Howdy folks
I've got sort of a strange query. Really interested in getting some thoughts from you all
So, it seems to me to be an ongoing conversation on the whole "DSP vs passive crossover" debate, and the solution some have landed on is one that I find very intriguing; that is, a hybrid crossover that utilizes both.
In such speakers, you typically see a passively filtered mid and tweeter, with a DSP filtered woofer.
There are many reasons for such an implementation, but among them is that we are allegedly less sensitive to unwanted digital transformations of the signal at lower frequencies. A term I often hear thrown around in these contexts is "digital artifacts". My experience with DSP is limited, but the argument seems to me to go like this: "While DSP offers more refined tuning, the overall fidelity still does not match that of passive filters. There is a subtle but unmistakable tonal character imparted onto the overall sound when digitally manipulating the signal (another term I hear is "digital sheen") that makes passive filters still the best option at the highest levels of hifi". (This is not to suggest that passive crossovers don't also add their own little flavor, but I guess the argument goes that it's less objectionable in this case)
Now, I don't know if this is true, but let's just say it is. Some people take this idea and say, "ok, but any timbral anomalies will not be audible, or at least perceptible at lower frequencies", thus, hybrid crossovers.
I have 2 questions. My first is this: why? Why can't we hear traces of DSP at lower frequencies, yet we can at higher ones?
My guess is that it's because the woofer (let's just say it's only playing 300Hz and below) is outside of the critical range that really defines the timbre of music.
So here comes my second question. If this is true, (and I'm not saying it is), could the same concept also be applied to the extreme top end? Would you still be able to detect this digital sheen in a speaker that's mid is filtered passively, with a tweeter filtered via DSP? Let's say crossover of 7KHz.
Why exactly you would do this? I don't know. I can think of a few things, but it's definitely a weird idea with few use cases, presupposing that it's a sound idea at all
Anyways, I am eager to hear your responses. Thanks to those that read through all of this
I've got sort of a strange query. Really interested in getting some thoughts from you all
So, it seems to me to be an ongoing conversation on the whole "DSP vs passive crossover" debate, and the solution some have landed on is one that I find very intriguing; that is, a hybrid crossover that utilizes both.
In such speakers, you typically see a passively filtered mid and tweeter, with a DSP filtered woofer.
There are many reasons for such an implementation, but among them is that we are allegedly less sensitive to unwanted digital transformations of the signal at lower frequencies. A term I often hear thrown around in these contexts is "digital artifacts". My experience with DSP is limited, but the argument seems to me to go like this: "While DSP offers more refined tuning, the overall fidelity still does not match that of passive filters. There is a subtle but unmistakable tonal character imparted onto the overall sound when digitally manipulating the signal (another term I hear is "digital sheen") that makes passive filters still the best option at the highest levels of hifi". (This is not to suggest that passive crossovers don't also add their own little flavor, but I guess the argument goes that it's less objectionable in this case)
Now, I don't know if this is true, but let's just say it is. Some people take this idea and say, "ok, but any timbral anomalies will not be audible, or at least perceptible at lower frequencies", thus, hybrid crossovers.
I have 2 questions. My first is this: why? Why can't we hear traces of DSP at lower frequencies, yet we can at higher ones?
My guess is that it's because the woofer (let's just say it's only playing 300Hz and below) is outside of the critical range that really defines the timbre of music.
So here comes my second question. If this is true, (and I'm not saying it is), could the same concept also be applied to the extreme top end? Would you still be able to detect this digital sheen in a speaker that's mid is filtered passively, with a tweeter filtered via DSP? Let's say crossover of 7KHz.
Why exactly you would do this? I don't know. I can think of a few things, but it's definitely a weird idea with few use cases, presupposing that it's a sound idea at all
Anyways, I am eager to hear your responses. Thanks to those that read through all of this
I use all DSP crossovers based on a very heavily modified DCX2496, MiniDSP SHD Studio and RME ADI-2 PRO FS R. Performance is good across the board, but I can't hear most of the top octave, but those who can haven't complained.
I have no digital sheen at all, hide the electronics and most people including audiophile types will not figure it out. It wasn't very easy, and I have spent the better part of a year figuring this out.
Speaker system is Iconic 165-8G in 11cu ft diy Onken boxes, TAD TD-4001 on Yuichi A-290 clones, and FaitalPro HF10AK on STH100 horns. Amplification is a mix of class D for woofers and 300B DHT on the mid and HF horns. The system is 4-way stereo with powered subs operating below 45Hz. Rather small room just < 200 sq ft
I have no digital sheen at all, hide the electronics and most people including audiophile types will not figure it out. It wasn't very easy, and I have spent the better part of a year figuring this out.
Speaker system is Iconic 165-8G in 11cu ft diy Onken boxes, TAD TD-4001 on Yuichi A-290 clones, and FaitalPro HF10AK on STH100 horns. Amplification is a mix of class D for woofers and 300B DHT on the mid and HF horns. The system is 4-way stereo with powered subs operating below 45Hz. Rather small room just < 200 sq ft
Well it’s not the DSP that rational people are complaining about. The DSP just does the number crunching. Like a CPU. I mean you can do Digital Signal Processing on a Raspberry pi, SHARC, or general purpose CPU or GPU.
I believe what people are complaining about are the DACs that are connected to the end of them. They want the cutting edge DACs.
Of course, there are other people who believe that analog signal processing is better. Anything you can do with DSP you can do with ASP. Just takes a lot more parts and a slight mismatch when compared directly to the DSP. The May give it special effects. No one will can argue with special effects.
I believe what people are complaining about are the DACs that are connected to the end of them. They want the cutting edge DACs.
Of course, there are other people who believe that analog signal processing is better. Anything you can do with DSP you can do with ASP. Just takes a lot more parts and a slight mismatch when compared directly to the DSP. The May give it special effects. No one will can argue with special effects.
That's awesome. I am excited to experiment with it. I hope to someday soon pick up a MiniDSP Flex Eight, and I already have an 8 channel power amp I bought from diya user john65b. I am betting that pending my design skills, things will sound pretty good 🙂
This is my focus. I'm curious about what range we are sensitive to the special effects
Well I have the Flex 8 since day 0 release. And it’s perfect for my needs. But as soon as it came out even the DSP pro camp are already complaining about it. Why no balanced output? (There’s no space on the back panel for 8 XLR or TRS).
Why not a state of the art DAC (it’s uses a boring 8 channel DAC- ESS9018 from around 2015 for a claimed SNR @128dB(A), THD+N @ -111dB; what about the latest AKM or ESS etc for a real THD+N of over -120dB we armchair experts will ask)
Well DIY Audio is increasingly niche market and these DSP makers who are selling to end consumers at retail pricing have to also support AND customer enquiries that goes with it; now that’s a USD $100K outlay right straight up, plus ongoing recurring costs…
If people want to complain about DSP not being good enough for them, please stop complaining annd start being productive. Design your own 152 channel DSP.
Get started; we will buy!
https://www.akm.com/us/en/about-us/news/2023/20230720-ak7709/
MOQ only 900 pieces.
Why not a state of the art DAC (it’s uses a boring 8 channel DAC- ESS9018 from around 2015 for a claimed SNR @128dB(A), THD+N @ -111dB; what about the latest AKM or ESS etc for a real THD+N of over -120dB we armchair experts will ask)
Well DIY Audio is increasingly niche market and these DSP makers who are selling to end consumers at retail pricing have to also support AND customer enquiries that goes with it; now that’s a USD $100K outlay right straight up, plus ongoing recurring costs…
If people want to complain about DSP not being good enough for them, please stop complaining annd start being productive. Design your own 152 channel DSP.
Get started; we will buy!
https://www.akm.com/us/en/about-us/news/2023/20230720-ak7709/
MOQ only 900 pieces.
I think it is less that people complain and think it's bad. Lots of people love DSP, but love passive even more.
Odd that people complain about 9018 DAC. Implemented well that's a phenomenal sounding chip. Some of my favorite ever DAC's use the 9018
Okay, so as a predominantly loudspeaker designer designer, I can say that a lot of funky stuff is going on in the high frequencies. Like one of the things we are learning in C21 is that the space around the tweeter or compression drivers have a big difference. I’m talking about diffraction, do use waveguides horns shaped baffles etc. They can all affect the top end not response. Not to mention that slight movements in your head can cause difference in SPL and phase etc (HRTF)
I’d really like see a well designed study where the exact same transfer function of DSP is compared to ASP or passive crossover and the listener gets to compare without having to move from her seat and make a call which one is preferable.
I'm wary of conjecture on the subject. It is better to identify your problems.
Either technique can successfully bring out the speaker if it is otherwise designed properly, and once that happens you can hear differences for what they are, if they exist. Listening to some of the conjecture that has been written about passive vs active, it's not clear that proper matching has been achieved or that speaker issues have been eliminated before comparisons are made.
Either technique can successfully bring out the speaker if it is otherwise designed properly, and once that happens you can hear differences for what they are, if they exist. Listening to some of the conjecture that has been written about passive vs active, it's not clear that proper matching has been achieved or that speaker issues have been eliminated before comparisons are made.
Yes but are they doing DSP right? You can’t switch in a few textbook crossovers into a DSP and call it a day. Or just measure on-axis.
You really have to design your way around filter shaping with passive crossovers and ASP first.
Everyone I know has moved to DSP and active amplifiers… well… like decades ago. They never looked back.
And if they release speakers with passive crossovers, well that’s just because of needing to go to market. And market acceptance.
I’ll happily do passive crossovers for anyone who wants to pay. But in my own system I use active crossovers and multichannel amplifiers of appropriate for intended use. If you feel like you need an anvil of Class A amp for your tweeter, I’m not going to argue.
Fair point. I'd like to someday experiment myself, but until that time I am taking others' words on it
That would be awesome. I wonder if such a study exists already, unbeknownst to us
Let me put it this way.. sometimes it is worth pursuing the issues you know to be a problem, in this case making sure you cross properly after careful acoustic design, and not concern yourself with issues which are not proven. If the time comes, you can deal with it then.. your acoustic and crossover design work won't go to waste.
DSP is not the panacea for all ills, you need a good if not perfect speaker system and a good treated room, or at least as good a room as you can make it.
Lots and lots of near field and room response (listening position and elsewhere) measurements, lots of small tweaks - minimalism is often best, over corrected may look great on that response graph but might not sound right.
Knowing where to stop is important, but having a good understanding of the performance limitations of the components you are using and the room is key.
I have been fully DSP for about a year now, and until about 2 months ago when things finally clicked I really thought I was barking up the wrong tree. 🤣
I have fully active for 7 years and the DSP was a big step up from the analog line level XOs I designed in 2017 for the system. I will not be going back
Lots and lots of near field and room response (listening position and elsewhere) measurements, lots of small tweaks - minimalism is often best, over corrected may look great on that response graph but might not sound right.
Knowing where to stop is important, but having a good understanding of the performance limitations of the components you are using and the room is key.
I have been fully DSP for about a year now, and until about 2 months ago when things finally clicked I really thought I was barking up the wrong tree. 🤣
I have fully active for 7 years and the DSP was a big step up from the analog line level XOs I designed in 2017 for the system. I will not be going back
I've use two different DSP crossovers with my homebuilt hybrid ESL speakers and I've never noticed any anomalies in the high frequencies. My old ears can't hear much past 7K anyway but my tuning mic displays the highs on the RTA screen all the way up to the 20kHz cutoff, so I know the highs are there and the portion I can hear sound wonderful.
Based on my experience, I can't imagine whatever problem you describe could be an artifact of using a DSP.
My first DSP was a Behringer DCX-2496 and I now use a DBX Driverack Venu 360. Both units perform well and sound great but the interface and auto-EQ'ing features on the DBX are much easier and faster to use.
Based on my experience, I can't imagine whatever problem you describe could be an artifact of using a DSP.
My first DSP was a Behringer DCX-2496 and I now use a DBX Driverack Venu 360. Both units perform well and sound great but the interface and auto-EQ'ing features on the DBX are much easier and faster to use.
Assuming a sufficiently high sample-rate is used, the problems of a DSP technically lie at the lower frequencies (vs. higher ones). This is true for both FIR and IIR filters. As frequency drops, FIR filters require a high number of coefficients (high order) and computations while IIR filters run into stability problems due to quantisation error buildup, often necessitating extra bits of processing even upto double precision/64-bits for very low frequencies.
Though there are multiple solutions for the above issues (both FIR and IIR), they are seldom used in audio, which is generally not considered a critical application. The audibility of these errors depends on their severity which is very low (with proper implementation) and is perhaps another reason for the same.
EDIT:
For FIR filters:
https://dsp.stackexchange.com/quest...th-low-cut-off-compared-to-sampling-frequency
For IIR filters:
https://www.earlevel.com/main/2003/02/28/biquads/
There's also the difference between fixed-point (n-bits) vs floating-point processing.
Last edited:
This is good information, all. Thank you.
Maybe "digital sheen" does not exist.
For argument's sake, let's say it does.
Would its audibility be hypothetically likely to be variable based on frequency?
If so, would the critical range (not a hard and fast range, but let's say 350Hz-7KHz) being passive and the rest being active result in a system without any audible digital sheen? Or would it just be at the lower range?
Maybe "digital sheen" does not exist.
For argument's sake, let's say it does.
Would its audibility be hypothetically likely to be variable based on frequency?
If so, would the critical range (not a hard and fast range, but let's say 350Hz-7KHz) being passive and the rest being active result in a system without any audible digital sheen? Or would it just be at the lower range?
Hi, it's quite easy to hear artifacts on some mp3 quality settings or some old youtube videos, I think the Tidal losless quality was audible as well when I tried it with headphoness. Not quite a same thing, but case in that quite small errors can be audible. I haven't compared DACs but if people say they sound different I believe to some extent, 15 years ago I had sound interface which didn't make very good sounding recordings and switching to RME interface truly was audible, but I don't know if it was converters or analog parts that made the difference, likely both had some role, perhaps something else.
Greatest benefit of DSP, in my opinion, is that it gives another degree of freedom for loudspeaker design. Together with modern software like VituixCAD I think crossovers are not an issue at all, all you have to worry about is to have as good acoustic construct as possible, which measures well and sounds good without issues, without any worry about xo. Crossover is just a process, while it has quite big impact on sound there is no reason you could not achieve best possible xo for the construct.
DSP allows you to evaluate effect of the crossover, how important it is to tune various things and to what degree of accuracy and so on. You can explore things, find out what matters, measure a lot and learn to avoid/see errors in measurements. It allows a process where you learn a lot. Perhaps you didn't care any of this and just want to listen to music and return to passive speakers sooner than you thought 🙂
If you find it worth your time, when you are finally done with your speaker design and it is as good as ever, you might find out the processing could be replaced with passive parts which you can now do if you wish, or just keep it.
If you hear digital sounds, then figure out if there is better sounding solution available. But now you know you need it, don't worry about it beforehand. If you do carbon copy of passive xo with DSP, don't bother. It's the whole system enabled by DSP that makes a difference, if and how you utilized the design freedom.
Greatest benefit of DSP, in my opinion, is that it gives another degree of freedom for loudspeaker design. Together with modern software like VituixCAD I think crossovers are not an issue at all, all you have to worry about is to have as good acoustic construct as possible, which measures well and sounds good without issues, without any worry about xo. Crossover is just a process, while it has quite big impact on sound there is no reason you could not achieve best possible xo for the construct.
DSP allows you to evaluate effect of the crossover, how important it is to tune various things and to what degree of accuracy and so on. You can explore things, find out what matters, measure a lot and learn to avoid/see errors in measurements. It allows a process where you learn a lot. Perhaps you didn't care any of this and just want to listen to music and return to passive speakers sooner than you thought 🙂
If you find it worth your time, when you are finally done with your speaker design and it is as good as ever, you might find out the processing could be replaced with passive parts which you can now do if you wish, or just keep it.
If you hear digital sounds, then figure out if there is better sounding solution available. But now you know you need it, don't worry about it beforehand. If you do carbon copy of passive xo with DSP, don't bother. It's the whole system enabled by DSP that makes a difference, if and how you utilized the design freedom.
Last edited:
For my experience , some dsp have bad sounding dacs inside with a lot of opamp and aluminium coupling caps in both output and input side. Is it better with resitors, inductors and capacitors used for the passive filters ? I say for the low i prefer DSP as big inductors killed the sound and you need time delays . And for the compressions driver i tend to prefer an analog crossover than a digital one as i have found no dacs inside the dsp sounding as good as my external dac.