Equaliser Support for JL Audio TwK-88 DSP

[gg48gg]

Hello and thank you for REW! What are the chances we could get support for the JL Audio TwK-88 DSP?
It uses 10 non symetrical PK filters with a gain range of -6 to 12dB. The cuts seem to have a narrower bandwidth than "generic" PK filters.
More detail about this filter type can be found here:

 

[John Mulcahy]

I have added the TwK-88 for the next beta releases. 10 filters, Q range from 0.25 to 10, gain range from +6 to -12 dB, frequency range 20 Hz to 20 kHz. Looks correct, but I don't have a unit to verify it.

 

[gg48gg]

Thank you John. Will you be building it with half the bandwidth for cuts as the gentleman who posted the above video showed?
in other words:
Gain Bandwidth = centre frequency/Q
Cut Bandwidth = centre frequency/Q/2

 

[John Mulcahy]

Bandwidth depends on gain, it is implemented in the same way as the device.

 

[gg48gg]

Wow! You've been a busy man! Thank you very much for the update at the speed of light! I hope you got some sleep!

When you are rested, here are a couple of follow up questions.

1) I would love to know how to calculate the Q and Bandwidth manually.Would you be able to provide the equations for bandwidth =, and Q =?

2) If you would like me to take any (mic) measurements from the device, I am happy to if you provide instructions for me to do that.

3) The TWK-88 settings are held in a plain text xml configuration file. While I am not a java developer, I could generate some perl code to load/parse/modify the xml with the REW generated filters, or I can provide sample xml files. Let me know if you are interested in this.

For amusement, I thought I would share something with you that may apply. It is a saying that I wrote down recently and I refer to
it frequently as a reminder not to become over-focused or spend too much time unnecessarily:
"There is significant cost and no guaranteed reward for going above and beyond, so choose your efforts wisely."

 

[John Mulcahy]

1) I would love to know how to calculate the Q and Bandwidth manually.Would you be able to provide the equations for bandwidth =, and Q =?

See the help.

2) If you would like me to take any (mic) measurements from the device, I am happy to if you provide instructions for me to do that.

Nope, all good thanks.

3) The TWK-88 settings are held in a plain text xml configuration file. While I am not a java developer, I could generate some perl code to load/parse/modify the xml with the REW generated filters, or I can provide sample xml files. Let me know if you are interested in this.

Sounds like a way you might while away some coronavirus lockdown hours, I already have a 492-item todo list

 

[gg48gg]

Thank you! Predicted SPLs from the auto-eq and actual measurements are in-fact spot on. This is awesome.

 

[gg48gg]

A potential small issue I am seeing is that when selecting the TwK-88 in REW preferences as the default equalizer, selecting a measurement and hitting the EQ button seems to always default back to "Generic".

 

[dotnet]

The discussion Q <-> BW conversion is interesting, if a little unsettling. I thought that there were a fixed relationship between those. I've always used this formula to get from Q to BW in octaves:
2 * asinh(1/(2 * Q))/ln(2)

Now it seems like the definition of bandwidth depends on the equaliser (centre_freq/Q, or sqrt(gain) * centre_freq/Q, etc.)??

How can I find out which definition the AUNbandEQ (Apple AudioUnits) applies, that I'm using?

 

[John Mulcahy]

A potential small issue I am seeing is that when selecting the TwK-88 in REW preferences as the default equalizer, selecting a measurement and hitting the EQ button seems to always default back to "Generic".

The equaliser selection is retained in the measurement, so you need to save the measurement after changing it. You can change the default equaliser selection on the Equaliser tab of the preferences.

 

[John Mulcahy]

The discussion Q <-> BW conversion is interesting, if a little unsettling. I thought that there were a fixed relationship between those. I've always used this formula to get from Q to BW in octaves:
2 * asinh(1/(2 * Q))/ln(2)

Now it seems like the definition of bandwidth depends on the equaliser (centre_freq/Q, or sqrt(gain) * centre_freq/Q, etc.)??

How can I find out which definition the AUNbandEQ (Apple AudioUnits) applies, that I'm using?

The formulations used vary, though ultimately they are equivalent. The different formulations do result in different values for the notional Q or bandwidth figure though. If you measure the shapes the equaliser of interest produces for some filter settings you can compare them to the shapes produced by REW's equaliser options to see which matches and then from the help see which formulation is being used.

 

[gg48gg]

The equaliser selection is retained in the measurement, so you need to save the measurement after changing it. You can change the default equaliser selection on the Equaliser tab of the preferences.

Understood. I am just trying to prevent having to select the equaliser type for each new measurement. The preferences I set don't seem to be used when starting EQ on a new measurement.

 

[John Mulcahy]

Preferences are applied when the measurement is made.

 

[dotnet]

The formulations used vary, though ultimately they are equivalent. The different formulations do result in different values for the notional Q or bandwidth figure though. If you measure the shapes the equaliser of interest produces for some filter settings you can compare them to the shapes produced by REW's equaliser options to see which matches and then from the help see which formulation is being used.

Thank you for that link, it gives me some background I was missing. I guess I have to find out which definition of bandwidth the AUNbandEQ employs (and whether it even uses second order IIR filters). That said, the differences in calculated bandwidth are rather small, I'm not sure that they even matter. For example, this is the latest set of filters I'm using:

With "BW" being the bandwidth calculated using the sinh formula, and "BW new" being sqrt(2)/Gain.

Unfortunately, I have no way of comparing the filter shapes (the AUNbandEQ doesn't show them). Is there a way of playing back measurement sweeps from file? I could then play the sweeps with the EQ in place, and compare the actual room response with the target curve.

 

[John Mulcahy]

Measuring with file playback.

 

[dotnet]

Whoa, what application is that?!

So, this is a 5.20 feature, I'm guessing?

 

[gg48gg]

Whoa, what application is that?!

So, this is a 5.20 feature, I'm guessing?

Yes, new measurement window with lots more options and features. Pretty fantastic. Thank you John.