Convert measurement to cal?

[Fran Guidry]

This last week or so has been a great learning experience. I'm still wandering in the fog but occasionally find the road.

My aim is to evaluate the changes in my recording/monitoring space as I add OC703 panel treatment. As I've read posts here to expand my understanding, I've learned that I should be able to create a calibration file that reflects my entire chain, including speakers, mic, preamp, and audio interface.

In order to get the best measurement of the total system, I fed some wire out the window and rigged up my speaker (Dynaudio BM6p) and mic (DPA 4061 omni) outside on a 10 foot ladder.

Here's the result:



I've applied 1/6 octave smoothing.

Now, I've exported this measurement file as text, renamed it to a .cal file, and pulled it into REW as my calibration file.

But ... I've read about reducing the impulse response window to eliminate first reflections from corrupting the measurement.

Here's one post with JohnM giving a formula for reducing the IR window.

http://www.hometheatershack.com/foru...tml#post122775

Since my ladder was 10 ft, I can use a 20ms window, giving a 50hz cutoff. I've saved this measurement as a .cal file and it shows somewhat different values in the low end than the 500ms window.

Since I'm recording acoustic guitar exclusively, I lose interest below about 80hz <grin>. Any recommendations on which .cal file I should use between the 20ms and 500ms versions? Should I edit out the values below 50hz?

Am I on the right track with my efforts to create a .cal file from this measurement?

Thanks for any insight.

Fran

[brucek]

I've learned that I should be able to create a calibration file that reflects my entire chain, including speakers, mic, preamp, and audio interface.

Can you explain what this means, what you want to actually measure, and your desired result?

brucek

[Fran Guidry]

brucek wrote:

Can you explain what this means, what you want to actually measure, and your desired result?

brucek

Hi, Bruce,

My desired result is to evaluate my room treatment.

So what I want to measure is the frequency response of my room.

My logic went something like this:

We run a calibration measurement on our soundcard, then use that to remove the effects of the soundcard from the measurement.

We apply a calibration curve for the input transducer, then use that to remove the effects of the mic from measurement.

I don't have a calibration curve for my mic and I haven't found a frequency response curve so I can create my own calibration file. But I can generate a calibration file for the mic just like I did for the soundcard, but I have to include the speaker in the calibration since I don't have a calibration file for that either.

... Wow, the light just came on to some extent - I should have run the calibration measurement procedure instead of the room measurement tool when I had the speaker outdoors ...

In any case, I made a jig to hold the mic 1 foot from the speaker, ran mic and speaker cables out the window, and set the speaker on a 10 foot ladder over 10 feet away from the house then I ran the measurement process.

So now I have a "room measurement" with no room.

If I convert that into a calibration file, room measurements made against that file will include calibration for the mic, the preamp, the soundcard, and the speaker, so the measurement would include only the effects of the room.

Exporting the measurement file to .txt appears to create a list of frequencies and sound levels, just like the list in a calibration file. Changing the extension to .cal makes the file appear in the list of calibration files in the settings/soundcard screen. When I select my converted file REW reports that it loaded a calibration file.

So my thinking is that with that in place I can now start installing room treatment and take measurements as I add and move panels, seeing only the room and the effect of the panels, with my recording/measureing system factored out of the results.

Fran

[Fran Guidry]

I've continued to explore and think about calibration vs measurement files. When I examined a native .cal file compared to my .txt file I realized that the .cal file is adjusted to 0dB, while my .txt file reflects the 75dB target level.

The result of this is that my measurements made with the converted .txt file in place for calibration are centered around 0dB rather than 75dB.

If I want my graphs to be comparable to the usual ones posted here I should put the speaker back outside and run the Settings/Soundcard/Calibration/Measure procedure. Or would I get the same basic result by pulling my converted file into a spreadsheet and subtracting 75dB from each measurement??

Fran

[brucek]

but I have to include the speaker in the calibration since I don't have a calibration file for that either.

Theres the rub. There's a failure in your logic.

If you want to evaluate your room, it must include the equipment that generates the sound. It's the test equipment that requires its influence to be negated. This means that once we have removed the computer, soundcard and the microphone from influencing the result, the response of the system can be considered valid.

You don't have a calibration file for your microphone. Then you must make one. It's quite simple if you can obtain a graph of its response.

brucek

[Fran Guidry]

brucek wrote:

Theres the rub. There's a failure in your logic.

If you want to evaluate your room, it must include the equipment that generates the sound. It's the test equipment that requires its influence to be negated. This means that once we have removed the computer, soundcard and the microphone from influencing the result, the response of the system can be considered valid.

You don't have a calibration file for your microphone. Then you must make one. It's quite simple if you can obtain a graph of its response.

brucek

Bruce, thanks for taking the time to discuss this.

Our total system is:

Soundcard out
Power amp
Speaker
Room
Mic
Mic Pre
Soundcard in

Each of these has some variation from linearity. We compensate for that non-linearity by creating a file that describes the frequency response of the item.

REW provides a procedure for creating such a file. We use this procedure and a loopback connection to create a cal file for the soundcard out and soundcard in.

In the usual case I've seen discussed here we assume linearity for the power amp and mic pre, that is, we do not usually include them in any calibration, although there have been discussions of padding to allow calibration to include the mic pre.

Since most of us are lacking anechoic space and reference transducers, we don't have a way to calibrate the mic or speakers, so we construct a mic calibration file by examining the frequency response curve of the mic and building a list of values that represent the deviation from flat response.

Then with these calibrations in place, we measure the combined response of the speaker and room.

But if we had an accurate frequency response curve for our speaker, we could create a cal file to compensate for it and eliminate it from the measurement, so we would be measuring only the room.

If I'm right so far, then the next step might be:

Eliminate the room
Put all the components including the speaker and mic in the "loopback."
Measure the combined response of the system all at once.
Using that measurement as a cal file, return to the room and measure.
The result is the room response rather than the speaker/room response.

We don't need the frequency response of the soundcard out and in to create a cal file, we measure and have a combined response. We don't know if our nonlinearities are in the output or input and we don't care because we compensate for the combined response.

By the same logic, we don't need the frequency response of the mic and speaker individually, because if we can get them out of a room we can measure their combined response, and then compensate for their combined response.

And, as I understand the discussions here, we can "get them out of a room" by placing the speaker some distance away from any reflective surface, placing the mic 1 foot from the speaker, and narrowing the impulse response window so our measurements do not include the first reflection. A 10 foot ladder gives us a 20 foot reflection path, allowing a 20 ms IR window.

I'm starting to doubt that I really need to go through all these hoops, but I got caught up in the power of this wonderful REW program and got carried away.

Fran

[brucek]

Fran,

You're missing the point somewhat. Let's look at it another way, so I can illustrate the folly of including your system equipment in a calibration file.

Imagine I have a simple amplifier and a speaker in a room. The amplifier and speaker have a horrible response, which is only exacerbated by the horrible room. It's a bad situation.

I tell the engineer that I want to add equalization (either electronic or physical treatment) to make the response of this poor system very flat, so when I play my CD's through it, it sounds flat.

The engineer attaches a piece of test equipment (sound source and microphone) that is absolutely flat (because he has calibrated it off site), and measures the response in the room and offers equalization (either electronic or physical treatment) to render the response of this poor system very flat. Now when the user plays any CD's through his system it sounds perfect.

Everyone is happy.

Do you see how incorrect it would have been for the amplifier or speaker to have been included in his calibration file? It's the number one rule in engineering measurements, that your test equipment is neutral. Believe me, I've been doing it for 35 years.

brucek

[Fran Guidry]

Bruce, you're a wonderfully patient person. Thanks very much.

My last thought, I promise.

What if, in the scenario you describe, the goal is to isolate the impact of room response, rather than to create overall flat response?

Rather than hoping to listen to a CD and have it be wonderful, the goal is to reduce room effects to a minimum, because the purpose of all this is to be able to record an instrument playing in that room. If our hypothetical room treatment compensates for problems in the power amp and speakers that compensation is stamped on the recording.

I've clearly taken us into philosophical hair splitting, and I apologize. And thanks again for your patience and knowledge.

Fran

[brucek]

What if, in the scenario you describe, the goal is to isolate the impact of room response, rather than to create overall flat response?

It's not possible with your scenario.

Think of it another way.

We decide we want to isolate the room effect only.

(Not that you would be able to do it), but let's suppose we do it your way and include all our test equipment and our sound system in a calibration file, such that it is completely flat from 0Hz-20KHz.

Then we test the room and install treatment so we end up with a flat response.

Now what happens when I play music with that imperfect speaker and amplifier in that room. The response will be off by an amount equal to the imperfect response of the speaker and amplifier.

