Agree mostly. There's nothing to let go.. passive is IIR. I don't do anything different between passive and active, except more soldering.
I've found acoustic issues that are hard to track down, give themselves away partially in the timing. Taking sound output and converting to FIR arbitrarily, is like pressing 'reset'.
I can't say I feel the same way. The hardest part is knowing/learning what you want it to do. But then again, my road is rather different than most.
As long as you are willing to spend the time, there's lots of options out there.
Just don't try to fix a bad room or badly designed speakers with DSP. Its not a miracle cure but it does have a lot of potential. It often gets a bad rep from mis-use or a lack of understanding. It is a tool, and like most tools you do need to learn how to use it.
Personally, I'd never use it to replace a passive crosover and stop there. That would leave a lot of potential unused.
As long as you are willing to spend the time, there's lots of options out there.
Just don't try to fix a bad room or badly designed speakers with DSP. Its not a miracle cure but it does have a lot of potential. It often gets a bad rep from mis-use or a lack of understanding. It is a tool, and like most tools you do need to learn how to use it.
Personally, I'd never use it to replace a passive crosover and stop there. That would leave a lot of potential unused.
I'm inclined to think you're speculating there. I can equal a passive crossover with active but not improve on one. For me, the acoustics come first.
I might have phrased the question a bit "confusing". What I was trying to ask is that with miniDSP, there is a FIR filter for the tweeter, one for the woofer, and EQ? How does EQ fit in that picture? It is probably separate from the filter?
Let's say now I have finished programmed the FIR filters for the tweeter and woofer. Now I want to do a bit of EQ, for example the low frequencies. How do I load the EQ curve, obtained from REW, into the miniDSP module?
Last edited:
Hi AllenB,
+10 to Wesayso's posts..
And agree with you on one supreme point...acoustics always come first 🙂
Past that, I can't agree at all with what I perceive is your take on IIR vs FIR...
You appear to think IIR can capture acoustic nuances that FIR can't.
Pls correct me if i am misunderstanding you....
Because IIR capturing more 'acoustic nuances' than FIR is simply impossible...
IIR filters are a completely contained subset of filters, within FIR filter overall capability...period, amen.
That says it all imo.
So any and all IIR acoustical nuances found using IIR, are 100% available with FIR, and with FIR offering even more 'acoustic nuance' capturing capability.
I really should ask, what do you think FIR is??
Maybe I'm missing your point??
Ok, now I'm sure there's a miscommunication somewhere.
Do we agree on what IIR means?
References please...
There is a huge difference between short and long tap FIR in sound quality.
I would choose IIR over short FIR, especially for low frequency.
I would choose IIR over short FIR, especially for low frequency.
DSP: MiniDSP 4x10
Amps: 2x stereo F5 for mids/highs and 2x mono NC400 for lows
Speakers: 3 way with sealed 10" SEAS woofer, 6" B&C mid. 1" SEAS tweeter
Speakers were acquired from the designer of a now out of production brand. The original design was line level analog based. The specific crossover details were provided.
Using miniDSP advanced options I was able to look at the biquad functions that each crossover slope represented. I learned to build the filters by stacking and compiling these biquads until I got a reasonable approximation of how I wanted the response to look.
Initially I used TrueRTA and an ECM8000 to check the response live with pink noise running through the system.
Recently I tried the miniDSP UMK-1 mic and set it up with REW. I ran several sweeps and dialed in the response. I use the available PEQs to make adjustments.
I also only play with things about once a year to make sure I don't wind up with a shifting reference. Over the three years I've been refining this setup I have been extremely happy with the sound.
There is a tutorial on the minidsp website that explains the integration with files generated in REW.
Auto-EQ tuning with REW
Amps: 2x stereo F5 for mids/highs and 2x mono NC400 for lows
Speakers: 3 way with sealed 10" SEAS woofer, 6" B&C mid. 1" SEAS tweeter
Speakers were acquired from the designer of a now out of production brand. The original design was line level analog based. The specific crossover details were provided.
Using miniDSP advanced options I was able to look at the biquad functions that each crossover slope represented. I learned to build the filters by stacking and compiling these biquads until I got a reasonable approximation of how I wanted the response to look.
Initially I used TrueRTA and an ECM8000 to check the response live with pink noise running through the system.
Recently I tried the miniDSP UMK-1 mic and set it up with REW. I ran several sweeps and dialed in the response. I use the available PEQs to make adjustments.
I also only play with things about once a year to make sure I don't wind up with a shifting reference. Over the three years I've been refining this setup I have been extremely happy with the sound.
There is a tutorial on the minidsp website that explains the integration with files generated in REW.
Auto-EQ tuning with REW
Analog Devices ADAU-1452 Eval board with on board ADC and DAC with I2S/Toslink in/out Sigma Studio suite for programming. PC USB DAC is SMSL Sanskrit 10 with AK4490. Amps include Alpha Nirvana (40W) Aspen Glass Harmony (50w) SE Class A for fullrange/tweeter, custom TPA3255 amp with PFFB for midbass/woofer. Sometimes Dayton APA150 for woofer. Speaker varies... UMIK-1 with Cross Spectrum Labs calibration. REW for measurement.
But I love the pure clean sound and the simplicity of passive crossovers as they only need 1 amp.
But I love the pure clean sound and the simplicity of passive crossovers as they only need 1 amp.
Last edited:
Good morning, andy2!
I read new posts and I got the impression that You haven't studied how minidsp dsp softwarelooks and works. 2x4 and 2x8 (for 4x10HD) multiway versions are basically similar. They have graphical display that show the effect of your settings immediately. But you cannot download your own measurement (like with Hypex HFD). That's why one must do lots of measurements after changing settings.
Tutorials with images of Minidsp software and REW
Stereo 2 Way Xover
miniDSP tutorials
With Hypex I found it easier to download straight response for start (comes with download package). Minidsp software has that as default and user cannot change it. Overall, HFD is much more difficult to learn than minidsp.
I read new posts and I got the impression that You haven't studied how minidsp dsp softwarelooks and works. 2x4 and 2x8 (for 4x10HD) multiway versions are basically similar. They have graphical display that show the effect of your settings immediately. But you cannot download your own measurement (like with Hypex HFD). That's why one must do lots of measurements after changing settings.
Tutorials with images of Minidsp software and REW
Stereo 2 Way Xover
miniDSP tutorials
With Hypex I found it easier to download straight response for start (comes with download package). Minidsp software has that as default and user cannot change it. Overall, HFD is much more difficult to learn than minidsp.
Last edited:
Which is why I said, you've got to learn what it is you want from it...
Just copying an IIR crossover in an active setup will get you just that... might as well stick to passive if you do. But that does not mean we don't listen when using active.
I could never have gotten where I am with passive means only. Your description:
says it all... You've got to be willing to spend way more time, letting go of passive knowledge, and learning what does what so you can shape the response and know/learn what to expect.
Learn what your ears pick up, learn what not to touch. It takes time, but it's well worth it. Indeed, it can be hard to figure out, but not impossible.
You want examples? Just go through my thread... It isn't about pretty graphs, it is about how those pretty graphs can sound. Just hammering a response into shape gets you nowhere. I see people do something like that most of the time and then wondering: why doesn't this sound any better? The first post in my thread contains links to reviews by others than me, as I am biased. Were they just being polite? 🙂
I use multiple forms of DSP to better the sound in my room. Hiding it's real properties, room treatment is involved. Shaping the sound to please the ears. Most of my time was spent learning what I wanted. (yes, I've read the theories of Toole, Geddes, Danley, Dunlavy etc... to get an idea of what it is I wanted. Add some Griesinger in the mix for room sound theory and now we're getting somewhere useful, many more can be added to this list)
I'm not breaking the laws of physics and I'm playing with psychoacoustics. That's why I keep saying: You've got to learn what it is you want the processing to do. And learn what isn't possible, what the limits are. But there are less limits in FIR processing. And lots of ways for it to go wrong. Which is why it does take time and holds no real relation anymore to passive theory. Let go of that. Even though I do use passive components (lol).
Tap count that I use: 65536. Plenty of resolution. But first the room + speaker must be able to work together. Plan it that way.
One more thing: it sounds pretty great in more than just the one sweet spot.
Last edited:
We agree on many things.
One point, we need (in this context) to stop saying passive or active.. it's IIR or FIR. Anything that you can do IIR digital active, you can also do passive because they are the same. I understand that some feel it is much harder, but the fact remains.
One point, we need (in this context) to stop saying passive or active.. it's IIR or FIR. Anything that you can do IIR digital active, you can also do passive because they are the same. I understand that some feel it is much harder, but the fact remains.
I can agree with that. 🙂
While the method is slightly different, the same rules apply. I'll refrain from further elaboration about FIR, unless the OP is interested in that.
There are many variants of Active, all have their place and purposes.
While the method is slightly different, the same rules apply. I'll refrain from further elaboration about FIR, unless the OP is interested in that.
There are many variants of Active, all have their place and purposes.
Cool...keep studying the miniDSP tutorials like Juhazi said...which i can tell you are doing as your questions keep getting more refined 🙂
Yes, make a separate FIR filter for each driver, where the goal is flat mag and phase and good summation.
To bring further EQ into the equation there are two main paths.
One is to have a completely separate EQ that spans the signal before the signal is split to the drivers.
The second is to embed the EQ or house curve etc into each of the FIR files for the two drivers. Which is what I think your last question gets at....
Put the house curve or low end EQ's into REW, when it generates the EQs for import into rephase.
IOW, let REW both flatten mag and build in the extra EQs, for one FIR file per driver.
If all the EQ applies to one driver only, such as low end boost, only the woofer FIR file will need the extra EQ embedded.
If it's a house curve that spans both drivers, put it in both FIR files...
Dang miscommunication issues 🙂
Sure, let's exchange some simple definitions to get on the same page...
I take IIR to mean infinite impulse response, recursive filters.
That can be implemented either passively, active analog, or active DSP.
So IIR does not on it's own imply an implementation method.
I think this jives with further posts from you ???
On the other hand FIR, finite impulse response, non recursive filters, can only be implemented via DSP.
So FIR does on its own imply DSP as the implementation method.
A FIR filter can be IIR or linear phase, or any combo thereof.
Although many automatically associate FIR with linear phase, that's not an appropriate association. Not at all.
I dare say most commercial applications of FIR tuned speakers are more about embedding IIR EQ's into the processing than achieving linear phase.
FIR implemented IIR allows an essentially unlimited number of EQ's, whereas regular IIR, analog especially, and even straight DSP IIR , has a relatively meager count.
So by extension, I don't think it's correct to draw a line between IIR and FIR.
I think it's much more correct to view IIR as a fully contained subset within FIR.
That's why for me, simple logic says FIR can only extend IIR's capability to capture acoustic nuances.
Because basically, FIR allows independent adjustment of time and frequency response, whereas IIR does not (from any practical viewpoint).
Maybe that's what also needs defining...what are acoustic nuances ?🙂
Anyway, hope this helped convey my logic and some sorting of the definitions/issues..
The misunderstandings in this thread implore just such definitions! So for the record...
1. A crossover can be active or passive, the difference being if it is placed before of after the power amplifiers. Passive crossover design is not the "same" as any form of active crossover design because as well as the (complex) frequency response target, there is also the impedance to take account of too.
2. Active crossovers can be analogue or digital, where an advantage of digital implementations is the ability to provide time delays, although traditionally all-pass filters have sufficed for analogue active crossovers.
3. Digital active filters can employ a variety of techniques that can be divided into two main categories, namely non-recursive filters that have a necessarily Finite Impulse Response and recursive filters that have an Infinite Impulse Response. Whilst analogue filters have an infinite impulse length too, they are neither FIR or IIR - whilst digital filter design techniques can emulate analogue filter design methods, there are different design constraints.
But to FIR and IIR filter differences... FIR filters are constructed exclusively from zeros, whereas IIR filters have at least one pole. An IIR filter can be split without any loss in definition into an FIR all-zero part and the all-pole recursive part, but IIR filters are certainly not a "subset" of FIR filters. The one-sided response (in time) of analogue and IIR digital filters precludes linear phase filtering that is a unique advantage of FIR digital filters (in most but not all audio applications - remember it is the summed output that is to be considered linear phase or not).
In ye olde fashioned times, where processing speed and memory where somewhat constrained (such as in the TI TMS32010), FIR filters were so limited as to cause problems. Nowadays 64K filters and the like ameliorate many such concerns. Further sub-sampling or sparse FIR filtering techniques are also available such as might be considered in acoustic applications with looonnngggggg impulse responses. But the infinite length of an IIR filter means they will always likely be more efficient in their implementation.
Ultimately, however, crossover and equalizer design is an engineering problem and leads to the optimisation of the best tools available (we hope!). The notion of "acoustic nuances" due to different filtering techniques just adds further misunderstanding to an already confused picture and to a thread that could be a useful point of reference.
I would also add (without wishing to criticise the rest of this particular contribution or move further off-thread) that none of the points in this thread are "psychoacoustic" because they are measurable and objective in their nature.
Last edited:
In my own thread , while it still is being measured and quantifiable, I'm playing with psychoacoustics. 🙂 I did point in that direction for those interested enough to find out (see my first post in this thread) without wanting to clutter up this one.
As said: I could not have done that (being the playing part with psychoacoustics) without FIR filters and/or linear phase EQ.
I couldn't have done all that with a MiniDSP either. The list of equipment used is in that first post.
It still is a thread about Active DSP, right? There are more options than miniDSP alone or sticking with IIR/FIR for crossover work only.
Last edited:
I'm using my own developed ladspa host for jack GitHub - bobo1on1/bobdsp
It has a webinterface in which I can add and remove ladspa plugins, adjust the controls live, and I can set the routing for jackd. I still plan to do a release for it, if I can find the time.
At my computer desk I have an extra desktop pc with an ASRock J4105M mainboard and Asus Essence STX II 7.1 sound card, powered by a toroid transformer and a wide input pico psu. It's connected to 3 stereo LM1875 amplifiers to drive the front woofers and tweeters, and rear speakers. There is also a stereo LM3886 amplifier to drive the subwoofer and center speaker. Audio is sent to this computer over ethernet using pulseaudio when I'm running linux, and when gaming I'm using the scream project on windows: GitHub - duncanthrax/scream: Virtual network sound card for Microsoft Windows
All filtering is done using ladspa plugins, the crossover between the front woofers and tweeters is 4th order linkwitz-riley.
The subwoofer is filtered using a 4th order linkwitz-riley lowpass, the other speakers are tuned to a Q of 0.7 and -3db point of 90 hertz, and filtered using a 2nd order butterworth at 90 hertz, which turns the speakers into 4th order linkwitz-riley highpass filters.
In my living room I've got a home theater pc with a self built dual TDA7850 amplifier built in, and two Asus Xonar D2X sound cards running from the same Crystek CCHD-575 oscillator. I have already built 3-way front speakers, and I will hook them up soon. All filtering for that setup will be done in the same way using ladspa plugins.
It has a webinterface in which I can add and remove ladspa plugins, adjust the controls live, and I can set the routing for jackd. I still plan to do a release for it, if I can find the time.
At my computer desk I have an extra desktop pc with an ASRock J4105M mainboard and Asus Essence STX II 7.1 sound card, powered by a toroid transformer and a wide input pico psu. It's connected to 3 stereo LM1875 amplifiers to drive the front woofers and tweeters, and rear speakers. There is also a stereo LM3886 amplifier to drive the subwoofer and center speaker. Audio is sent to this computer over ethernet using pulseaudio when I'm running linux, and when gaming I'm using the scream project on windows: GitHub - duncanthrax/scream: Virtual network sound card for Microsoft Windows
All filtering is done using ladspa plugins, the crossover between the front woofers and tweeters is 4th order linkwitz-riley.
The subwoofer is filtered using a 4th order linkwitz-riley lowpass, the other speakers are tuned to a Q of 0.7 and -3db point of 90 hertz, and filtered using a 2nd order butterworth at 90 hertz, which turns the speakers into 4th order linkwitz-riley highpass filters.
In my living room I've got a home theater pc with a self built dual TDA7850 amplifier built in, and two Asus Xonar D2X sound cards running from the same Crystek CCHD-575 oscillator. I have already built 3-way front speakers, and I will hook them up soon. All filtering for that setup will be done in the same way using ladspa plugins.
Attachments
Some good fundamental definitions 1-3. 🙂
And I certainly defer to your knowledge regarding the intricacies of actual filter construction...i know nothing of poles and zeros, etc....and don't want to Lol
I think it's fair to broadly classify dsp processors as either capable of IIR only, or FIR (which includes IIR capability.)
So from a pragmatic, functional point of view, it also seems fair to think of IIR filters as a subset of FIR.
That's where I'm coming from, ....all the while knowing there is plenty of technical minutia to negate a formal set/subset relationship.
OK, thank you. You may have clarified what I was looking for.
Based on your post, EQ can be applied to each individual channel, for example one for the tweeter, and one for the woofer. And of course, there is also one FIR filter for the tweeter and one FIR for the woofer.
Your post also implies that EQ can be applied to the front end signal before going into each individual filter for the tweeter/woofer.
Attachments
Last edited: