Feature requests for loudspeaker design

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
Of course...REW is for room EQ, it is not purpose built for loudspeaker design. However, it has many features beyond EQ functions, and only a few more features to be fully there in my mind.

1. Full 2-channel measurement. Currently REW supports a semi-dual channel measurement where one channel is used as a timing reference. With that in mind, it should not be much of an addition to provide a complete 2-channel measurement. Record both inputs fully, and normalize the measurement input to the reference. This allows for a self-calibrated measurement, where all non-linearity of equipment upstream of the reference input is compensated for. This allows for compensation of nonlinear amplifiers, addition of filters and protection capacitors with very minimal impact on the measured frequency response. Addition of this feature would provide REW the same measurement functionality of loudspeaker design software like ARTA or SoundEasy.

The next 2 requests are available by simply exporting all REW measured data to VituixCAD, but I'm sure a lot of people would find the features built-in to REW a great benefit.
2. Polar response graphs. For each measurement, assign an angle, allowing for generation of a polar heat map type of plot.
3. "Spinorama". Using the same set of polar data, allow for generation of CTA-2034-A chart with in-room response and power response.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
Using a reference channel does not compensate for non-linearities, it can only handle linear effects. The loopback calibration does the same thing but has the advantage of being based on the channel being used for measurement rather than a different channel. Polar maps are on the todo list, but quite a long way down the list.
 

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
What I meant by nonlinearities is "linear distortion". If you have any meaningful difference between left and right channels on your audio equipment, that equipment should not be used for measuring, listening or recording. Loopback calibration as it is written in the REW documentation is only calibrating for the response of the audio interface, missing the amp, any filters you place in the loop (for example a protection capacitor that I mentioned above), the long leads running across the room...True loopback calibration would be completed at the speaker terminal, of course it can be done with just a couple resistors to feed back speaker level signal to line level input.

The Full dual channel configuration provides a lot of benefits, not only for loudspeaker measuring but for circuit analysis as well, it is essentially a differential measurement.
 

jschwender

Member
Joined
Mar 16, 2021
Messages
140
Location
GERMANY
More  
Front Speakers
Nubert digital pro
Other Speakers or Equipment
Philips dss940
Please, can you explain what you mean by protection capacitor?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
You can include anything you want in the loopback measurement. Using another channel as the loopback response reference introduces errors due to channel differences, such as the effects of component tolerances.
 

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
Please, can you explain what you mean by protection capacitor?
A capacitor in series with the driver, providing a basic 1st order filter so you don't sent bass frequencies to a delicate ribbon tweeter for example.
 

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
You can include anything you want in the loopback measurement. Using another channel as the loopback response reference introduces errors due to channel differences, such as the effects of component tolerances.
The error is of absolute level only, which can easily be calibrated, and most irrelevant for a loudspeaker measurement. As a simple calibration process, I have individual input gain knobs on my USB interface, so I can run realtime measurement with both inputs and adjust the gain to match if I really need that sort of comparison. An automated calibration process already exists in REW for impedance measurement for example, where both inputs are utilized precisely as I suggest. I am only suggesting to be able to use both inputs in the same manner for acoustic measurement. Resistor tolerance is really out of the question for matching with individual gain on each input.
 
Last edited:

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
Not using the other channel as the response reference was a design decision made in 2011 to improve accuracy. Before that (REW versions 3 and 4) the loopback channel was used as the response reference but it was limited by the channel matching, which is often poor at very low frequencies due to the tolerances of coupling caps. ARTA is excellent software though, it has many features REW lacks.
 

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
People just need to stop using UCA202 thinking it is competent measurement gear.. ;)

Thanks for your consideration, I had always wondered why REW would not complete a full 2 channel measurement so it’s nice to have some context. Unfortunately I am likely to keep using ARTA for most of my measurement needs. REW has a nicer view for distortion analysis, and of course all the EQ functions so it still has its use cases for me.
 
Last edited:

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
Just a bump on this request. I fully believe that full dual channel measurement (such as pre-2011 REW) and semi-dual with timing reference only can co-exist within REW, to satisfy the needs of loudspeaker designers such as myself, as well as people with poor hardware with wild differences between channels.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
I'll take a look at it. Simply using the loopback channel as the data source for deconvolution (as that very old code did) isn't great as the harmonic distortion data would no longer be valid. A possibility might be to use the linear response of the loopback channel to apply magnitude and phase adjustments to the measurement input, preserving the measurement distortion data. It would become an automatically-generated soundcard calibration file, in effect, which could be removed or replaced afterwards using the existing options to change cal data.
 

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
Thank you! Yes, essentially an automatically generated soundcard cal with each measurement is what I am after.

Sorry, your comment on the deconvolution for distortion data may be a little over my head, FWIW the other dual channel measurement systems I've used, ARTA and SoundEasy, both resort to a single channel measurement for harmonic distortion.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
OK. Note that this wouldn't make REW a dual channel FFT system in the sense the term is used for ARTA, SMAART and the like. In such systems the second channel is used to provide the measurement system with the stimulus and a transfer function view is generated live from averaging what in effect is the ratio of the FFTs of the data on the measurement and stimulus inputs, with a corresponding coherence trace that indicates where the result is reliable.
 

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
I'm not sure I understand. My intent would be a "calibrated" impulse response of the stimulus input against the reference input. If I'm understanding what you've posted, the dual channel implementation in mind would not result in a corrected impulse response, but rather the raw stimulus input impulse with a calibration applied to the FFT result. This is not ideal, since exported impulse response would require the accompanying compensation, but it's a start.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
It was just a clarification of terminology, since "dual channel FFT" has a very specific meaning in the context of a measurement system.

I have had a longstanding todo item to add a feature to merge cal files with the IR, so I'll implement that.
 

teunvanderveeken

Registered
Joined
Oct 18, 2020
Messages
1
Hi John,I don't want to hijack this thread but have a question related to that asked by the thread starter.
First post so some explanation: I mainly do PA systems engineering for small companies that want to get the maximum out of their systems.
Previously i've always used HolmImpulse combined with Smaart to do system engineering where HolmImpulse would be used for single channel capture to linearize the driver using the loudspeaker controller (EQ, FIR, IIR Allpass) and investigate optimal crossovers considering distortion etc. Afterwards, I would open Smaart to easily be able to do time alignment between drivers using the continuous pink noise dual channel FFT (Transfer function mode). Additionally, i would use Smaart Transfer function using multi channels to estimate the optimal crossover frequency regarding directivity of the separate drivers as well as the integrated box using multiple mics.
I've recently taken a dislike towards HolmImpulse since it has provided some issues and noticed REW has all if not more features compared to HolmImpulse (except the live IR windowed gating which i do miss to easily filer out reflections instead of needing to recalculate).
I noticed REW has an option for multi input channel capture which for 100USD is worth the investment in my opinion, but i'm really missing the dual channel continuous noise FFT to make this a single go-to tool for all purposes i intend.
Are there any plans for such a function to be integrated in REW?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
In the long run yes, dual channel FFT is on the todo list and would be part of implementing a measurement method using a noise stimulus. When that might happen is another matter, but not soon. Note that you can drag the window indicators at the top of the Impulse graph to get a live preview of what effect the window change is having, but you need to be zoomed out sufficiently for the window limit indicators to be within the span of the plot.
 

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
Hi John,I don't want to hijack this thread but have a question related to that asked by the thread starter.
First post so some explanation: I mainly do PA systems engineering for small companies that want to get the maximum out of their systems.
Previously i've always used HolmImpulse combined with Smaart to do system engineering where HolmImpulse would be used for single channel capture to linearize the driver using the loudspeaker controller (EQ, FIR, IIR Allpass) and investigate optimal crossovers considering distortion etc. Afterwards, I would open Smaart to easily be able to do time alignment between drivers using the continuous pink noise dual channel FFT (Transfer function mode). Additionally, i would use Smaart Transfer function using multi channels to estimate the optimal crossover frequency regarding directivity of the separate drivers as well as the integrated box using multiple mics.
I've recently taken a dislike towards HolmImpulse since it has provided some issues and noticed REW has all if not more features compared to HolmImpulse (except the live IR windowed gating which i do miss to easily filer out reflections instead of needing to recalculate).
I noticed REW has an option for multi input channel capture which for 100USD is worth the investment in my opinion, but i'm really missing the dual channel continuous noise FFT to make this a single go-to tool for all purposes i intend.
Are there any plans for such a function to be integrated in REW?
Hi teunvanderveeken,
I'm not a Smaart user, the price tag of that software is a bit much for myself, however with the combination of REW and VituixCAD you can complete all loudspeaker design needs for the low low price of free, although I'm sure the developers would welcome a donation.

For loudspeaker design, continuous FFT is simply not required. The dual channel sweep measurements available in the latest 5.20.10 preview is ideal for the task. With the aid of VituixCAD, full "spinorama" of the speaker power response and directivity index is available and all the crossover tools as well whether passive, IIR, or FIR filters. The crossover transfer function in VituixCAD can even be simulated in software using APO EQ so you can hear the result first hand with any multichannel audio device.

I am very familiar with VituixCAD and am happy to help you with it, but I would start by reading through the help file, specifically the "How to get started working with VituixCAD" section and the measurement guide for REW.
 

sm52

Member
Joined
Mar 14, 2019
Messages
863
Is there a difference in the graphs between the old ways and the new one?
 

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
You are asking "what's the difference between dual and semi-dual measurement?"... When done properly, all non-linearity in the amplitude response is effectively calibrated out with a dual channel measurement. This provides some benefit for loudspeaker design, such as being able to include a protection capacitor for delicate tweeters without affecting the measured response (apart from raised noise floor at low freq). As well, the response of the output filter of many load-dependent class-D amplifiers can be effectively removed from the measured response.

Take measurements yourself and see the result. Observe the calibration result to see how much difference the reference channel is affecting the response.
 

sm52

Member
Joined
Mar 14, 2019
Messages
863
I can't do it yet. Do you have saved measurements taken with one of the old methods (loopback timing reference or acoustic timing reference) and the new method of the same speaker with multiple drivers? I'm interested to see.
 

dcibel

Member
Thread Starter
Joined
Sep 10, 2017
Messages
151
For normal measurements with competent hardware, the measurements should not vary by much between timing reference only and timing + calibration. However, I will provide a more extreme example. I simply placed a 3" wideband driver on my desk for a quick and dirty measurement. The response itself is unimportant, just the difference between the two traces. This driver has an Fs around 120Hz, and I placed a large capacitor in series with it to act as a basic high pass filter, which provides a high Q peak around the driver Fs.

The purple trace is the "single channel" uncalibrated measurement, showing the raw response of the driver and capacitor interaction. The orange trace is the "dual channel" measurement, using timing + calibration, which is the same as the response of the driver without the capacitor in place, apart from the elevated noise floor at very low frequencies. Here, the effects of the capacitor are effectively removed by the automated calibration process.

53760


Here's what that calibration curve looks like:
53761
 
Top Bottom