This occurs because we included the sound equipment in the calibration of our test equipment. You can't do that.

brucek

[Tonmeister]

Hello Fran, hello brucek,

there is always one bad thing in equalizing a soundsystem without tuning the room: the first sound wave would be EQed to the frequency response matching all reflections in the room. So first aim has to be to correct the room with hardware material (diffusers, resonance plates, Helmholtz resonators and other stuff). When then room reflections and reverb time are smooth, you may EQ the electronic equipment at a second step.

Fran, ist that what you want to do?

Yours

Olaf

[Fran Guidry]

Tonmeister wrote:

Hello Fran, hello brucek,

there is always one bad thing in equalizing a soundsystem without tuning the room: the first sound wave would be EQed to the frequency response matching all reflections in the room. So first aim has to be to correct the room with hardware material (diffusers, resonance plates, Helmholtz resonators and other stuff). When then room reflections and reverb time are smooth, you may EQ the electronic equipment at a second step.

Fran, ist that what you want to do?

Yours

Olaf

Yes, Olaf, I don't intend to EQ my system. My only interest is measuring and improving the response of my room.

I have 13 broadband absorber panels installed now, and just made 9 more. This will allow me to install panels over the listening position (which is also my recording position) and cover additional corners. I'm hoping for another increment of improvement.

Fran

[Tonmeister]

Okay Fran,

first let me ask you: what ist the longest diagonal in your room? This value is for the deepest frequency your room would be able to reproduce correct. It hast to be, that minimum one half-cycle hast to fit in there.

Frequencies below this value never would be reproduced correct in that room in case of acoustic limitations.

Second it is a good idea, to destroy reflections as good as possible using diffusers and non-parallell walls.

Third aim is to level reverberation time to a good value. You will find clues on the www, especially at the website of the AUDIO ENGINEERING SOCIETY [AES], the German "Institut fuer Rundfunktechnik" [IRT] or at the "Verband Deutscher Tonmeister" [VDT]. Look for MULTICHANNEL SURROUND SOUND SYSTEMS AND OPERATIONS, Reference Room (AES TC-MBAT Infomationsdokument: Mehrkanal-Surround-Systeme und Anwendungen)

Sincerely
Olaf G.

[Fran Guidry]

Tonmeister wrote:

Okay Fran,

first let me ask you: what ist the longest diagonal in your room? This value is for the deepest frequency your room would be able to reproduce correct. It hast to be, that minimum one half-cycle hast to fit in there.

Frequencies below this value never would be reproduced correct in that room in case of acoustic limitations.

Second it is a good idea, to destroy reflections as good as possible using diffusers and non-parallell walls.

Third aim is to level reverberation time to a good value. You will find clues on the www, especially at the website of the AUDIO ENGINEERING SOCIETY [AES], the German "Institut fuer Rundfunktechnik" [IRT] or at the "Verband Deutscher Tonmeister" [VDT]. Look for MULTICHANNEL SURROUND SOUND SYSTEMS AND OPERATIONS, Reference Room (AES TC-MBAT Infomationsdokument: Mehrkanal-Surround-Systeme und Anwendungen)

Sincerely
Olaf G.

Olaf, thanks so much for taking time to share your knowledge.

Here's a model of my room:



The corner to corner length (excluding the hallway) is approximately 24 feet (7.3 meters or so). I'm really lucky to have such a generous sized room for my home studio. I suppose my room should be OK down to 24 hz or so, yes? I'm only interested in recording acoustic guitar, so I really only need to be able to capture and playback down to 80 hz or so, depending on how my guitar is tuned.

Unfortunately, my wife frowns on my making the walls non-parallel. I'd _really_ like to knock down some walls and raise the ceilings but she's not at all interested in spending our money that way. So far I'm using broad band absorbers (4" thickness of Owens Corning OC703 compressed fiberglass) to treat reflections and attempt to reduce low frequency cancellations and reinforcements.

In the past when I've tried to access information at the AES it's only available for members (or at least that's what I remember) and my German language skills are utterly nonexistent. I have some suggestions for BCC papers but haven't delved into them yet.

Thanks again for your advice.

Fran

[DrWho]

In concept, making a cal file out of your speaker measurement is a great way to isolate the frequency response effects of room treatment in your room.....

However, the problem is you're only measuring on one-axis and your speaker is spraying sound in many different directions. The sound you measure at the listening position is the culmination of all that sound bouncing around the room and landing at the listening position. The tonal balance of the sound to the sides of the speaker isn't going to have the same frequency response as the sound measured on-axis. Because of this, the frequency content of the reflections are different.

