As a matter of interest, here are the noise measurements of the miniDSP 4x10HD analog in/analog out. I can't see the problem IF you have your gain structure correct. For any unit, get the gain structure wrong and you have problems. Yes I'd still prefer analog volume on the output, but I can add that.
Review: MiniDSP 4x10HD – Neurochrome
Note I have no issues with people chosing the X-over tool they are happiest with, be that passive, line level passive, active, DSP etc. And the key point as many have made is that you need to know what you are doing. DSP lets you **** it up much faster if you don't
Review: MiniDSP 4x10HD – Neurochrome
Note I have no issues with people chosing the X-over tool they are happiest with, be that passive, line level passive, active, DSP etc. And the key point as many have made is that you need to know what you are doing. DSP lets you **** it up much faster if you don't
How many people still think that making a speaker with LR4 is picking that named crossover from their favorite DSP tool drop down box and be done with it.
Or for the analog variant filling in the 4 ohm or 8 ohm in an online crossover calculator and getting those suggested values in inductors and capacitors and really believe they've done a good job.
It isn't until we look at the acoustical slopes of the drivers involved that we get any closer to an ideal crossover. Still lots that can go wrong because we'd need good measurements too. Be it with DSP or passive, IIR or FIR.
The problems mostly aren't the tools or the means, but the believe they've done a good job and somehow getting a horrible result. Don't blame the tools, check your work and try again!
There are pitfalls for every way to cross our drivers, be it passive or active. Indeed with active it is even easier to get it wrong. Especially with those convenient named crossovers available. It makes a large group confident they have a LR4 crossover or whatever named crossover they picked from a list. In passive solutions things aren't much better. How many people claim to have tried a first order crossover because they only used a single capacitor? Getting it right isn't an easy task, certainly not a passive first order crossover!
So in order to have a valid comparison between any combination, IIR vs FIR or passive vs active... there will be a long list of things to check to find out if it is even remotely comparable or valid as a true comparison.
After making sure of all that, then we can talk about the differences in processors, EQ etc. or the brands of capacitors and types of inductors and resistors...
It never ends...
Or for the analog variant filling in the 4 ohm or 8 ohm in an online crossover calculator and getting those suggested values in inductors and capacitors and really believe they've done a good job.
It isn't until we look at the acoustical slopes of the drivers involved that we get any closer to an ideal crossover. Still lots that can go wrong because we'd need good measurements too. Be it with DSP or passive, IIR or FIR.
The problems mostly aren't the tools or the means, but the believe they've done a good job and somehow getting a horrible result. Don't blame the tools, check your work and try again!
There are pitfalls for every way to cross our drivers, be it passive or active. Indeed with active it is even easier to get it wrong. Especially with those convenient named crossovers available. It makes a large group confident they have a LR4 crossover or whatever named crossover they picked from a list. In passive solutions things aren't much better. How many people claim to have tried a first order crossover because they only used a single capacitor? Getting it right isn't an easy task, certainly not a passive first order crossover!
So in order to have a valid comparison between any combination, IIR vs FIR or passive vs active... there will be a long list of things to check to find out if it is even remotely comparable or valid as a true comparison.
After making sure of all that, then we can talk about the differences in processors, EQ etc. or the brands of capacitors and types of inductors and resistors...
It never ends...
That is very true, and certainly with dsp, it's easier to do that. Good dsp crossovers can sound very good, but it's as hard as doing it analog (be it active or passive). And both should be based on the specs of the driver, not random on theoretical formula's.
Btw, my first order serial passive crossover for my main set has 9 components (per side) in it... that is because it's studied and based on the real response and impendance graphs that are measured in the box, not just a textbook crossover. And that is the way to do it.
Btw, my first order serial passive crossover for my main set has 9 components (per side) in it... that is because it's studied and based on the real response and impendance graphs that are measured in the box, not just a textbook crossover. And that is the way to do it.
Last edited by a moderator:
Those days...
I have a couple of Echo interfaces + breakout boxes laying around, later generation PCI-cards.
Close to 17 years old, still very good though.
This is another classic from the Aardvark era:
Attachments
Last edited:
Oh, where do you get dsp processing is an issue?
PC's, even a rasPi, can handle all the taps we can eat.
I don't think I have an issue with my reference FIR. I use Equilibrium, which is one of the recording / mastering standard.
Anyway, I can clearly hear the difference between 60,000 to 200,000 in my room even with ABX test. I think something wrong with your theory. A Chord engineer also claims 164,000 taps is not enough for their converter. I actually use the same FIR for something like brick wall in addition to active crossover. Do you think is this the reason?
I really wish 6,000 is enough for me, because extremely long tap causes hefty delay and uses up processing power.
Last edited:
Completely agree with you Wesayso,
It's alot of work and measurement to achieve true LR4 acoustic xovers.
There's only two ways to do it that i know of, and both require driver by driver tuning.
First is the most often used method i guess, start with the raw acoustic driver response and keep nudging it with filters till you measure an LR4 acoustic rolloff out of the driver.
Then and only then, can you put the drivers together for LR4 summation.
Which should need no more than lining up acoustic centers' timing with delays, along with matching levels and polarities.
If if takes further adjustments than those 3 things, it says you didn't achieve an LR4 acoustic response on one driver or another or both, ...ime.
A lot of trial and error like that.
And then when you want to compare to a different xover order, you have to go through the entire driver by driver nudge process again to match the new acoustic target order.
And then reassemble the full speaker again, which at least should be able to use the previously found delays and levels .. if not, something is wrong. (Polarities may need changing though.)
How many folks go thru all that? I wonder bigtime....
(There is an easier, second way to achieve LR4, one that I've nearly preached about many times that works with either IIR or FIR.....but i'll stay off that soap box today
)So i can climb on a different soap box with my next post or so haha
Specs of the driver(s) are a determining factor, another is the complexity of the system.
To quote Dieter Rams:
“less but better”.
nigelwright7557 sticks to full range drivers.
While some of these do sound pretty good, linearity, max. output and extension at both ends of the range are often severely compromised.
The best ones aren't exactly cheap either.
So, what comes next?
Technically a 3-way, due to the BMS Coaxial, but with a suitable compression driver, the system can also work in 2-way config.
It's the pseudo DIY loudspeaker system that stole the show, 8 years ago in my country. An example of a fairly straightforward (digital) active concept with moderate use of DSP (i.e. horns/drivers are time aligned without DSP).
It's basically a tribute to the old WE-Altec/RCA cinema systems. A G.I.P. descendant of these classics was also demoed. The tube powered, passive G.I.P. is at least 15x more expensive, quite complicated to setup, quite good sounding, but no match for the DIY system.
Attachments
Last edited:
Hi Plasnu, i don't want to possibly offend or go down a rabbit hole that may have flames at the bottom.
In outspoken honestly, just as i've come to recognize how much audiophoolery there in home audio, it appears there is just as much studiophoolery in the studio marketplace.
I can only believe in credible engineering nowadays, backed up by listening.
And some common sense that recognizes there are so many relatively insignificant things we tend to stress about or think makes a difference just because we can ABX them.
In outspoken honestly, just as i've come to recognize how much audiophoolery there in home audio, it appears there is just as much studiophoolery in the studio marketplace.
I can only believe in credible engineering nowadays, backed up by listening.
And some common sense that recognizes there are so many relatively insignificant things we tend to stress about or think makes a difference just because we can ABX them.
Just look at the number of threads about 'cables' on gearslutz.
Thanks for your reply btw!
So for a 2-way with an asymmetric / hybrid crossover FIR is preferable, or is this a borderline case?
FIR linear phase filters work better as complementary that way most of the ringing is cancelled out. That doesn't work when the slopes are asymmetric.
I would say if your slopes are 24dB or less the potential benefits of a linear phase crossover are quite small. Every use case is different, to me it's more a matter of how steep you want your crossover slopes to be rather than how many ways the speaker is that would determine if an FIR is warranted, or if you are chasing a perfect impulse response.
Linear phase crossovers can be very easy to implement. Flatten the drivers for two octaves either side with PEQ, time align the drivers, bang a complementary LP crossover of whatever type in, done. The same thing can be done with an LR4 IIR crossover, that was the Linkwitz way and it works well too if the drivers have enough response outside their pass bands.
I would say if your slopes are 24dB or less the potential benefits of a linear phase crossover are quite small. Every use case is different, to me it's more a matter of how steep you want your crossover slopes to be rather than how many ways the speaker is that would determine if an FIR is warranted, or if you are chasing a perfect impulse response.
Linear phase crossovers can be very easy to implement. Flatten the drivers for two octaves either side with PEQ, time align the drivers, bang a complementary LP crossover of whatever type in, done. The same thing can be done with an LR4 IIR crossover, that was the Linkwitz way and it works well too if the drivers have enough response outside their pass bands.
I've always found the SL approach interesting ( and it ended up with some fearsomely complex analog crossovers by some metrics). It couldn't be further from the 'norm' of working with the natural acoustic slopes of the drivers but I couldn't say there was anything wrong with it, just a hardcore EE approach to the problem.
And brick wall xo you can cross where it end just if it sound good, above 90db/oct.
Vertical responses might have just a small 2db error on xo point
Last edited:
What sampling rate is that filter using as it makes a big difference to the relative number of taps. At 44.1 to 48K, those are long filters at 192K not so long.
164,000 taps for an anti aliasing filter isn't enough, hmm, really hard to separate the marketing BS from engineering reality with some companies.
A brick wall at low frequencies could well need a lot of taps to not cause unwanted issues elsewhere.
This is why throwing out, more taps is better causes confusion and may be met with opposition.
Off topic:
Roland,
Emagic! When Logic was the most advanced daw for being creative with hardware synth ( modular approach to midi sequencing was very powerful). Bought it for pc just before they announce they'll quit the platform.
Their hardware was good, i liked their midi interface a lot as they was almost immune to note stolen and crash from midi real time controler stream. Audiowerk was good soundcard for the time.
Plasnu, yes the Aardvark clock was very good: i once was part of a demo of it with digidesign 888 and sony pcm3348hr. Once clocked almost zero difference between both converters! Impressive given the horror 888 was, nice real upgrade for a ProTools. It was standard in all rooms i've been in Paris circa 2000.
I use multiple Aardvark soundcard: aark 20/20 for the pc i talked about and 4xQ10 with a dedicated silent pc for remote recording: 32in/out with 1,5ms latency at 96/24 including 16 phantom preamp and 16 preamp for dynamic mic/ line or DI input ( all preamps sound good despite being semipro and the DI are great). Direct monitor on out, wordclock input...All this in a 8 unit rack. It is bulky in comparaison to what is availlable atm but it is easy to operate and move ( easier to move than my 02r anyway). Very nice recorder with powerful editing ( it run older Sequoia or Nuendo) and almost as flexible for routing than the 02r.
Roland,
Emagic! When Logic was the most advanced daw for being creative with hardware synth ( modular approach to midi sequencing was very powerful). Bought it for pc just before they announce they'll quit the platform.
Their hardware was good, i liked their midi interface a lot as they was almost immune to note stolen and crash from midi real time controler stream. Audiowerk was good soundcard for the time.
Plasnu, yes the Aardvark clock was very good: i once was part of a demo of it with digidesign 888 and sony pcm3348hr. Once clocked almost zero difference between both converters! Impressive given the horror 888 was, nice real upgrade for a ProTools. It was standard in all rooms i've been in Paris circa 2000.
I use multiple Aardvark soundcard: aark 20/20 for the pc i talked about and 4xQ10 with a dedicated silent pc for remote recording: 32in/out with 1,5ms latency at 96/24 including 16 phantom preamp and 16 preamp for dynamic mic/ line or DI input ( all preamps sound good despite being semipro and the DI are great). Direct monitor on out, wordclock input...All this in a 8 unit rack. It is bulky in comparaison to what is availlable atm but it is easy to operate and move ( easier to move than my 02r anyway). Very nice recorder with powerful editing ( it run older Sequoia or Nuendo) and almost as flexible for routing than the 02r.
Yes they were pretty complex, I know I built one
Very few drivers these days have desirable natural acoustic slopes to work with, most are breaking up badly by the time they roll off. Yes this is like what Mark does except steeper, IIR does not make any sense for this.
Thanks, that's what I guessed.
I prefer not to use drivers with less then 1 octave usable response outside the intended pass band.
This especially applies to cone drivers, unless there's a well behaved roll off - a rarity.
Last edited: