I agree with most of what you wrote @Draki. But you can never get a passive solution as well as a DSP solution. It just is not pactical to eq for many different room standing waves or reflections that interfere between channels, it's just too much.
With DSP is can be done as a matter of routine.
Look at which type of speaker is appreciated by recording people and professionals: the Kii-3s, the Duch & Dutch 8c, Grimms' Ls1.
Not accidentially, these are speakers that can be placed very close to or against a rear wall; Martijn Lensink from Dutch & Dutch calls it 'room judo'- use the room against itself!
And you only can pull it off with support from DSP. The result is a speaker that sounds good in any room, with inpeccable placement and resolving power. You just can't do it passively, not by a long shot!
Jan
With DSP is can be done as a matter of routine.
Look at which type of speaker is appreciated by recording people and professionals: the Kii-3s, the Duch & Dutch 8c, Grimms' Ls1.
Not accidentially, these are speakers that can be placed very close to or against a rear wall; Martijn Lensink from Dutch & Dutch calls it 'room judo'- use the room against itself!
And you only can pull it off with support from DSP. The result is a speaker that sounds good in any room, with inpeccable placement and resolving power. You just can't do it passively, not by a long shot!
Jan
Last edited:
I designed digital filters for a living for many years. I found that the errors in digital filters often happen at low frequencies in systems running at high sample rates. If you look at the coefficients of low frequency filters generated for high sample rates they are either very close to zero or very close to one. If there isn't sufficient mathematical precision in the MAC (multiply accumulate) operations a numeric truncation occurs. This results in a different frequency and damping ratio for the filter than what was desired. This happens in floating point and fixed point implementations. The smallest number that can be added to 1.0 really matters in these systems. Single precision floating point might not cut it at a high sample rate. If you take a miniDSP 2x4HD and implement a 50 Hz 4th order low pass filter and then measure it using a USB audio interface and Arta or REW or what ever you have, you will find that the measured frequency response does not match that pretty picture in the miniDSP configuration software. The filter will roll of initially and then it will just diverge and maybe flatten out, not producing the deep attenuation as you go up in frequency. Experienced careful designers do verification testing on their design work, realize this can happen and work around it. So no, these filters have no problem working text book perfectly at high frequencies. In budget systems with hybrid crossovers, the passive crossover is less expensive to implement for the high frequency drivers because the cap and coils are small and cheap relative to another DSP and amplifier channel. A quality 4th order low frequency passive crossovers with baffle step compensation and some bass equalization requires large caps and coils that are more expensive than DSP and an amplifier.
Last edited:
It's not really less is more. It's more a question of where to stop.
Eg you have a notch on freq response. Most new users will try to bring it up using a bell positive eq. Outcome is it might not sound goodif it is sbir related ( as the issue exist only at the place where the microphone is located). Worst if you needed something like +6db gain then you lost 6db of headroom...for nothing! And given our brain is not very sensitive to cuts... better leave it alone at first.
As it have been already expressed once 'sky is the limit' you have to be careful about what you do and why.
One good thing to do imho is to read about automated/routine based software and try to understand what is done and why. It's not the finest you can do but it'll give you an idea of what should be done and what not.
Then read what 'poweruser' ( experienced) do and why. There is a bunch here: Wesayso, Mitchba, Cask05, Mark100 and more...
If you can follow a course about measurements technique and how to interpret them, it'll help tremendously.
Eg you have a notch on freq response. Most new users will try to bring it up using a bell positive eq. Outcome is it might not sound goodif it is sbir related ( as the issue exist only at the place where the microphone is located). Worst if you needed something like +6db gain then you lost 6db of headroom...for nothing! And given our brain is not very sensitive to cuts... better leave it alone at first.
As it have been already expressed once 'sky is the limit' you have to be careful about what you do and why.
One good thing to do imho is to read about automated/routine based software and try to understand what is done and why. It's not the finest you can do but it'll give you an idea of what should be done and what not.
Then read what 'poweruser' ( experienced) do and why. There is a bunch here: Wesayso, Mitchba, Cask05, Mark100 and more...
If you can follow a course about measurements technique and how to interpret them, it'll help tremendously.
Last edited:
What a detailed comment. Thank you. Admittedly, I don't understand a lot of what you said. Is there any way for a layman such as myself to identify whether a DSP solution has these low frequency problems before buying? Is it an issue that buyers should avoid at all costs? Or should we just remain cognizant of that and do our due diligence throughout the design process to account for it? MiniDSP Flex Eight for example, just because that's the unit I am planning on buying 🙂
Acoustic corrections are inevitably made to some defined area in space.... a single point, or average of any number of points.
Whatever helps that defined area in space, usually comes at the expense of other areas.
Couple that simple fact, with overly explicit corrections, ......most often made with higher Q filters....
and while risking computational errors with IIR or lack of freq resolution with FIR....
well,.... less can indeed be more 🙂
The MiniDSP Flex is an excellent product. I used one in a design for a customer a few months ago. It is in full time use in their recording studio monitors. You will be happy with it. As I mentioned before, you might measure the input to output transfer function of your very low frequency filters to verify they are doing what you want and see if you might want to make adjustments. I have not tested very low frequency filters, 50 Hz and below, on the Flex, only on the 2x4HD. I have also used the Hypex plate amps with built in DSP. If you don't already have multiple channels of amplification those two and three channel plate amps are a bargain. You can see them at the Madisound or Hypex websites.
Hostility? What is there to be hostile about? It is more a case of baffled mild amusement. Now you can't draw on your own technical knowledge/understanding to address a topic like this which is by far the most reliable way to answer a technical question. But you can identify to a reasonable extent people and sources that have the technical knowledge to answer the question reliably and you can identify to a reasonable extent people and sources that lack technical knowledge (or worse have it but are avoiding using it) and are making appeals to authority, FUD, telling fibs, emotive meaningless phrases, etc... all the things we have been forced to become familiar with in the modern world. It is going against this that is baffling and mildly amusing because I am pretty sure most people can see it but choose to reject the odds and be "open minded" to the uninformed being right. Why?
What would you say is the more damning factor? Poor design practices? Or increased computational errors as you increase the complexity of the DSP filters?
I can't speak for others, but in my case at least, I hope I don't give the impression of gravitating towards those whose claims lack any data driven backing. When I say open minded, I don't mean to suggest that we should all ignore the facts when confronted with anyone with a penchant for flowery (and often cryptic) language, and blindly abandon all technological advancement at the slightest hint of naysayers with a talent for spreading FUD. All I mean to suggest is that there are qualities to music reproduction that maybe we haven't yet learned to quantify, so I think there is still a place for subjective impressions. Perhaps not at the expense of technical knowledge, but I don't think it is fair to dismiss someone until you've done your due diligence in proving them wrong. I don't necessarily give credence to claims that can't easily be substantiated, but I don't immediately dismiss them on that basis either. As someone who is still learning, I can't quite as easily distinguish someone who is just fibbing, FUD'ing, etc. from someone who is identifying something real that deserves to be discussed, even if we can't point to a specific measurement that explains what's going on. So, I guess your amused bewilderment is deserved. I am learning a lot from this thread though, if it's any consolation to you.
Last edited:
I have looked at the Hypex stuff as well. It looks excellent, and I hear they are close to releasing a newer version with FIR filters and the new ncorex amps. Consider my breath bated!
I think I will stick to a separate unit though. Even if the conclusion I come to is that DSP filters best passive filters in every way that is important to me, I still want the flexibility to separate the speakers from the amp without much effort. Drilling an amp to a cabinet is just too committal for me, as someone who is constantly tinkering and swapping stuff out. Anyways, thank you for your impressions on the Flex Eight. Would be cool if Hypex released a multichannel kit with DSP, akin to their new stereo Nilai kits
I think you are primarily thinking about IIR DSP here. In that case your statement is true for 90-95% of the filters out there, but not all and there are some other issues that come in to play. For some types of filters, when you try to implement them as an analog active circuit, there are cases where it just doesn't work or the circuit also produces undesirable gain or loss. For example some circuits that implement the biquadratic filter (the Linkwitz Transform) have this characteristic. Also lowpass notch filters produce gain in the passband because that is inherent in the transfer function. This reduces headroom. These limitations are removed with IIR DSP and processing headroom is for all intents and purposes infinite when using double precision.
Also, last time I checked you cannot do FIR filtering very well in the analog domain... and that is a type of DSP.
As someone mentioned earlier in the thread, the sound quality aspect of DSP has everything to do with the DACs and little to nothing to do with the DSP processing itself. The one exception I can think of is lower sample rate processing and filtering or EQ above about 10k Hz. This is because the Nyquist limit influences the high-frequency behavior of IIR DSP filters even with pre-warping. For this reason I like to use a 96k Hz sample rate when I can.
Last edited:
Briguy, why would Hypex bother doing something like that: their core business is amplifier.
If you want dedicated dsp the market is already avorflooded with dedicated brands for many years ( pro use this kind of toys for something like 30yaers...).
If you don't want something 'integrated' and like to change stuff on a regular basis forget about a loudspeaker management system ( dsp are categorised under this name in proworld) and use a computer with a good soundcard and converters: it's the most flexible and open to evolution choice you can make imo.
About technology used: don't be worried it works, it's 30 years old tech! The way you implement it is much more a concern.
If you want dedicated dsp the market is already avorflooded with dedicated brands for many years ( pro use this kind of toys for something like 30yaers...).
If you don't want something 'integrated' and like to change stuff on a regular basis forget about a loudspeaker management system ( dsp are categorised under this name in proworld) and use a computer with a good soundcard and converters: it's the most flexible and open to evolution choice you can make imo.
About technology used: don't be worried it works, it's 30 years old tech! The way you implement it is much more a concern.
All of these points that you elaborate are only inherent to hardware DSP. Try software DSP instead! A $250 mini PC, some free software, and a good quality DAC is better and more flexible than most hardware DSP processors (apart from the increased latency).
Here are a few pictures of my Hypex install on a pair of Magneplanar 3.7i speakers. No place to put them in a planar speaker. I built boxes for the plate amps and just connected banana plugs to the hypex wires. After some mods to the Maggies, they just plug in. So with separate enclosures these amps can be easily moved to another speaker if desired.
I don't find latency to be different between a computer and my dsp. Of course i don't use consumer gear but i hope even this kind of digital gear is now definitely not plagued anymore by latency!
But still if you want something stable i would look at pro gear: it's the first and foremost parameter, stability ( if it isn't you can't work so no money!)..
But still if you want something stable i would look at pro gear: it's the first and foremost parameter, stability ( if it isn't you can't work so no money!)..
Clever solution! I bet those sound killer. I might have considered that, except that I already bought a multichannel amp. Anyways, I love the look of a stereo rack, so I'm content this way anyways
My favorite speakers I've ever owned have been the Magnepan .7's, except for maybe my Selah Tanzanites that i own now (Rick Craig). Absolutely stunning sound. Just lacked a bit of bass and highs, and the dynamics left a bit to be desired. But I loved the wall of sound, and the clarity was the best I've ever heard in a speaker. Stock 3.7i's were my dream speaker, before I fell down this whole diy rabbit hole haha
How do they sound? Specifically curious about how you like the dynamics, as that was my biggest complaint about the .7's
I also felt that EVERYTHING sounded huge. Tubas sounded huge, as they should have, Hammond organ sounded huge, as it should have, but mandolin also sounded huge.. which I wasn't as much a fan of
I’m astonished that no one has so far talked about AudioLens or Acourate which are just in another league to any hardware based DSP solution you care to mention. Don’t take my word for it, read MitchCo’s articles on the subject.
But Accourate isn't only dsp oriented, it implement some room correction targeted function too.
Audiolense was a bit less evolved about this point last time i took a look at it, it might have changed since idk?
It depends what you want to do too, 'room correction' is another thing in my view.
And i would not say way ahead of hardware dsp: Qsc Core are just monster modular dsp tool. It depend of your needs and will imho.
And cost are not the same too: a second hand hardware ( even a Qsc Core) is way less than Acourate software. There is freeware availlable which does more or less same things for room correction ( but less user friendly imho, price of being freeware).
But yes, Mitchba's ebook is a must read, as well as his review of Audiolense! Game changer for me.
And Mitchba's Hangloose Convolver is worth taking a look at, very clever software which solve a lot of issue one could face implementing dsp through computer ( and fairly priced given what it does).
Audiolense was a bit less evolved about this point last time i took a look at it, it might have changed since idk?
It depends what you want to do too, 'room correction' is another thing in my view.
And i would not say way ahead of hardware dsp: Qsc Core are just monster modular dsp tool. It depend of your needs and will imho.
And cost are not the same too: a second hand hardware ( even a Qsc Core) is way less than Acourate software. There is freeware availlable which does more or less same things for room correction ( but less user friendly imho, price of being freeware).
But yes, Mitchba's ebook is a must read, as well as his review of Audiolense! Game changer for me.
And Mitchba's Hangloose Convolver is worth taking a look at, very clever software which solve a lot of issue one could face implementing dsp through computer ( and fairly priced given what it does).
Last edited: