Hi, I’m using 2 way MEH, with a externally mounted 1” CD tweeter, controlled by a Xilica XP. I’ve used the Xilica for over a year, countless hours adjusting FR, driver time alignment, and implementing digital crossovers. The Xilica is IIR filers only, no FIR.
I cannot keep the time alignment consistent, and my woofers come in 2ms late and lag to 6ms by 100hz. I’m seeing this in REW spectrogram . When I try to delay mid driver more I get a “bulge” near the crossover point.
Wondering if a more flexible DSP would give me the tools to get better time alignment and Phase consistency. I have no experience software or electrical engineering, so I’m a little leary of some of the supposedly most flexible options, Audiolense, Acurrate. I’m researching things like Lake and Linea processors or even Trinnov or Deqx.
I’m just looking for platform experience and advice, for 2 channel only, no home theater.
I know this is a very vague question but hopefully some of you can give experience of the DSP route you took.
Thank you for any perspective!
I cannot keep the time alignment consistent, and my woofers come in 2ms late and lag to 6ms by 100hz. I’m seeing this in REW spectrogram . When I try to delay mid driver more I get a “bulge” near the crossover point.
Wondering if a more flexible DSP would give me the tools to get better time alignment and Phase consistency. I have no experience software or electrical engineering, so I’m a little leary of some of the supposedly most flexible options, Audiolense, Acurrate. I’m researching things like Lake and Linea processors or even Trinnov or Deqx.
I’m just looking for platform experience and advice, for 2 channel only, no home theater.
I know this is a very vague question but hopefully some of you can give experience of the DSP route you took.
Thank you for any perspective!
Hi,
Dolby Lake Dlp4d12 owner. I won't recommend it because it show his age on some area. That said i'm not ready to change it for close future...
If i had i would go pc/ soundcard, software like accurate and convolver plug in like Mitchba's Hangloose for FIR processing...
From your description to me it's not an issue with your processor but rather what you do with it imo.:
Bulging? This is where eq are your friend: tweak them until it get back to flat.
Delay, ms range is more than ok to deal with your issue imho.
Meh, get in touch with Mark100, he might be of great help to tune your processing imho.
Dolby Lake Dlp4d12 owner. I won't recommend it because it show his age on some area. That said i'm not ready to change it for close future...
If i had i would go pc/ soundcard, software like accurate and convolver plug in like Mitchba's Hangloose for FIR processing...
From your description to me it's not an issue with your processor but rather what you do with it imo.:
Bulging? This is where eq are your friend: tweak them until it get back to flat.
Delay, ms range is more than ok to deal with your issue imho.
Meh, get in touch with Mark100, he might be of great help to tune your processing imho.
What crossover transfer function are you using? Not sure I understand the fixation on IIR filters, people far smarter than me seem to think FIR filters used correctly are just fine. Electrical and acoustical crossover slopes will likely be quite different and some experimentation with slopes, delay and phase shift may be required to get a smooth transition.
I have a 4 way system that is run entirely by DSP distributed between a very heavily modified Behringer DCX2496 for the mains which are 3-ways, and a MiniDSP SHD Studio for XO between subs and mains and for Dirac. The DCX provides some rudimentary EQ, and manages delays between the drivers in main speaker system, and the Studio handles the delays and other sub related issues as well as providing a digital pre-amp, and a roon ready streamer.
Short of a DEQX I don't have many options.
Things to consider would be a DBX Drive Rack or a MiniDSP 2 x 4HD?
I have a 4 way system that is run entirely by DSP distributed between a very heavily modified Behringer DCX2496 for the mains which are 3-ways, and a MiniDSP SHD Studio for XO between subs and mains and for Dirac. The DCX provides some rudimentary EQ, and manages delays between the drivers in main speaker system, and the Studio handles the delays and other sub related issues as well as providing a digital pre-amp, and a roon ready streamer.
Short of a DEQX I don't have many options.
Things to consider would be a DBX Drive Rack or a MiniDSP 2 x 4HD?
I really like QSC's Q-SYS.
Now that used prices on ebay are falling back to earth, a Core110f is a killer platform for DIY, ime.
The flexibility and feature set that the q-sys designer software has, is mind boggling.
Whole 'nuther step up from fixed architecture, conventional type processors.
Now that used prices on ebay are falling back to earth, a Core110f is a killer platform for DIY, ime.
The flexibility and feature set that the q-sys designer software has, is mind boggling.
Whole 'nuther step up from fixed architecture, conventional type processors.
