REW & RePhase same EQ settings producing different curves

[daftcombo]

Hi,

I'm a bit puzzled. I have noticed a different behaviour of REW & RePhase parametric EQ.

I've generated a flat impulse to show it on the most simple example.

Check the two pictures and explain me what's going on, please.

I used 6 EQ with Gain = -2 dB and Q = 2.

Rew shows a 4 dB dip. RePhase a 8 dB dip.

REW is v5.20 Beta 8.
Rephase is v1.3.0.


Cheers.

 

[flipflop]

REW looks way more jagged than it should be, so I would disregard that.

FWIW EQ APO + Peace GUI shows ~-7 dB:

I believe there is a way to calculate how big the dip should be, but I can't seem to find the formula anywhere.

 

[DonH56]

The size of the dip depends upon the filter topology used and the resolution of the plot. The formula also depends upon the topology (type) of filter. Too many unknowns for me, though likely someone here knows the filters REW and RePhase use in their implementations so could run the numbers (probably IIR biquads? Or do they implement FIR filters? Butterworth, Chebyshev, SS, ??? I don't know.)

 

[daftcombo]

APO & RePhase seems to agree against REW here. But still 1dB difference between RePhase and APO now!!

And how do we know the behaviour of miniDSPs in that regard?

 

[pos]

rePhase implements different EQ types that you can choose using the drop down menu in the EQ tab.
Most common ones are constant Q (default ones, and probably the most commonly used among software and hardware processors implementations on the market) and proportional Q (less common, but notably used in miniDSP products and analog device sigmaStudio).

AFAIK REW does implement constant Q EQs when "rePhase" is chosen as the target EQ device. This is what rePhase considers them to be when importing them. Maybe @JohnPM can chime in and confirm.

When implementing biquads you will also have differences when approaching Nyquist frequency.
This is why rePhase typically uses a very high sampling frequency (100 000 times the cutoff frequency of the given EQ/filter) when building EQs/filters using biquads, to avoid such problems.
rePhase does not implement constant Q and proportional Q EQs using biquads though: magnitude and phase formulas are directly deducted from the transfer function.

By the way @daftcombo, I see you are using rephase version 1.3.0.
Although this would not change the situation you are in, you could update to a more recent version (current is 1.4.3) that would, among other things, let you edit EQ points directly on the response curve

 

[daftcombo]

Hi,

AFAIK REW does implement constant Q EQs when "rePhase" is chosen as the target EQ device.

I had "RePhase" chosen as the Equaliser in REW when I did the test. I switched to "Generic" but it was almost the same.

Although this would not change the situation you are in, you could update to a more recent version (current is 1.4.3) that would, among other things, let you edit EQ points directly on the response curve

Thanks, I have that new version too but I don't like the new tricks for the moment. I'm used to the older.
There is also a newer Beta of REW available.

 

[pos]

Thanks, I have that new version too but I don't like the new tricks for the moment. I'm used to the older.

You don't have to use them if you don't like them: you can still use the faders.
Or is the help/info message too intrusive on the graph window? That is something I can adjust in a future version.
Thanks for the feedback

 

[RayDunzl]

That is something I can adjust in a future version.

Oh...

You're Mr rePhase...

(I'm a little dense, sometimes)

 

[JohnPM]

AFAIK REW does implement constant Q EQs when "rePhase" is chosen as the target EQ device. This is what rePhase considers them to be when importing them. Maybe @JohnPM can chime in and confirm.

Yes. rePhase equaliser selection.

 

[daftcombo]

You don't have to use them if you don't like them: you can still use the faders.
Or is the help/info message too intrusive on the graph window? That is something I can adjust in a future version.
Thanks for the feedback

Yes, the yellow glowings everywhere & the info message distract me a bit. But I guess it's just me & that other people can find the right-click zoom and other graph manipulations convenient.
I like the "Throughout Bank" tool though.

 

[daftcombo]

Yes. rePhase equaliser selection.

Thanks, but how is it so different from what RePhase does then?

 

[JohnPM]

REW doesn't produce a response like that you initially showed, which suggests there may be something odd with your "flat impulse". The REW response for that filter set is shown below. In RePhase you appear to have selected linear phase filters, IIR filters cannot be linear phase so you are not comparing like with like.

 

[daftcombo]

REW doesn't produce a response like that you initially showed, which suggests there may be something odd with your "flat impulse". The REW response for that filter set is shown below. In RePhase you appear to have selected linear phase filters, IIR filters cannot be linear phase so you are not comparing like with like.
View attachment 25976

Thank you for testing it. Could you send me the impulse you used, please?

 

[JohnPM]

Here it is.

 

[daftcombo]

Here it is.

Thank you. Well, I have to say that now I have the same graph as yours (around -7.3dB, like in RePhase). So my impulse was definitely weird!

I don't understand how it is possible. Can't it be a matter of bitrate?

 

[daftcombo]

Here is the impulse I used.

 

[JohnPM]

That IR is at 44.1 kHz sample rate, it isn't going to be a good candidate for examining a filter response that extends to 40 kHz. Note that you don't need to load a measurement to look at filter responses in REW, the EQ window can be used without any measurements loaded.

 

[daftcombo]

I see. How do you know that the filter extends to 40 kHz? Can it be changed?

 

[JohnPM]

From the filter response, which you have been posting. You chose the filter values, with Q of 2 and a highest centre frequency of 20 kHz, that's the response they give. If you want a different response pick different filters. If you just want something that shelves down at the top end a single shelf filter would do the job.

 

[daftcombo]

That was just an example. But actually, to tame the brightness of a speaker, I use a G=-6dB, Q=0.7, F=14 000Hz linear phase EQ in the form of a 44.1kHz .wav impulse created with RePhase and convolved in Foobar2000 (after resampling everything to 44.1kHz _ though most of my music is in 44.1kHz/16bit already).
Seing the weird curve such a 44.1kHz impulse has in REW when EQed, I am now worried about what Foobar2000 Convolver does behind my back when fed with my 44.1kHz impulse...