Interesting DSP info

I think we all know by now what this section is for.
iansr
Old Hand
Posts: 415
Joined: Thu Nov 08, 2007 10:44 pm
Location: West Riding

#1 Interesting DSP info

Unread post by iansr »

I think most people would agree that DSP has the potential to be incredibly useful in relation to both speaker crossovers and room correction. It can achieve things which are simply impossible if you restrict yourself to the analogue domain. The widespread perception however is that all that utility and flexibility comes at a cost, which is the alleged degradation in sound quality that results from any ADC and the inevitable DAC. In my experience MiniDSP kit does nothing to allay these concerns. Some years ago I performed some experiments with a MiniDSP 4x10HD and confirmed to my own satisfaction that it was far from transparent. Furthermore my experiments seemed to confirm that the issue was the sample rate converter that converts even digital inputs to a standard sample rate / bit depth used by the DSP modules. (If you use the analogue output then that just makes things even worse!)

As a result of these experiments I lost interest in DSP, that is until I started reading a series of articles by Mitch Barnett about Acourate and AudioLense.
These are software DSP solutions (ie they run on a PC) and are incredibly powerful in terms of what they can do - they are just in a different league to hardware based DSP and I include in that Trinnov and DEQX. if you are not already familiar with them then I suggest you read up on them.

So we come to the interesting bit; I did have a concern that Acourate and AudioLense might be similarly afflicted by a sample rate conversion process. However I have recently discovered that at least some convolution engines (inc JRiver, AudioLense’s convolver and Hang Loose Convolver) leave the original digital signal in its “native” state and it stays unconverted all the way to the DACs. What happens instead is that the filters are converted to the same sampling frequency / bit depth as the music signal so as to facilitate the convolution. This is the ideal solution.

I’m pretty excited by what these software packages can do and I shall be trying out AudioLense late this year.
"Its good enough for Government work."
iansr
Old Hand
Posts: 415
Joined: Thu Nov 08, 2007 10:44 pm
Location: West Riding

#2 Re: Interesting DSP info

Unread post by iansr »

Well given the number of responses this obviously wasn’t as interesting as I thought ! :roll:

I am inclined to think that a lot more hi fi enthusiasts would be interested in AudioLense and Acourate if (a) they understood enormous advantages these software packages bring to the party and (b) there wasn’t such widespread stigma associated with digital audio. (Plus all the historical passive crossover baggage that the industry is saddled with . . . ). The irony of course is that 99% or more of music is now recorded, mixed and mastered in the digital domain. Anyway, by of encouragement to delve more deeply into this stuff, here is a quote from renowned mastering engineer, Bob Katz:


“Acourate is the first DRC that I can thoroughly recommend. The resulting sound is unquestionably equal or superior to the uncorrected loudspeaker in all respects: Transparency is equal, there is no perceived loss with Acourate. It is truly an audiophile-quality system that even die-hard audiophiles and analogphiles need not be afraid of. Everything else about Acourate makes the corrected loudspeaker sound superior: Stereo imaging and soundstage are more exact and the sweet spot in the center is effectively widened. Tonality is greatly improved and the frequency response extends perceptually flat from 20 to 20 kHz. Transient impact is superior and there is no loss of headroom and no perceived noise, when the gain staging is done correctly. Some of the technical reasons for Acourate's superiority: 64-bit calculation throughout, properly dithered to 24-bits at the end of the chain; no degrading ASRC circuit, the sample rate that goes out is the same as what comes in. No overcorrection, a unique breakthrough in psychoacoustic analysis beats any previous third- or sixth- octave techniques for estimation of the audible effect of the room and loudspeaker combination, and a variable calculation window ensures accurate frequency response. For the first time with any correction system, I felt no need to change or tweak any filters or add any filters to the circuit. Superior target design, this is the most ergonomic part of the program and allowed me to zero in on the ideal high frequency rolloff for my system in a very short time. I haven't found need to change the target since the first day I designed it. Superior impulse response and phase response. Superior crossover implementation, linear phase and with the most accurate curve. Basically, they do everything right and I've only scratched the surface in this description of its superior abilities. Superior to any other DRC I have used or tested.”

PS. Note he references the SRC issue that I refer to in the OP
"Its good enough for Government work."
User avatar
IslandPink
Amstrad Tower of Power
Posts: 10041
Joined: Tue May 29, 2007 7:01 pm
Location: Denbigh, N.Wales

#3 Re: Interesting DSP info

Unread post by IslandPink »

I missed your first post on this, but i am interested in where you get to.
I share your frustrations with the typical implementations of DSP.
In my case it isn't offset to any extent by accepting a lot of modern recordings as being OK. Most of my record collection dates from pre-1985. However in the modern era I tend to find that CD's sound better than LP's.
"Once you find out ... the Circumstances ; then you can go out"
vinylnvalves
Old Hand
Posts: 1099
Joined: Sat Jan 26, 2013 12:08 pm

#4 Re: Interesting DSP info

Unread post by vinylnvalves »

I saw the first post, but didn’t at the time have anything to add. ( I could have added the popcorn meme :wink: )

Audiolense is a tool like a lot of others it’s a convoluter. I know Chris uses the tools inside REW to write his FIR files. I have tried a few they all have there own tricks, yes they can generate you a billiard table flat response with linear phase.
I have shared my experiences of linear phase before, when LINN demonstrated it a show, yes it does make a difference, if you like it is another matter.

The challenge for me and others is the number of taps your DSP chip can handle, which limits the equations you can use. The number of taps increases inversely with frequency, so modifying sub 100hz is very power hungry.

I generally use drivers that are well behaved with linear responses, so only use my limited DSP power to generate conventional crossover filters, so dont use notch filters etc, just mild contouring.

If your looking at going down the whole linear phase rabbit hole, you will need a dedicated PC running the likes of Jrivers to do this. I don’t thing the new hypex fusion amps have that much DSP capability either. I maybe out of date with what the likes of Powersoft amps and the like can do... Chris will correct me if I am wrong.

A recent article on Diyaudio about William Cowans latest mega bucks horn speaker, had him stating they had tried linear phase and hadn’t gone down that route, although the speakers use DSP to get the delays for the MEH’s right.

A side observation to Marks comment, I was listening to Tusk this evening on LP, via my DSP driven system, I was surprised how much better it sounded on vinyl, so DSP isn’t just for modern music.
iansr
Old Hand
Posts: 415
Joined: Thu Nov 08, 2007 10:44 pm
Location: West Riding

#5 Re: Interesting DSP info

Unread post by iansr »

Steve
Yes you need a PC or a Mac running a convolution engine to apply the filters generated by Acourate and Audiolense. So in my case that PC will run JRiver (which has an embedded convulsion engine) ED I’ve just seen this auto correction but I’m going to leave it in for entertainment purposes and it will also become my digital music source / streamer. The DSP filters use 65K taps - that’s a hell of a lot but that’s a reflection of the power and functionality of this software. You can play around with your own choices of target response curves and BTW you certainly don’t want billiard table flat. Just referencing FR and linear phase is greatly understating the sophistication and functionality of these software packages - I can’t begin to cover it all here so if I’ve piqued your interest then I’d recommend Mitch’s YouTube video presentation :
"Its good enough for Government work."
vinylnvalves
Old Hand
Posts: 1099
Joined: Sat Jan 26, 2013 12:08 pm

#6 Re: Interesting DSP info

Unread post by vinylnvalves »

Yes I have 9800 taps to play with, so limited to only HF stuff with FIR’s. I do have a second Nadja ( spare) but don’t want the hassle of syncing and turning on multiple devices, I like to be able to turn on a couple of boxes and that’s it. Been there with a 3 valve amps to turn on, happy to compromise with my multichannel amp ( it is 4 amps in one box really).
iansr
Old Hand
Posts: 415
Joined: Thu Nov 08, 2007 10:44 pm
Location: West Riding

#7 Re: Interesting DSP info

Unread post by iansr »

I totally get the convenience thing and I’m with you. I wouldn’t dream of using a PC to play music unless I could run it headless and use a tablet as the interface. Multiple amps is a bit of a pain but I think it is going to be worth the hassle in what it delivers sonically. If my expectations of AudioLense are not fulfilled then I’ll be going at least semi-active with a Pass active analogue crossover.
"Its good enough for Government work."
vinylnvalves
Old Hand
Posts: 1099
Joined: Sat Jan 26, 2013 12:08 pm

#8 Re: Interesting DSP info

Unread post by vinylnvalves »

With a horn loaded system, with high DI, room modes shouldn’t be an issue, so a hybrid system may be workable, using a hypex plate amp for the lower octaves
User avatar
IslandPink
Amstrad Tower of Power
Posts: 10041
Joined: Tue May 29, 2007 7:01 pm
Location: Denbigh, N.Wales

#9 Re: Interesting DSP info

Unread post by IslandPink »

vinylnvalves wrote: Mon Mar 13, 2023 10:54 am With a horn loaded system, with high DI, room modes shouldn’t be an issue,
Directional at 200Hz ..... :)
"Once you find out ... the Circumstances ; then you can go out"
Post Reply