I used Behringer deq2496 and MiniDsp with IIR EQ for fullrange drivers. Assisted with arta and atb pc Pro measurement System.
Works perfect - as IIR and fullrange drivers linearize amplitude and phase at the same time.
Works perfect - as IIR and fullrange drivers linearize amplitude and phase at the same time.
Thanks for the replies so far. I am interested in FIR filters, I definitely get phase “non linearity” at the crossover point from woofer to mid, (600hz), and like I said a 6-8 ms growing delay from 500- 50 hz.
Interested in Q-sys, like you say it seems extremely flexible.
Currently I have my Xilica between my preamp and amps, which leads me to wonder about the AD/DA conversation. Wondering if people have noticed quality differences in this area?
I know I’m asking pretty basic questions, but there’s a lot of options, and most it seems (?) not primarily designed for home hifi.
Interested in Q-sys, like you say it seems extremely flexible.
Currently I have my Xilica between my preamp and amps, which leads me to wonder about the AD/DA conversation. Wondering if people have noticed quality differences in this area?
I know I’m asking pretty basic questions, but there’s a lot of options, and most it seems (?) not primarily designed for home hifi.
Yes there is difference with converters in hardware units, and why i won't change Lake soon, the converters are still very good, there is AES availlable if i want to upgrade to something different too. There is Dante too but it's older version and kinda limited.
FIR are locked to what is availlable in the software too ( xover, no way you could achieve global phase correction in the unit*).
That said with Qsys you have Dante so it is open system to a lot of options for converters.
No it's not designed for 'home hifi' whatever it mean ( i suppose you talk about pro line level and balanced impedance).
It depend what kind of amplifier you'll use but recent ClassD with no 'preamp' gain stage ( input buffer and maybe a small pad in front ) works ok using the full dynamic scale ( in the range 21/24dbu is often encountred - having +1/+4 dbu with 20 db headroom).
*: not really an issue as you can use the Lake as if it is 12 independant DAC( with powerfull eq as bonus for each) and 4 input ADC. With a pc as source you can then implement FIR through a convolver driving the unit thanks to digital soundcard ( i use RME AES within the pc). Or use a global 2 track global FIR correction in 'regular' mode ( contour/mesa).
The ADC inputs can be taylored to analog source too, from +26dbu to +10dbu full scale. Cover almost any needs smoothly.
FIR are locked to what is availlable in the software too ( xover, no way you could achieve global phase correction in the unit*).
That said with Qsys you have Dante so it is open system to a lot of options for converters.
No it's not designed for 'home hifi' whatever it mean ( i suppose you talk about pro line level and balanced impedance).
It depend what kind of amplifier you'll use but recent ClassD with no 'preamp' gain stage ( input buffer and maybe a small pad in front ) works ok using the full dynamic scale ( in the range 21/24dbu is often encountred - having +1/+4 dbu with 20 db headroom).
*: not really an issue as you can use the Lake as if it is 12 independant DAC( with powerfull eq as bonus for each) and 4 input ADC. With a pc as source you can then implement FIR through a convolver driving the unit thanks to digital soundcard ( i use RME AES within the pc). Or use a global 2 track global FIR correction in 'regular' mode ( contour/mesa).
The ADC inputs can be taylored to analog source too, from +26dbu to +10dbu full scale. Cover almost any needs smoothly.
Last edited:
With MEH using IIR filters you usualy need asymmetric crossover slopes (IE 4th order and 6th order) in order to get broad enough time alignment. If you upload measurments of the tweeter and woofer sections on axis with absolute time reference I can have a look at what would work. FIR will allow you to just hammer it flat in the time and frequency domain with great results though!
Nice options: Symetrix (Radius AEC available used and has DANTE), QSC
Cheap option: Biamp Tesira (also has USB interface built in which avoids need for DAC)
Realy realy bougie: Outline Newton
Less complicated/more live sound: https://linea-research.co.uk/asc48/
Nice options: Symetrix (Radius AEC available used and has DANTE), QSC
Cheap option: Biamp Tesira (also has USB interface built in which avoids need for DAC)
Realy realy bougie: Outline Newton
Less complicated/more live sound: https://linea-research.co.uk/asc48/
With MEH using IIR filters you usualy need asymmetric crossover slopes (IE 4th order and 6th order) in order to get broad enough time alignment.
That’s interesting, I usually check the slopes with REW and adjust to create symmetrical slopes. Generally using 4 th order. I will try some asymmetry.
I’ve looked at the Linea, it kind is seems a lateral move from Xilica, but with their proprietary “FIR” filters. I will investigate Symetrix, very interested in QSC Qsys.
