UMIK-1, REW adventure

[Ron Texas]

I bought a UMIK-1 calibrated microphone. Mine came from Parts Express although you can get them directly from MiniDSP or Jeff Bozo's company. It costs 90 something dollars delivered. I used REW to do measurements and generate PEQ filters and typed the parameters into the parametric EQ in JRiver. The improvement in sound was very noticeable. Electric bass and female vocals picked up a lot of definition and detail.

The author is a member here. The software is amazing. It's not fully automated, but it did not take me long to get results. It appears to be one of those things that the more you know the more you can get out of it.

My system has a pair of LS50's on 24" stands about 20" in front of a big wall unit which is 18" deep, 36" high and 10' long. The speakers are 7' 4" apart and the listening position is at the tip of an equilateral triangle. The couch is 2' from the back wall. There is a 12" down firing sub off to the right side and outside of the triangle. The room is a finished attic with an irregular ceiling and wall to wall carpeting.

I took measurements placing the mike on two old speakers covered with a small wool blanket. The mike was where my head would be. It faced straight ahead at a spot midway between the speakers. (note: the USB cable supplied with the mike, is 10' and was not long enough. I had to use an extension.) The regular config file was used, not the 90.

Measurements were up to 200 hz (400 was tried, but made no difference.) I used variable smoothing and generated filters for "full range loudspeaker". The only other change was to limit boost adjustments to 3 db. If changed to 4db the results were very different.

Is there a better way to do this like doing left, right and sub separately and combining them? Should I point the mike directly at the speaker being measured?

Perhaps I am lucky, because the music sounds much better now...

 

[DonH56]

There is a long thread on AVS about using REW.

You should invest in a boom mic stand, e.g. https://www.sweetwater.com/c396--Microphone_Stands

Most folk use the 90 degree cal file and mount the mic pointing straight up so it senses all speakers equally without (or less of) the mic's directionality changing the response.

WIWFM - Don

 

[Snarfie]

I bought a UMIK-1 calibrated microphone. Mine came from Parts Express although you can get them directly from MiniDSP or Jeff Bozo's company. It costs 90 something dollars delivered. I used REW to do measurements and generate PEQ filters and typed the parameters into the parametric EQ in JRiver. The improvement in sound was very noticeable. Electric bass and female vocals picked up a lot of definition and detail.

The author is a member here. The software is amazing. It's not fully automated, but it did not take me long to get results. It appears to be one of those things that the more you know the more you can get out of it.

My system has a pair of LS50's on 24" stands about 20" in front of a big wall unit which is 18" deep, 36" high and 10' long. The speakers are 7' 4" apart and the listening position is at the tip of an equilateral triangle. The couch is 2' from the back wall. There is a 12" down firing sub off to the right side and outside of the triangle. The room is a finished attic with an irregular ceiling and wall to wall carpeting.

I took measurements placing the mike on two old speakers covered with a small wool blanket. The mike was where my head would be. It faced straight ahead at a spot midway between the speakers. (note: the USB cable supplied with the mike, is 10' and was not long enough. I had to use an extension.) The regular config file was used, not the 90.

Measurements were up to 200 hz (400 was tried, but made no difference.) I used variable smoothing and generated filters for "full range loudspeaker". The only other change was to limit boost adjustments to 3 db. If changed to 4db the results were very different.

Is there a better way to do this like doing left, right and sub separately and combining them? Should I point the mike directly at the speaker being measured?

Perhaps I am lucky, because the music sounds much better now...

I did the same with Foobar2000 incombination with the Mathaudio room EQ addon. My impression was that (taking my peronal roomaccoustic in consideration) that lows mids an highs where suddenly inbalance. The problem with my room acoustics is that the higer frequency range from 200 to 20.000 hz wher overamplified/scatterd (sometimes a 8db difference) . When using Mathaudio the whole freqency range was from one on the other moment ballanced (strait line between 20 an 20.000 hz). I enjoying now much more especially my older Cd's like Abbey Road from the Beatles the bass of Paul Mccartney is suddenly apearing way more stronger. Think roomacoustis software is way under estimated.

 

[Ron Texas]

Most folk use the 90 degree cal file and mount the mic pointing straight up so it senses all speakers equally without (or less of) the mic's directionality changing the response.

WIWFM - Don

After thinking about it the 90 degree orientation makes more sense to me when measuring multiple sources at once. I tried it and the filters were a bit different. Sound is very slightly different. Overall quality is similar.

 

[Ron Texas]

I did the same with Foobar2000 incombination with the Mathaudio room EQ addon.

That's something I want to experiment with.

 

[Cortes]

I've used rew filter in foobar with convolver. Up to now, a full disaster, I get only sterile sound.

What's your target curve in rew?.

After the creation of the rew filters, how do you measure the real response with the filters working?.

 

[Daverz]

I bought this Pyle mic stand

https://www.amazon.com/gp/product/B00SVRLPYY

Sturdy and a big improvement over trying to prop the little tripod that the UMIK-1 comes with on the edge of my listening chair at the right height.

I aim the mic at the midpoint between the speakers and check all the distances with a laser ranger, then play back left and right sweeps created by REW from my LMS server.

 

[Ron Texas]

The tutorial written by Jerry Austin says measure right+sub then left+sub and average them together. This gives a different result than all three at the same time. That surprised me. I moved the speakers a bit after doing the separate measurements. The difference between l and r are smaller now. It isn't possible to get them the same as the room does not allow a symmetrical layout. Separate side measurements sound just a bit better.

 

[Snarfie]

That's something I want to experiment with.

