Should I use filter settings or impulse response REW

[Urbanized]

I bought myself a calibrated Sonarworks mic and did some measurements. In the EQ section of REW I performed an EQ correction on the room response. Now I want to export the final setting to use it with a parametric eq- or a convolution VST plugin in my DAW (my room is not treated). Which of the 2 methods would work best / most reliably?

How does this method differ from companies like Sonarworks, Dirac etc?

 

[ernestcarl]

I bought myself a calibrated Sonarworks mic and did some measurements. In the EQ section of REW I performed an EQ correction on the room response. Now I want to export the final setting to use it with a parametric eq- or a convolution VST plugin in my DAW (my room is not treated). Which of the 2 methods would work best / most reliably?

How does this method differ from companies like Sonarworks, Dirac etc?

@dominikz might have a link in his sig for a comprehensive review of those room correction software — or just do a search. I don’t quite know what would work more reliably for a DAW as I don’t use one. However, some folks have mentioned they prefer an external hardware solution to avoid software issues e.g. miniDSP, IK ARC (the new one — see reviews online).

It’s more common to apply per channel equalization. And when they average, it’s to get better spatial representation. Above 300Hz or so you better be careful esp. if you don’t have (quasi)anechoic measurements. Most of the (semi)automated programs apply some kind of averaging and/or frequency dependent windowing.

 

[dominikz]

I bought myself a calibrated Sonarworks mic and did some measurements. In the EQ section of REW I performed an EQ correction on the room response. Now I want to export the final setting to use it with a parametric eq- or a convolution VST plugin in my DAW (my room is not treated). Which of the 2 methods would work best / most reliably?

How does this method differ from companies like Sonarworks, Dirac etc?

Both the impulse response (FIR) + convolution or using IIR PEQ with settings produced by REW can give equivalent results if done correctly - assuming you only plan to use causal filters.

There are some things to be careful with though:

• FIR/convolution can introduce more latency compared to simple IIR PEQ. This is primarily depending on how much time there is in the filter impulse response file before the IR peak - keep this as short as possible for least added latency. Latency is an issue in a DAW context, especially if you do live tracking.
• If you need certain types of phase correction (e.g. non-causal filters to compensate for loudspeaker crossover phase wrap) you will have to use FIR - but as said that can add very significant latency and is IMHO not worth it (from audibility perspective).
• Note that PEQ IIR filters may have slightly different implementations so their shapes might not always match perfectly to what you calculate in REW. FIR/convolution on the other hand is fully described by the exported filter impulse response, but how well it tracks the desired filter response will depend on the number of "taps" available in your convolution engine - the more taps the closer the actual filter response will be to the desired theoretical filter.
• Regardless if you use FIR or IIR, in a DAW context you will need to add a plugin to your monitoring bus. Be careful not to export your mixes/masters with the room correction plugin applied! Perhaps you can use a different monitoring bus or you DAW has some options specific for this.
• An alternative is to implement EQ somewhere outside of a DAW - usually in a HW box. E.g. some studio monitors have built DSP/PEQ capability (e.g. Adam A-series, Neumann KH-series, Genelec 83xx series), some audio interfaces have PEQ/room EQ in their on-board DSP (e.g. RME TotalMix SW has a 3-band PEQ per output, and they also just added room EQ capability to some interfaces) and there are also dedicated DSP devices (e.g. miniDSP 2x4/Flex/SHD(OpenDRC-DI, IK Multimedia ARC Studio, t.racks DSP 4x4 mini...).
  A benefit of this approach is that you don't have to fiddle with master bus EQ in your DAW, and you get the same room correction both inside and outside of your DAW.

In short - either approach (IR export or PEQ config export) normally works very well for the typical use with REW-calculated correction filters, so you should use whichever is easier to implement.

Hope this helps!

 

[dominikz]

How does this method differ from companies like Sonarworks, Dirac etc?

In my experience you can implement just as great sounding correction filters with manual correction based on REW as you can do with Dirac, Sonarworks or similar DRC SW.
The main benefit of Dirac and such is they are simpler to start using compared to REW - but they also offer less control.

After doing extensive testing of various room correction SWs and approaches (see link in my signature) I ended up using filters generated in REW.

 

[dominikz]

I bought myself a calibrated Sonarworks mic and did some measurements. In the EQ section of REW I performed an EQ correction on the room response. Now I want to export the final setting to use it with a parametric eq- or a convolution VST plugin in my DAW (my room is not treated). Which of the 2 methods would work best / most reliably?

