REW - Room Measurements Users Discussion Thread

[playdough]

10 hours ago, OzJustin said:

One question I had is why does the ringing at 500Hz and 1100Hz show on the waterfall graph but not the spectrogram? Aren't they both showing a sound that is taking longer for the energy to decay than other frequencies?

Sometimes a constant magnitude finger coming out of the graph like that can be, a fridge, a PC fan or something in the room resonating or an actual speaker resonating. Those present here are far into the noise floor ang not an issue, but work has to be done to identify them.

 

Sometimes a measurement with no speakers should be done so as you can see the actual noise floor. 

Fans and fridges will show up in these.

The noise floor measurement above shows the din of the city in the sub regions, fridges and fans off and no dogs barking, windows closed.

Low bass is a challenge at the best of times and finding closer (multiple)_ mic position measurements can help get more of an average. The graph will change quite a lot with below 100Hz through different positions in the room due to nodes of summation and cancellation within the boundaries of the room. I tend to EQ consistent and obvious peaks/troughs across many measurements across the lounge. 

If you can't get the fundamental right, take the sub out to the yard and try measure it again, for conceptual information.

I'll be doing bass integration in REW and other methods in that convoluted Acoustics thread of mine as the PSE144 project goes together, might be worth a look if not ventured there as yet.

 

playdough

 

 

[playdough]

If you study the noise floor graph, you will note the 2 peaks <100Hz, this is the leakage into the Lounge and it tends to accumulate/cancel (pressure node) just the same as a sub making those frequencies within the area.

Hard to fix, but that is where bass integration work in pre/during/post measurements and fine tuning the EQ.

 

Obviously there are automated system set ups to "do the work for you" although, you can have a good go manually first, it's discovery. 

[OzJustin]

On 11/01/2024 at 1:59 AM, Keith_W said:

 

Because of different windowing. If you go to your spectrogram, click on the top right "controls" button (it looks like a gear). Change "Window" to match your waterfall. You should see the 500Hz/1100Hz ringing reappear. 

Sorry but what do I need to change to 'match' the waterfall?

 

My waterfall controls are on Fourier mode, while my Spectrogram controls are Fourier mode and Gaussian window type.

[OzJustin]

On 11/01/2024 at 11:57 AM, playdough said:

If you study the noise floor graph, you will note the 2 peaks <100Hz, this is the leakage into the Lounge and it tends to accumulate/cancel (pressure node) just the same as a sub making those frequencies within the area.

Hard to fix, but that is where bass integration work in pre/during/post measurements and fine tuning the EQ.

 

Obviously there are automated system set ups to "do the work for you" although, you can have a good go manually first, it's discovery. 

Some interesting observations playdough.

 

Re the 500Hz and 1100Hz magnitude fingers, the noise floor showing up could make sense. I don't recall my exact room usage at the time of those measurements but it's likely I had an AC unit running on the left side of the room, a projector fan in the middle of ceiling, and PC fans running in the right side alcove area. Maybe that is what is being picked up - would make sense given the slightly different fingers across my left, centre and right speaker measurements.

 

Re my noise floor - so you're thinking my 45Hz and 70Hz slow decay is from bass energy reflecting off the lounge in the centre of my room? Obviously I can't do much about the lounge, but your suggesting it's possible to EQ out to speed up the decay?

 

Won't the 45Hz and 70Hz apparent room mode issue (as Keith pointed out) from my speakers disappear once I set my crossovers above that threshold e.g. 120Hz for subs to take over, or will the same reflections occur but from a multi sub source as opposed to speakers? 

 

I'm still keen on any feedback from members on the 80-150Hz cancellation issue across my three main speakers as it has been bugging me for a while (see my raw response graph below for these speakers and all crossovers set at 40hz with ARC off). My surround left and right also have a dip 80-200Hz range but my rear and Atmos speakers are ok. I've moved speakers, lounge and subs with no success. Is there anything I can do to get the speakers performing in this range? My fall back position is to assign crossovers for these speakers at 130Hz and set the High Frequency Extension (HFE) at 250Hz on my receiver. That should mask the lack of speaker bass.

 

[davewantsmoore]

2 minutes ago, OzJustin said:

I'm still keen on any feedback from members on the 80-150Hz cancellation issue across my three main speakers as it has been bugging me for a while (see my raw response graph below for these speakers and all crossovers set at 40hz with ARC off). My surround left and right also have a dip 80-200Hz range but my rear and Atmos speakers are ok.

You'd have to take measurements of the distances and compute the cancellations frequencies, but it is almost certain cancellation bounce from the floor, walls, ceiling distances.

 

There isn't anything that can be done, except move the listener, move the speaker, move the surfaces.   Treatment in this frequency range will be very extreme, and almost certainly over damp other frequencies.

 

EQ is the last resort, but also likely most effective solution... you just have to be mindful how the EQ will affect multiple seats, and how it won't be able to solve sharp cancellations.

 

Overlapping multiple speakers through this range can help to even out any cancellations... but that is getting quite advanced to do manually.   Systems like Dirac Live, can do it "automagically".

[Keith_W]

1 hour ago, OzJustin said:

Sorry but what do I need to change to 'match' the waterfall?

 

My waterfall controls are on Fourier mode, while my Spectrogram controls are Fourier mode and Gaussian window type.

 

 

 

In the two pictures above, I changed the windowing from 500ms to 800ms. You can see the difference. 

[Keith_W]

1 hour ago, OzJustin said:

Re the 500Hz and 1100Hz magnitude fingers, the noise floor showing up could make sense. I don't recall my exact room usage at the time of those measurements but it's likely I had an AC unit running on the left side of the room, a projector fan in the middle of ceiling, and PC fans running in the right side alcove area. Maybe that is what is being picked up - would make sense given the slightly different fingers across my left, centre and right speaker measurements.

 

I gave his post a "like" because I think he is right on the money. TBH I didn't notice it myself until he pointed it out, but if you look at the fingers of the waterfall graph at 500Hz/1100Hz, it remains constant amplitude out to the edge of the graph. If you don't feel like redoing the measurement without speakers, you could extend the windowing of the waterfall to 1000ms and see whether those fingers remain constant amplitude. 

 

1 hour ago, OzJustin said:

Re my noise floor - so you're thinking my 45Hz and 70Hz slow decay is from bass energy reflecting off the lounge in the centre of my room? Obviously I can't do much about the lounge, but your suggesting it's possible to EQ out to speed up the decay?

 

Reducing the amplitude definitely also speeds up the decay. You need to selectively lop off those peaks and the decay will automagically come down. 

 

1 hour ago, OzJustin said:

Won't the 45Hz and 70Hz apparent room mode issue (as Keith pointed out) from my speakers disappear once I set my crossovers above that threshold e.g. 120Hz for subs to take over, or will the same reflections occur but from a multi sub source as opposed to speakers? 

 

It is not clear to me what is your reason for doing this? I generally suggest against setting the XO for the sub too high, because then they start to become directional, especially if your sub generates harmonic distortion. Whether you will see the same room modes depends on where your subs are in your room. 

 

1 hour ago, OzJustin said:

I'm still keen on any feedback from members on the 80-150Hz cancellation issue across my three main speakers as it has been bugging me for a while (see my raw response graph below for these speakers and all crossovers set at 40hz with ARC off). My surround left and right also have a dip 80-200Hz range but my rear and Atmos speakers are ok. I've moved speakers, lounge and subs with no success. Is there anything I can do to get the speakers performing in this range? My fall back position is to assign crossovers for these speakers at 130Hz and set the High Frequency Extension (HFE) at 250Hz on my receiver. That should mask the lack of speaker bass.

 

Try to EQ that region back to flat by boosting those frequencies. Don't worry about getting it perfectly flat, just try to get that broad shelf back up. 

 

May I ask you: (1) what is the size of your room, (2) what is the configuration of your system - i.e. active or passive XO, using a PC as a source? What other sources do you use? 

[playdough]

2 hours ago, OzJustin said:

 

Re the 500Hz and 1100Hz magnitude fingers, the noise floor showing up could make sense. I don't recall my exact room usage at the time of those measurements but it's likely I had an AC unit running on the left side of the room, a projector fan in the middle of ceiling, and PC fans running in the right side alcove area. Maybe that is what is being picked up - would make sense given the slightly different fingers across my left, centre and right speaker measurements.

 

Correct.

2 hours ago, OzJustin said:

Re my noise floor - so you're thinking my 45Hz and 70Hz slow decay is from bass energy reflecting off the lounge in the centre of my room? Obviously I can't do much about the lounge, but your suggesting it's possible to EQ out to speed up the decay?

Look at my graph and yours

Note, both look almost the same sub 100Hz, almost exactly the same with my noise floor actually quite loud late in the afternoon, not really noticed but it's there.

EQing Subs is a world of pain to decipher when learning.

My explanation may  be different from others that do a great job.

I did mine again  today, wrote it up in the Acoustic Thread as well !

I do a close measure of the sub raw then EQ it to a decent thing.

Please, er, was done with an RTA, but you will get the gist of it, Forum pls don't be too hard on me for showing the RTA Readings,,,,,,,

Sub RAW no filter or EQ

Sub EQ applied, mine just happens to play 10 to 240Hz in the band pass and voice up to the PSE144 Point Source Elliptical Speaker System.

EQ and filter applied to the Sub/low bass, crossover set to 250Hz

 Now the result, just over a meter from the port aperture or about as far as I get with the mic,

 

before the room starts to push up accumulation nodes <30Hz and 50Hz forming that  hole *we see common to our graphs, then peak again +6dB centered at 125Hz, which is what I believe your sub and mains will do.

 

Our Lounge dimensions  must be similar to a degree.

 

Combined measurement close to listening position, which is far and away different to what the sub was reading  only 2 meters away ffs BUT, actually sounds crazily ok, albeit a little 20Hz heavy, which is tolerable, I'm a bass nerd.

 

I'll do a little to try nail down the peaks, with EQ from here in REW and UMIK1 however I have little faith it can look all that much better due to the ROOM ACOUSTICS, sounds great. leave it !!!!

Yet to set up target curve across this brand spanking 8 channel 4 way speaker set,  @davewantsmoore has divulged  a PSE144 EQ instruction THANKS MATE,,, sheet I didn't know existed (a little excited about working through that) 

The graph shows the full frequency response 2 dB per vertical step, so roughly within 10dB   getting there. Sounds bright and accurate, clinical albeit 20Hz bass heavyy........

 

Hope that helps, as I keep suggesting there are some Easter Eggs for you in my Thread, you haven't looked at, recommended

 

playdough

 

 

 

Edited January 12, 2024 by playdough

[playdough]

@OzJustin

Stick at it Mate. 

Oh, might have done a few REW measurements as well,,,,,,,,,,,only the Dog knows

[playdough]

4 hours ago, davewantsmoore said:

You'd have to take measurements of the distances and compute the cancellations frequencies, but it is almost certain cancellation bounce from the floor, walls, ceiling distances.

 

There isn't anything that can be done, except move the listener, move the speaker, move the surfaces.   Treatment in this frequency range will be very extreme, and almost certainly over damp other frequencies.

 

EQ is the last resort, but also likely most effective solution... you just have to be mindful how the EQ will affect multiple seats, and how it won't be able to solve sharp cancellations.

 

Overlapping multiple speakers through this range can help to even out any cancellations... but that is getting quite advanced to do manually.   Systems like Dirac Live, can do it "automagically".

Spot on Dave.

[OzJustin]

3 hours ago, Keith_W said:

 

I gave his post a "like" because I think he is right on the money. TBH I didn't notice it myself until he pointed it out, but if you look at the fingers of the waterfall graph at 500Hz/1100Hz, it remains constant amplitude out to the edge of the graph. If you don't feel like redoing the measurement without speakers, you could extend the windowing of the waterfall to 1000ms and see whether those fingers remain constant amplitude. 

I think you are both on the money. I just tried a window out to 2000ms and the amplitude remains the same. There must be something in the room that is causing those issues at specific frequencies. Given they are very narrow (and likely not much I can do about if they are AC, projector or PC related) I gather I won't really notice them anyway.

3 hours ago, Keith_W said:

 

Reducing the amplitude definitely also speeds up the decay. You need to selectively lop off those peaks and the decay will automagically come down. 

 

 

It is not clear to me what is your reason for doing this? I generally suggest against setting the XO for the sub too high, because then they start to become directional, especially if your sub generates harmonic distortion. Whether you will see the same room modes depends on where your subs are in your room. 

 

Try to EQ that region back to flat by boosting those frequencies. Don't worry about getting it perfectly flat, just try to get that broad shelf back up. 

I've been running MSO and a miniDSP to EQ my subs but am not sure I have that manual EQ capability in my Anthem receiver to knock off the peaks in my speaker channels.

If you look at my Anthem ARC Genesis corrected graphs (refer below file) you'll see what I'm dealing with for various speaker channels (main left and right, centre and both surrounds). Even with ARC EQ, my speaker bass response is significantly down from 80Hz to 150-200Hz depending on speaker. I've tried moving my MLP forward up to 1.5m and backward up to 1.5m and left about 40cm, speakers forward and left but haven't been able to resolve the bass dip. This is why I'm now looking to set various crossovers around 120-130Hz so my subs can handle the missing bass. I'm out of ideas for how to resolve via speaker/MLP placement, room treatment etc.

 

ARC Genesis Results 7.4.4 MRX1120 with MSO 28.12.23.pdf

 

 

3 hours ago, Keith_W said:

May I ask you: (1) what is the size of your room, (2) what is the configuration of your system - i.e. active or passive XO, using a PC as a source? What other sources do you use? 

Here's a floor plan so you can see my room shape, size, seating (ignore the back row as very rarely used), speaker and sub locations. Also included is a design showing my installed acoustic treatment (note I'm yet to install the 3D diffusers).  

 

My home theatre setup in a nutshell - 7.4.4 channel:
Speakers - VAF DC-X35, DC-X G4Mk2, DC-X (early gen), DC7, Krix Tropix x4 (as Atmos) - all inbuilt crossovers

Subs - Dual SVS PB-2000 Pro + PB-2000 + HoverEze sub platform (W12S4 x4) + miniDSP 2x4HD+Umik-1 (all time aligned and EQ'd via Multi-Sub Optimiser (MSO))

Receiver - Anthem MRX1120 (onboard amp powering all 11 channels, no power amps in use)

Sources - Panasonic DP-UB420GNK + Apple TV 4K 2019 + Windows 10 PC + Logitech Harmony Companion as room control

 

 

[Keith_W]

I doubt if the MiniDSP or the Anthem is doing very much bass correction. The MiniDSP, depending on which model, has at least 1024 taps per channel. I don't know what your Anthem has. This is what happens if you reduce 65536 taps per channel down to 1024: 

 

 

Red line: correction filter for subwoofer (65536 taps). Brown line, the exact same filter but reduced to 1024 taps. The marker is positioned at the end of the first bin, which is 43Hz. In this case, the sampling rate was 44100. So, 44100/1024 = 43Hz. The next bin is at 86Hz, and so on. You can see that 1024 taps is nowhere near enough to make bass corrections. 

 

So how many taps do you need? Here is the same filter, but truncated down to different taps: 

 

 

I thought that you weren't running any room correction, but it appears that you are. It is possible that your MiniDSP doesn't have the resolution to lop off the bass peaks because it is limited by taps. 

 

If you want to see the verification measurement, here it is: 

 

[OzJustin]

Looking at the waterfall graphs for each of my subs, I do have some slow decay of bass frequencies there too but they are at different frequencies from my speakers. I guess this makes sense as the point of origin of the bass frequency is different for each speaker and sub. If these bass frequencies are all bouncing off my lounge etc then it's probably beyond my (and my equipment's) EQ capabilities so I might have to leave that for now. I assume a good room calibrator could tackle this kind of issue?

 

@Keith_W Thanks for that. I think we're at cross purposes now. I'm not familiar with 'taps' but my understanding is the miniDSP 2x4HD uses gain and delay blocks etc with 8 input and 8 output PEQs available for EQ. MSO makes use of these various PEQs and sets the Q value, frequency, boost or cut etc to flatten out the response across all of my desired seating positions. It has actually been very effective when you compare pre and post bass response across my five seats. For my Anthem receiver, I don't know how many PEQ filters etc ARC has access to but Anthem is pretty highly regarded before you start getting into Trinnovs etc.

 

Regardless, the lack of bass 80-200Hz across my main and surround speakers is being controlled by my Anthem receiver and it has boosted those frequencies as much as it feels safe but there is still a significant gap. My miniDSP is only controlling and EQ'ing the LFE channel and any bass frequencies sent to it below the crossovers of the main/surround speakers. If my speaker crossovers were set at 80Hz (so my subs and miniDSP controlled frequencies 0-80Hz) then I still have a significant lack of bass 80-200Hz where my main speakers should be playing. Through this investigation and all of the kind feedback from members like yourself, I've reached the conclusion that it must be a strange room dimensions, reflection/cancellation type issue that I'm unable to resolve with seating and speaker positioning or room treatment (at least within the limitations of my knowledge). Again, maybe a good calibrator could get to the bottom of it? This has got me to the point of changing my speaker crossovers so the speakers in question will only handle to about 130Hz and then I'll get my subs to take over below that. It seems to be the best fix available to me at present.  

[Keith_W]

This is the REW room sim. I input as many parameters as I knew about your system. 

 

 

And this is the FR graph you posted. I note it's L, C, R only, with no subs mixed in. 

 

 

It looks like a close approximation to what you have measured, including predicting that dip between 80Hz - 120Hz in this case. Bear in mind that REW does not know what correction has been applied. It makes a number of assumptions (speakers are flat, subwoofers are flat), and of course I don't know some parameters of your room so I assumed it. You will get a better sim if you play around with it yourself. 

 

There is nothing wrong with trying a slight boost using your MiniDSP to try to fill in that broad Q gap between 80-120Hz, in fact that's the first thing I would try. There should be enough resolution in those frequencies for it to work. It may or may not fill it, or you might suffer some distortion if you tried it, I don't know. Those are things I would be looking for in the verification measurements. But it won't cost you any money to try. 

 

Yes, you could hire a calibrator but where is the fun in that

 

Oh yes, a "tap" is a delay/gain pair. This is an imperfect analogy, but you can think of it as a PEQ band. If you have 1024 taps, and a 44100 sampling rate, this means each "band" is 44100/1024 = 43Hz wide. If you have 65536 taps, each band is 44100/65536 = 0.67Hz wide.

 

The downside of longer taps is worse latency, particularly if you are using FIR filters - and this is a particular issue if you are using it for HT because it will mess up lip sync. I don't do HT at all, but if I did I would use IIR filters and ask myself whether I should sacrifice accurate correction for low latency by reducing tap length. For music it is less of an issue, I do get some latency when changing tracks (quite noticeable actually) but I would rather have accurate correction with music. 

[OzJustin]

@Keith_W - Really appreciate you going out of your way with the room analyser in REW. That does look reasonably accurate with what I'm experiencing. A big flat spot!

 

The issue though is that the bass absence is in my speakers not subs. My speakers are being controlled by my Anthem receiver so I have no ability to boost anymore. My miniDSP is only correcting my subs and I don't have the 80-200Hz dip in my sub response. 8-150Hz across all my five seats is pretty flat based on my summed sub measurements. It's only the speakers that aren't performing 80-200Hz.  

 

Here is my sub only modelled bass response through MSO across my five seats. MLP is +-2dB 10-200Hz with the outer seats only performing a bit worse in the 150-200Hz range.

[Keith_W]

No trouble for me at all, after all it's all the same ... i'm sitting in front of the computer! I'm doing some spring cleaning at the moment so it's a nice break from moving all that rubbish around (and finding heaps of stuff to put on SNA and Gumtree!). 

 

That sim looks nice, but remember a sim is a sim, but reality is something different

 

If your anthem refuses to boost those frequencies, you could try cutting all the other frequencies instead. You WILL lose maximum volume by doing this, but only you can decide if this is an acceptable compromise. 

[davewantsmoore]

On 12/1/2024 at 9:06 PM, Keith_W said:

I doubt if the MiniDSP or the Anthem is doing very much bass correction. The MiniDSP, depending on which model, has at least 1024 taps per channel. I don't know what your Anthem has. This is what happens if you reduce 65536 taps per channel down to 1024: 

 

This can be a bit confusing for people..... and you see some of the potentially bad takeaways from this sort of comment repeated over and over.

 

Keiths comment is strictly limited to a "FIR" filter where the phase is being modified at low frequencies.

 

The miniDSP (and similar) can still correct the low frequencies with very high precision, if it does not alter the phase.

 

 

Another thing to note specifically here for the miniDSP is that the number of taps exposed by the miniDSP plugins is quite a lot less than the DSP hardware in the box can actually do.

 

When running Dirac Live on the miniDSP hardware, there are lots and lots more "taps" available..... this is in addition to the use of "mixed phase" filters, where the number of taps shown in the charts Keith posted are not necessarily required (they become the worst case requirement).

 

 

 

I hope that makes some sense.   If anyone doesn't understand, or doesn't think this is right... then please ask, and I'll try and explain better or any more specifics.

[davewantsmoore]

20 hours ago, Keith_W said:

but if I did I would use IIR filters and ask myself whether I should sacrifice accurate correction for low latency by reducing tap length. For music it is less of an issue, I do get some latency when changing tracks (quite noticeable actually) but I would rather have accurate correction with music. 

 

Unsure here.

It sounds like you are saying "use IIR .... AND reduce tap length to avoid latency"

 

If you are using IIR type filters, there are no taps.   There is no "tap length" and there no latency.    IIR filters have high (infinite) precision for the amplitude correction, and do no correct the phase.

 

 

 

It is only when we start to alter the phase, that we need the "taps" (memory) in the DSP, and incur the resulting latency.

[davewantsmoore]

19 hours ago, OzJustin said:

Here is my sub only modelled bass response through MSO

[Keith_W]

On 14/01/2024 at 11:46 AM, davewantsmoore said:

When running Dirac Live on the miniDSP hardware, there are lots and lots more "taps" available..... this is in addition to the use of "mixed phase" filters, where the number of taps shown in the charts Keith posted are not necessarily required (they become the worst case requirement).

 

I was not aware that MiniDSP with Dirac used mixed phase filters, I thought it was a choice of IIR or FIR. I just had a look at Mitch Barnett's Dirac walkthrough - although that walkthrough is for the PC version, I would imagine that the MiniDSP version is the same. So yes, I agree that it can do a decent job of bass correction, but without any phase correction.

 

I can also see how it would be advantageous - (1) it can be run on hardware with limited processing power like a MiniDSP or an AVR, (2) it actually does a decent job with FR correction, and (3) it is low latency which is especially advantageous for an AVR. 

 

However, it will have the disadvantages of IIR filters - no excess phase correction, so no correction of non-minimum phase behaviour. Not to forget, any correction of the amplitude also affects the phase. 

 

I myself have only ever used FIR filters so I have no first-hand experience of the audible result of an IIR correction. All I can say is that if I mess up the phase correction in my FIR filter, it muddies the bass considerably even if the FR looks OK. I would love to watch someone creating a crossover from scratch and performing room correction with a Dirac/MiniDSP so that I can better understand what it can or can't do. 

 

Apologies to anyone who thinks this is gobbledegook, happy to try to explain in plain English if required. 

[davewantsmoore]

2 minutes ago, Keith_W said:

I was not aware that MiniDSP with Dirac used mixed phase filters, I thought it was a choice of IIR or FIR.

You don't "choose".

 

Dirac Live computes its correction, and computes an independent amplitude, and independent phase (hence "mixed phase").

 

2 minutes ago, Keith_W said:

I just had a look at Mitch Barnett's Dirac walkthrough - although that walkthrough is for the PC version, I would imagine that the MiniDSP version is the same. So yes, I agree that it can do a decent job of bass correction, but without any phase correction.

 

miniDSP with the miniDSP plugin:

 

• Limited taps for "FIR" filters (limited resolution at low frequencies, when correcting phase --- ie. cannot run long impulse responses)
• Can do IIR filters (infinite amplitude resolution, minimum phase)

 

miniDSP with the Dirac Live plugin:

 

Many many (32k) ?, I'd have to check specific miniDSP / SHARC model) more "taps" available than what you quoted the miniDSP does.    This is to do with the Dirac Live software/firmware using the DSP in a mode which makes use of more of the memory.   Effectively the miniDSP plugin runs in the easy sandbox on the DSP... where as the Dirac code takes over all the DSP resources).   MiniDSP could do the same for their own plugins, but they don't.

 

2 minutes ago, Keith_W said:

I can also see how it would be advantageous - (1) it can be run on hardware with limited processing power like a MiniDSP or an AVR, (2) it actually does a decent job with FR correction, and (3) it is low latency which is especially advantageous for an AVR. 

 

However, it will have the disadvantages of IIR filters - no excess phase correction, so no correction of non-minimum phase behaviour. Not to forget, any correction of the amplitude also affects the phase. 

 

There are two incredibly big/important point to make here.

 

The first is......  that if we look at the things we are correcting, with speakers, and sometimes with rooms ..... very often, the correction of phase and amplitude go hand in hand... ie. you want the "minimum phase" correction.

 

If we are correcting speaker, and we measure something that looks like it has an amplitude / phase divergence ..... there are one of two things going on (kinda the same thing, but viewed different ways) .... either we have not measured the speaker correctly ("mistake") .... or we are viewing a problem (such as diffraction), which shouldn't be corrected based on one axial measurement.

 

This is a very big problem in modern speaker design / speaker correction workflows.

 

The second is .....  "Excess phase" correction is almost universally considered to be inaudible.   Whether we are talking about a speaker, or about a room .... the excess phase, is almost universally inaudible, vs the amplitude correction.

 

2 minutes ago, Keith_W said:

I myself have only ever used FIR filters so I have no first-hand experience of the audible result of an IIR correction. All I can say is that if I mess up the phase correction in my FIR filter, it muddies the bass considerably even if the FR looks OK.

 

There quite a lot which could be causing this (inadvertently, or not)... but it is quite simple to test these general principles I am talking about.

 

You can create two convolution. 

 

One which corrects the speaker amplitude to a target and a minimum phase target.

Another which correct the speaker amplitude the identical target and a 0 degrees phase.

 

They will sound identical.

 

 

 

I actually do use FIR type filters, but....

 

the benefits of using them are vastly(!!!) overstated (to the degree that most of that stated benefits do not even exist)

 

it is very very easy to mis-use them (although to be fair, this largely applies to any/all EQ)

[playdough]

 Flex series is fairly powerful, with this new platform.

1 hour ago, davewantsmoore said:

miniDSP with the Dirac Live plugin:

 

Many many (32k) ?, I'd have to check specific miniDSP / SHARC model) more "taps" available than what you quoted the miniDSP does.  

