It seems that you know a lot about DSP and cross overs, but it don't help that you think that DIY'er are stupid and don't know what they are doing. You even claim that everybody using a DCX2496 are stupid - even soundengineers (who will never use a Behringer if there are other possibilities), and your comments on how PA systems wotk, clearly shows that PA this is not you strong side.
Regarding audiophile use. Have you even seen the audiophile mods to the Behringer? Have you ever considered that audiophiles using the Behringer don't necessarily just use high order slopes? The Behringers strengths is cheap DSP power that can be modified to audiophile quality.
Behringer wasn't first with this product group (and they will never be).
Thanks to you other guys for a very interesting thread, keep up the good work.
Last edited:
Though I would love to see 192khz supported I have a doubt that your filters could cope with the processor overhead, well your processor I mean. doubling the frequency rate quadruples the processor requirements. Though I should probably read up a bit on what DSP you are using.
Either way I think this is a very cool project. My mac coding skills aren't what they were and python is just something I play around with. But I could offer to test mac stuff.
Also any updates to this project?
Matt
Either way I think this is a very cool project. My mac coding skills aren't what they were and python is just something I play around with. But I could offer to test mac stuff.
Also any updates to this project?
Matt
Hi Matt,
The DSP could still do quite a few things at 192 kHz.
What you write about doubling the sampling frequency and quadrupling the processor load is not always true. For IIR filtering for example, a double sampling frequency 'only' yields double processing load. Same for delay lines management.
On the other hand, what you're writing is indeed true for non-sparse FIR filters, such as convoluting with a sampled real-world impulse response, because both the buffered portion of the incoming signal and the system impulse response are twice longer.
At this stage, here is the status of the project - thanks for asking.
- Python was rejected a long time ago, somewhere between page 3 and page 7 of this thread
The selected framework is C++ and wxWidgets for the user interface and full assembly for the dsp firmware.- A big part of the dsp code is ready.
- The UI is in progress, with synchronized Windows and Linux builds.
- There are talks about a design for a new board with code-compatible DSP and some goodies.
Cheers,
chaparK
I doubt the others will welcome you thanks. You are incorrect with your assumptions. This is offending, and deserves to be flagged.
I think you offend the users of Behringer DCX2496 (and the like). I know by fact, that audiophiles who uses this filter modify them with digital in/out or better analog in/outputs. They don't use steep slopes, they create new filters via excel spreadsheets, and they actually measure the output, and tweak untill they have the desired results (phase, frequency, gain etc.). Just like any other audiophile DYI'er would do with a passive XO.
Hey amigos,
Please peace. Dance some samba, have a couple of wodka. If that doesn't clear the issue, please use PMs so that this thread stays focused on innocent filters.
Cheers,
chapaK
He did but who cares....I have used the DCX2496 for many years and Im extremely happy. Im not concerned about subjective opinion of some audiophiles since that is a dirty word to me anyways
I have followed this thread for a while and Im always interested in new options. I have the MiniDSP and Im testing it now but Im hoping something happens in a thread like this too.
There is room for ALL products and everyone has 100% control over their OWN choices. Lets keep the thread on the right topic, please!
Chapak, is there a funding issue?? Maybe some donations are needed?
Last edited:
Thanks buddy, but there's no such thing.
It's all about time . I was coming alone saying hey i have these boards let's do something. Here's what i put on the table, a good hardware base and i can do a custom dsp firmware according to agreed specs - if realistic cause my job gives me already more programming than a normal human could wish. Well i'm ok to give away the board for what the unit would cost just in parts and pcb when batched in hundred. Then the whole thing could become an open-source project or one could send me a PM that he has 5 kids to feed and so let's plan a better world $
$ but in both cases he should legitimately put something in.Then it's about timing because if there's no deadline there's no fun. Together we can bop it out in a short while, while alone i'll make it like a slow ballad.
Hm now i'm thinking again about your offer. "Ah! les sirenes!"
Cheers,
chapaRK
Such new option would be the Lipshitz-Vanderkooy filter (the delay-based one) like described here : AES E-Library: A Family of Linear-Phase Crossover Networks of High Slope Derived by Time Delay . I've done it already using SynthMaker, it takes 5 minutes to program on such platform. Do you know SynthMaker ? If you dislike SunthMaker, you can implement it partially using a DCX2496, using the DCX2496 for the needed delay lines, with an external analog lowpass and an external analog substractor. Anyway, using my SynthMaker implementation, I've been positively impressed by the results when chosing a Bessel 4th order. You get a 4th order low pass, a 2nd order highpass (actually, you can get slightly better by slightly oversizing the nominal delay value), the reconstructed phase is perfectly linear (inherent by construction) (you know the meaning of this ? really ? ), and because of the Bessel filter, the relative phase shift between lowpass and highpass is very small. For audiophiles hating high-order filters, this sounds like ideal. If you tolerate more relative phase shift between the lowpass and the highpass, you may opt for a dual Butterworth lowpass (two Butterworth 2nd order lowpass in series), and you may then get a 3rd order highpass instead of a 2nd order lowpass. From what I remember. Once you will see this at work you wil agree with me that 99.9% of soi-disant crossover specialists still have no idea what DSP can bring to audio crossovers apart from throwing more power in tweeters using very high slope highpass filters. Or hiding erratinc woofer/medium responses in the highs using very high slope lowpass filters. I'm deeply sorry about this. Now, how many people will a) want to read this Lipshitz-Vanderlooy AES proceeding and b) pay the few dollars for dowloading the .pdf and c) invest the time that's needed for translating this into a DSP program onto a DSP platform and d) make objective measurements and e) make subjective quality assessments in comparison with Linkwitz-Riley and the likes ? There is no free lunch !
Last edited:
As you may have noticed most people don't compensate the transducers own responses correctly. How many people do take the time extracting the transfer function of their unfiltered transducers ? Nobody because they don't own measurement gear ? Okay, but why isn't there a Bode Plot modality in the digital crossover ? Because a mike input is something too expensive ? Come on ! No no, they don't provide this because 98% of the customers won't even read the user manual. The remaining 2 percent will be 1% reading the manual but not applying the procedure correctly and maybe 1% reading the manual and applying the procedure correctly.
Measuring the unfiltered transducers introduces painful question : what is the required bandwith where we will exactly compensate the transducers ? Is it when the target output is above -12 dB (hence more than 1/4 in voltage compared to the other transducers), or shall we go for -20 dB (hence more than 1/10 in voltage compared to the other transducers). You need to make intelligent choices. Like basing on the natural response of the drivers, and keeping adding more filtering, instead of being forced to raise levels before you do some filtering. If you see what I mean.
So, in a nutshell, when is the last time you could enter the Bode plot of your unfiltered transducer (gain and phase), select a Lipshitz-Vanderkooy delay-derived linear phase crossover, and let the crossover generate IIR coefficients using a built-in bilinear p->z transform ?
What is the market for such device (how many units a year ?), and what would be the end-user price for the needed SMT-equipped PCBs and software ?
And, being audiophile-graded, what would be the end-user price for two options : a) the AES/EBU input fitted with a sample-rate converter doing also the phase de-jitting, and b) the volume control using an infrared remote.
Last edited:
Im confused about your posts. Are you making general statements or are you advancing the thread. Im not interested in subjective opinion.
1. I think anyone that is in the DIY game with speakers is measuring something, who cares about anyone that doesn't.
2. Im not paying money for an AES link and no I do not know anything about synthMaker and I have the DCX (3 of them), they are 100% fine for its requirements, all who hear my system love it, I love it...Audiophiles are not invited though they sound and smell funny This is thread is for an alternative solution, I do not want pro audio XLR and gain issues involved with this solution. As for anyone thinking anyone is stupid...those people need to put there last monthly income statement into the discussion and we will see who is actually smart and who isnt
3. Who posts "soi-disant crossover specialists", "Bode plot of your unfiltered transducer" or even "transducers" at all Im sorry but Im never impressed with talk like that (congrats on being a scrabble master though), just simplify it several notches because its a waste of time and creates more confusion then necessary.....being a little more normal and using common terms is always a good thing.