How does this method differ from companies like Sonarworks, Dirac etc?

Lastly, looking at the frequency responses you posted, I'd advise you not to use any filters above approximately 300Hz. Correcting higher frequencies based on in-room measurements can sound worse than no correction.

 

[Urbanized]

@dominikz might have a link in his sig for a comprehensive review of those room correction software — or just do a search. I don’t quite know what would work more reliably for a DAW as I don’t use one. However, some folks have mentioned they prefer an external hardware solution to avoid software issues e.g. miniDSP, IK ARC (the new one — see reviews online).

It’s more common to apply per channel equalization. And when they average, it’s to get better spatial representation. Above 300Hz or so you better be careful esp. if you don’t have (quasi)anechoic measurements. Most of the (semi)automated programs apply some kind of averaging and/or frequency dependent windowing.

Coincidentally, I was reading about the new ARC 4 hardware. I'll wait until enough people have bought it to get reliable reviews.

I have done a left and right monitoring measurement and want to make an eq of both separately.

I'll see if I can find some reviews at dominikz.

 

[ZolaIII]

Both, PEQ including manual for first two room mode spikes and up to 1 KHz no smoothing and then convolver to let's say 96 KHz for all and not to insane amount of taps to keep latency acceptable and again psy PEQ to highs in the end of needed.

 

[Urbanized]

Both the impulse response (FIR) + convolution or using IIR PEQ with settings produced by REW can give equivalent results if done correctly - assuming you only plan to use causal filters.

There are some things to be careful with though:

• FIR/convolution can introduce more latency compared to simple IIR PEQ. This is primarily depending on how much time there is in the filter impulse response file before the IR peak - keep this as short as possible for least added latency. Latency is an issue in a DAW context, especially if you do live tracking.
• If you need certain types of phase correction (e.g. non-causal filters to compensate for loudspeaker crossover phase wrap) you will have to use FIR - but as said that can add very significant latency and is IMHO not worth it (from audibility perspective).
• Note that PEQ IIR filters may have slightly different implementations so their shapes might not always match perfectly to what you calculate in REW. FIR/convolution on the other hand is fully described by the exported filter impulse response, but how well it tracks the desired filter response will depend on the number of "taps" available in your convolution engine - the more taps the closer the actual filter response will be to the desired theoretical filter.
• Regardless if you use FIR or IIR, in a DAW context you will need to add a plugin to your monitoring bus. Be careful not to export your mixes/masters with the room correction plugin applied! Perhaps you can use a different monitoring bus or you DAW has some options specific for this.
• An alternative is to implement EQ somewhere outside of a DAW - usually in a HW box. E.g. some studio monitors have built DSP/PEQ capability (e.g. Adam A-series, Neumann KH-series, Genelec 83xx series), some audio interfaces have PEQ/room EQ in their on-board DSP (e.g. RME TotalMix SW has a 3-band PEQ per output, and they also just added room EQ capability to some interfaces) and there are also dedicated DSP devices (e.g. miniDSP 2x4/Flex/SHD(OpenDRC-DI, IK Multimedia ARC Studio, t.racks DSP 4x4 mini...).
  A benefit of this approach is that you don't have to fiddle with master bus EQ in your DAW, and you get the same room correction both inside and outside of your DAW.

In short - either approach (IR export or PEQ config export) normally works very well for the typical use with REW-calculated correction filters, so you should use whichever is easier to implement.

Hope this helps!

Latency can be reduced by cutting out the quiet parts in the convolution file. At least if that's what you mean.

I was planning to use the Convology XT VST plugin on my master bus and when exporting songs I will turn it off. I think I'll go for the EQ variant, because convolution is too much hassle.

Those DSP monitors are too expensive. Maybe I will buy the just released IK Multimedia ARC 4 box in the future. As long the reviews/experience of people is good at that time.


Thanks.

 

[Urbanized]

In my experience you can implement just as great sounding correction filters with manual correction based on REW as you can do with Dirac, Sonarworks or similar DRC SW.
The main benefit of Dirac and such is they are simpler to start using compared to REW - but they also offer less control.

After doing extensive testing of various room correction SWs and approaches (see link in my signature) I ended up using filters generated in REW.

