Kii Three / D&D vs. PSI Audio actives - DSP vs. analog crossover

[phase_accurate]

Very dangerous are those "they must have done it for a reason" arguments.. That is exactly what the brands want you to think.

Regarding the PSI tweeter I know the reason from first hand:
They wanted to have a dome tweeter with good SPL capability without using ferrofluid. So they went for a very small air gap which improves efficiency and voice-coil cooling. But voice coil diameters can't be manufactured to any arbitrary tolerance one would desire. So they have three different categories of pole-piece and top-plate pairs which they select accordig to the measured actual diameter of the voice coil when they manually assemble the tweeter. IIRC the claimed difference in long-term SPL capability is around 6 dB compared with other FF less dome tweeters.

Regards

Charles

 

[mark100]

My questions were probably not understood. For example, although I have expressed my experience with additional AD conversion being detrimental, I do not care about it because all my material is digital and I will be feeding DSP speakers (if I get such) via digital input. Also, I have no doubts that potential of DSP is much bigger than analog active or passive, but again my questions are not about advantages of one over the other. Everyone understands the technology, my questions are about the quality of parts and quality of implementation in the examples given. Based on my own experience I know how significant impact jitter can have and how different digital inputs can sound, for example sending the same signal from the same source to the same DAC over USB and network, the differences can be enormous, you do not need any extensive listening, a few seconds of sound and you hear it right away. I am saying this in attempt to explain what I am after. On the other hand I also know that excessive analog processing has detrimental impact on transparency, there is no single pre-amp that is transparent enough IME. So when you consider the subpar digital circuits and DA in DSP based speaker (subpar compared to standalone DACs for example, due to the price point), or all the processing in analog crossover (examples given are Bryston analog crossover and PSI Audio and ATC active speakers), what can be expected to have less impact on transparency, what will better preserve minute details, the DSP/DA used in presently available products (Kii, 8c) or analog active (PSI Audio, ATC)? If someone had experience in building active speakers for decades, and has gained experience with solutions available on the market, or has gained experience working with specific examples of the products (many sound engineers have) then such person can provide well informed opinion. But what you hear mostly is just a drivel, someone built one or two systems, and believes it is the best that can be achieved.

Good questions and understood i think....

I just think there are too many variables being juggled/considered all at once together, to form any kind of assessment about their relative contributions to the whole product.
How can one begin to start to compare dsp vs active analog in general, quality of drivers, and quality of electronics be they digital of analog, all at the same time?
Ime/imo, you have to substitute/change one variable at a time to form any valid conclusions. ... like swapping to dsp on the PSI Audio, or active analog on the Kii. it's pure speculation till you isolate variables imo.

 

[SashaV]

Regarding the PSI tweeter I know the reason from first hand:
They wanted to have a dome tweeter with good SPL capability without using ferrofluid. So they went for a very small air gap which improves efficiency and voice-coil cooling. But voice coil diameters can't be manufactured to any arbitrary tolerance one would desire. So they have three different categories of pole-piece and top-plate pairs which they select accordig to the measured actual diameter of the voice coil when they manually assemble the tweeter. IIRC the claimed difference in long-term SPL capability is around 6 dB compared with other FF less dome tweeters.

Regards

Charles

Thank you for informative answer Charles.
I would really appreciate your subjective opinion on all those speakers you said you had opportunity to hear, specifically the three-way PSI (I assume it was not the new A23-M?), the Grimm Audio LS1, ATC SCM 100 ASL, and PMC BB5. It would help me considering that I am familiar with ATC 100 ASL and to some degree with BB5, but not with PSI Audio or Grimm.
And since you seem to have good knowledge about PSI Audio, do you have any opinion about A23-M and their new mid driver?

 

[mark100]

Until now I have not heard a strong argument why DSP should be superior, apart from "ease of use".

DSP can perform precise timing, typically to 1 sample, so 0.02ms at 48kHz sampling rate.
This equates to being able to time acoustics centers to within 1/4 inch.
While at the same time make adjustments that span many inches.

DSP with FIR can implement linear phase xovers, at practically any order.
Which means speaker designs are not constrained to first order xovers to avoid any phase rotation or group delay.

Try to do either of those with active analog.....


DSP can implement a far greater number of minimum phase filters than active analog.
Perhaps this goes in the 'ease of use bucket', but when the effort of matching filter counts becomes so humongous,
it gets a little ridiculous to say active analog can match DSP imo.
If using FIR, filter count is functionally unlimited.


