Measuring after correction curves are generated?

[robbnj]

Hi all.
'Sorry if this has been asked a dozen times and is a total neophyte question, but I just can't seem to word my search correctly to get an answer:
Is it possible to create a correction, and then immediately test how that correction affects the measured response/output of a speaker?

I know I can create and add a convolution in JRiver, or apply a curve I created through APO system-level EQ, but I can't run the REW sweep through JRiver, and I'd like to avoid APO for fear I will forget it is active, and run corrected sweeps while testing various speakers.
I was thinking along the lines of creating and saving a correction and then being able "apply curve to output sweep" in REW, and run the test again.

Am I totally in left field with this idea?

 

[ddude003]

I am thinking you would want to check out the Generator to create a file that contains the sweep tones with start and stop (timing) tones to play on JRiver while you are taking your measurements and after to see what your correction looks like using convolution... This would include your entire kit before and after correction... https://www.roomeqwizard.com/help/help_en-GB/html/siggen.html

 

[robbnj]

That's what I was hoping to avoid as it's kind of clunky, but if I have to do it, then I guess I have no choice :-)

 

[ddude003]

I am wondering how having a separate convolution engine in REW or elsewhere in your testing signal chain would reflect the reality of using your own JRiver, APO and whatever signal chain... You may as well just use some more REW maths to simulate your response...

 

[robbnj]

I am wondering how having a separate convolution engine in REW or elsewhere in your testing signal chain would reflect the reality of using your own JRiver, APO and whatever signal chain... You may as well just use some more REW maths to simulate your response...

If I read your post correctly: I would assume that if REW could apply the correction to a sweep, the correction would only be applied to that sweep and not to everything the system played when REW was closed down. For this thread's purpose, I was thinking of how some programs out there (Audacity, LP de-click/pop progs, JRiver) have a "preview" where you can listen to an applied filter on-the-fly, and it would be nice if REW did the same.

In other words:
1, Run/measure a sweep, see the response
2. Generate a correction
3. Run the sweep again with correction applied, see the response

instead of
1. Run/measure sweep, see the response
2, Generate a correction
3. Export a correction
4. Generate a sweep file in REW and save it
5. Open the sweep in Jriver and apply the correction
6. Save the corrected sweep
7. Open the corrected sweep in REW and measure, see the response

Just a few extra steps if you only do one sweep, but if you are doing 5 or 10 or more (that's nothing for someone who is doing speaker design), those few extra steps become a LOT.

 

[jtalden]

I may not be following your concern, but...
The Denon, B&O, and JRiver in your system will not impact the SPL; only the filter will do that.
So:
Measure and create filter in REW EQ window.
Choose to look at the measurement, or the predicted response, or both together as shown below.

If you then save and import the filter into JRiver and remeasure, there will be no difference from the REW predicted response unless the room influence was different.

 

[ddude003]

@robbnj I think your off a bit with the steps... I do it like this...
1. Generate sweep file(s) with REW... This only needs to be done once...
2. Run the sweep file through my kit/signal chain while measuring the sweep and seeing the response...
3. Generate a correction file and load it into my convolution engine (Roon, LiquidSonics or CamillaDSP as part of my kit)...
4. Run the original sweep file again thru my kit/signal chain with correction applied by my convolution engine, see the response...
Extra credit...
Compare the measured corrected response with the REW predicted response as @jtalden points out above...

Of course this is for room correction and not speaker design...

 

[robbnj]

Thank you both for the reply. ddude, I think you are a little more on track with where I am headed, but missing one thing: Even as a tinkerer, I would do multiple correction curves and want to test each one out. That means exporting to (in my case) JRiver, applying and saving, thun opening in REW and testing all over again.

I'm wondering if it's maybe not possible to make the REW program do what I'm thinking. I think it would be a big timesaver for someone really digging into correction curves for a serious home system, and really simplify the process. But, I'm no software engineer and this program is amazing to begin with...

Edit To Add: I wasn't wrapping my brain around the idea of playing a "nude" sweep file with the convolution applied by JRiver.I was thinking that REW had to play the sweep in order to capture and analyze it.
But that sill leaves the user having to do those extra steps instead of just having REW apply the correction to the sweep and re-measure. Again, I know nothing about programming and how difficult (if not impossible) this may be.

 

[jtalden]

As @ddude003 suggested, once you have a correction filter there is no need to measure again. If it is a wav filter just drag it into your original REW measurement file and use trace arithmetic (A*B) on each measurement you want to "convolve". You will also need to adjust the levels back down to the original levels if you wish to overlay the A*B results with the originals. It goes pretty quickly after doing it a couple of times.

 

[ddude003]

I'm wondering if it's maybe not possible to make the REW program do what I'm thinking. I think it would be a big timesaver for someone really digging into correction curves for a serious home system, and really simplify the process. But, I'm no software engineer and this program is amazing to begin with...

Most users of REW would agree that REW is an amazing bit of software...
And here is a bit of an example of the effort it takes to throw together a bit of convolution software... I'll use CamillaDSP as an example, as it is Open Software and is available for anyone to download and use or change to their hearts content... Here are the statistics of the amount of lines of "code" it uses... CamillaDSP itself is written primarily in Rust, the rest are utilities to build, test and use CamillaDSP as well as some documentation... Lets just say it is no small effort...

And it looks like Equalizer APO is of similar size and effort primarily written in C++...

 

[ddude003]

By the way... I don't think you even need to re-measure the room over and over if you are just experimenting with room curves... There is no reason to go back to my step 2 above... I often create several correction filters with my step 3 above... Then you have a handful of filters to load into your DSP/Convolver and try...
You might take a look at Mitch's Hang Loose Convolver at https://accuratesound.ca/products.html... I think it will allow you to load up to 6 correction files, and seamlessly switch between them in real-time...

 

[robbnj]

I appreciate the responses but think I’ve done a poor job of explaining what my idea is, or maybe REW already does it and I just don’t realize it.
But I thank you guys for trying!

 

[ddude003]

I know what you are asking @robbnj... You want REW to apply the filters it creates to REW's output sweep tone(s)... Maybe you could ask the author of REW for this enhancement...

By the way, have you looked at the number of Hardware and Software Equalizers, Digital Signal Processors, Convolvers and other utilities REW already supports?

 

[robbnj]

I

I know what you are asking @robbnj... You want REW to apply the filters it creates to REW's output sweep tone(s)... Maybe you could ask the author of REW for this enhancement...

By the way, have you looked at the number of Hardware and Software Equalizers, Digital Signal Processors, Convolvers and other utilities REW already supports?

I'm confident I have barely scratched the surface of what it can do. I have a tough time reading a guidebook and then trying to do.I liek to dive in and try stuff, and then learn new things and try them as I go.
Hypreractive, lol

I thought this was the REW support forum and it might get to him, but I will definitely try a shoutout.

 

[John Mulcahy]

There's no technical limitation that would prevent applying filters to the sweep, REW has what it would need amongst its 270,000 lines of code but there are practical issues in having an interface that would let users choose which filter set to apply and easily distinguish measurements made with filtering from measurements without. The result should match the REW prediction, however, which is generated by applying the filters to the measurement's impulse response, so there wouldn't be any practical benefit to offering that feature.

 

[skid00]

If you install eqAPO, you can easily export the generated EQ file as text, and copy/paste it into an eqAPO EQ file. You can have multiple files, and select any number of them at the same time, in eqAPO's GUI.

So you can have REW running, and then test/retest your curves. Yesterday, I tested with and without EQ at 1, 2, 3, 4, 5 feet (each speaker) and at my head's normal listening position, for both left and right speakers.

I have previously posted small examples of this.

 

[robbnj]

I posted this in the REW thread in another forum I belong to, and thought I would also share here since it involves direct use of REW.
Please remove if the share isn't appropriate or allowed.

So my UMIK-1 arrived and I had a few minutes to play while wife was away.
As promised, I did a comparison of the UMIK against the other mics I already had (and had compared against each other with surprising results).
It was far from "scientific, but I tried to make the test accurate and consistent as possible (at least enought o tell me what I wanted to know)
I tested each mic four times in a row. At the beginning of each test I confirmed the signal was coming from that particular mic, and on the 4th test I wrapped it in six layers of bath towel to make sure it was still where the signal was coming from. In other words, each response definitely came from the mic I was testing.
All were mounted on a tripod, 36" from the floor, pointed at the ceiling. They were arranged on a tripod in a line about 3" across, parallel to the front wall, as the apex of an equilateral triangle.
I calibrated REW using a calibrated RadioShack SPL meter.

The list of "competitors": UMIK-1 ($80); Dayton Audio IMM6 ($30); Audyssey "Eiffel Tower" ($40); 15+ year old Panasonic mic element from Digikey (probably $2.00 or less when purchased).
No calibration files were entered. This is just raw recording. 0-20KHz sweep.

My takeaway (just my personal evaluations and feelings; don't get angry if your results vary): These microphones are far more similar in what they pick up from my system than I ever expected or would have imagined. In my opinion that's a good thing because CONSISTENCY. But It also bums me out a bit because PRICE. I'd almost wonder if they aren't all the same capsule model in different clothing, they are THAT close. Let's put it this way: If you applied these as an EQ curve to your system and compared all four, I would be very surprised if you could tell them apart by listening.

For ME, I would have no problem using the cheapest of them to calibrate a system, once I dialed in a calibration curve for it, or saving that, I would have no problem using the cheapest calibrated one, which is the Dayton IMM6. I had read several complaints that the Dayton was lacking in it's ability to accurately pick up low frequencies. If that's the case, then it seems the Audyssey and the UMIK have a similar deficiency.
I will admit that the calibration file for the UMIK has data points than the IMM6, but I have to wonder how much that can matter when we are talking frequency responses seen in speakers in an average home setting.
And yes, the response curves are with Audyssey engaged. They're a mess, and are similarly messy when I used my IMM6 with the cal file loaded. I thought Audyssey shined in its handling of bass, but the graph is not all that different with or without Audyssey. I guess I have some troubleshooting to do.
Maybe a MiniDSP will be in my stocking next month...