REW Confused on proper measuring Car Audio - Home Theater Forum and Systems - HomeTheaterShack.com

 
Thread Tools
post #1 of 9 Old 09-17-16, 04:39 PM Thread Starter
New Member
 
Join Date: Sep 2016
Posts: 4
REW Confused on proper measuring Car Audio

It seems in reading different threads everyone has different ways they measure with REW. Some use the RTA some use the Spectrum mode. Some use uncorrelated pink noise and some use correlated pink noise. What is the proper method with using REW? All I have at my disposal is uncorrelated pink noise from the IASCA disc which is used at the SQ competitions. Am I okay with using the RTA Function? If so, what are the settings? Am I okay with using uncorrelated pink noise with using the RTA Function versus the Spectrum?


Thanks,
sswails is offline  
Sponsored Links
Advertisement
 
post #2 of 9 Old 09-17-16, 06:37 PM
Elite Shackster
Wayne Myers
 
Join Date: Jun 2012
Location: Lincoln, NE
Posts: 4,838
Re: REW Confused on proper measuring Car Audio

Quote:
sswails wrote: View Post
It seems in reading different threads everyone has different ways they measure with REW. Some use the RTA some use the Spectrum mode. Some use uncorrelated pink noise and some use correlated pink noise. What is the proper method with using REW?
By "proper" you must mean "most accurate" or "most effective" or "most useful."

Quote:
All I have at my disposal is uncorrelated pink noise from the IASCA disc which is used at the SQ competitions. Am I okay with using the RTA Function?
That is the way I would do it, uncorrelated Pink Noise with RTA function. REW's generator will generate pink noise too.

Quote:
If so, what are the settings?
A good setting to start with is:
Periodic PN (16384) from the generator, RTA 1/12th octave, 16384 block size, 4 averages, all the rest at default

Quote:
Am I okay with using uncorrelated pink noise with using the RTA Function versus the Spectrum?
Yes.
AudiocRaver is offline  
post #3 of 9 Old 09-17-16, 07:10 PM Thread Starter
New Member
 
Join Date: Sep 2016
Posts: 4
Re: REW Confused on proper measuring Car Audio

I actually found some posts from John that may have cleared up my confusion. It sounds that using the RTA function versus the Spectrum function is most effective with Pink PN not requiring any averages. Just move the mic around very slowly to get the averages around the headspace. I've seen some big differences and some not so big differences when people measured both ways. I've always used the uncorrelated pink noise and tuned for a nice downward curve and tuned by ear after.

I'm interesting in trying the Pink PN with the RTA Function. I see he mentions to use the same FFT for the RTA and the Pink PN. 65536 FFT with a 1/48 smoothing seems to be the recommendation. Although, I like the idea of trying it at 1/12 and going from there.

Now, with uncorrelated pink noise you get a nice downward curve - not flat. With using the Pink PN, would I see the same trend? Will I see the same 10db filter as with the uncorrelated? Thanks!
sswails is offline  
 
post #4 of 9 Old 09-17-16, 08:56 PM
Elite Shackster
Wayne Myers
 
Join Date: Jun 2012
Location: Lincoln, NE
Posts: 4,838
Re: REW Confused on proper measuring Car Audio

Quote:
sswails wrote: View Post
I actually found some posts from John that may have cleared up my confusion. It sounds that using the RTA function versus the Spectrum function is most effective with Pink PN not requiring any averages.
Always go with what John says over anyone else's advice. He is The Man!

Periodic PN gives the steady result that needs no averaging. It is peak limited (the crest factor (peak to average ratio) is limited),
and... it REPEATS!

Just to clarify a little, using RTA function will give a horizontal measurement plot with pink noise or with periodic pink noise, and yields the same amount of detail per frequency band, i.e. per octave, or per smoothing band. I prefer periodic pink noise because of its "steady" nature for readings, especially SPL readings with the smaller block size. The SPL reading will hardly vary, if at all. Whatever settings you use, the block size in the generator and in the analyzer must be the same, or you can get insane results. With REW (later versions), the FFT block size in the generator automatically tracks the setting you have entered in the analyzer.

The spectrum analyzer mode will generally give a downward-tilted plot toward higher frequencies, depending on the source signal. It is not used nearly as much in audio, in my experience. Others may disagree. RTA with pink noise and selected smoothing and the desired number of averages gives tremendous flexibility.

There is no right, only what is most useful. Use 1-oct smoothing to see the over freq balance, 1/48-oct to see the fine detail that is "hidden" by the smoothing. 12th-oct or 1/6th-oct is a good setting for seeing what most listeners are likely to be able to hear. For automatic EQing, it is usually wise to use more smoothing, 1/6-oct, maybe even 1/3-oct, so you don't get a crazy number of filter values and/or a sterile sound (be careful not to over-EQ), or VAR smoothing, which gives more detail at LF, where modal EQing prevails, and less at HF, where broader EQing of trends is more common. The newer Psychoacoustic and ERB smoothing modes are useful in getting more detail of what the ear actually perceives. Both put more emphasis on peaks and less on dips in the response, as the ear does. I find myself using the Psychoacoustic most of the time any more.

Averages (0,2,4,8...) allow you to quiet down all the motion in the plot, or see the raw analysis detail with full motion, all depending on what you are looking for. As with any kind of averaging, detail is lost. But the brain sees the result more easily. Periodic PN helps with this too.

Quote:
Just move the mic around very slowly to get the averages around the headspace.
The moving mic method (MMM) is a useful way to get a picture of the whole headspace in one plot. Try this: plug two mouse receivers into your PC, use one mouse for regular navigation, the other for a remote. Put a piece of black tape over the laser of the "remote" mouse, and hold it in one hand while holding the mic in the other. Now you can start/stop the analysis when you want to. Set averaging to "forever," hold the mic at the starting point, click the left mouse button on your "remote," move the mic around slowly and randomly for 30 seconds, and Voila! you have an acoustically well-averaged plot of the FR of your headspace!

Quote:
I've seen some big differences and some not so big differences when people measured both ways.
Probably due to user error, i.e., amount of MMM movement, randomness, speed, etc. It is not a perfect method, there are ways to mess it up. But it is very useful and pretty easy to master.

Ah, I just got what you meant. It is probable there will be significant difference between a MMM plot and any other single-position or single-measurement method. The spacial averaging of the MMM will de-emphasize - as the ears and brain tend to do automatically - the modal variations at low frequencies, which slow up big time in a single-measurement plot.

Quote:
I've always used the uncorrelated pink noise and tuned for a nice downward curve and tuned by ear after.
"tuned for a nice downward curve" - sounds like a combination of personal preference, tweeter evasion, and "too much HF stuff bouncing around."

Quote:
I'm interesting in trying the Pink PN with the RTA Function. I see he mentions to use the same FFT for the RTA and the Pink PN.
Do or die. Actually, it will hurt no one, but the results will not make sense if you do not match the block sizes. As mentioned, this is automatic in later versions, the generator setting will automatically track the analyzer setting (but not the other way around).

Quote:
65536 FFT with a 1/48 smoothing seems to be the recommendation. Although, I like the idea of trying it at 1/12 and going from there.
see suggestions above

FFT block size, while the math is beyond most of us (except John), is gives another way of getting more detail at the sacrifice of time response. More averaging is a result of the analysis, so the averaging setting is usually left at 0 with longer FFT block sizes.

Bigger block size > More LF detail and slower response

Quote:
Now, with uncorrelated pink noise you get a nice downward curve - not flat. With using the Pink PN, would I see the same trend?
A "nice downward curve" is a subjective choice, probably fairly common in car work because of the constraints of car acoustics, and because a lot of car audio people like bass response that measures on the Richter scale. I am not sure why that would automatically be the result with the conditions you mention. Uncorrelated pink noise is a necessity for FR measurements with multiple channels at once, vs channel balancing with correlated noise.

Periodic PN gives a nice, steady result because of its repetitive nature and low crest factor (as previously mentioned). It is VERY convenient to work with.

It should give a similar result and be easier to read.

Quote:
Will I see the same 10db filter as with the uncorrelated? Thanks!
Not sure I understand this question.



By the way, welcome!
AudiocRaver is offline  
post #5 of 9 Old 09-18-16, 10:37 AM Thread Starter
New Member
 
Join Date: Sep 2016
Posts: 4
Re: REW Confused on proper measuring Car Audio

Thanks, your insight is very useful.

When I was checking out the Physcoacoustics that appears to be only available in Spectrum which is not so much intended for audio response? I was reading the RTA mode is more accurate in measuring a response and shooting for a particular house curve. In car audio we don't tune for a flat response. If we did it would sound horrible. I've attached a couple examples and you will note the downward target response. The first being the old JBL standard house curve or pretty close to it. The other is an example of what we tend to see in some of the best sounding cars.

I'm guessing with using the Pink PN, we can shoot for the same sort of target curve. From my understanding, you can tune to flat with white noise and when measured using pink noise would actually measure with a downward curve. Speaking in the car audio world.

I'm so used to using uncorrelated pink noise that I know what curve to shoot for and tune by ear after. It sounds like if I did use uncorrelated pink noise with the RTA mode it really poses no real concern with measuring? This Pink PN just sounds like it's a much better way to measure with more accurate results on what I'm hearing?
Attached Thumbnails
REW Confused on proper measuring Car Audio-target-house-curve.jpg  

REW Confused on proper measuring Car Audio-targethousecurve2.jpg  

sswails is offline  
post #6 of 9 Old 09-18-16, 10:51 AM Thread Starter
New Member
 
Join Date: Sep 2016
Posts: 4
Re: REW Confused on proper measuring Car Audio

Just for giggles, I compared Pink PN to Pink Noise using the signal generator running through my laptop speakers. The results were identical. Not sure if the Pink Noise in REW is uncorrelated or not?
sswails is offline  
post #7 of 9 Old 09-19-16, 01:41 AM
Elite Shackster
Wayne Myers
 
Join Date: Jun 2012
Location: Lincoln, NE
Posts: 4,838
Re: REW Confused on proper measuring Car Audio

Quote:
sswails wrote: View Post
When I was checking out the Physcoacoustics that appears to be only available in Spectrum which is not so much intended for audio response?
Details!

You are correct, Psy smoothing is only available with Spectrum mode in RTA. I normally use the sweep mode, where it can be applied to any curve. 6th or 12th oct will work.

Quote:
I was reading the RTA mode is more accurate in measuring a response and shooting for a particular house curve.
Not more accurate, maybe simpler, since you can easily play back recorded PN or PPN on you CD player. Sweep mode with a recorded sweep is kinda complex to pull off.

Quote:
In car audio we don't tune for a flat response. If we did it would sound horrible.
Forgive me playing the skeptic here, but it sounds like there are a lot of undefined and unstated assumptions in that statement. But for now, we'll stick to your questions.

Quote:
I've attached a couple examples and you will note the downward target response. The first being the old JBL standard house curve or pretty close to it. The other is an example of what we tend to see in some of the best sounding cars.
I believe you. I just don't believe you. Of course I am totally kidding. (90%?)

Quote:
I'm guessing with using the Pink PN, we can shoot for the same sort of target curve. From my understanding, you can tune to flat with white noise and when measured using pink noise would actually measure with a downward curve. Speaking in the car audio world.
Sounds like a few assumptions and shortcuts, BUT, if it works, DO IT! Your logic is sound, you clearly have more experience with cars than I do.

Quote:
I'm so used to using uncorrelated pink noise that I know what curve to shoot for and tune by ear after. It sounds like if I did use uncorrelated pink noise with the RTA mode it really poses no real concern with measuring?
Correct, no concern.

Quote:
This Pink PN just sounds like it's a much better way to measure with more accurate results on what I'm hearing?
Pink PN advantages are low crest factor, not a big deal for most applications, and a rock steady reading with 8192 FFT Block Size.

Quote:
sswails wrote: View Post
Just for giggles, I compared Pink PN to Pink Noise using the signal generator running through my laptop speakers. The results were identical. Not sure if the Pink Noise in REW is uncorrelated or not?
Let's get clear on what correlated means. If you have one speaker playing, correlated vs. uncorrelated has no meaning.

If you have two speakers playing the exact same signal, then there is 100% correlation between them, they are identical and they are in sync and would line up perfectly on a scope. With matching balance and timing, the psychoacoustical brain would create a "phantom image" midway between the speakers. But if you tried to take a measurement, you would probably get messy phase cancellations. The ears can do what a single microphone can not.

If you have two speakers playing and the two signals to them are unrelated, then those signals are uncorrelated. Uncorrelated can be accomplished by using one signal and delaying one of the signal paths by a second or two so the sound coming out of the two speakers is perceived as uncorrelated by the brain.
AudiocRaver is offline  
post #8 of 9 Old 09-19-16, 04:57 PM
REW Author
 
JohnM's Avatar
John
 
Join Date: Apr 2006
Location: UK
Posts: 6,561
Re: REW Confused on proper measuring Car Audio

A few points, partly repeating posts above:
  • REW's pink noise is random (which I guess you could call 'uncorrelated'). It is produced by passing uniform random (white) noise through a -10 dB/decade shaping filter. If it is averaged for long enough the spectrum is pink, 'long enough' at very low frequencies is a long time
  • REW's pink PN is a synthesised signal which is perfectly pink within the chosen sequence length. The RTA needs to have the same block length as the generator for this to work, if the RTA block is longer than the generator's gaps will be visible in the spectrum where the generator has no content, if it is shorter it behaves more like random pink noise except it doesn't keep converging as the averaging is increased
  • Always use the rectangular window setting when measuring with a noise source, this is essential if using PN
  • To make system frequency response measurements with any form of pink noise the analyser needs to be in an RTA mode for a flat response to produce a horizontal line, in spectrum mode a flat response will produce a line that slopes down at 10 dB/decade as frequency increases
JohnM is offline  
post #9 of 9 Old 09-19-16, 11:54 PM
Elite Shackster
Wayne Myers
 
Join Date: Jun 2012
Location: Lincoln, NE
Posts: 4,838
Re: REW Confused on proper measuring Car Audio

Quote:
JohnM wrote: View Post
A few points, partly repeating posts above:
  • REW's pink noise is random (which I guess you could call 'uncorrelated'). It is produced by passing uniform random (white) noise through a -10 dB/decade shaping filter. If it is averaged for long enough the spectrum is pink, 'long enough' at very low frequencies is a long time
  • REW's pink PN is a synthesised signal which is perfectly pink within the chosen sequence length. The RTA needs to have the same block length as the generator for this to work, if the RTA block is longer than the generator's gaps will be visible in the spectrum where the generator has no content, if it is shorter it behaves more like random pink noise except it doesn't keep converging as the averaging is increased
  • Always use the rectangular window setting when measuring with a noise source, this is essential if using PN
  • To make system frequency response measurements with any form of pink noise the analyser needs to be in an RTA mode for a flat response to produce a horizontal line, in spectrum mode a flat response will produce a line that slopes down at 10 dB/decade as frequency increases
Great detail to help tie up this thread, John, thanks for posting.

Did some research on signal correlation.
  • With two separate signals, correlation can range from 1 (the two signals are identical) to 0 (random, no relationship between the two at all, nothing in signal 1 gives any clue about what is happening in signal 2) to -1 (signal 1 and signal 2 are opposites, identical but 180 degrees out of phase).
  • With one signal, correlation means repetition, that one part of the signal can be used to predict some other part of the signal. Also known as Autocorrelation. Zero correlation means zero repetition, total randomness, no part of the signal gives any clue about what is going on in any other pat of the signal.
AudiocRaver is offline  
Reply

Bookmarks

Tags
audio , car , confused , measuring , proper , rew

Quick Reply
Message:
Options

Register Now




PLEASE COMPLETE ALL REQUIRED FIELDS BELOW... THANKS!

REQUIRED FIELDS ON THIS PAGE
YOU MUST COMPLETE ALL OF THESE

Username
Password
Confirm Password
Email Address
Confirm Email Address
Random Question
Random Question #2




User Name:
Password
Please enter a password for your user account. Note that passwords are case-sensitive.

Password:


Confirm Password:
Email Address
PLEASE READ BELOW PRIOR TO ENTERING AN EMAIL ADDRESS!

ATTENTION!

YOU MUST ACTIVATE YOUR ACCOUNT!

Activation requires you reply to an email we will send you after you register... if you do not reply to this email, you will not be able to view certain areas of the forum or certain images... nor will you be able download software.

AN INVALID EMAIL ADDRESS WILL CAUSE YOUR ACCOUNT TO BE DELETED!

See our banned email list here: Banned Email List

We DO NOT respond to spamcop, boxtrapper and spamblocker emails... please add @hometheatershack DOT com to your whitelist prior to registering or you will get nowhere on your registration.


Email Address:
OR

Log-in









Human Verification

In order to verify that you are a human and not a spam bot, please enter the answer into the following box below based on the instructions contained in the graphic.



Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page



Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML is not allowed!
Trackbacks are Off
Pingbacks are Off
Refbacks are Off

 


For the best viewing experience please update your browser to Google Chrome