using FIR. I want to preserve phase as much as possible. Is there a way to manually create FIR filters and use them with Jriver convolver?

best filters.jpg
 
Last edited:

prerich

Well-Known Member
May 21, 2012
246
10
923
Nyal and I have been working with REW. I am stunned at how powerful REW is compared to what I knew before using XTZ room analyzer. Nyal generated some excellent filters using several measurements we took using REW Filter Adjust. I attached a screenshot of my jriver PEQ tab. I added a couple of small low shelf filters just for fun. :cool: All of these filters are below 200hz. I think they sound better than any other filters I've ever tried.

My understanding is that Jriver uses IIR filters and dithers them to 64 bits. I have no problem with the sound. I think it sounds great and I can't hear any degradation using these filters. However, I am interested to know what the same filters would sound like using FIR. I want to preserve phase as much as possible. Is there a way to manually create FIR filters and use them with Jriver convolver?

View attachment 11432
Other than using Dirac or Audiolense which must be purchased, the only other way that I know of that's free is using DRCDesigner. FIR filters also work in the Time Domain, and are genrerally believed to be better than IIR filters (although both have their advantages and disadvantages IMHO).

Fyi, I thought you already had Audiolense or am I mistaken ;) :)
 

dallasjustice

Member Sponsor
Apr 12, 2011
2,067
8
0
Dallas, Texas
I sold Audiolense a few years back. At the time, I didn't like it because it was full range. I've never been satisfied with full range target based DSP. I understand Audiolense has made some very significant improvements since then. Isn't it possible to high pass the target so I can limit the DSP to the frequency range where it's really useful in my system? Maybe I will try it again. I just didn't want to drop the coin on it again and not use it again. At least Bernt will transfer the license if things don't work out; I think that's pretty exceptional in this business.

Other than using Dirac or Audiolense which must be purchased, the only other way that I know of that's free is using DRCDesigner. FIR filters also work in the Time Domain, and are genrerally believed to be better than IIR filters (although both have their advantages and disadvantages IMHO).

Fyi, I thought you already had Audiolense or am I mistaken ;) :)
 

prerich

Well-Known Member
May 21, 2012
246
10
923
I sold Audiolense a few years back. At the time, I didn't like it because it was full range. I've never been satisfied with full range target based DSP. I understand Audiolense has made some very significant improvements since then. Isn't it possible to high pass the target so I can limit the DSP to the frequency range where it's really useful in my system? Maybe I will try it again. I just didn't want to drop the coin on it again and not use it again. At least Bernt will transfer the license if things don't work out; I think that's pretty exceptional in this business.
If I'm not mistaken, yes you can high pass the target. I also think (don't quote me on this), if you buy the surround version, you can eq just the sub and choose the frequency range that the test tone will play. There's aslo an enitre help page for DRCDesigner http://www.alanjordan.org/DRCDesigner/DrcDesignerHelp.html
It's free...but I think its just for 2 channel.
 

dallasjustice

Member Sponsor
Apr 12, 2011
2,067
8
0
Dallas, Texas
I only use 2 channel. I gave up on the subs. I did blind tests comparing jriver volume control against the totaldac's internal 69 bit dithered datapath. TotalDac beat Jriver's volume control easily on all three tracks in the blind test. It was a real ear opener for me. I never thought different dithering schemes could really make a difference. It did make a difference in ways I hadn't expected and were easy to identify once I knew what to listen for.

Thanks again.
 

edorr

WBF Founding Member
May 10, 2010
3,139
14
36
Smyrna, GA
I only use 2 channel. I gave up on the subs. I did blind tests comparing jriver volume control against the totaldac's internal 69 bit dithered datapath. TotalDac beat Jriver's volume control easily on all three tracks in the blind test. It was a real ear opener for me. I never thought different dithering schemes could really make a difference. It did make a difference in ways I hadn't expected and were easy to identify once I knew what to listen for.

Thanks again.

Interesting. At what levels of attenuation did you do your shootout? Conventional wisdom is says degradation is a function of digital attenuation level, and below 10dB the degradation is negligible. Did you find a bigger gap between the TotalDac and Jriver attenuation at higher attenuation levels?
 

dallasjustice

Member Sponsor
Apr 12, 2011
2,067
8
0
Dallas, Texas
I agree, it was very interesting and unexpected. I also thought dither was simply a function of the number of bits dithered and number of decibels attenuated. It's not that simple. I've spoken to Vincent Brient, totaldac, about this. Dither is much more complex and it is an artform to do it right. I am not saying Jriver does it wrong. However, if dither is done right inside of a DAC's datapath, then it should be superior to an equal number of bits dithered volume control in the software on the server. This is certainly true with Jriver compared to totaldac. Jriver dithers to 64 bits. Totaldac's internal datapath for dither is 69 bits. I've also read Bob Katz regarding different dither implementations. I was able to confirm some of the observations he's made comparing better and worse dither schemes. The difference is better space, depth and less hardness. It is repeatable and easy to test with a friend in a true blind ABX scenario. It's not a huge difference, but noticeable.

Scotty Warren and I selected 3 different tracks. 2 were high rez and 1 was redbook. Each used different amounts of attenuation ranging from -15db to -9db. Every time, it was easy to pick out which volume control was being used.

Yes, the gap was a little bigger using greater attenuation.

Interesting. At what levels of attenuation did you do your shootout? Conventional wisdom is says degradation is a function of digital attenuation level, and below 10dB the degradation is negligible. Did you find a bigger gap between the TotalDac and Jriver attenuation at higher attenuation levels?
 

mojave

Well-Known Member
Oct 29, 2010
251
0
321
Elkhorn, NE
I own Audiolense and you can limit the correction to the highest frequency range you desire. It also has the ability to adjust the volume level of the uncorrected range so that it matches the corrected range. Otherwise the corrected range could be -3 dB from the uncorrected, for example.

This probably wasn't the case, but there are two possible ways that JRiver's dither wasn't in use or setup properly:

1. What version of JRiver were you using? Version 18.0.181 added dither automatically for all bitdepth down conversions. Prior to that it was an option and needed to be selected in Options > Audio > Output Mode Settings to make sure dither was activated.

2. Does the TotalDAC have an ASIO driver? If using ASIO, some 24-bit devices present themselves as 32-bit devices. This can be verified in JRiver's Audio Path. If the device presents itself via ASIO as a 32-bit device and you are using internal volume control, you need to check "Device uses only most significant 24-bits" (in Device Settings) so that JRiver dithers to 24 bits. Otherwise JRiver will dither to 32 bits and the DAC will truncate to 24 bits.

Bob Katz was the impetus behind JRiver's dither.
 

dallasjustice

Member Sponsor
Apr 12, 2011
2,067
8
0
Dallas, Texas
Mojave,

I had read Katz' observations about dithering to 24 bits before I did this test. I set it up exactly the way he recommended. I know Matt added the dither to 24 bits feature he requested even though Matt didn't think it would matter using an ASIO driver.

I used jriver 18 only 3 weeks ago; the most up to date version. I was using an RME HDSPE AIO soundcard and going out AES/EBU in both cases (no totaldac driver used). I checked the dither to 24 bits box, like Katz recommended when using Jriver volume control.

When I used totaldac volume control, I totally disabled the Jriver Volume control.

There's a difference. It's real.
 

mojave

Well-Known Member
Oct 29, 2010
251
0
321
Elkhorn, NE
You certainly did a good job of making sure everything was setup right. :)

Back to your first question, in REW you can export the filters impulse response as a wav file and load into JRiver's convolution engine. I've never done it, but have read of others using it and it saves time from manually entering the filters. I'm not sure if this method is FIR.

There is a free software called rePhase that allows you to create FIR filters for use in JRiver's convolution engine.
 

dallasjustice

Member Sponsor
Apr 12, 2011
2,067
8
0
Dallas, Texas
Mojave,

I don't blame you for being skeptical. I was also surprised at the difference I heard.

You possess a bountiful font of DRC knowledge! Thanks for the millionth time! :D

Maybe I could get Scotty Warren to jump in here and post some comments on what he heard between the 2 volume controls.
 

Scott W

Well-Known Member
Apr 20, 2010
336
163
1,600
Texas
www.suprahifi.com
Here I am :)

Michael asked me to pick three tracks that I was familiar with. He then volume matched them(except one, more on that later) and played the tracks twice, one with and one without JRiver volume control. He did not tell me which was which and asked me to pick out the one I liked best. In all three instance I chose the tracks without the JRiver volume control. In fact on the last track Michael forgot to volume match the tracks and the track with JRiver volume control was louder, but I still chose the non JRiver volume control track as best. For me I hear a greater sense of spaciousness in soundstage and around instruments in the image. Also there is an edge with the JRiver Volume control that sounded a bit unnatural to me.
 

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
Applying dither can be very tricky, especially for oversampled data converters. It is a demanding science blended with art to choose the appropriate dither model for the signal of interest, be it audio, radar, or telcom system. At least IME (which is heavily radar and telcom oriented for full disclosure).

I am not as sure about the choice of FIR vs. IIR, especially for oversampled audio systems. Makes a big difference in some scenarios where pulse integrity is critical, but even then computational efficiency oftens wins. - Don
 

prerich

Well-Known Member
May 21, 2012
246
10
923
You certainly did a good job of making sure everything was setup right. :)

Back to your first question, in REW you can export the filters impulse response as a wav file and load into JRiver's convolution engine. I've never done it, but have read of others using it and it saves time from manually entering the filters. I'm not sure if this method is FIR.

There is a free software called rePhase that allows you to create FIR filters for use in JRiver's convolution engine.
Mojave, I do it this way and it works great!!!! However the filters are still IIR filters (added a Linkwitz Transform on the sealed subwoofers using PEQ2 after convolution - I love it)!
 

j_j

New Member
Jun 25, 2013
325
0
0
In the Rain
home.comcast.net
Applying dither can be very tricky, especially for oversampled data converters.

If they are multibit this isn't so hard, assuming you know the intended final resolution. Of course, if you're also doing noise shaping (which most delta-sigma convertors are going to do) it becomes a bit tougher.

There's an ADC/DAC tutorial at www.aes.org/sections/pnw/ppt.htm It would help to have the audio but it was only recorded as an AES talk and they charge money for those.
 

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
I found noise (dither) bandwidth to be an issue for DS converters, both for performance and stability. I spoke with Gabor Temes about it, and I am pretty sure "tricky" was one of the words he used. :)
 

dallasjustice

Member Sponsor
Apr 12, 2011
2,067
8
0
Dallas, Texas
I found noise (dither) bandwidth to be an issue for DS converters, both for performance and stability. I spoke with Gabor Temes about it, and I am pretty sure "tricky" was one of the words he used. :)

Good thing my DAC isn't SDM, i guess. :D
 

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
Can still be an issue for converters without noise shaping, but generally a poor dither choice will not cause the bloody things to go unstable...

Off-topic aside re. SDM: I began a mission decades ago to switch the world to delta-sigma instead of sigma-delta. A DS modulator first does differencing (delta), then summing (sigma). According to Gabor the original paper was mistranslated...
 

dallasjustice

Member Sponsor
Apr 12, 2011
2,067
8
0
Dallas, Texas
There is a free software called rePhase that allows you to create FIR filters for use in JRiver's convolution engine.
This looks like it's the ticket. I could simply keep the iir filters I already like and use this software to adjust phase where needed. The only problem is that I think it requires use of the miniDSP hardware. I wouldn't want to throw a SHARC chip into the mix. :D
 

NorthStar

Member
Feb 8, 2011
24,305
1,323
435
Vancouver Island, B.C. Canada
IIR filters -> Infinite Impulse Response (Frequency Response and Time Domain and ...?)
FIR filters -> Finite Impulse Response (Frequency Response and Time Domain)

Parametric Equalization -> Frequency Response and 'Q' (width)

Is Time Domain (Phase & Delay) important? If so why use a Parametric EQ?
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing