Ninfendo said:
The Affirmer will most likely have adjustable latencies because there are tradeoffs and compromises to be made.
You should be able to get a very good DRC with under a frame latency (30msec).
Thuneau,
Can you explain these compromises, do you mean the number of taps?
I have played with http://convolver.sourceforge.net/ when adjusting the partitions you can get very low latency at the expense of CPU load. Will the affirmer work similarly?
Since Intel is releasing Quad Core CPUs before year end a multithreaded convolver should get amazing performance on such CPUs.
Can you explain these compromises, do you mean the number of taps?
I have played with http://convolver.sourceforge.net/ when adjusting the partitions you can get very low latency at the expense of CPU load. Will the affirmer work similarly?
Since Intel is releasing Quad Core CPUs before year end a multithreaded convolver should get amazing performance on such CPUs.
You should try BruteFIR, too.
In simple terms, you have two ways of doing a convolution (digital filtering) :
- you compute it the oldskool way (ie. multiply-sum the incoming signal with the filter impulse response). If you do room correction, your impulse response will be quite long (maybe 10-50k samples, called taps), so you'll have to do this number of multiply-accumulate operations for each sample. So, for a 50k taps filter at 48 kHz sampling rate stereo, you get 50k*48k*2 = 4.8 billions of multiply/accumulate per second. So, this method is only valid for very short filters.
- you take a chunk (of size N) of the input signal, FFT it, and do the filtering on the FFT, then un-FFT it and get your filtered signal.
FFT is about O(N log N) operations, and filtering is O(N), so this is insanely faster than doing it the oldskool way.
However a FFT of a chunk of size N is about twice (2 / log 2 really) as fast as a two FFTs of chunk N/2.
And your latency is about 2N samples.
So, lower latency means smaller chunks which means more CPU power needed.
But latency is only a concern if your audio has to be in sync with something that can't be delayed, like video from a DVD player for instance.
You can live with 1 second latency (which is huge) if you don't have video : the music will simply come out 1 second after pressing play. If you play video from a device or software that allows you to delay the video by the audio latency amount, it's exactly the same.
In simple terms, you have two ways of doing a convolution (digital filtering) :
- you compute it the oldskool way (ie. multiply-sum the incoming signal with the filter impulse response). If you do room correction, your impulse response will be quite long (maybe 10-50k samples, called taps), so you'll have to do this number of multiply-accumulate operations for each sample. So, for a 50k taps filter at 48 kHz sampling rate stereo, you get 50k*48k*2 = 4.8 billions of multiply/accumulate per second. So, this method is only valid for very short filters.
- you take a chunk (of size N) of the input signal, FFT it, and do the filtering on the FFT, then un-FFT it and get your filtered signal.
FFT is about O(N log N) operations, and filtering is O(N), so this is insanely faster than doing it the oldskool way.
However a FFT of a chunk of size N is about twice (2 / log 2 really) as fast as a two FFTs of chunk N/2.
And your latency is about 2N samples.
So, lower latency means smaller chunks which means more CPU power needed.
But latency is only a concern if your audio has to be in sync with something that can't be delayed, like video from a DVD player for instance.
You can live with 1 second latency (which is huge) if you don't have video : the music will simply come out 1 second after pressing play. If you play video from a device or software that allows you to delay the video by the audio latency amount, it's exactly the same.
Tweeter protection
I have up until now only used active filtering for the subwoofer and midrange. I would now like to replace my current passive crossover for midrange and tweeter. I am a bit worried about blowing the tweeter by mistake when removing the caps. Does anyone use some sort of tweeter protection? Can someone recoomend something? The tweeter I use are ScanSpeak 2905/9500. The crossover frequency will be around 2500 Hz.
I have up until now only used active filtering for the subwoofer and midrange. I would now like to replace my current passive crossover for midrange and tweeter. I am a bit worried about blowing the tweeter by mistake when removing the caps. Does anyone use some sort of tweeter protection? Can someone recoomend something? The tweeter I use are ScanSpeak 2905/9500. The crossover frequency will be around 2500 Hz.
Ninfendo,
I still use a single large cap in front of any ribbon mid or tweeter. I had a few 3" quasi ribbon tweeters die while directly connected. I dont think it's the small DC offset that kills the tweeter, but rather the turn-on thump from the amp.
I have some Seas Millienium tweeters that I suspect might be able to handle the turn-on thump, but I'm not feeling like damaging expensive tweeters.
To be honest, I can only slightly hear the difference between no-cap and a single large foil cap. I'm using the Dayton foils from PartsExpress.
I still use a single large cap in front of any ribbon mid or tweeter. I had a few 3" quasi ribbon tweeters die while directly connected. I dont think it's the small DC offset that kills the tweeter, but rather the turn-on thump from the amp.
I have some Seas Millienium tweeters that I suspect might be able to handle the turn-on thump, but I'm not feeling like damaging expensive tweeters.
To be honest, I can only slightly hear the difference between no-cap and a single large foil cap. I'm using the Dayton foils from PartsExpress.
Daveis said:
I was thinking about getting a 10 uF Hovland Musicap (that's the largest one you can get, alternatively picking a slightly smaller one) and then connect it in parallel with a cheap cap, i.e 20 uF SCR cap.
Hi
Looking at the pictures of the Audiotrak Profigy 7.1XL card
(on the Korean site) I say that either a cheeky photographer or Pr. agenthad put one B&B OPA2134 Opamp in one of the sockets and a JRC in the other. As the Prodigy 7.1 Hifi is said to have the same feature; IE Socketed opamps, perhaps we will see lots of
claims and counter claims as to a quality increase with particular
combinations. How about double socketed OPA627s running class A biased with custom RF shields. Now all Audiotrak has to do is
repair its reputation in the software realm. From a crossover point of view it will make PCs in the sound chain that more accepted which bodes well for the future.
Regards
AnthonyPT
Looking at the pictures of the Audiotrak Profigy 7.1XL card
(on the Korean site) I say that either a cheeky photographer or Pr. agenthad put one B&B OPA2134 Opamp in one of the sockets and a JRC in the other. As the Prodigy 7.1 Hifi is said to have the same feature; IE Socketed opamps, perhaps we will see lots of
claims and counter claims as to a quality increase with particular
combinations. How about double socketed OPA627s running class A biased with custom RF shields. Now all Audiotrak has to do is
repair its reputation in the software realm. From a crossover point of view it will make PCs in the sound chain that more accepted which bodes well for the future.
Regards
AnthonyPT
I'm currently looking for a soundcard upgrade. The price difference between RME Fireface 800 and Multiface II is rather big, but both seem to offer the same amount of analog outputs. How does these compare in terms of sound quality?
Ninfendo said:
The Fireface is a completely different animal to the multiface 2.
You've got 10 inputs and outputs, 18 digital in's and outs. You've also got 4 mic pre's, better quality ADC and DAC's, 42bit attenuation and overall a more complete and better sounding interface. Its also aimed at DAW's with STMP option.
What can I say Thunau!
I've finally got around to setting up the Frequency Allocator(FA) and am stunned at the quality and the price.
I've been flicking back and forth between the Waves+Voxengo setup, DEQX and your own software. The DEQX sounds pretty sterile and processed when you flick back and forth, still sounds great but I'd simply can't bear to use it after hearing what FA does.
Extremely easy to use, lots of power to tweak to Nth degree but still approachable. More over the sound is the best I've heard from any of my digital implementations.
Buy ARTA, FA and an ECM8000 and you'll be knocking out world class speakers in literally days and for only a few hundred dollars, its incredible really.
I'll be selling my DEQX and re-investing the money into the PCXO once again!
I've finally got around to setting up the Frequency Allocator(FA) and am stunned at the quality and the price.
I've been flicking back and forth between the Waves+Voxengo setup, DEQX and your own software. The DEQX sounds pretty sterile and processed when you flick back and forth, still sounds great but I'd simply can't bear to use it after hearing what FA does.
Extremely easy to use, lots of power to tweak to Nth degree but still approachable. More over the sound is the best I've heard from any of my digital implementations.
Buy ARTA, FA and an ECM8000 and you'll be knocking out world class speakers in literally days and for only a few hundred dollars, its incredible really.
I'll be selling my DEQX and re-investing the money into the PCXO once again!
ShinOBIWAN said:
Wow,
Thanks for a good review. May I quote you?
I never heard the DEQX so I can't comment on the differences, but I like the sound on the Allocator through my RME/Alesis Interface better than the custom active analog crossover I built (optimized in lspCAD- no generic LR4 filters here). And that's coming through an older digital to analog converter (Alesis AI-3) as compared to the output of a high end Sony SACD into OPA2134 opamp based crossover.
Don't get me wrong, the analog crossover sounds really good, but the digital implementation with linear phase is better yet.
ShinOBIWAN said:
Are you using the RME Fireface 800 or the 400 ?
Culd you enlighten us if you know the converter types ?
BR
Morten
I'm currently having trouble accessing thuneau.com . . . is this just a transitory thing or has the site been taken off-line? Thanks!
Thunau said:
No problem at all Thunau, probably best to correct all my dodgy spelling and gramma first 🙂
cph2000 said:
Looks to be good for measuring speakers and generating FIR filters. It seems like you'd have to use the generated filters from this with something like Brutefir.
I'm currently using Thuneau's Allocator Lite for it's IIR type filters and very low latency. I like that I can run a DVD software player and have audio-video synch. A 50,000 tap FIR filter certainly isnt going to let me do that.
Still, this looks promising enough to try. The $430 USD price is a bit much, but then nothing else I've seen does the convolution and log sweeps in a single package for Windows.
Daveis said:
I've only had a brief look but my impression is a good one. Its a bit head scratching at first since there's not that much documentation on how to work the whole thing but Uli has since added some tutorials which do an OK job of giving you a rough idea of what to do.
What's most impressive is the rate Uli is adding features to the package, it now does DRC, target curve, driver correction and crossover generation as well as a couple of other little utilities. All those make this one of the most complete packages available IMO. You pay for all that power though but 340Euro is OK value considering all things relative.
Definitely one to watch but I don't consider the documentation anywhere near sufficient for the asking price. I'd like to see a nice and thorough PDF manual that give thorough explaination of functions and also step by step instructions on how to perform particular tasks as well as a detailed trouble shooting section. I've emailed Uli with questions and he answers promptly so his support is good.
- Status
- Not open for further replies.