However, the thinking is still flawed because the purpose of acoustical treatment is to deal with the time-arrival / frequency balance of all the later arriving reflections. As the reflections arrive later and later, our ears perceive that reflection as shifting from a tonal imbalance to a distinct reflection. The significance is that a late enough arriving reflection may induce a big peak/dip in the frequency response, but it won't be perceived as a tonal imbalance. Because of this, adding EQ to make it measure flat actually creates a new tonal imbalance and doesn't address the audible impact of the later arriving reflection.

So all that said, I would highly recommend diving into the ETC and exploring the audible correlations of that plot.

[brucek]

The significance is that a late enough arriving reflection may induce a big peak/dip in the frequency response, but it won't be perceived as a tonal imbalance. Because of this, adding EQ to make it measure flat actually creates a new tonal imbalance and doesn't address the audible impact of the later arriving reflection.

I would think that if we are EQ'ing at our listening / measurement position, and the gating of the impulse response is set to the point where the signal enters the noise, that all reflections (including late arrivals) would be taken care of?

brucek

[Tonmeister]

Hello again,

planning a good studio for reliable (sound engineering) listening is something, that deals with many complex processes. Nevertheless your overall transmission from source to loudspeaker should be as flat as possible, to ensure, that the tonal balance of the first arriving (not reflected) sound is correct. All electronic equalizing is a sellout, that makes the original sound bad to get the reflected soundmix better. So it is always the first aim to tune your room, not EQing electronically. REW is a good tool helping withal, but use it in the right way:

First take a look at the WATERFALL tab - you see low frequency resonances of your room that might be antagonized by a combination of inclined resonance plates (some said 7° or more, no you don'n need to break down your walls! ;-)) and Helmholtz resonators.

Second zoom deep into your IMPULSE tab - from 2.5ms to about 15ms you will see first reflections from floor, walls and ceilling. Try to scatter them with diffusers and again inclined resonance plates. Incoming wave angle is equal to leaving wave angle, use a mirror, to locate the best place for the diffusers! It has to be, that REW will show no discrete peaks between 0.5ms and 15ms!

Third make a RT60 measurement (Topt, one third octave bands). It shold be 0.25*(V / V0 ) with V=volume of your room and V0=100m³ (cubic meters). For example: a Room of 7 meters length, 5 meters widh and 3 meters height should be RT60 at 0,2625s from 200 Hz to 4 kHz and +/-0.05s tolerance.

If you have done those three things, you may have a look at the tab MEASURED and if your over all frequency response is not within 4dB tolerance (+/-2dB) from 40 Hz...16 kHz, then go to FILTER ADJUST and EQ your setup... but not until having done step one, two and three!

Yours


Olaf G. Guenther

[DrWho]

brucek wrote:

I would think that if we are EQ'ing at our listening / measurement position, and the gating of the impulse response is set to the point where the signal enters the noise, that all reflections (including late arrivals) would be taken care of?

brucek

Just as a very simple example....say you have a single reflection that arrives 30 seconds late. If you make your window 45s long, you're going to see dips and peaks in the frequency response. Do you think an EQ is going to get rid of the 30 second delay sounding like an echo?

[brucek]

Do you think an EQ is going to get rid of the 30 second delay sounding like an echo?

But his room isn't large enough to produce a reflection of 30 seconds. Sound travels ~1100 feet/second, so his room would produce reflections inside the ears gate time of ~50msecs and should not be perceived as two discrete sounds. Is that not correct?

So, all the reflections will be gated by REW and produce a response graph and waterfall that can easily be adjusted with treatment and completed with EQ if needed.

brucek

[DrWho]

The Haas Window is non-linear with frequency. Also, the smearing effect still continues well inside the Haas Window. EQ doesn't fix the smearing (even though the smearing isn't heard as two distinct sounds).

The shape of the ETC has a huge effect on what is perceived as well. Specular reflections reduce intelligibility and ruin imaging...same with sparsely distributed reflections. You want the reflections to decay in amplitude over time too....don't want swelling or flat spots, single spikes, etc...

Long story short, EQ'ing a room with a poor ETC to measure a flat frequency response is not going to sound as good as a room that has properly treated the reflections....even though the frequency response of the latter could be quite similar to the EQ'd system.

In a studio setting, you want to try real hard to achieve an ITD that is as long as possible as this will dramatically improve the referencing of the mix. And then usually, you want the Haas Kicker (the first reflection) to be very apparent so that it and all the reflections following are perceived as separate from the direct sound.