1. I think anyone that is in the DIY game with speakers is measuring something, who cares about anyone that doesn't.
2. Im not paying money for an AES link and no I do not know anything about synthMaker and I have the DCX (3 of them), they are 100% fine for its requirements, all who hear my system love it, I love it...Audiophiles are not invited though they sound and smell funny
This is thread is for an alternative solution, I do not want pro audio XLR and gain issues involved with this solution. As for anyone thinking anyone is stupid...those people need to put there last monthly income statement into the discussion and we will see who is actually smart and who isnt 3. Who posts "soi-disant crossover specialists", "Bode plot of your unfiltered transducer" or even "transducers" at all
Im sorry but Im never impressed with talk like that (congrats on being a scrabble master though), just simplify it several notches because its a waste of time and creates more confusion then necessary.....being a little more normal and using common terms is always a good thing.
Last edited:
1. Do you feel interest in the Lipshitz-Vanderkooy AES publications and in the Eric Baekgaard AES publications about linear phase crossovers ? If your answer is yes, please go on the AES library as advised.
2. Do you feel it is mandatory to embed the driver correction into the crossover transfer functions ? If your answer if yes, why do you go ad-hominem against me instead of discussing the possibility of adding a mike input along with the corresponding software module ?
3. Do you apply a standard listening test for assessing the perceived quality of each crossover arrangement ? What are your conclusions ? You felt in love with Linkwitz-Riley (selective but reconstructed phase is highly non linear), you felt in love with the 1st order (reconstructed phase can be linear), or you own a special receipt you would like to share with us ?
4. You seem not to be aware. SynthMaker is a de-facto standard platform for testing audio filtering, audio crossovers, and many other things related to audio. If would be completely foolish to program a DSP with fresh ideas inside, without prototyping the whole stuff using SynthMaker and a quality multichannel audio USB subsystem. It only takes a few minutes.
5. The world is changing, Doug. Audio DSP is not anymore what it used to be. You can do it for free on any decent PC and you can do it for pennies using Freescale Kinetis ARM Cortex-M4 controllers clocked at 150 MHz.
6. A dedicated DSP hardware needs thus to deliver at least 3x the processing power of a Cortex-M4 at 150 Mhz, but unfortunately this is overkill for a 3-way or 4-way audio crossover implementing a few IIRs. If you want a 1024-FIR doing the room correction and if you want a whole new department dealing with dynamics (compression or expansion), that's another story. A dedicated DSP would be mandatory.
7. One reason why you would use a dedicated DSP hardware is because of the multiple physical I2S interfaces going to the 3 or 4 stereo DACs, for dealing with 6 or 8 channels. You may thus think you need at least 3 or 4 I2S ports. Unfortunately the most recent DACs, even high-end stereo, are now equipped with a multichannel TDM interface. TDM is a must nowadays. It could be TDM can be implemented using a Freescale Kinetis ARM Cortex-M4 controller clocked at 150 MHz, using the built-in SPI or I2S along with DMA.
8. The Freescale Kinetis ARM Cortex-M4 has built-in MAC, built-in floating point unit, adequate internal RAM and EEPROM. It may kill entry-level dedicated audio DSPs, especially when you realize that you also get, for free, a PC connectivity thanks to USB.
9. It is written into the sky : your next digital 6-way digital crossover will be a tiny box designed to operate with a PC, to be considered as USB stereo sound, 49 dollars, or the same DSP engine fitted with a different input arrangement (analog or SPDIF), say 89 dollars as stand-alone unit, also connectable using USB on a PC for the setup. It is going to be tiny, easy, inexpensive and versatile.
10. Now you must feel very confuse. Sorry again.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.