Indeed, REW has a lot of possibilities than I can imagine. I'm not sure if I should also EQ the small dips/peaks manually, or just the big ones. And up to which frequency range should I stop eq? One says up to 200Hz, or 500Hz and the other 800Hz.

 

[Urbanized]

Both, PEQ including manual for first two room mode spikes and up to 1 KHz no smoothing and then convolver to let's say 96 KHz for all and not to insane amount of taps to keep latency acceptable and again psy PEQ to highs in the end of needed.

I think I'll stick to PEQ only

 

[ZolaIII]

I think I'll stick to PEQ only

Well PEQ can't help you much over let's say 2 KHz and Q over 20~25, simple Q is no more adequate there. It's not hard to create inverted convolution or apple it (cuple minutes of work only). It really depends what you are getting in uper mids and highs if you can or can not pass without it.
What's the application and playing on what?
Edit: it looks like you do need convolver and wide Q PEQ sloping afterwards. Try to position speakers better before all (horizontal angle ° towards target) so that you get better response in highs.

 

[dominikz]

Indeed, REW has a lot of possibilities than I can imagine. I'm not sure if I should also EQ the small dips/peaks manually, or just the big ones. And up to which frequency range should I stop eq? One says up to 200Hz, or 500Hz and the other 800Hz.

This will require some experimentation, there are no rules written in stone, so there will naturally be some different opinions.

If you follow the approach illustrated in dr. Toole's book, you should not EQ based on the in-room-response above the rooms transition frequency; which is typically in the 200-400Hz range for most residential rooms.

Getting the exact frequency is difficult but ultimately also not that important, because the transition from perceiving the combined room+loudspeaker response (at low frequencies) to hearing the loudspeaker direct sound and room sound as two separate phenomena (in higher frequencies) is gradual anyway.

This is also why I'd always recommend to use 'variable' smoothing when calculating EQ filters - it will give no smoothing at low frequencies (where you need to correct peaks) but do a lot of smoothing at high frequencies (which you should not touch, or possibly only with low-Q/wide filters). I also prefer to use MMM instead of sweep measurements as basis for EQ - IMHO MMM is quicker and directly gives results that are comparable to anechoic PIR (above approx. 1kHz).

Regarding which peaks/dips to correct and which not I'd say the following:

• You should not correct any dips with EQ; ideally you should instead optimize loudspeaker and MLP placement to minimize them. Note that the dips are mainly caused by the distance of loudspeakers from nearby walls/ceiling/floor - i.e. SBIR.
• Peaks should be brought down by PEQ, and I'd suggest to EQ all peaks under about 300Hz. You can probably get similarly good-sounding results by EQ-ing only the big peaks (as those are the most audibly offending) - e.g. in my destop system I use only 3-bands of PEQ and still get a good result - but if you have the option to run more bands of PEQ it's definitely worth to try.
• Experiment a bit with target curves - namely try a (slight) downward slope and/or a small boost <200Hz instead of a fully flat curve. This is to account for the natural rising directivity (with frequency) of conventional front-firing loudspeakers.

Perhaps you will also find some of these posts useful:

1. On SBIR and tips for improving loudspeaker/subwoofer positioning
2. How to do MMM and calculate EQ in REW
3. Room EQ FAQ

 

[Urbanized]

Well PEQ can't help you much over let's say 2 KHz and Q over 20~25, simple Q is no more adequate there. It's not hard to create inverted convolution or apple it (cuple minutes of work only). It really depends what you are getting in uper mids and highs if you can or can not pass without it.
What's the application and playing on what?
Edit: it looks like you do need convolver and wide Q PEQ sloping afterwards. Try to position speakers better before all (horizontal angle ° towards target) so that you get better response in highs.

I already moved the monitors and desk, no change.

I downloaded the Convology XT free plugin and hope it will work/accept the REW impulse response file. But for me the most important is the low frequency area, let say under 300Hz. But it would be great if I'll be able to eq the whole freq. area without artifacts (is this a correct English word?)

 

[Urbanized]

This will require some experimentation, there are no rules written in stone, so there will naturally be some different opinions.

If you follow the approach illustrated in dr. Toole's book, you should not EQ based on the in-room-response above the rooms transition frequency; which is typically in the 200-400Hz range for most residential rooms.

Getting the exact frequency is difficult but ultimately also not that important, because the transition from perceiving the combined room+loudspeaker response (at low frequencies) to hearing the loudspeaker direct sound and room sound as two separate phenomena (in higher frequencies) is gradual anyway.

