Next level Active DSP Crossover

Driverack PA2 ? Or the VENU360, not sure on the noise level for horns though..

1707158639578.png


1707158614328.png
 
gigantic, what is the phase response of your system ? Would you post a screen shot from REW of spl with the phase.

The reason I ask is that phase response is historicly ignored in horn loaded Klipsch speakers due (I believe) to the difficulty of driver time alignment with passive crossovers.

ChrisA (Cask05) brought phase response into focus here https://community.klipsch.com/index...y-effects-of-quasi-linear-phase-loudspeakers/
here's the Phase Response & SPL. I don't think I have it formatted correctly, it's a bit of a mess:
IMG_7968.jpeg


Here's the spectrogram:
Screen Shot 2024-02-05 at 3.24.58 PM.png

and finally, group delay:
Screen Shot 2024-02-05 at 3.14.14 PM.png


there's a little room for improvement, but it's lot better than passive. I'm still a bit of a novice with REW and don't find the interface to be particularly intuitive.
 
Last edited:
That phase looks like you need to zero the "IR delay" from the actions menu on the upper right, then select "shift IR". If you have prior versions of REW than the last couple of months, that menu item can be found on the "Controls" menu instead.

Additionally, I'd recommend putting a lot of absorption on the floor between the microphone and the loudspeaker (and I hope you're taking measurements from 1m in front of the loudspeaker--no farther away), with the width of the absorption about 6 feet wide. Point the microphone at the loudspeaker--not at the ceiling, and make sure you're using the correct microphone calibration file for pointing at the loudspeaker if you're using a UMIK-1.

All these things will clean up the phase response within REW considerably.

Additionally, I'll again ask for you to zoom in on your vertical scale to minor divisions of 1 dB and phase divisions more like 180 degrees (full scale). The group delay minor divisions should be no larger than 1 ms.

Chris
 
Last edited:
...An other conclusion that I have made is that there will be his from the horn whatever you do...
Just FYI: I just put my head into each of my front three loudspeakers (K-402 horns in each case). No hiss at all that I can hear. So if you use balanced connections and quiet preamp, DSP crossover and amplifiers, there should be no hiss at all. YMMV.

There is a very small 60/120 cycle hum from the one balanced/unbalanced connection from my preamp to the amplifier driving the TAD TD-4002s...front left/right HF compression driver channel) but that's because I chose to stay with the First Watt F3 driving those two channels--it doesn't have a balanced connection, so I have to use a Jensen PI-2XX transformer to reduce common mode noise on those two channels.

Chris
 
Read into the thread that I linked to, above (the K-forum thread on Subconscious Auditory Effects of Quasi-Linear Phase Loudspeakers) starting at page 3. The sequence of steps to do it can be found here: https://community.klipsch.com/index...ar-phase-loudspeakers/page/4/#comment-2388972
If I'm reading the posts correctly, you are simply using the PEQ and/or shelving filters to limit out of band energy much like a "crossover" would, but just by attenuating the frequencies of interest in the stop band? I understand the natural bandwidth and roll-off comments.

Also, with these settings in mind, listening at >110dB may not be a good idea (driver wise for FM/AM).
 
...Also, with these settings in mind, listening at >110dB may not be a good idea (driver wise for FM/AM)...
I don't know why so many audio guys are so focused on "way too loud" scenarios. I don't think that's an issue. (I also don't drive my autos with my foot on the floor everywhere I go, either.)

Remember that my loudspeakers are generally professional cinema models, so the drivers and the horns are pretty robust (K-402s, KPT-KHJ-LFs, etc.). And my ears are a lot more important to protect, and I don't believe that 110 dB/1m is something that I'd recommend if you're trying to retain your hearing. In any case, if you're worried about out-of-band power, make sure that your stop band PEQs and shelf filters give you the kind of attenuation you think is necessary for their survival. It's like any type of crossover filter scheme--no different.

If I were using ElectroVoice T-35 (i.e., Klipsch K-77 tweeters), without fast limiter correction, just note that they can't take more than 4 w, and I believe that there is a time limit for how long they can endure even that. In any case, setting the limiters in the DSP crossover would be a good idea if you're thinking of using your home loudspeakers for PA or nightclub/venue duty.

Chris
 
I'm currently multi-amping my LaScala clones, using a Dayton Audio DSP-408 active crossover. There is a lot to like about it in terms off functionality: 4 channel in, 8 out; 10 bands of parametric eq per out, up to 24db Linkwitz-Riley/Butterworth/Bessel filters and a 20hz-20kHz response. It had a decent UI, with bluetooth/iOS connectivity and a windoze interface, as well. However, I would like to improve the noise floor, is has a slightly perceptible hum that is present when music isn't playing. It's entirely single-ended, but I would prefer balanced, at least as an option. Finally and this is the biggest issue for me: the DAC is limited to 48Hz, 28-/56 Bit processing, which regardless of whether it's audible or not, effectively nullifies my Schiit Audio Bifrost 2's 192Hz, 24 bit capabilities for hi-res audio file playback.

What options are available to maintain the functionality aspects that I like about the Dayton Audio DSP 408 (ins, outs, 10 band PEQ, filters & frequency response, ui and relative affordability) with better processing, at least 96Hz/24 bit, but preferably 192Hz, 24 bit, balanced inputs and outputs? I know there are pro options from companies like Ashly that are more expensive than I'd prefer and more affordable pro options from dBX, Behringer and others skimp on PEQ, or ins/outs and/or have poor UI functionality... I'd prefer to keep in the $500 ballpark (I'm ok with used components) and this is a big ask, have either iOS or Mac OS UI.
There's the newish Minidsp Flex HTx with 8 balanced inputs, 8 balanced outputs and 32 bit 400 MHz DSP processing (plus all the usual minidsp features). Claimed 127 dB SNR. One downside is that its ~$1000.
https://www.minidsp.com/products/ht-series/flex-htx
 
Sorry, I was so slow writing that my post here abowe did time expired. So I'm not able to edit or delete it. Here under is my complete message.
-------------------------------------------------------------------------------------------------------------------------------------------------------


I know the pain of finding the right set-up for horn speakers. The hiss or noise does ruin the otherwise sparkling sound. I have not noticed so great difference between 48 and 96 decoding but FIR sounds much more open than IIR in my system.

Another conclusion that I have made is that there will be hiss from the horns whatever you do. 110db/1w level of efficiency is hard to manage. So I use L-pads until I find another better solution.

I have a Najda preamp/signal processor / DAC. This is a very nice processor with analogue, Toshlink, USB input 8 channel DAC output etc. Najda is also very quiet. It can do IIR or FIR processing. Not very powerful but see my solution to this in the attached picture.

The biggest advantage is that it is easy to use for anyone in the family. I think that this is the biggest challenge that is often forgotten in threads like the one we have here. I feel that this is the level of usability we need to demand and seek for. I know that it is a challenge with the offering that is available today.

Najda is sadly not any more available as Nick the developer has been declared lost somewhere in South America on a motorcycle holiday. Used or even never-used boards do show up now and then. They are not expensive typically under 500€.

My interest here is to find a solution to be able to use longer FIR filters. Mini DSP, Hypex FusionAmp and others like my Najda do seldom support more than 1000-2000 taps. This is only enough to EQ over 800z. I however am stretching it a little and also use FIR filters for my midrange drivers 170Hz to 1000Hz but this demands some compromises. The base or sub can not be touched with FIR and these processors.

So now we get to the question how about using a Raspberry with Motu Mk5. Wirrunna suggested this and it is tempting. CamillaDSP, jRiver comvolution and also Accurate Sound Hang Loose Convolver support this. But my problem is that I think that Raspberry is expensive. I paid 100€ + the SSD for my music PC a Lenovo M92p - Intel i5-3470T 2.90GHz - 8Gb RAM - Windows 10Pro - SSD 500Gb. This is in a nice box with several USB ports and can convolve any filter I have made. Raspberry seems to cost much more.

Here is a Hang Loose test that also tests how Raspberry manages FIR convolution. It seems to work but will it also do all other tasks like streaming etc?
So will everybody focus on Raspberry to bring us a family-friendly, powerful all need-supporting solution? Or is there maybe a more powerful board computer (non-Windows or MAC) or something else that would answer at least some of the demands that we would express?
It might not look so as Mitch writes at ACR "I am working with a company that develops custom digital audio I/O boards compatible with Raspberry Pi." But let's hope that one solution opens the door for competition so that the HiFi world can move into higher taps rates.

Here are some pictures of my speaker project. The goal with these was to build classic JBL speakers to be able to learn how to measure speakers and build FIR filter XO and EQ. I was working as a professional with PA systems in the 70s and 80s. Then we did only have golden ears. Now we have REW. It is almost like having heaven. I hope you all can understand this.
Speaker drivers are 15" JBL 2234 closed box, 10" JBL 2123 and horn JBL lense #5006815 with driver JBL 2408H-2 or SB Audience Rosso-65CND-T.
 

Attachments

  • Najda SP2 setup.jpg
    Najda SP2 setup.jpg
    62.7 KB · Views: 41
  • JBL SP2 .jpg
    JBL SP2 .jpg
    105.1 KB · Views: 40
  • Kotona SP2.jpg
    Kotona SP2.jpg
    147.1 KB · Views: 33
  • JBL SP2 etätyö.jpg
    JBL SP2 etätyö.jpg
    165.2 KB · Views: 24
  • 03-IMG_3617.jpg
    03-IMG_3617.jpg
    390.5 KB · Views: 30
Last edited:
  • Like
Reactions: 2 users
My interest here is to find a solution to be able to use longer FIR filters. Mini DSP, Hypex FusionAmp and others like my Najda do seldom support more than 1000-2000 taps. This is only enough to EQ over 800z. I however am stretching it a little and also use FIR filters for my midrange drivers 170Hz to 1000Hz but this demands some compromises. The base or sub can not be touched with FIR and these processors.

Yep, the quest for longer FIR filters, in a way we can use them, becomes a common goal.

How many taps per channel, at what sampling rate are you looking for?
 
Spalmgre has mentioned Acourate already and there is also Audiolense. These will give you more taps than you know what to do with ;)
@Cask05; Chris it was fun reading about your audio journey but I did think that these software based DSP solutions might have saved you a lot of grief. Have you thought about trying them even now?
 
@Cask05; Chris it was fun reading about your audio journey but I did think that these software based DSP solutions might have saved you a lot of grief. Have you thought about trying them even now?
Yes, but understand that my needs are pretty much outside the needs of the OP's in this thread. I'm looking at bypassing the need for a AVP, etc., such as a multichannel DAC and a pretty capable computing platform that can handle 6 input channels and ~12 outputs, with pretty significant FIR correction and 96 kHz sampling speed. Know any threads on that subject?

Chris
 
How do you do you EQ tuning work in Camilla? IIR, FIR, or combo?
Great results, however you're doing it
All three - https://www.diyaudio.com/community/...overs-room-correction-etc.349818/post-7213813
Ultralite Mk5 has higher residual noise than the Okto, about 7 uV vs 3.5 uV which may be more noticeable with higher sensitivity drivers and/or higher gain amplifiers.
My modified K-Horns have an Eliptrac mid/hi horn with B&C DCX464 111db@1W sensitivity and I can't hear any hiss with my head in the horn mouth. The Motu UL5 feeds a pair of SMSL SH-9 headphone amps for mid and hi.

Having dropped ChrisA in it with a link to his post on the effect of Phase, I think it appropriate to include two of his earlier posts:
https://community.klipsch.com/index...ametric-equalizer-peq-settings/&ct=1599687416
https://community.klipsch.com/index...ne-time-delays-between-drivers/&ct=1596844825

spalmgre asks about taps in the fIR filters. I tried 8192 taps per channel and eventually settled on 4096 as one of my inputs is analog from the TV and 4096 taps has almost no delay so you barely notice lip sync delays between listening on the tinny rear facing TV speakers and through the tri-amped K-Horns. While experimenting, I tried 64000 taps on the bass channel and 4096 on the others and had to set the delay to something like 29 seconds. Obviously unacceptable for TV watching.

Back to the question "Next Level DSP Crossover", CamillaDSP and Motu UltraLite Mk5 as ticked all the boxes for me.
https://www.audiosciencereview.com/forum/index.php?threads/rpi4-camilladsp-tutorial.29656/
 
  • Like
Reactions: 1 user
How many taps must the professor be able to do. This will not be a problem once the processors used are capable enough. FIR needs more power than IIR but it seems that a relatively low-power CPU can do the job for even longer taps. So this question will go away once the industry steps up to next-level processors. That would be using processor boards like Raspberry or similar solutions. Maybe we also can have some kind of software package that can run on several platforms. CamillaDSP is a good startingpoint her. It just seems like too much effort to get running.

So how many taps do we need? Mitch at Accurate Sound only semas to speak for 65 536 tapps. I have got along with 1000 taps in the hi-end. So the sweet point is somewhere in between. But if you want to do bas region XO and EQ then you will need many taps. I am using a hybrid solution by running the base speakers through miniDSP 2/4 (the original)and doing IIR this way. It has its advantages but then there is one more box to manage.

Previous posts in this thread have been given some advice about how to measure to be able to configure filters. I have done at least 1000 measurements with REW during about seven years. I have all the time had difficulty to be able to measure the same change in the reproduction curve as the simulation in REW or rPhase implies.

I did the standard one-meter measurement to linearise the drivers. Then moved to the listening area and measured to do some room correction.
Only two months ago did I finally understand how to do it right. At least so that I get a consistent result that can be repeated.

1) I did move loose chairs tables and pillows away from the listening area. I even moved the sofa when it was possible. This is to minimise reflections. These will of course return to the sound once in normal listening mode. But it helps to get better understandable measurement results.

2) All measurements were made at the listening area witch in my case is a 3-person sofa.

3) Hi and mid drivers are measured separately without XO and EQ.

4) Base drivers were measured together as mono. This is as I assume that the base in music is mixed as mono. This approach takes out a lot of wariables getting the base sound right.

5) All drivers were measured with the microphone moved to 7 different spots. This adds up to 35 measurements ((Hi L/R+ Mid L/R + Base mono) *7) . If you want to make it easy for you then take a look at Impala free measurement software. It measured as REW but it is very nice to work with when measuring in several spots. The Impala measurements can then be imported to REW.
I have not found much gain from doing more measurements. All measurements are "time aligned" in REW and then "vector average" to produce e good understanding of how the driver behaves in the room. Several measurements will minimise reflections so doing this further away from the speaker does not seem to harm the result.

6) Then I EQ in REW and export the filter settings and the measurement to rePhase. I have also tried the "Convolution with Inversion" method.

7) Then I listen for several days to find the right volume level between the drivers. I also produced several Hi driver FIR filters with different room curves. Switching between these not only determines how the hi-end will sound. The base level is much a product of how hot the Hi end is.

8) I have iterated the 1 to 7 steps at least 3 times during the last year after moving into our recent flat. And now I am pretty sure that I can't do better. But I will do one more try once I get hungry again. Each iteration is quite a lot of work and takes several hours.

9) Yesterday I started to think could I do something more without starting all over again? I got inspired after reading this thread.
So I tied some MMM (moving microphone measurement) with noise and RTA measurement. This is fast and easy to do. But doesn't produce the phase curve.
So this gave me a good understanding of where I'm at now. And you guess it right if you have been there. It is never perfect so now I am planning for my next EQ session.
 
Last edited:
This is something I did not know to do. The Umic-1 microphone correction file does correct the frequency curve but not the phase curve.

Here is how to add phase correction to your correction file. Look at the video from 00:05:04.
This channel's videos are very interesting and introduce new ways to use REW. You can even do FIR directly from REW.

Phase Match Your Speakers with AllPass Filters

 
  • Like
Reactions: 1 user
I really feel its important for others that post here to know that I do apologize in case someone feels picked on or otherwise personally challenged by my comments. That's not at all what I was intending. Just the opposite--my comments are directed towards the equipment, not the people.

But I do to strongly request a bit of heightened sensitivity to the differences between direct radiator and horn loaded loudspeakers. These differences will not simply be "wished away" by those having no direct experience with fully horn-loaded loudspeakers (fully horn-loaded generally means their bass bins, too). The driving electronics that have to operate in signal amplitude regions well below that of direct radiators--even very large "higher efficiency" direct radiators--are operating in ways that the bulk of the direct radiator crowd simply hasn't experienced (but they offer their opinions as if they have, nevertheless).

The absence of audible modulation distortion is a defining feature of horns (something that many/most direct radiating enthusiasts are completely ignorant of), and one that regularly exposes defects in electronics performance at very low signal levels (i.e., clarity) in a way that truthfully no other type of loudspeaker technology can.

I'm not trying to over-emphasize the subject, but when it occurs, my personal intent is to clearly communicate my experiences to those who think those differences are simply eye wash. They're not, I can assure you. I don't think that the physics is going to change anytime soon (as PWK sometimes mentioned), but those that think only in terms of small, low efficiency direct radiating loudspeakers are going to continue to commit those same errors of inductive logic when it comes to horn loading and their experiences with direct radiating loudspeakers.

Since Klipsch Corp. has relatively recently pushed out its DIY crowd from its forum website (which occurred definitively in the fall of 2022...a marketing management decision I firmly believe--not just one notable design engineer we all know by name), I think it's natural to find many of those DIY refugees here, along with the existing refugees of Altec and JBL 1950s-1980s horn-loaded loudspeakers also inhabiting this forum, as well as those enthusiasts of other smaller company's products that produce high quality/hi-fi horn-loaded loudspeakers.

Chris
 
Last edited:
  • Like
  • Thank You
Reactions: 4 users