• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DSP is bad!

See digital technology and dsp changed audio so much for better that i can only feel sad for people like that.
 
levimax said:
DSP is a tool which can be used or abused.

Reading some of jj's work on room correction I tried using "ERB Smoothing" on the measurements before I created the correction filters in REW, previously I had used "Var smoothing". Of course this resulted in a much lighter and smoother correction filter which measured much less smooth than either the manual filters I created using Var smoothing in REW or what ever filters DIRAC DLBC cooks up. DIRAC by far measured the "smoothest" in room response but I preferred the "ERB smoothed" correction filters.

While it is now possible for almost anyone to get a MIC and make measurements of their system before and after correction I am not sure that is a great thing as it pushes companies like DIRAC and DIY filter makes to try to hammer their in room response as flat as possible which while it may look nice on the measurement graph may be suboptimal to listen to. Human's ability to "hear through the room" as well as humans hearing limitations, both of which are taken into account to some extent by ERB smoothing, needs to be considered when creating DSP filters as more is not always better.
Click to expand...

Great post...or at least one I find myself in full agreement with ! :)

And good to see you trying different degrees of correction filters. That has been an area of major focus for me when using FIR, for quite a while.
One piece of software I use for FIR generation allows smoothing the original measurement like you describe. And then it also allows different degrees of smoothing the FIR filter applies, to user specified bands segmenting the filter's passband.
Another one does all that, and adds in a host of windowing options

I learned in the last year, to look more closely at the filters impulse response....the electrical response. I learned the cleaner it looks, the less jagged, less oscillatory, and still provide the desired acoustic transfer function....well, simply said...the better it sounds !

In analyzing filters' impulse responses, I've been finding that the degree of smoothing applied in attempting to clean up impulse, has been making less of a difference than the length of the FIR filter. It helps to use the shortest FIR filter than gets the acoustic job done, ime.

I think for muti-way speaker setup, this begs for each driver section having its own specific FIR filter length that gets progressively shorter as frequency increases.
For global FIR across the entire speaker, that's not possible.

So for global, I think a frequency dependent FIR filter length is needed, one that applies progressively fewer taps as frequency increases..
A great big long FIR filter able to handle the bottom end, applied across the entire speaker, has bad juju potential, ime/imo.

This is my real reason for posting now....to highlight what I think is a root problem with many (most) of the DSP applications we are seeing. Either DIY, or automated commercial.

Strongly encourage folks to make transfer functions and grab impulse responses, of the electrical filters being applied. It amazes me how different they can be, and still give acoustic measurement results that rather look the same.
 
DSP is obviously very good in the right hands. Not sure what else you would deploy at reasonable pricing to help with low end?
 
gnarly said:
In analyzing filters' impulse responses, I've been finding that the degree of smoothing applied in attempting to clean up impulse, has been making less of a difference than the length of the FIR filter. It helps to use the shortest FIR filter than gets the acoustic job done, ime.
Click to expand...
Very interesting.... I was always worried about getting enough taps for LF accuracy and never thought that more taps had a downside beyond processing power limitation and latency.

Beyond the impulse responses visual differences you said you have been able to hear a difference in the sound of a "long FIR filter" vs a "short FIR filter"? What does it sound like?

Edit: I see @gnarly already said he preferred shorter filters.... so changed to ask what the differences were that he hears.
 
Last edited:
dsp = desirable sound pleasure , audio behringer DCX 2496
 
Last edited:
If a manufacture has neither the knowledge nor the inclination to implement modern design it is just easier to disparage it.
Keith
 
He has a Phd? But he dosnt know basic EE. "The idea is straightforward: compensate the electrical delays introduced by inductors" No such thing.
"He argues that preserving the timing relationships among those impulses matters more than chasing a ruler-flat frequency response." Time domain and freq. domain are interchangeable. Fundemental to signal processing, analog and digital.
 
The recordings are heavily processed with DSP in ways that affect phase, rooms and room reflections affect phase, physical misalignment of drivers and passive crossovers affect phase! And he's telling us that DSP which uses FIR and/or mixed phase filters are the chief murder suspect?

So much better to listen to those 10 db standing waves at 50 and 150 hz in my room than use Diract to correct them, not to mention the group delay on my SVS subs. Right!
 
phoenixdogfan said:
So much better to listen to those 10 db standing waves at 50 and 150 hz in my room than use Diract to correct them, not to mention the group delay on my SVS subs. Right!
Click to expand...
Minimum phase filters to tame standing wave peaks is one of the best uses of DSP and almost always results in a preferred result.

Trying to use DSP to correct group delay in a small sealed already heavily DSP'd sub like the SVS (not minimum phase) not so much and may do more hard than good.
 
levimax said:
Very interesting.... I was always worried about getting enough taps for LF accuracy and never thought that more taps had a downside beyond processing power limitation and latency.

Beyond the impulse responses visual differences you said you have been able to hear a difference in the sound of a "long FIR filter" vs a "short FIR filter"? What does it sound like?
Click to expand...

Yeah, me too. I worked forever to get more taps for low end work. Rather humbling to find out that they aren't really needed, when using more optimal low end techniques.

I'd describe the sound as a little clearer, with more distinct separation between multiple vocalists, and instruments more defined. It's like the background is quieter for everything to heard against.
I kinda theorize that impulse oscillations are some unrecognized form of low level noise or maybe time smearing. Dunno, but do know I hear things more clearly.

Besides minimizing the FIR filter length, the other equally important task for cleaning up FIR filters electrical impulse, seems to be the selection of the crossover frequency and its order.
Out of pass band EQ work that the filter needs, to match the target acoustic crossover, increases impulse oscillation quite a bit. It's cleaner to just use a higher order crossover, to avoid needing out of band EQs imbedded into the FIR filter. However too high an order lengthens the FIR filter, and the oscillations. Right now, if I had to choose one order as optimal for any xover anywhere, I'd probably pick 60dB/oct. Anyway, I digressed for those interested in muti-way DIY speaker building....

One thing I do strongly believe in now, is to not only look at acoustic measurements like we do, but also look at the electrical filters measurements. Regular impulse, ETC, and step, all give a picture of what's being applied in time.
Worth doing for straight minimum phase work too I think, as post ringing can also be minimized.
 
gnarly said:
Yeah, me too. I worked forever to get more taps for low end work. Rather humbling to find out that they aren't really needed, when using more optimal low end techniques.

I'd describe the sound as a little clearer, with more distinct separation between multiple vocalists, and instruments more defined. It's like the background is quieter for everything to heard against.
I kinda theorize that impulse oscillations are some unrecognized form of low level noise or maybe time smearing. Dunno, but do know I hear things more clearly.

Besides minimizing the FIR filter length, the other equally important task for cleaning up FIR filters electrical impulse, seems to be the selection of the crossover frequency and its order.
Out of pass band EQ work that the filter needs, to match the target acoustic crossover, increases impulse oscillation quite a bit. It's cleaner to just use a higher order crossover, to avoid needing out of band EQs imbedded into the FIR filter. However too high an order lengthens the FIR filter, and the oscillations. Right now, if I had to choose one order as optimal for any xover anywhere, I'd probably pick 60dB/oct. Anyway, I digressed for those interested in muti-way DIY speaker building....

One thing I do strongly believe in now, is to not only look at acoustic measurements like we do, but also look at the electrical filters measurements. Regular impulse, ETC, and step, all give a picture of what's being applied in time.
Worth doing for straight minimum phase work too I think, as post ringing can also be minimized.
Click to expand...
Oh no a new rabbit hole to go down :)

See below, first graph is step response for my "old" correction filter generated in REW using VAR smoothing and 96 KHz to just above Schroeder. (I used to upsample everything to 96 KHz so I could use the same FIR room correction filter for all content but stopped that after I realized the resampling was causing digital clipping issues).

The second graph is my "new" light handed correction filter generated by REW from the same measurement and same target but using ERB smoothing and 44.1 KHz. I definitely prefer listening to the "light handed" correction even though if I measure the corrected response at the listening position it is considerably "less smooth" looking.

I am not experienced in correlating "looking at a step response and correlating it to what I hear" but just looking at the graph of my "old" filter it looks scary. I am starting to think "over doing" DSP is a very real and common thing.

FIR VAR Smoothing 96 KHz.png


FIR ERB Smoothing 41 KHz.png
 
Last edited:
I nearly lost my mind looking into thousands of graphs and figuring out what that really means. Not that difficult for 2 channel, but once you ramp that up to like 15 channels and e.g 4 subs it becomes a torture and fruitless exercise.

Settled for the decent FQ response and lowest decay time, and disregarded the rest. Happy ever since...
 
levimax said:
Oh no a new rabbit hole to go down :)

See below, first graph is step response for my "old" correction filter generated in REW using VAR smoothing and 96 KHz to just above Schroeder. (I used to upsample everything to 96 KHz so I could use the same FIR room correction filter for all content but stopped that after I realized the resampling was causing digital clipping issues).

The second graph is my "new" light handed correction filter generated by REW from the same measurement and same target but using ERB smoothing and 44.1 KHz. I definitely prefer listening to the "light handed" correction even though if I measure the corrected response at the listening position it is considerably "less smooth" looking.

I am not experienced in correlating "looking at a step response and correlating it to what I hear" but just looking at the graph of my "old" filter it looks scary. I am starting to think "over doing" DSP is a very real and common thing.

View attachment 477903

View attachment 477904
Click to expand...
Why are you using % instead of db? So your saying the filter with all the overshoot is better? It dosnt look as acurate.
 
Anyone else feel DSP is an overpriced graphic equalizer ? And similar results can be had with one. Obviously you cant do so much like speciality low/high pass filters etc, but for generally tuning speakers and to a room or preference of sound aren’t they much of a muchness.
 
Westsounds said:
Anyone else feel DSP is an overpriced graphic equalizer ? And similar results can be had with one. Obviously you cant do so much like speciality low/high pass filters etc, but for generally tuning speakers and to a room or preference of sound aren’t they much of a muchness.
Click to expand...
They add maybe $100 to the price of an amp. How could that be called overpriced?
 
Cbdb2 said:
Why are you using % instead of db? So your saying the filter with all the overshoot is better? It dosnt look as acurate.
Click to expand...
In REW under the impulse tab I guess % is default? If it isn't I changed it on accident. I prefer listening to the second filter which uses much smaller corrections due to ERB smoothing before creating the correction. What I don't like about looking at the graph of the first filter is how long it rings. As I mentioned I don't have a feel for looking at a step response and correlating it to what I hear.
 
You must log in or register to reply here.

Similar threads

J
  • Locked
Why Cables CAN Sound Different: NOW with Proof?
Replies
8
Views
1K
RickS
RickS
digitalfrost
  • Locked
Top Audio Engineers Admit Ignoring Hi-Res Streaming Specs and Mastering 2x Louder Than Recommended
Replies
6
Views
1K
DavidMcRoy
DavidMcRoy
J
NDH 30 and LCD-X: soundstage? linearity? EQing? stereo imaging blindness?
Replies
1
Views
2K
DVDdoug
DVDdoug
Back
Top Bottom