My first REW Measurement - DIY Speakers - Did I get speaker polarity wrong?

[lucian]

Did u measure both speakers on the same output channel? If not maybe switch the channels and see if the problems travel with channel.

I recently had a broken ground line at one channel of a cinch cable. That sounded very much off too. Bought new speakers just to find out it was just the cable. So the cable guys are actually right - cable matters.

 

[Keith_W]

Putting the speaker issue aside, is your room highly reflective? Lots of glass, bare walls and ceiling, hard floor? Because the only way I can get something like that in my room is to maximize my reflections. And I do have a lot of them.

There is no need to ask, you just need to download his MDAT and look at the ETC to see a tonne of early and loud reflections.

As to whether those reflections are creating that upward tilting frequency response: the answer is no, they are not. Apply a FDW of 1 cycle and you will see.

At this stage I would suspect a measurement error. Please confirm the following:

- you are using ASIO or WASAPI Exclusive mode (if you are using "Java" mode, make sure EXCL is selected). If you are using WASAPI Shared, the signal goes through the Windows mixer and might be subject to all sorts of weirdness from sample rate changes.
- Read this post by @JohnPM about Windows assuming that your mic is noise cancelling.
- you did not mention what mic you are using and what interface. Please state what you are using so that we can check that it is appropriate for taking measurements.

If all else fails, I would suggest you borrow a microphone that is known to work and take some measurements.

 

[nikb]

Putting the speaker issue aside, is your room highly reflective? Lots of glass, bare walls and ceiling, hard floor? Because the only way I can get something like that in my room is to maximize my reflections. And I do have a lot of them.

If you can take three measurements it can help tell if this is room reflections. Point the speakers just to the outside of the mike, then straight into the room, then half way between. If this is a reflection issue, that upper frequency tilt should change.

Change, not disappear... unless you are very lucky.

Yes, pretty much this. The room is pretty bare and sparsely furnished. Brick wall and a conctrete ceiling. Only one small rug.
I will make measurements with the speakers pointed as you suggested.

I think a big rug would be a good investment. Not sure, what I can do about the walls, though.

 

[Sokel]

Yes, pretty much this. The room is pretty bare and sparsely furnished. Brick wall and a conctrete ceiling. Only one small rug.
I will make measurements with the speakers pointed as you suggested.

I think a big rug would be a good investment. Not sure, what I can do about the walls, though.

What mic are you using to take those measurements?
Calibration file loaded ok?

 

[Sokel]

What mic are you using to take those measurements?
Calibration file loaded ok?

..and I'm awnsering to myself, yes, seems ok:



...on the other hand...

That can't be right, usually the drivers used by Troels are high quality, no way they have so much distortion.
Also enormous noise floor. Too much gain somewhere?

 

[Juhazi]

What microphone are you using? Is it chosen in Preferences table? A laptop's built-in mic would produce that kind of response

 

[nikb]

What microphone are you using? Is it chosen in Preferences table? A laptop's built-in mic would produce that kind of response

I bought the microphone online some 10 years ago for some room-correction experiments (different room).

It came with calibration files for 0° and 90°, a preamp and the microphone itself. On the preamp it says to turn up gain half way, balance to the middle and pan to the center. That is what I did. I did have to turn on phantom power on the pre to get any input. It is turned off on the interface.

 

[Sokel]

I bought the microphone online some 10 years ago for some room-correction experiments (different room).

It came with calibration files for 0° and 90°, a preamp and the microphone itself. On the preamp it says to turn up gain half way, balance to the middle and pan to the center. That is what I did. I did have to turn on phantom power on the pre to get any input. It is turned off on the interface.

View attachment 486856 View attachment 486857 View attachment 486858

Seems like a sanity check is needed.
Can you hit RTA window, with the following settings and the mic online and show us what it looks like?
No sound must be played back during the listening.

(Y-axis is SPL units)

 

[nikb]

At this stage I would suspect a measurement error. Please confirm the following:

- you are using ASIO or WASAPI Exclusive mode (if you are using "Java" mode, make sure EXCL is selected). If you are using WASAPI Shared, the signal goes through the Windows mixer and might be subject to all sorts of weirdness from sample rate changes.

My interface is a motu m2.

I am using my desktop computer running Arch linux. No asio on linux, but usually sound just works. Not using any linux low latency sound drivers like Jack or PipeWire right now (the closest thing to asio) as I did not think that was needed. There might be some resampling going on in the beackground.

 

[Pareto Pragmatic]

There is no need to ask, you just need to download his MDAT and look at the ETC to see a tonne of early and loud reflections.

Given questions on settings and such, I do prefer to wait to make sure all that is worked out before I look at any measurements. "First" measurement are often useless due to various things that happen at the start of the learning curve. You yourself suspect measurement error, right? So telling me to look at the data seems... premature.

My reason for suggesting what I did is that it can be done easily with skills already possessed. The multiple measurements seem reliable, if not valid.