32-bit floating-point 400MHz Analog Devices SHARC DSP, wow is quite the update on say a miniDSP 2x8, with a 172MHz - 28/56bit DSP Processor.

 

FIR filters and parametric EQ on the input channels, and parametric EQ, crossovers, advanced biquad programming and delay on each output channel. 

FIR filters are specified using a large array of numbers. In the case of the OpenDRC, for example, there are 6144 coefficients (or "taps") per channel. In the case of the Flex (without Dirac Live), there are a total of 4096 taps assignable across the four output channels. Generation of this large array of numbers must be done in a separate program, such as rephase. Some of the miniDSP processors that support FIR filtering have the filters on the output channels, while some have them on the input channels. 

FIR filtering has some important characteristic that IIR filters lack. Which of these can be used depends on whether the FIR filtering is done on the input or output channels, so check your requirements before ordering. 

FIR filters can implement linear-phase filtering. This means that the filter has no phase shift across the frequency band.

FIR filters can alter phase without altering amplitude.

FIR filters can be used to correct frequency-response errors in a speaker or speaker driver to a finer degree of precision.

However, FIRs can be limited in resolution at low frequencies, and the success of applying FIR filters depends greatly on the program that is used to generate the filter coefficients. Usage is generally more complicated and time-consuming than IIR filters.

 

IIR filters are the most efficient type of filter to implement in DSP (digital signal processing). They are usually provided as "biquad" filters. For example, in the parametric EQ block of a miniDSP processor, each peak/notch or shelving filter is a single biquad. In the crossover blocks, each crossover uses up to 4 biquads. Each band of a graphic EQ is a single biquad, so a full 31-band graphic EQ uses 31 biquads per channel

 

Actual Specs. 

FIY, copied from miniDSP.

 

 

[davewantsmoore]

1 hour ago, playdough said:

 Flex series is fairly powerful, with this new platform.

32-bit floating-point 400MHz Analog Devices SHARC DSP, wow is quite the update on say a miniDSP 2x8, with a 172MHz - 28/56bit DSP Processor.

 

FIR filters and parametric EQ on the input channels, and parametric EQ, crossovers, advanced biquad programming and delay on each output channel. 

FIR filters are specified using a large array of numbers. In the case of the OpenDRC, for example, there are 6144 coefficients (or "taps") per channel. In the case of the Flex (without Dirac Live), there are a total of 4096 taps assignable across the four output channels. Generation of this large array of numbers must be done in a separate program, such as rephase. Some of the miniDSP processors that support FIR filtering have the filters on the output channels, while some have them on the input channels. 

FIR filtering has some important characteristic that IIR filters lack. Which of these can be used depends on whether the FIR filtering is done on the input or output channels, so check your requirements before ordering. 

FIR filters can implement linear-phase filtering. This means that the filter has no phase shift across the frequency band.

FIR filters can alter phase without altering amplitude.

 

miniDSP also really paint the false (general) picture for people that "FIR is better than IIR".

 

Unfortunately it is a complex topic, so hard to unpack here without going down rabbit holes.

[davewantsmoore]

1 hour ago, playdough said:

 Flex series is fairly powerful, with this new platform.

32-bit floating-point 400MHz Analog Devices SHARC DSP, wow is quite the update on say a miniDSP 2x8, with a 172MHz - 28/56bit DSP Processor.

 

It really lets you do nothing more than run FIR type filters..... which is quite useful for a "room correction" type system like Dirac Live...... but not so much for speaker filters (driver linearisation, crossovers, EQ, etc.) ..... cool to have of course.

 

The real step-change in the newer products is the converter performance.    The just release FlexHT boxes I saw were touting 127 odd dB of dynamic range.    Crazy.

[playdough]

3 minutes ago, davewantsmoore said:

"FIR is better than IIR".

That's not a quote 

1 hour ago, playdough said:

IIR filters are the most efficient type of filter to implement in DSP

That's the quote.