For me, those three items are unquestionably real world DSP superiority's over active analog, for building and tuning speakers

 

[hifijim]

Very dangerous are those "they must have done it for a reason" arguments.. That is exactly what the brands want you to think. A whole lot of it often boils down to one thing: marketing. To separate yourself from others, be unique, be different, stand out. Sometimes there even is a viable engineering reason for something as well, but definitely not always.

I was just writing a response which made this exact point... If speaker company A has a successful active monitor which uses DSP, and company B wants to introduce a new active monitor to compete, their marketing department (in some cases just a single person) will poll the market by talking to dealers and customers... if they find that there is a market for an active monitor with full-analog circuitry to satisfy those customers who do not want any digital filtering, then company B will offer an active monitor with analog circuitry... It is about what you can sell and still make a little money...

 

[TNT]

It seems that if you have high quality audio processing (filters etc..) software and no extra AD/DA hop, any eventual degradation from this software is far outweighed by the gain of linearising the FR of a speaker. My take is that such software beats any physical R, C and L in transparency. So if you can do the filtering in the digital domain before your DAC(s) in a proper way, it is for me the vinning proposition. Any day of the week.

I didnt care for the Behringer EQ box nor did the first minidsp box stay for long in my system. They made its sound off and grey. Maybe it was the analog parts??

First time I heard the Kii I was very impressed. Second tiem a year later, a little bit less so. But I wouldn't mind test a a pair in my home. The D&D didn't impress me at all.

//

 

[plasnu]

I didnt care for the Behringer EQ box nor did the first minidsp box stay for long in my system. They made its sound off and grey. Maybe it was the analog parts??

//

Cheap digital equipments sound bad not only because bad analog and conversion parts, but mostly because their digital part is horrible. Very hard to design a good sounding digital audio processing program. That is why mastering engineers have been exclusively using Weiss EQ, even it is usually digitally i/o connected. I know they sometimes use a Behringer's cheap analog equipment, but not their digital equipment. Please try high quality EQ such as Weiss or Equilibrium, and you will know what I do mean, and I have tested 100+ digital EQs. Very rare to find a transparent one.

 

[TNT]

Doing EQ (biquads) in my Soerkris DAC makes me happy.

But I'm contemplating using Camilla (thread on this site!) on a Pi4...

//

 

[mark100]

Cheap digital equipments sound bad not only because bad analog and conversion parts, but mostly because their digital part is horrible. Very hard to design a good sounding digital audio processing program. That is why mastering engineers have been exclusively using Weiss EQ, even it is usually digitally i/o connected. I know they sometimes use a Behringer's cheap analog equipment, but not their digital equipment. Please try high quality EQ such as Weiss or Equilibrium, and you will know what I do mean, and I have tested 100+ digital EQs. Very rare to find a transparent one.

Hi plasnu, normally i'm pretty much in sync with your posts, but i think this one is nonsense.
In the pure digital domain, EQ filters and crossovers are just numeric implementations..with ever increasing standardization / conformity i think.
If mastering engineers are sometimes using cheap Behringer analog equipment and shunning Behringer's all-digital stuff....well...they have it exactly backwards.

 

[voodooless]

In the pure digital domain, EQ filters and crossovers are just numeric implementations..with ever increasing standardization / conformity i think.

Not really. There are various implementations out there with variing bit depths for calculations, floating point vs integer calculations, different accumulator designs. Devices have different processing capacity, making things possible on some devices vs others (for instance FIR needs a shitload of memory which the cheaper devices don't have). Is the gain structure properly setup? Then there is the question of sample rate conversion which can be done in various qualities, and then the final step in converting to the output bitrate. Is dithering applied? What kind? Is volume control properly done. In effect, you can do many things in many ways, and not all are equally good. Some things are down to software, others are limited to the specific hardware architecture.

 

[plasnu]

Hi plasnu, normally i'm pretty much in sync with your posts, but i think this one is nonsense.
In the pure digital domain, EQ filters and crossovers are just numeric implementations..with ever increasing standardization / conformity i think.
If mastering engineers are sometimes using cheap Behringer analog equipment and shunning Behringer's all-digital stuff....well...they have it exactly backwards.

mark100, I 'm too pretty much in sync with your posts, but I think I have to argue about this matter with you...

It is a big myth that all digital processing sounds the same. Top digital designers and companies keep it secret how to make their digital processing engine sound better than their competitors, so I can only guess why this is happening just based on occasional interviews of those designers on professional audio magazines. What they say is, gain structure, dither, upsampling, all do matter, and they don't even start talking about FIR step length and window, etc.

The professional mixing engineers who regularly use digital EQ always advice that we have to carefully set the input level of digital EQ, because zero digital dB is not the optimal operational point of digital EQ in many cases. They say blindly sending zero dB signal to digital EQ is probably making sound worse. But do you know how to correctly reduce gain digitally? Now we have to consider noise shaping etc. and we don't even know if we should dither/upsample the signal or not before EQ. Digital processing is the same as analog, it's not easy, we are simply ignorant about it since it is a black box. Only I can say is some codes sound much better than the other codes in my experience.

I don't know any mastering guy who is (was) actually using Beringer's digital processor, but maybe someone does. Howie Weinberg was secretly using Behringer's Edison, a $100 analog psychoacoustic processor, and it made him big late 90's. My point here is, cheap analog processor can sound subjectively good (warm, powerful, or whatever), but cheap digital equipments sound nothing but cheap digital, which is eventually, useless.

 

[mark100]

mark100, I 'm too pretty much in sync with your post, but I think I have to argue about this matter with you...

It is a big myth that all digital processing sounds the same. Top digital designers and companies keep it secret how to make their digital processing engine sound better than their competitors, so I can only guess why this is happening just based on occasional interviews of those designers on professional audio magazines. What they say is, gain structure, dither, upsampling, all does matter, and they don't even start talking about FIR step length and window, etc.

The professional mixing engineers who regularly use digital EQ always advices that we have to carefully set the input level of digital EQ, because zero digital dB is not the optimal operational point of digital EQ in many cases. They say blindly sending zero dB signal to digital EQ is probably making sound worse. But do you know how to correctly reduce gain digitally? Now we have to consider noise shaping etc. Digital processing is the same as analog, it's not easy, we are simply ignorant about it since it is a black box.

I don't know any mastering guy who is (was) actually using Beringer's digital processor, but maybe someone does. Howie Weinberg was secretly using Behringer's Edison, a $100 analog psychoacoustic processor, and it made him big late 90's. My point here is, cheap analog processor can sound subjectively good (warm, powerful, or whatever), but cheap digital equipments sound nothing but cheap digital, which is eventually, useless.

Thx plasnu,
I think it's probably just a matter of our different perspectives.
Mine is from that of using EQ to fix/tune speaker designs, and adjust overall system tonality....
There EQ's are almost classic textbook implementations.....simply designed to do whatever Q or BW, at whatever freq, and at what gain.
I'm guessing yours is that of studio...where EQ's are as much of an 'effect' as anything else???
I can see how greatly things can vary there. I hear the live sound guys constantly swapping notes on effects, but hardly ever on straight EQs.
I've never played with anything other than vanilla EQs.

Another guess is most DIYers are looking for good EQ's for their speaker builds, and there i do believe many digital boxes are fairly close equivalents.

 

[plasnu]

I'm guessing yours is that of studio...where EQ's are as much of an 'effect' as anything else???

In mixing studios, they use transparent EQ (they call it surgical EQ) for correction, and use classic analog EQ emulation (API, NEVE or such) as more like an effect. Both type of EQs are used in mastering studios.

I was talking about surgical EQ in my posts.

 

[mark100]

In mixing studios, they use transparent EQ (they call it surgical EQ) for correction, and use classic analog EQ emulation (API, NEVE or such) as more like an effect. Both type of EQs are used in mastering studios.

I was talking about surgical EQ in my posts.

Ok, thx for clarification.
I believe surgical EQ is becoming equal if, not nearly already there, when speaking about a pure digital in - digital out device,...... across numerous products priced high to low.

 

[plasnu]

Mark 100, as I said, I'm not 100% sure everything I said is correct due to my very limited knowledge about digital audio processing, but most people on professional audio forum seem to agree that each surgical EQ will sound different even when high pass filter with the same function applied.

We can probably measure the difference with a free audio measuring software called Audio DiffMaker, which compares 2 different digital audio files and tell us how same the 2 digital files are. I saw someone was using this software to compare DAC and ADC, and I remember that the result was very interesting. I don't know how trustful this software is, though...

 

[phase_accurate]

Yes, there are lots of things that can go wrong with DSP if not done with a lot of care. With IIR filtering there are the so-called limit cycles (I wouldn't even think of using my AD DSP with single precission processing only) and warping of the frequency axis. The latter can be fought with higher sample rates at the cost of making the first problem bigger at lower frequencies. With FIR crossovers it is easy to achieve perfect temporal response on axis. If not done properly you will have much worse temporal response odd-axis than with any analog solution.

YES - digital has the potential to be better than analog (and in some cases it definitely is) but one can definitely not take it for granted.

I use the solution inbetween: I do the prototyping of analog active crossovers with DSP and build the final active crossover in analog fashion.

I do only have a restricted amount of time available for DIY and I don't want to use that to develop a high quality DSP solution which would probably take me ages to complete.

And buying a finished high-quality DSP solution isn't DIY anymore …….

Regards

Charles

 

[AllenB]

Until now I have not heard a strong argument why DSP should be superior, apart from "ease of use".

Yes, it's been a bit thin on the ground.

(I am guessing):
a) Good drivers with decent crossover and amps will always sound better than lower quality drivers no matter how much you try to do with DSP

For this comparison you need two speakers/rooms that are equally flawed. The best way to do this is to use the same speaker with each type of crossover.

Each crossover would have to do the exact same thing. The process might be more efficient if the speaker had a solid acoustic design to begin with, because some distortions (but not all, and not always), cover up other distortions. (And whoever implied an analogue crossover has to be first order to avoid audible group delay, pull the other one.)

But that is NOT delay! It's phase shift. Not the same.

Quite, but Siegfried does show an all-pass filter.

 

[mark100]

Mark 100, as I said, I'm not 100% sure everything I said is correct due to my very limited knowledge about digital audio processing, but most people on professional audio forum seem to agree that each surgical EQ will sound different even when high pass filter with the same function applied.

We can probably measure the difference with a free audio measuring software called Audio DiffMaker, which compares 2 different digital audio files and tell us how same the 2 digital files are. I saw someone was using this software to compare DAC and ADC, and I remember that the result was very interesting. I don't know how trustful this software is, though...

My bad plasnu, i think i was making too lite an issue of the differences in EQ implementations. I mean heck, look at the different formats REW provides, and my go-to software FirDesigner has a EQ matching mode, for Powersoft and Linea Research etc..

I guess i never think past speaker tuning, where one measurement to the next usually varies far more than anything I've seen with different processors.

But as to what we can hear, this thread has given me an experiment to try...
I'm going to replicate a set of EQ's from a Massive Passive i still own, via FIR, and see if i can hear a diff on some good Stax headphones, and maybe also on a pair of active analog Meyer boxes.
Massive Passive vs dsp FIR w AD/DA conversion is about as far apart an EQ implementation as i can think of ...

 

[mark100]

Hey Scott, if you're still monitoring this thread, and as you are into tape, you might enjoy this:
https://worldradiohistory.com/Archive-All-Audio/Archive-Recording-Engineer/80s/Recording-1981-12.pdf

A nice gent recently posted it on one of the prosound forums.
The ads alone are worth looking at...
A real turn back in time

 

[wesayso]

For this comparison you need two speakers/rooms that are equally flawed. The best way to do this is to use the same speaker with each type of crossover.

Absolutely.

Each crossover would have to do the exact same thing. The process might be more efficient if the speaker had a solid acoustic design to begin with, because some distortions (but not all, and not always), cover up other distortions. (And whoever implied an analogue crossover has to be first order to avoid audible group delay, pull the other one.)

While this would test analog vs digital, it would not make use of the advantages that digital could bring to the table.

Why even go digital in such a case. Use the tools for what they can do, if you feel the need to, anyway. There's nothing wrong with passive or active analog. But digital makes sense for things that are much harder to do in an all analog way. That doesn't mean you'd resort to using inferior drivers etc. (unless there's more money in such a thing for companies that market these)

If you don't use what digital has to offer, I see no need to go there. Why mimic a passive setup, except to see if it can hold up. But it would still be very hard to make them behave exactly the same in every aspect for such a test to work.

I love what digital can do for me, because it wouldn't even have crossed my mind to try to do that passive or analog active. To each his own.

I've heard enough analog setups to know it can sound just wonderful. Though I expect to see many more digital solutions in the coming years. Not all will be equally successful. Nor were their passive predecessors...