That’s interesting, I usually check the slopes with REW and adjust to create symmetrical slopes. Generally using 4 th order. I will try some asymmetry.
I’ve looked at the Linea, it kind is seems a lateral move from Xilica, but with their proprietary “FIR” filters. I will investigate Symetrix, very interested in QSC Qsys.
If a bulge i rising as XO you delay one of the drivers, it indicates that the two drivers was not in phase at XO.
I'm not so sure that broad (as in several octaves) is needed - if you use a dB calculator and add say 90 dB and 78 dB you will se that its a very small addition...
90dB + 78db = 90,3dB (i.e. one octave from XO in a 2nd order filter)
https://www.noisemeters.com/apps/db-calculator/
//
I'm not so sure that broad (as in several octaves) is needed - if you use a dB calculator and add say 90 dB and 78 dB you will se that its a very small addition...
90dB + 78db = 90,3dB (i.e. one octave from XO in a 2nd order filter)
https://www.noisemeters.com/apps/db-calculator/
//
A qsys Core110f will run 8 channels of 6144 taps of custom FIR for multiway. I don't think it can run more channels at that tap count, or more taps per channel.
But if you drop down to 4k taps per channel, a lot more channels are available.
If the goal is global FIR (something i'll advised ime), it can do two channels of 9216 taps...
It also has it's own FIR high-pass and low-pass with adjustable orders, that use considerably less processing resources that custom FIR.
Qsys tell you real quick when processing capability has been exceeded.
Discontinued Cores that use I/O cards, like the 250i, 500i, and 510i in increasing order of power, can run greater taps counts and channels.
I have a Core510i that has no problem with 15 channels of custom FIR @ 16K taps. 16k taps per channel is the current qsys max allowed.
Qsys is 48kHz. I've found i no longer care about higher sample rates, or AD/DA conversion quality, after having compared the qsys processors/cards/network amps to a RME baby pro, Teac UD-501, and the Linea Research ASC-48. It's all too low in significance to matter imho. Although that said, I wouldn't tun down 96kHz. (i would turn down higher sample rates though, I think they are 95% silly)
The Core110f's USB is 16 bit however, I do wish it were 24 or higher...that's about my biggest gripe with it.
I'm using qsys Dante nowadays with Audinate's Virtual Sound Card on my PCs, so don't care much about USB anymore.
If you're interested in Qsys, I'd suggest downloading Q-sys Designer, and build some designs. They will load into a PC , just wont pass audio etc....takes a Core to compile to...
The online training series is one of q-sys biggest advantages. That and mind boggling flexibility / capability. Along with being able to build any kind of remote you want, for any kind of either general controls, or elaborate comparison testing.
A day past edit time...sorry didn't have time to proof
Because that line was about as clear as mud....
Meant without typo: "If the goal is global FIR (something ILL advised ime), it can do two channels of 9216 taps..."
Consider what Danley himself recommended and the work of Chris A - no beyond first order fixed crossover slopes at all. Or as Danley put it - no filter with a man's name attached to it. So nothing more than a first order. I have found those are not needed either.
I am using the Klipsch K402 as per Chris A's design.
When you look at the drivers by themselves you will see there is lots of crossover slope already there - you will mainly need notch filters for the high pass of the woofers - they most likely are dropping like a rock already. A fixed slope there is only going to make things worse. Looking at the natural response of the drivers is the first step. That will tell you everything you need to know.
Fix the out of band response with PEQs. When done my individual responses look as much like a textbook response as can be reasonably implemented. Your timing issues will take care of themselves.
I use a fixed slope on the subwoofers where the delay comes in handy.
I am using the Klipsch K402 as per Chris A's design.
When you look at the drivers by themselves you will see there is lots of crossover slope already there - you will mainly need notch filters for the high pass of the woofers - they most likely are dropping like a rock already. A fixed slope there is only going to make things worse. Looking at the natural response of the drivers is the first step. That will tell you everything you need to know.
Fix the out of band response with PEQs. When done my individual responses look as much like a textbook response as can be reasonably implemented. Your timing issues will take care of themselves.
I use a fixed slope on the subwoofers where the delay comes in handy.
@rickmcinnis , I have same speaker. You’re just using peq to adjust the drivers natural roll off? I’ve tried this before and then moved on, maybe I should try again. How consistent is your phase through the crossover? How’s your spectrogram look with K402? Mine always shows delay near the crossover and lots of very early reflection, 2-10 ms, at same point. Are you using Xilica? Tried any other DSP?