Unreliable measures are worthless. Reliable measures, even if of low validity, can be used to look for differences. A bad survey won't tell you what a population thinks, but if it is reliable the changes over time (different measures) can have use value, enough to see if something is more or less popular at least. No way to tell how popular, but more or less, that works with a reliable survey.

In audio terms, I can use an uncalibrated mic to track differences in frequency response between measurements for many parts of the frequency response. Not with great precision, and certainly not full spectrum. But if I wanted to see how an EQ at 500hz performed, I could do it.

Not at 100hz, as I recall. Big mismatch on either side of that on the uncalibrated mic I used when I used it. Which was when I was trying to learn how to use REW, to see if it was something I was willing to pursue.

It was, and a calibrated mic was purchased shortly thereafter. Which is also when I learned just how MANY ways settings can be off.

 

[nikb]

Seems like a sanity check is needed.
Can you hit RTA window, with the following settings and the mic online and show us what it looks like?
No sound must be played back during the listening.

(Y-axis is SPL units)

Here you go:

1. No input gain was used on the motu

2. With max input gain

 

[Sokel]

Here you go:

1. No input gain was used on the motu

View attachment 486865 View attachment 486871

2. With max input gain

View attachment 486873 View attachment 486872

Seems about right with no gain.
The one with gain though is huge, you don't apply gain twice to it I hope (one from the mic preamp and one from the Motu).

Try one sweep without gain, as shown and see how it looks like.

 

[nikb]

I had to turn the interface up all the way to get rid of the messages that the signal is too low.
Also I need to read up on the rta window.

 

[nikb]

Seems about right with no gain.
The one with gain though is huge, you don't apply gain twice to it I hope (one from the mic preamp and one from the Motu).

Yes I had gain applied twice (preamp & interface). Have to strictly not do that

 

[Pareto Pragmatic]

running Arch linux

I often worry about new people being overwhelmed with technical talk. No worries with you!

 

[nikb]

I often worry about new people being overwhelmed with technical talk. No worries with you!

I'am actually having a blast, due to you and all the other people with helpful suggestions here!
Although my head already hurts a bit...

I am hoping to come closer to trustworthy measurements, but there is so much to learn still

 

[Keith_W]

I bought the microphone online some 10 years ago for some room-correction experiments (different room).

It came with calibration files for 0° and 90°, a preamp and the microphone itself. On the preamp it says to turn up gain half way, balance to the middle and pan to the center. That is what I did. I did have to turn on phantom power on the pre to get any input. It is turned off on the interface.

View attachment 486856 View attachment 486857 View attachment 486858

I asked you what microphone you are using, you didn't answer. I am squinting at the picture of the mic you posted, I think I see the word "ELECTRET" on it. The fact that it requires its own preamp makes me suspect it is an electret condenser mic.

The difference between "electret condenser" and "true condenser" mics is that the former has a pre-charged capsule and requires less power to polarize. Older electret microphones can lose the charge and may not remain linear. Most of us use omnidirectional condenser (non-electret) microphones for taking audio measurements.

I suggest you borrow a microphone and compare.

 

[kemmler3D]

I would concur that this looks like a mic issue first and foremost... The exact nature of the issue is another question but @Keith_W has a good idea IMO. If you can get your hands on a UMIK or other known good mic we will make some progress.

 

[nikb]

I asked you what microphone you are using, you didn't answer. I am squinting at the picture of the mic you posted, I think I see the word "ELECTRET" on it. The fact that it requires its own preamp makes me suspect it is an electret condenser mic.

The difference between "electret condenser" and "true condenser" mics is that the former has a pre-charged capsule and requires less power to polarize. Older electret microphones can lose the charge and may not remain linear. Most of us use omnidirectional condenser (non-electret) microphones for taking audio measurements.

I suggest you borrow a microphone and compare.

Sorry, I did not quite understand what you were asking about. Now I think I understand better: my microphone is a ECM40 FMJ. On the body it says "Stageline ECM-40 electret measuring mike". So to me it looks like an electret condenser microphone, in which case I should try to compare it with a known good mic.

On the other hand, as others have pointed out, this will probably not lead to a 20db drop (unless its totally broken). I am back suspecting the room in combination with really bad positioning of speakers and listener. Just playing with the REW room sim.

 

[kemmler3D]

Sorry, I did not quite understand what you were asking about. Now I think I understand better: my microphone is a ECM40 FMJ. On the body it says "Stageline ECM-40 electret measuring mike". So to me it looks like an electret condenser microphone, in which case I should try to compare it with a known good mic.

On the other hand, as others have pointed out, this will probably not lead to a 20db drop (unless its totally broken). I am back suspecting the room in combination with really bad positioning of speakers and listener. Just playing with the REW room sim.

A broad 20dB drop like that, with different speakers and positions within the room, is very hard to produce just with bad acoustics or positioning. In fact I wouldn't know how to do it, if I was trying to do it. Bass will tend to build up and produce some spikes regardless of how much high frequency is reflected, so the big drop we see seems more like an artifact of measurement.

Since you have got the same result with known good speakers, unless everything sounds like a broken tin can to you in that room, it's probably the mic or some other piece of the measurement chain.