Let us know when you do an comparison. With mathaudio room eq the position of your mic in vertical. Depending if you listen close monitoring as i do on a bit more than 1 meter distance i have 9 measuring points each 25 cm apart dead center. I am using Superlux 999ECM mic.

 

[March Audio]

I've used rew filter in foobar with convolver. Up to now, a full disaster, I get only sterile sound.

What's your target curve in rew?.

After the creation of the rew filters, how do you measure the real response with the filters working?.

Is your target curve flat? That would definitely lead to a sterile sound. In room measurements should not be flat, only flat when measured in an anechoic chamber.

Typically a 10 dB slope from 20Hz to 20kHz works well.

You can save the REW measurement sweep as a wave file in the generator tab. Then Play it back through foobar.

 

[Cortes]

@march,

yes I've tried with flat curve. I have to try with the type of target you say.

The issue with play it back with foobar is that (as fas as I know) rew does not measure the new output.
I'll try Equalizer Apo to see whether I can measere the corrected response.

BTW the fact that REW only does frequency correction, and not fir as Dirac and other commercial products do, is really a limiting factor for good room correction?.

Ah, a warning to potential buyers of umik-1. When I purchased it I thought I would gain knowledge through rew, and later I would purchase a more advanced soft to use with it. However, later I've learned that umik-1 only works with Dirac & Rew. It does not work with other software, such as sonarworks (it needs another recording frequency) or ARC 2 from what I've read.



Best.

 

[Kal Rubinson]

However, later I've learned that umik-1 only works with Dirac & Rew. It does not work with other software, such as sonarworks (it needs another recording frequency) or ARC 2 from what I've read.

AFAIK, the shoe is on (off?) the other foot. Umik-1 is a pretty generic USB mic although of fixed frequency. OTOH, software packages can, and do, restrict what mics and at what frequencies they will work. From miniDSP:

"Which software is compatible with UMIK-1?
The UMIK-1 is essentially a USB audio interface. It will be discovered in your PC/Mac as a USB sound card the same way your USB DAC or other USB audio devices will be discovered. In other words, any software supporting standard Audio input device (WDM/ASIO) will work with the UMIK-1. Here are some links of recommended software:

• Room Eq Wizard (REW) / Recommended since plug&play support for SPL reading
• Arta / Great commercial package with lots of features
• HolmImpulse / Another great well known acoustic measurement software
• Smaart / A well known commercial audio package for live sound

Have another package you'd like to recommend? Email us and we'll add it to the list. "

 

[Ron Texas]

You can save the REW measurement sweep as a wave file in the generator tab. Then Play it back through foobar.

Cool!

 

[Daverz]

Nice thing about the UMIK-1, being a digital mic, is that the correction file for your mic includes the sensitivity of the mic (reading of the mic at 94 dBSPL). REW reads this value out of the correction file, so you don't need an SPL meter in REW to calibrate the SPL level.

 

[goodkeys]

If you are tech savvy I can recommend Python Open Room Correction (PORC) to generate the correction filters:

https://github.com/bstegmaier75/porc

It's free, but it runs on command prompts - no GUI. So it takes a bit of reading on how to use it. The filters it generates are top notch though. For me the results with it were on par with Dirac Live (much simpler to use, but cost around 600$), and better than Sonarworks 3.

For hassle-free systemwide correction I am using the PORC generated filters in Equalizer APO (Windows only):

https://sourceforge.net/projects/equalizerapo/

 

[March Audio]

@march,

yes I've tried with flat curve. I have to try with the type of target you say.

The issue with play it back with foobar is that (as fas as I know) rew does not measure the new output.
I'll try Equalizer Apo to see whether I can measere the corrected response.

BTW the fact that REW only does frequency correction, and not fir as Dirac and other commercial products do, is really a limiting factor for good room correction?.

Ah, a warning to potential buyers of umik-1. When I purchased it I thought I would gain knowledge through rew, and later I would purchase a more advanced soft to use with it. However, later I've learned that umik-1 only works with Dirac & Rew. It does not work with other software, such as sonarworks (it needs another recording frequency) or ARC 2 from what I've read.



Best.

The sloping curve is due to in room non anechoic measurement. There are a few similar variations I will post some plots later.

I have had good results with simple non Fir EQ. Just don't make more than Broad adjustments above 200Hz. If you want to get sophisticated try audiolense or Acourate software.

If foobar is performing the convolution it REW will measure the corrected output when you play the measurement sweep wave file. Umik works like a USB soundcard, so should work with any software that doesn't require beskope hardware.

 

[Ron Texas]

Default setting in REW provides a sloping curve.

 

[RayDunzl]

The tutorial written by Jerry Austin says measure right+sub then left+sub and average them together. This gives a different result than all three at the same time. That surprised me.

Left, Right and Both (blue).

Not averaged.

Though neither speaker has a frequency response dip at 48Hz, when they play together, their reflected waves are out of phase by 180 degrees at that frequency. The room is rectangular with the left rear corner open, so there are two room lengths at play.

At least, that's what I think based on what I see.

The simple average (green) gets it wrong, the vector average gets it right (violet), but should be 6dB higher, like the blue from above (that would be summed, not averaged, maybe @JohnPM can think about that for a future release.


The simple (green) and vector (violet) averages:

Low Frequency Phase, left and right - 180 degrees opposed around 48Hz

 

[Ron Texas]

I found the Vector Average button. It makes a suck-out at 139 hz look a lot worse on the graph.

 

[JohnPM]

When using vector averaging make sure the measurements are suitably time aligned or have a common timing reference. The trace arithmetic sum is a vector sum, provided the A and B inputs both have impulse responses or phase data.