The LF resolution of the FF+NF curve ( the NF measurement) seems low. No? Hmm...
Some continued experimentation.
I took measurements of the tall tower speakers (in the earlier picture) using the slide technique (moving away from the speakers). I took 8 measurements, this time moving back 30cm between each measurement. I decided to include more distance between each measurement so as to make sure that at least the first reflection (as observed in the impulse response) does not land in the same spot between consecutive measurements. Over the measured distance, the first reflection was observed between 1.8ms (furthest away measurement) and 3.8ms (closest to the speaker measurement).
I again too vector average and RMS average. I so exported each impulse in to Audacity - reduced the amplitude of each measurement - and then mixed them into a single track and imported the track into a WAV impluse.
Four pictures have been included below (all why psychoacoustic smoothing):
1) RMS Average vs. Vector Average (no SPL adjustment)
2) RMS Average vs Vector Average (SPL adjusted to overlay one over the other)
3) RMS Average vs. Audacity Import
4) Vector Average vs Audacity Import
Which one should I be relying on? It will be impossible to take ground plane measurements to verify anytime soon 🙁. Does my speaker have a dip at 200hz? or no? Either way, I can't rely on my hearing to tell me whether it's there or not.
I took measurements of the tall tower speakers (in the earlier picture) using the slide technique (moving away from the speakers). I took 8 measurements, this time moving back 30cm between each measurement. I decided to include more distance between each measurement so as to make sure that at least the first reflection (as observed in the impulse response) does not land in the same spot between consecutive measurements. Over the measured distance, the first reflection was observed between 1.8ms (furthest away measurement) and 3.8ms (closest to the speaker measurement).
I again too vector average and RMS average. I so exported each impulse in to Audacity - reduced the amplitude of each measurement - and then mixed them into a single track and imported the track into a WAV impluse.
Four pictures have been included below (all why psychoacoustic smoothing):
1) RMS Average vs. Vector Average (no SPL adjustment)
2) RMS Average vs Vector Average (SPL adjusted to overlay one over the other)
3) RMS Average vs. Audacity Import
4) Vector Average vs Audacity Import
Which one should I be relying on? It will be impossible to take ground plane measurements to verify anytime soon 🙁. Does my speaker have a dip at 200hz? or no? Either way, I can't rely on my hearing to tell me whether it's there or not.
Last edited:
Um, what are you even worrying about? Seriously. If you move your microphone left or right 30cm you will have greater variations in your measurements. What many people do not think about it that a measurement is a snapshot of the device, the room and the placements of the Device, the microphone in the room. You have great measurements there.
I still do not know why you use Audacity. You can do all the math functions within REW. And as it is staying native within the measurement system it will be treated in a way that John meant in the first place. I have measured the same item for testing with different software and the same mic and gotten slightly different results as you have here. A lot of what is happening in measurements is related to the stuff under the hood. And while you know that a db or so ripple is meaningless. You do not have any large wide dips a db or so that would be audible. And even this, audible under limited circumstances. Our ears are marvelously forgiving after 10 minutes or so.
Mark
I still do not know why you use Audacity. You can do all the math functions within REW. And as it is staying native within the measurement system it will be treated in a way that John meant in the first place. I have measured the same item for testing with different software and the same mic and gotten slightly different results as you have here. A lot of what is happening in measurements is related to the stuff under the hood. And while you know that a db or so ripple is meaningless. You do not have any large wide dips a db or so that would be audible. And even this, audible under limited circumstances. Our ears are marvelously forgiving after 10 minutes or so.
Mark
Good thinking on wider spacing the "data points".
Is it possible to upload the REW file of all 8 measurements? That may provide some insight, but the sad truth is that in room measurements have a lot of reflections to overcome. As we're heading into winter where I live, and I have a speaker project in progress, I may be working on a way to make this work in doors this winter.
Is it possible to upload the REW file of all 8 measurements? That may provide some insight, but the sad truth is that in room measurements have a lot of reflections to overcome. As we're heading into winter where I live, and I have a speaker project in progress, I may be working on a way to make this work in doors this winter.
Oh, thought about, but failed to type. You can simulate your room modes quite accurately with REW. This allows you to deduce what may be a room mode and what may be an actual measurement. Getting to know your available tools is a powerful piece of the measurement puzzle.
Mark
Mark
You made me think a little here. You can quite accurately simulate room modes. There area number of quite accurate methods available and free software including REW that will do a good job of this. Knowing those room modes in the first place allows you to identify them in the measurements. More importantly using the tools in REW I have been able to locate my mic and the Device Under Test in an advantageous position. I do this almost every day. And what I am finding is a fair degree of correlation between these multiple measurement methods and ground plane or anechoic measurements. Enough of a similarity that you can design with these inside measurements as long as you know and understand their limitations. That can be said of every method. An anechoic measurement has little actual resemblance to what we listen to in a room. So this in itself must be taken under consideration. Floor bounce and ceiling bounce effects are part of this understanding. Without making allowances for these effects you can measure either gated or anechoically and end up with a loudspeaker that is very unnatural sounding in a typical room.
Another 2 cents to add in. I have been working on Data sheets and reports for a few different items and these things have been front and centre in my weeks work.
Mark
Modal simulation is not at all accurate for typical rooms. Accurate Sims for real situations is very difficult.
Earl I have posted measurements and simulations of my measurement room. They coincide quite well. Simple room, 7 feet wide 15 long and 8 high. I'm not typing for the sake of typing. I use this regularly. Also I'm not pretending it covers every and all situations. Just saying it is useful to show what may be a room bump or dip versus a loudspeaker bump or dip. It has gotten myself and many clients out of the woods into the clearing where you can see what tree is what within the measurements.
We don't have to simulate the entire room.
I would be already very happy in simulating the first floor/ceiling/wall bounce.
That being said, simulating is worthless without rectifying.
Which bring us in a annoying loophole.
I have been thinking and brainstorming more from a practical approach.
In theory, one could get a cardiode mic, point it the other direction, so the nul is facing the speaker.
In that way you get everything EXCEPT the direct sound.
That way in theory, one could just extract the reflections from the original measurement.
It is in fact, totally retarded why we use omnidirectional microphones, since we are only interested in the direct sound of the loudspeaker.
So a super hyper uber cardioid mic would make a lot more sense. 🙂 😀
Cardioid mic. Hmmm. I can see as many problems as I can see benefits. The super directional idea again has trouble written all over it. You could process even an omni to give you rejection of the off axis results. That is basically what beam forming is doing is it not? My greatest weakness is my mathematical abilities. But I can appreciate what is being done via the math in the beam forming ideas. Klippel is giving a best fit type of a measurement. As is REW. Neither are totally interchangeable, but I have rarely seen the identical loudspeaker measured on two systems line up perfectly anyway. Mics are tools. They get us an indication of what is happening they are not infallible. I ran on Friday a series of distortion tests from 6mm nearfield to 1 metre keeping the same SPL level as I moved away. It was interesting. The nearfield looks really good. And that is how we get specs from manufacturers. But it doesn't reflect the reality at our listening position.
Secondary thought. How much processing would you need to undertake to allow a super-cardioid mic to first of all be flat on axis, and then be comparable to an omni directional mic?
Mark
Secondary thought. How much processing would you need to undertake to allow a super-cardioid mic to first of all be flat on axis, and then be comparable to an omni directional mic?
Mark
I was just looking at Thomann the other day, but most cardioid mics are already fairly flat, or just totally flat in freq resp.
The only thing that is not quite linear, is the directivity itself.
Although for measurements that doesn't matter, since higher = better .
But yeah, you can make a cardioid out of one omni with some data processing.
That doesn't even need to be on they fly.
We basically just need cardioid behavior up to about 300-350Hz or so.
The rest can be stitched with a regular time window.
It's just that the longer you can get that time window, the more frequency resolution you have.
As far as I know, Klippel does something similar, although for some weird reason they have their time window super high (around 1kHz or so if I remember correctly).
Which makes little sense to me.
Hey Mark - trust me, Boss, I totally get it 🙂. I guess my inquiry is more intended for the sake of learning rather than to achieve audio perfection. I'm pretty sure I will not be able to tell the difference between the two curves.
I do agree with your comment on the unnecessary use of Audacity. The last picture included in my post above clearly proves that the whole Audacity process results in a near-identical response to the REW Vector average. I believe @aslepekis came to the same conclusion at the bottom of post #593.
I'm just trying to get an understanding of how we can obtain better low-frequency response measurements when the ground-plane method is not a near-term viable option. So I wanted to know if there is any reason to believe that the Vector Average is a better indicator of a speaker's anechoic performance than the RMS Average, or vice-versa. Or maybe both are incorrect?
@aslepekis - I couldn't attach the file here (due to file size) - so I'll try to obtain your contact info via PM and send it to you via email.
If you have a perfectly rectangular room that is acoustically isolated from other rooms (no open doors, windows, etc.) and the room is lightly damped (and especially uniform in distribution,) then you meet the assumptions of the room modelers. But as you deviate from these assumptions the results will be less accurate. "Most" real rooms will violate several of these assumptions, but a room could be setup to work well.
But, actually, I could see making an "ideal room" and then modeling its LF acoustics and using that to find the actual sound source strength. This could work to very LFs if depending on the accuracy of the model. It's an interesting problem, but not simple to work through.
Last edited:
I would think more along the lines of a "shotgun" mike, which has a ver6y narrow field of coverage. It takes several elements to do this, but it is doable.
Also, all these "averaging" techniques - best case - reduce the reflections, they do not eliminate them. Elimination works best always.
Or just a few with more post processing.
Keep in mind that we don't need to have immediate live sound.
When put on a (robot) arm done with stepper motors, you can do it with one monopole basically and change the phase+gain accordingly.
The reason why I said "super hyper uber cardioid mic", was because you want to make sure you don't get any direct output from the source (speaker).
My room is a shoe box. Not the ideal proportions. But easily modeled. and closed off as you stipulated. Open rooms, or rooms with openings I agree. Have fun modeling this! I have posted a few ties the dimensions. 7 feet wide 15 feet long. 8 foot ceilings. And by a quirk of foundation placement the floor slopes on a downward grade 4 inches over 7 feet. It doesn't create a lot of help, but it has some effect. Using this room regularly has taught me through experimentation where the best places are to have your producer of sound. And I started off this tomfoollery by modeling the room in REW.
Mark