Crosses at ~525 Hz to a Celestion Axi2050 using first order IIR crossover filters and no channel delays. As always, measurements are taken at 1 m. All measurements here use Xilica XP series processors.
Here is one that uses only a low pass on the woofer channel at 550 Hz (the prototype in my listening room) to a BMS 4592 ND (dual diaphragm 2" compression driver) :
That little drop out at 6.2 kHz was the result of an experiment that I was running at the time, and is no longer there in the current DSP settings.
Anything beyond about 1.1 ms is physically due to reflections outside of the horn (17 inch depth)--and are assignable to early reflections outside the loudspeaker's mouth
You may have to change your preamp/AVP and/or your DAC if you're having trouble. Make sure you measure at 1 m with plenty of absorption on the floor between the microphone and the loudspeaker, about 2 m wide.
Chris
I am using xilica Solara.
I do not see how the use of PEQs would cause early reflections.
I do not see how a fixed rate slope could ever decrease delay at the crossover but I readily admit to being a dilettante when it comes to this stuff.
I tend to go with the Earl Geddes idea that frequency response is all important and there is a mild 1dB dip at the nominal crossover point. Of course, one can fill it if they want to - I have not.
There is a bit less than 120 degrees of phase change from 100 Hz t0 10KHz.
There does not seem to be much choice of boxes - I have considered trying a computer based DSP but I do not want to use the same computer for that as I use for playing the files. There is not much available for AES EBU digital in and out sound cards that work with LINUX (hate the thought of a gigantic WINDOWS OS) so I think I am stuck. Not that I am not happy with xilica but one always has nagging doubts about gear such as this.
I do not see how the use of PEQs would cause early reflections.
I do not see how a fixed rate slope could ever decrease delay at the crossover but I readily admit to being a dilettante when it comes to this stuff.
I tend to go with the Earl Geddes idea that frequency response is all important and there is a mild 1dB dip at the nominal crossover point. Of course, one can fill it if they want to - I have not.
There is a bit less than 120 degrees of phase change from 100 Hz t0 10KHz.
There does not seem to be much choice of boxes - I have considered trying a computer based DSP but I do not want to use the same computer for that as I use for playing the files. There is not much available for AES EBU digital in and out sound cards that work with LINUX (hate the thought of a gigantic WINDOWS OS) so I think I am stuck. Not that I am not happy with xilica but one always has nagging doubts about gear such as this.
“Crosses at ~525 Hz to a Celestion Axi2050 using first order IIR crossover filters and no channel delays. As always, measurements are taken at 1 m. All measurements here use Xilica XP series processors.”
Chris and Rick, regarding crossovers, I’ve wondered, if you construct a slope with PEQ, would there be any phase difference between a “named” crossover of the same slope?
Another, possibly basic question. Is phase influenced by early, strong reflections?
Also, with 1st order XO at 525hz, you’re only going to be down 12 db at 130hz, but the 2” driver doesn’t even go near that low, I’m confused.
Thanks for any guidance. Ted
Chris and Rick, regarding crossovers, I’ve wondered, if you construct a slope with PEQ, would there be any phase difference between a “named” crossover of the same slope?
Another, possibly basic question. Is phase influenced by early, strong reflections?
Also, with 1st order XO at 525hz, you’re only going to be down 12 db at 130hz, but the 2” driver doesn’t even go near that low, I’m confused.
Thanks for any guidance. Ted
Yes.. PEQs create much smaller induced phase shifts per their "minimum phase system" property (which is essentially zero over the limits of their effects), while named crossover filters always have an all-pass component of ~90 degrees of overall phase lag of the lower frequency drivers vs. higher frequency drivers per order of the filters used.
Only the phase measurements are affected (i.e., noisy measurements)--not the underlying minimum phase from the MEH.
However, note that in your case, I never saw a measurement unaffected by digital issues which were obvious in the phase shifts and extreme rolloff of high frequencies above 10 kHz. Did you ever solve those measurement or electrical signal chain issues of the measurement gear?
You seem to be focused on the electrical performance of the crossover filter vs. the total electro-acoustic response of the driver/horn plus the crossover filter. Note that there are many multiway loudspeakers that exist but that do not use any kind of electrical crossover filters, relying instead on the natural response of the mounted drivers only to cross to essentially flat SPL response (especially in the PA market).
The focus here is on total response, not just the electrical portion of the signal chain. You seem to have forgotten the portion with the largest effect on total output: the driver+horn itself.
Chris
Last edited: