DiyMobileAudio.com Car Stereo Forum banner

1 - 20 of 46 Posts

·
Registered
Joined
·
1,912 Posts
Discussion Starter #1
I am looking for processors that have FIR filtering. The XDP-4000X I have on the way does but I am also interested in others such as Audison, Helix, Mosconi, etc. I know that PPI, Soundstream and other miniDSP based models might not since miniDSP does not have FIR filtering except for in their miniSharc which requires separate DACs. Before I spend hours researching the net today, has anyone already considered this? Thanks!
 

·
Registered
Joined
·
1,912 Posts
Discussion Starter #3
Bummer! From my research so far only the DRZ-9255 and XDP-4000X have FIR filtering and the 9255 requires workarounds to accept optical input. Interesting since phase is such a big issue in automotive environments.
 

·
Registered
Joined
·
5,098 Posts
Bummer! From my research so far only the DRZ-9255 and XDP-4000X have FIR filtering and the 9255 requires workarounds to accept optical input. Interesting since phase is such a big issue in automotive environments.
Phase is NOT an issue in a car environment because you can do so little about it. You're just not getting enough direct sound in a car. Meanwhile, response is a huge issue in a car. In a car phase is nothing more than timing, using same orders on HP/LP and some response tweaking around the xover points. To worry about phase beyond this in a car, is a waste of your time.

31 bands of GEQ per driver from a processor is more than enough. Resolution is important. A processor allowing you to eq in +/- 0.3 db is better than one that allows +/- 1db, and so on.
 

·
Registered
Joined
·
1,912 Posts
Discussion Starter #7
Phase is NOT an issue in a car environment because you can do so little about it. You're just not getting enough direct sound in a car. Meanwhile, response is a huge issue in a car. In a car phase is nothing more than timing, using same orders on HP/LP and some response tweaking around the xover points. To worry about phase beyond this in a car, is a waste of your time.

31 bands of GEQ per driver from a processor is more than enough. Resolution is important. A processor allowing you to eq in +/- 0.3 db is better than one that allows +/- 1db, and so on.
Cool, thanks for the info, that makes sense why we don't see FIR phase correct devices in cars anymore.

Funny enough, this is for a home application, hence my interest in it :)
 

·
Registered
Joined
·
385 Posts
I was under the impression that FIR was the standard digital filter design because of its advantages over IIR. I just assume most did. Serves me right for assuming.

There is this option with toslink: http://aplaudio.com/conc2/apl1/
 

·
Premium Member
Joined
·
4,235 Posts
There are also drawbacks of FIR. Getting good resolution in the bass requires either partitioning the filter or many taps. The more taps, the longer the filter, the more inherent delay and the more processing power required.

Without the partitioning, a bunch of the processing power is wasted on high frequencies where it isn't needed. For example, a 4096 tap filter would give you 19,980/4096 = 4.8Hz resolution, but half of that would be used between 10kHz and 20kHz. That's a pretty long filter. The MS-2 uses a partitioned filter with 512 taps below 1kHz and 512 above. That gives 2Hz resolution below 1k and 37Hz resolution above.
 

·
http://*****aplaudio*****/
Joined
·
123 Posts
There are also drawbacks of FIR. Getting good resolution in the bass requires either partitioning the filter or many taps. The more taps, the longer the filter, the more inherent delay and the more processing power required.

Without the partitioning, a bunch of the processing power is wasted on high frequencies where it isn't needed. For example, a 4096 tap filter would give you 19,980/4096 = 4.8Hz resolution, but half of that would be used between 10kHz and 20kHz. That's a pretty long filter. The MS-2 uses a partitioned filter with 512 taps below 1kHz and 512 above. That gives 2Hz resolution below 1k and 37Hz resolution above.
I would like to disagree with your point of view regarding FIR drawbacks. You could say that 10 years ago there was a massive processing power needed for FIR but even at that time not expansive FPGAs were available. Nowadays you do not have rights to say that FIR requires massive processing power.
The delay, introduced by FIR, completely depends on how you synthesize particular FIR.
If you use linear phase FIR – yes, you will have a delay equal to half of FIR impulse response length.
But, you will have no delay for minimum phase filter and, additionally, the phase correction of the minimum phase part of a loudspeaker transfer.
You can see this in example of correction of Quested monitors:
delay estimation and measurement

Instead of possible drawbacks, FIR filters have advantage that they do not have recursive feedbacks as IIR filters that are making seriously degradation to the processing quality.

But most important thing about FIR is that FIR is like a very sharp knife – you can do anything with it including cutting off your finger. And result of using of FIR completely depends on how accurately you prepare correction filter for it to create predistortions that will be compensated by distortions of loudspeaker, instead of creating (introducing) new distortions.
If you will try to equalize such „measured” curve, you will get into big trouble (1. attachment).

The way of measurement mic (or just mic) sound field perception shows more exact „problems” than our hearing and this circumstances causes appearance of problems that actually do not exist. This can be explained in such simple way. Let’s take loudspeaker stereo system on your desk, created of the best loudspeakers you can find. Feed same signal to both of them and listen to mono performance. Than move yourself from side to side. You will hear that mono image will move the same direction as you and nothing more. But if you put measurement mic in your position, it will show „problems” caused by interference from your two speakers. Same thing happens when you have just one speaker near the reflective surface. It means that you must measure performance of loudspeaker with consciousness of such an effect. It cannot be removed by simple smoothing of measured curve as it is in all ordinary analyzers.
Next picture shows true loudspeaker „problems” of the same loudspeaker in the same placement as in previous graph. The correction, that follows this curve, will give you excellent result (2. attachment)
 

Attachments

·
Premium Member
Joined
·
4,235 Posts
Of course, all of the above is true. One of the marketing angles for FIR is the use of linear phase filters. The second point, which could be summarized by saying, "any form of EQ is only as good as the interpretation of the measurement and the application of the correction" is true as well.

I think from a practicality standpoint, the discussion of recursive vs. nonrecursive is pretty esoteric. I've used both in the same car to arrive at the same correction results in the frequency domain and can't hear the difference.

Finally, throwing away the phase measurement and synthesizing using minimum phase is valid for correcting the response of a minimum phase system--but we don't listen to minimum phase systems. In that regard, the correction doesn't include phase correction and that feature is the one that often causes people to want FIR.

I guess my point is that all methods are valid, but the simple "one is better than the other" isn't really useful.
 

·
Registered
Joined
·
530 Posts
It was interesting comment about Alpine Imprint using FIR filter. My impression of Imprint tuning results were simply bad, despite application of almost all possible tips, which could be found on internet (and many night hours, of course). The way how APL is processing signal - rather different, but final results amazed me from the first time. I suppose that one of the advantages - a rather big number of filters (4096). I desagree with the opinion that 31 band EQ is enough in caraudio. At least not in my case. Another advantage - in measuring the power intensity response curve instead of regular RTA...
The most people reading such threads do not need to go deep in the theoretical investigations, They just need a simple DSP unit, which does not destroy the sound. APL is in this case a good choice. As confirmation to this - two first places at Eurofinals in Zalzburg, March 2014 (car of Sergey Dubinin, installer: Alexander Martyanov).

In the post above there was only one model mentioned: APL1 - for two channels. There are other models available from the same manufacturer. For example APL3 designed for six channels. This will be my next step in the ongoing investment project.
 

·
Registered
Joined
·
5,098 Posts
I disagree that 31 band EQ is enough in caraudio. At least not in my case. Another advantage - in measuring the power intensity response curve instead of regular RTA...
The most people reading such threads do not need to go deep in the theoretical investigations, They just need a simple DSP unit, which does not destroy the sound. APL is in this case a good choice. As confirmation to this - two first places at Eurofinals in Zalzburg, March 2014 (car of Sergey Dubinin, installer: Alexander Martyanov).

In the post above there was only one model mentioned: APL1 - for two channels. There are other models available from the same manufacturer. For example APL3 designed for six channels. This will be my next step in the ongoing investment project.
Dude I said 31 bands per driver. It would be great if we could have 1/6 oct cause our hearing is basically setup at that level. It would be a steep learning curve from 1/3 to 1/6 though. I hope you know that you don't need to flatten every peak that you can measure in order to get good sound and stop plugging APL whatever it is. Auto tune doesn't win championships:).
 

·
Registered
Joined
·
1,085 Posts
Joey, auto-tune also sells Pop records ( A LOT of Pop records haha..) :)
 

·
Banned
Joined
·
4,394 Posts
that there is a European product now available on a consumer market level, using FIR filters and winning in competitions, is a decidedly "good" thing.

I like being made aware of these things.

I also like seeing FIR positives and negatives discussed more, as it is a fuzzy logic area in audio, in that most people logically fuzz the details about why or why not.

If phase correction isn't an inherent feature of an FIR-based DSP engine, then I like knowing that, I was under the mistaken impression that using FIR filters were giving you an automatic positive in that it didn't mess with phase, or the phase you got in, is what you got out.

If that is not a fact, and that FIR filters **** your phase, (excuse my french) then what about IIR filters, do they just dig deeper in the phase ****ing?

and if you can't hear it, why is it so special?


that's the main thing, I guess. Somebody has to claim to hear it, like rise time in amplifiers, or slew rate on op-amps.

can you hear it?


is it better than before, is it worth another 200 bucks in DSP re-working, or can you get just as good without it?
 

·
Registered
Joined
·
299 Posts
I tested a pioneer odr processor with irr and fir filters on board in a car we did setup the car with both filter options one setting with fir filters and one without fir filters, the setting without fir filters sounded better to me! With the fir filters the sound was just if the dsp had to work to hard to get everything in line you could hear that!
The car was tuned using a measure meant system for the same curve and loudness and with the same crossover points.

Also I think in the car with the already strange phase curves off alone the speakers without a crossover with a irr filter and eq you can adjust the phase to get the different speakers better phase alined as with with a fir filter!
 

·
http://*****aplaudio*****/
Joined
·
123 Posts
Thanks, Andy,

„recursive vs. nonrecursive”
– it depends how your IIR is created. If it is using a double precision and not a simplest s-z transform, there is nothing to say about IIR (serious DAWs in studio). But DSPs in car audio head units is much different story. Alexander Martyanov (his nick is „martyanov” here and he has beautiful build log here and he is two times EMMA European champion Sergey’s engineer) told me his experience with usage of head unit equalizers and this experience is not good for that equalizers. And proximity of Nyquist frequency for FIR is nothing in opposite to IIR.

FIR equalizer may have ordinary parametric equalizer interface and may be used just as parametric.
I got a lot of compliments from pro (concert, studio) clients (I was not specially asking them) how „light” is sound of such FIR parametric. And they asked „can we get such quality into DAW plagin” : )

Most important advantage of FIR is that you have ability to use virtually any (unlimited) number of different equalizers same time but still having just a one FIR engine. For example 1) accurate loudspeaker correction equalizer, 2) parametric 1., 3) parametric 2., 4) time domain delay equalizer, 5) Hi and Low timbre for taste, 6) parametric 3. for final tweak, 7) parametric 4. for ... and so on...
And no degradation by switching on new additional EQ.

„we don't listen to minimum phase systems”

– 90% of problems in loudspeakers is minimum phase problems and can be ideally (amplitude and phase) corrected by minimum phase equalizer. Non minimum phase problems may be caused only by inaccurately tuned crossover. But let’s tune it as accurately as we can. Especially if we have possibility to „see” delays between different bands by use of TDA. And, if we are still not happy with such result, let’s use a frequency dependant delay correction. Sorry for returning to previous example of Quested monitors: delay estimation and measurement

„I hope you know that you don't need to flatten every peak”

- sorry, I did not read books and forums and now this : ), but it is true for 1. attachment picture in post #11. And it is not true for 2. attachment picture.

Just in few past years we have offers of dsp units with FIRs for car audio. But story is same as with professional BSS processors that offered FIRs much early (and it was much, much early offered by yours faithful servant). When I asked them how to use they FIRs, what to upload to them, the answer was – we don’t know.
Serious reason to use FIR is just in one case - when we have very accurate and detailed information (curve) of a loudspeaker performance. And 10 years experience shows that we can obtain this only by working in power domain (with some similarity with microwave engineering, where system parameters are distributed, not lumped and systems are described with much more poles then four, and we talk not about voltage and currant but about power – the loudspeaker is the same from a wave theory prospective)

And another serious reason to use FIR is because of FIR ability to follow any curve we like and as accurate as we like. We may spend hours of our time to try to recreate some curve by use of parametrics. And this will be just an approximation anyway, a raw approximation.

“If phase correction isn't an inherent feature of an FIR-based DSP engine, then I like knowing that, I was under the mistaken impression that using FIR filters were giving you an automatic positive in that it didn't mess with phase, or the phase you got in, is what you got out.”
- FIR is allowing to do anything with phase independently regarding amplitude. And it is completely your responsibility what to do. My first suggestion is to use minimum phase filters that are doing phase correction based of properties of minimum phase systems.

“…like rise time in amplifiers, or slew rate on op-amps.” - it is about non linear behavior (non linear distortions) of the system that is creating new additional components in signal. It is some kind opposite to linear distortions that is not creating new components but is changing relationship of existing components of the signal.
 

·
Banned
Joined
·
4,394 Posts
“If phase correction isn't an inherent feature of an FIR-based DSP engine, then I like knowing that, I was under the mistaken impression that using FIR filters were giving you an automatic positive in that it didn't mess with phase, or the phase you got in, is what you got out.”
- FIR is allowing to do anything with phase independently regarding amplitude. And it is completely your responsibility what to do. My first suggestion is to use minimum phase filters that are doing phase correction based of properties of minimum phase systems.

“…like rise time in amplifiers, or slew rate on op-amps.” - it is about non linear behavior (non linear distortions) of the system that is creating new additional components in signal. It is some kind opposite to linear distortions that is not creating new components but is changing relationship of existing components of the signal.

yes, I was unclear about my analogy.

my point is that there are some things we take for granted as having an audible effect and an accepted premise that we should be able to tell the difference between products that measure differently, but in real time we hardly can do this with accuracy.

when you say there is a big difference between FIR filters and IIR filters, I am sure people will interpret that as being a big audible difference, and I attempted to show that perhaps, it's an audible difference in the amount of significance, as slew rate might be in audible terms.


so it's just not clear whether an FIR filter is worth the amount of trouble it seems to take to implement them, and their promised superiority is not all that is claimed.
 
1 - 20 of 46 Posts
Top