Combining a digital dsp with analog crossovers are not always a good idea

[BDWoody]

it wasnt a blind test, but very experienced listener ( 3 ) did the listening.

Then why are you referring to it as if it meant anything at all?

 

[FrantzM]

All this has the flavor of what one would expect from those subjectivist magazines or fora...

Some would qualify this as ...
and they would be right

 

[kongwee]

That’s because “propagation bandwidth” is not a thing. It doesn’t exist

Yes, scientifically it is. Higher bandwith, faster rise time and slew time will be shorter. Just inverse mhz into milli or even nano second that you frequently see. It is not that difficult to understanding in engineering sense. I mean your fave class D amplifier are using mosfet that run up to 1.5mhz.

I made a mistake writing 600mhz than 600khz in previous post.

 

[voodooless]

Yes, scientifically it is.

No, it’s not. Just google the term and tell me if you can find the definition, it’s not there.

 

[kongwee]

No, it’s not. Just google the term and tell me if you can find the definition, it’s not there.

Yes, google and found them.

https://octopart.com/blog/archives/2017/10/op-amp-speed-bandwidth-slew-rate

Plus lots of them if you can understand those terms.

 

[RayDunzl]

Everything you put in between a source and a speaker makes the sound slightly worse .

Well, you gotta put something there.


I've gone to great lengths to make my sound worse.

I think I have succeeded, beyond my any reasonable initial expectations.

Strike 1: Digital sources, TOSLink and Coax, go to an Audio Authority 177 switch and can be output as TOSLink or coax as desired.

Strike 2: That signal goes to a Focusrite Clarett Pre USB where the USB from PC or analog sources can be digitized and mixed in, if desired. Sometimes it changes the sample rate, I suppose.

Strike 3: Whatever is left of the signal goes to a MiniDSP OpenDRC-DI, where either my own settings for EQ/Compression/Volume/Polarity/Delay or AcourateDRC FIR and IIR filters are applied to further mess with the signal. I definitely get the sample rates adjusted there if they aren't bad enough yet.

Strike 4: That goes to an old Behringer DEC2496 which gives a nice RTA of whatever is left, and can apply even more destruction if the signal hasn't been worsened sufficiently already.

Strike 5: Whatever is left of the digits goes to a Benchmark DAC 2 to turn it into a lossy analog electrical voltage wave, resampling internally to 211kHz, which must have been the worst frequency they could find,

Strike 6: That feeds a Krell KCT preamp which turns the voltage signal into a current signal, just to be sure every last bit of worsening can be attained. The volume gets set here, and just for laughs, I can reverse the polarity if it still sounds good.

Strike 7: The Zone 1 output goes as a current signal to a couple of dusty (well, everything here is dusty) old Krell monoblocks,, converted back to something resembling a voltage source, and finally fed to the passive 24 year old Martin Logan reQuests, which, just for fun, i suppose, stick a transformer into the signal path in front of the panels, just in case there is something left to make worse. Or the Zone 2 output can go to the JBL LSR 308, which don't understand the current signal, so a little more worsening can be applied turning it back into a voltage signal.

Strike 8: Oh, the cables are all cheap junk. I figure if the signal gets through at all, I'm good.

It really does sound bad, especially if I play some old music recorded live with a boombox or other primitive device sometime in the last century.

I even play Bad once in a while, to see if it sounds bad, just to make sure it is all working as badly as possible.

Strangest thing though, it still seems to measure pretty well in-room. Well, some parts worse than others, which is to be expected, I guess.

Even Audio Buddy probably said it sounded bad at some time when he came by.

 

[voodooless]

Yes, google and found them.

https://octopart.com/blog/archives/2017/10/op-amp-speed-bandwidth-slew-rate

Plus lots of them if you can understand those terms.

The words “propagation bandwidth” are nowhere to be found in that article. This is an exact science. You can’t just make up words and then pretend they mean something else..

 

[Chrise36]

At faktiskt.io, a swedish hifi forum they have tested the new minidsp in a very expensive system . The result didnt come out well soundwise.

faktiskt.io • Logga in

In my opinion , trying to combine a dsp with analog crossovers are not always a good idea . Everything you put in between a source and a speaker makes the sound slightly worse . It may be passive crossovers, cables , AD converters , digital gain structure and more .

Im surprised that the reviewers at faktiskt.io dont draw the conclusion or think about the reason why it might sound worse putting a dsp in a signal chain with passive loudspeakers .

My experience - Im not suprised at all. No gear , analog or digital are fully transparent to the ear . One of the worst offender is passive crossovers in loudspeakers .

To gain real benefits with a dsp, one should instead go fully active with the minidsp - with all the crossovers in the digital domain and no passive coils, resistors or capacitors in the signal path. The source ( for example a computer with a DDC) would then go directly to the digital input of the Minidsp SHD. No A/D:s involved. No external DAC outside the minidsp is needed.

This way one can gain real sound improvements .

As I said in the beginning - buying a dsp to correct an analog loudspeaker is often a waste of money or time . The first step for real improvements would instead be converting the passive crossover in the passive loudspeaker to be fully active in the digital domain , keeping the signal digital as long as one possible can, and doing the crossover in the digital domain without converting it first to analog.

Whats your experience ?

The crossover has nothing to do with dsp artifacts a driver alone acts as an analogue filter what would change with a crossover network?

 

[jhaider]

At faktiskt.io, a swedish hifi forum they have tested the new minidsp in a very expensive system . The result didnt come out well soundwise.

Were there any specific, supported, complaints? Or just audiophool circle-jerkery?

Everything you put in between a source and a speaker makes the sound slightly worse . It may be passive crossovers, cables , AD converters , digital gain structure and more .

That is obviously wrong. Anything added is just a degree of freedom. Used well those degrees of freedom can be used to improve fidelity. Used poorly they can make things worse. But the implied holy grail of straight line with gain from source to transducer will sound bad every single time.

Whats your experience ?

It takes time, and thought, but thoughtfully applying EQ to passive loudspeakers based on the right criteria (room sound power down low, anechoic or quasi-anechoic listening window higher up) has never made things worse for me.

 

[thewas]

Besides all the already mentioned problems of such sighted tests, a very poor implemented and measuring analogue or digital circuit or filter choice can deteriorate the sound audibly, that has nothing to do though with what the title of the tread implies, namely that the combination of a DSP and analogue crossover is more problematic than the sum of their problems. For example you can have a DSP crossover loudspeaker where the electronics are even worse than of an external DSP, only it is not so easy to find out as you usually have no base to compare to.

 

[Tangband]

Were there any specific, supported, complaints? Or just audiophool circle-jerkery?



That is obviously wrong. Anything added is just a degree of freedom. Used well those degrees of freedom can be used to improve fidelity. Used poorly they can make things worse. But the implied holy grail of straight line with gain from source to transducer will sound bad every single time.



It takes time, and thought, but thoughtfully applying EQ to passive loudspeakers based on the right criteria (room sound power down low, anechoic or quasi-anechoic listening window higher up) has never made things worse for me.

Well, I dont entirely agree - everything have drawbacks - pick your poison .

 

[HarmonicTHD]

Well, you gotta put something there.


I've gone to great lengths to make my sound worse.

I think I have succeeded, beyond my any reasonable initial expectations.

Strike 1: Digital sources, TOSLink and Coax, go to an Audio Authority 177 switch and can be output as TOSLink or coax as desired.

Strike 2: That signal goes to a Focusrite Clarett Pre USB where the USB from PC or analog sources can be digitized and mixed in, if desired. Sometimes it changes the sample rate, I suppose.

Strike 3: Whatever is left of the signal goes to a MiniDSP OpenDRC-DI, where either my own settings for EQ/Compression/Volume/Polarity/Delay or AcourateDRC FIR and IIR filters are applied to further mess with the signal. I definitely get the sample rates adjusted there if they aren't bad enough yet.

Strike 4: That goes to an old Behringer DEC2496 which gives a nice RTA of whatever is left, and can apply even more destruction if the signal hasn't been worsened sufficiently already.

Strike 5: Whatever is left of the digits goes to a Benchmark DAC 2 to turn it into a lossy analog electrical voltage wave, resampling internally to 211kHz, which must have been the worst frequency they could find,

Strike 6: That feeds a Krell KCT preamp which turns the voltage signal into a current signal, just to be sure every last bit of worsening can be attained. The volume gets set here, and just for laughs, I can reverse the polarity if it still sounds good.

Strike 7: The Zone 1 output goes as a current signal to a couple of dusty (well, everything here is dusty) old Krell monoblocks,, converted back to something resembling a voltage source, and finally fed to the passive 24 year old Martin Logan reQuests, which, just for fun, i suppose, stick a transformer into the signal path in front of the panels, just in case there is something left to make worse. Or the Zone 2 output can go to the JBL LSR 308, which don't understand the current signal, so a little more worsening can be applied turning it back into a voltage signal.

Strike 8: Oh, the cables are all cheap junk. I figure if the signal gets through at all, I'm good.

It really does sound bad, especially if I play some old music recorded live with a boombox or other primitive device sometime in the last century.

I even play Bad once in a while, to see if it sounds bad, just to make sure it is all working as badly as possible.

Strangest thing though, it still seems to measure pretty well in-room. Well, some parts worse than others, which is to be expected, I guess.

Even Audio Buddy probably said it sounded bad at some time when he came by.

OMG you poor thing. I feel sorry for you. Your ears must be bleeding. Be careful.

 

[fpitas]

My experience is that you have to be careful with EQ no matter whether the crossovers are passive or digital. Sometimes you flatten a peak, and it helps, and sometimes it's a lot more nuanced.

 

[jhaider]

Well, I dont entirely agree - everything have drawbacks - pick your poison .

You didn’t answer my question: Were there any specific, supported, complaints?

Also, what was methodology of applying correction?

As for everything having drawbacks, sure. Any manual process requires expertise and time to get right - that’s the drawback. Lack either, and results are likely to be suboptimal. Most of the valid complaints about processing come down to lack of user expertise - bad starting measurements, poor gain staging, asking tools to do what they cannot, etc. The rest of the valid complaints tend to be from shortcuts in the measuring or optimization process, or defective/incompatible hardware. Others are audiophool idiocy for the most part: bitching and moaning about non-issues such as sample rates or an extra A/D conversion for example.

But the bottom line is, without modern EQ tools it’s near impossible to get the bass right in a small room. Given that bass is estimated at 30% of preference, the logical inference is that anyone interested in fidelity should learn these tools and use them.

 

[Tangband]

You didn’t answer my question: Were there any specific, supported, complaints?

Also, what was methodology of applying correction?

As for everything having drawbacks, sure. Any manual process requires expertise and time to get right - that’s the drawback. Lack either, and results are likely to be suboptimal. Most of the valid complaints about processing come down to lack of user expertise - bad starting measurements, poor gain staging, asking tools to do what they cannot, etc. The rest of the valid complaints tend to be from shortcuts in the measuring or optimization process, or defective/incompatible hardware. Others are audiophool idiocy for the most part: bitching and moaning about non-issues such as sample rates or an extra A/D conversion for example.

But the bottom line is, without modern EQ tools it’s near impossible to get the bass right in a small room. Given that bass is estimated at 30% of preference, the logical inference is that anyone interested in fidelity should learn these tools and use them.

The sound quality with the minidsp in the signalpath were considered to sound slightly unnatural in the treble register.

 

[mdsimon2]

The sound quality with the minidsp in the signalpath were considered to sound slightly unnatural in the treble register.

Sure, based on sighted listening tests that were not level matched.

The only possible explanation I see is clipping due to the ASRC and 0 dB volume level on the miniDSP although I think that is unlikely.

Michael

 

[jhaider]

The sound quality with the minidsp in the signalpath were considered to sound slightly unnatural in the treble register.

Oh, so just stupidity without evidence. Got it.

 

[Weeb Labs]

The term "sound quality" and phrases such as "made the sound worse" always stand out to me as incredibly vague; especially when the objective is to assess the impact of an addition to the signal chain. "Sound quality" is not itself a characteristic and doesn't convey any useful information without a specific description of the perceived change.

Has the noise floor become audible? Is the output clipping? One channel out of phase or significantly delayed? Changes to the frequency response? Audible jitter? Sprays of harmonic shmoo?

 

[MAB]

trying to combine a dsp with analog crossovers are not always a good idea

Uh oh, here we go again with your favorite hypothesis. In the past, you have presented similar with no supporting data. Let's see if this time we get some evidence:

In my opinion

No.

In my opinion - The analog crossovers often hides the soundquality gains that can be had . One can ofcourse be happy with that .

No again, with a dose of snark it seems.

At faktiskt.io, a swedish hifi forum they have tested the new minidsp in a very expensive system . The result didnt come out well soundwise.

This is also not evidence. You didn't even try to summarize the thread for the broad ASR audience in English, so doubly not evidence. And I hope you appreciate that of all the citations, this weak in any language. Please read the ASR charter documents.

How about a physical model?

Everything you put in between a source and a speaker makes the sound slightly worse

No, unless you have gear that has distortion and noise above your ear's detection limit. OK, so you personally can hear the sonic signature of every piece of equipment inserted into an audio chain? No.

To gain real benefits with a dsp, one should instead go fully active with the minidsp

In the same post, you essentially contradict yourself. MiniDSP is OK by itself, but coupled with the passives pushes it over a cliff? Or some mysterious Digital-Analog interaction??? So No.

No gear , analog or digital are fully transparent to the ear . One of the worst offender is passive crossovers in loudspeakers .

So is it 'transparency'? Like, can you hear the difference between different capacitor types or inductor types? Can your ear measure the SINAD or PRaT of a 1uF MusicCap? No.

Whatever you propose here is extremely difficult to prove. But comments like this are difficult to leave:

Im surprised that the reviewers at faktiskt.io dont draw the conclusion or think about the reason why it might sound worse putting a dsp in a signal chain with passive loudspeakers .

Actually, your conclusion is the surprise. Some issues are not easily correctable by DSP. If you are trying to use DSP to battle a room mode, a resonance in a speaker, a breakup mode of a driver, an issue in the crossover... then for sure it is going to be a struggle with potentially no good solution. No idea what the people in this review were struggling with.

Whats your experience ?

Since you ask, I built a pair of Seas Thor in 2003. These have very challenging midbass drivers. I upgraded to larger transmission line and went through many crossover iterations, the Thor saga can be found all over diyaudio.com. Certainly the different crossover and cabinet iterations made large differences in sound. One thing was constant, the need to control the 4.7kHz resonance of the W18001 midbass driver. I spent $1500 on fancy passive components at one point, that was a total waste that didn't make a difference (well, I did have to add a adjust a couple of things to compensate for different DC-resistances of the boutique components). I later bypassed the passive crossover with an analog active, ending up with boards I got from Linkwitz Lab since they had all the functionality I needed to exactly duplicate the passive crossovers. Plus I super appreciated the supporting knowledge and documentation Linkwitz provided. And, right off the bat I was able to get the Thor to sound just like they did passive, except now active. The biggest gain seemed to be the headroom, normal volume seemed the same. I will grant that I likely never had the balance between mid and tweet the same as the passive, but gosh they were still sounded like Thor. The greatest thing about the active: I was able to change the crossover (relatively) easily. I tried all sorts of slopes and crossover points, with turnaround in minutes rather than the days it took me for a new passive. So I loved the analog active, so much fun. I ended up with an active crossover very similar to the super-Thor passive, except I did move the notch filter back to passive. Linkwitz even kindly helped me with the passive notch implementation suggesting a series circuit rather than the parallel. My reasoning at the time for the passive was based on a simple model of mechanical damping, and I admit the model in my head was naïve. Later, PuriFi wrote a nice paper with a more advanced understanding and detailed measurements confirming at least part of my reasoning:

https://purifi-audio.com/wp-content/uploads/2022/03/220211_R05-Notchfilter.pdf

Thanks to @thewas for bring this to our attention.
I later replaced the Linkwitz analog crossovers with DSP. I tried several units over a few years, and was typically able to reproduce exactly the sound of the Linkwitz crossovers, but each suffered from one or more performance issues. I finally settled on a MiniDSP 4x10 HD which was great, allowed me to exactly reproduce the function of the Linkwitz and added a volume knob and remote among other things. It wasn't as sexy as the nice crossover chassis I made with a Sonic Frontiers Line-3 preamp, but it was fun and easy. I want to emphasize that there is no way I could tell the difference between the Linkwitz and the MiniDSP when the same functional crossover was implemented. I also want to say that the notch filter stayed passive. Every time I tried to do an active notch, I either couldn't hear the difference or I couldn't get the Q of the active filter just right... I'm not sure I was hearing the superior performance of the passive filter as illustrated in the white paper I referenced above, but the passive notch sure worked well and there I think I now understand the physical reason why. I am pretty sure that I can make a great DSP notch and get nearly the same audible performance as the passive notch, but I don't see the DSP magically outperforming the passive notch. And I rather use the filter space to correct for other issues, so the passive notch remained and performed great.
So, my experience is once I had a good analog active crossover I would never turn back because of the ease and flexibility. And I never wanted to turn back from DSP once I was able to get a transparent DSP with performance better than my ears. Even then, we are only starting to see reasonable DSP that supports more than a 2-way system (MiniDSP Flex8 for instance is eagerly awaited by many who are still put off by more complicated crossover implementations). And, my experience is also studded by all sorts of money thrown at the problem. I overheated the Thor's tweeters experimenting with a very low crossover point, my bad since the Seas Millennium tweeters are not cheap. I had numerous heart attacks when glitches happened to the DSP that caused the filters to disappear, or I got the tweeters and woofers mixed up on the back of the Linkwitz crossover, or the numerous clicks and pops that I needed to manage, or the pre-flight checklist to make sure I didn't smoke a tweeter during power-on, or the pile of amps and stuff. In the end, after nearly 20 years I retired Thor. I had bought a pair of Genelec 8361A, and I wanted to focus my attention on room acoustics, bass management, and building subwoofers. The Genelecs do sound better and actually cost less than I spent on the Thor. That is my experience; fun and expensive and I learned stuff, but I would never recommend doing this unless it was to a fellow DIY enthusiast who had lots of money and time and didn't mind wasting some of it.

As I said in the beginning - buying a dsp to correct an analog loudspeaker is often a waste of money or time .

So, no. Owning a crappy speaker that cannot be improved is a waste of money and time.

The first step for real improvements would instead be converting the passive crossover in the passive loudspeaker to be fully active in the digital domain , keeping the signal digital as long as one possible can, and doing the crossover in the digital domain without converting it first to analog.

While this may be a fun project for a few people and will potentially lead to some great sound if the person has the time and knowledge and money to pursue this goal, it is really bad advice for 99% of the people who own a bad speaker that could be improved by replacing the crossover. And it's bad advice for 100% of the people who already own good speakers.

 

[fpitas]

While this may be a fun project for a few people and will potentially lead to some great sound if the person has the time and knowledge and money to pursue this goal, it is really bad advice for 99% of the people who own a bad speaker that could be improved by replacing the crossover. And it's bad advice for 100% of the people who already own good speakers.

Agreed. If you want to learn speaker design, it's a great hobby. Otherwise, you'll want to kill yourself pretty quickly.