This is also why I'd always recommend to use 'variable' smoothing when calculating EQ filters - it will give no smoothing at low frequencies (where you need to correct peaks) but do a lot of smoothing at high frequencies (which you should not touch, or possibly only with low-Q/wide filters). I also prefer to use MMM instead of sweep measurements as basis for EQ - IMHO MMM is quicker and directly gives results that are comparable to anechoic PIR (above approx. 1kHz).

Regarding which peaks/dips to correct and which not I'd say the following:

• You should not correct any dips with EQ; ideally you should instead optimize loudspeaker and MLP placement to minimize them. Note that the dips are mainly caused by the distance of loudspeakers from nearby walls/ceiling/floor - i.e. SBIR.
• Peaks should be brought down by PEQ, and I'd suggest to EQ all peaks under about 300Hz. You can probably get similarly good-sounding results by EQ-ing only the big peaks (as those are the most audibly offending) - e.g. in my destop system I use only 3-bands of PEQ and still get a good result - but if you have the option to run more bands of PEQ it's definitely worth to try.
• Experiment a bit with target curves - namely try a (slight) downward slope and/or a small boost <200Hz instead of a fully flat curve. This is to account for the natural rising directivity (with frequency) of conventional front-firing loudspeakers.

Perhaps you will also find some of these posts useful:

1. On SBIR and tips for improving loudspeaker/subwoofer positioning
2. How to do MMM and calculate EQ in REW
3. Room EQ FAQ

I used 1/6 as smoothing. So I should use the variable one?

I already moved the monitors and desk and little has changed.

Is the q of parametric EQ's all the same? So if I add a q of 3 in REW, will the q width be the same in, for example Fabfilter, after adding them in my DAW?

I did the measurements with the Sonarworks mic + calibration file. Is this combination reliable enough?

Thanks I will try your tips & urls!

 

[ZolaIII]

@Urbanized well I played with MConvolutionEZ as I already had it with MFreeformPhase (as it's cumulative plugins pack) that I intend to use to push between speakers crossovers anyway. It worked fine and with not much lag.
Regarding PEQ-ing first cuple room length influenced prominent spikes.

How effective would heavy ceiling treatment work here?

I agree with the previously mentioned comments that the wall of glass windows is more of a concern than the ceiling.

www.audiosciencereview.com

I use JRiver as base.

 

[dominikz]

I used 1/6 as smoothing. So I should use the variable one?

Room-induced resonances can be sharper than 1/6 octave, so variable smoothing is usually better suited for fixing them, as it does less smoothing in low frequencies.

I already moved the monitors and desk and little has changed.

The ~85Hz dip from your in-room response suggests to me one of two most likely possibilities:

1. The front faces of your monitors are ~1m away from the wall behind them. If so, I'd try to push the loudspeakers as close as possible to the wall behind - that would move the dip higher in frequency to where it can be more easily treated. If your monitors are back-ported leave at least ~5cm between the wall and the monitor for the port to function correctly.
2. You are using a subwoofer which is close to being out-of-phase with the monitors around 85Hz. If so, I'd try reversing the phase switch on the sub and remeasuring.

Is the q of parametric EQ's all the same? So if I add a q of 3 in REW, will the q width be the same in, for example Fabfilter, after adding them in my DAW?

Not always, several definitions exist - though in practice the deviations may end up being irrelevant; something you'll see when you remeasure the response with EQ applied.
On this topic I suggest to read Rane note 167 for a bit more generic information on Q definition differences and also the REW Equalizer Selection help article which explains the options you can configure in REW.

I did the measurements with the Sonarworks mic + calibration file. Is this combination reliable enough?

I have no experience with Sonarworks microphones so can't comment.
If you're concerned about this you may be able to find a 3rd party microphone calibration services in your area to validate the stock calibration curve. I know of Cross-Spectrum Labs in the US, and HiFi-Selbstbau in Germany who offer such services.

EDIT: If you decide to get your microphone calibrated by an independent party I'd suggest to get both the 0° and 90° calibration curves. 90° calibration curve with vertical microphone orientation is more suitable for in-room response measurements and room EQ (you can read the explanation here).
I wrote (and included some measurements) regarding independent calibration of a Dayton Audio mic in this post.

 

[Urbanized]

Okay guys, I think I know enough for now. Thanks for your support!