Home Theater Forum and Systems banner

comparing measurments/how accurate are the levels?

2K views 19 replies 5 participants last post by  bjs 
#1 ·
When using REW to compare say:left channel to right channel etc...how accurately does REW match the the measurements to the target level of 75dB and to each other?Are they just matched to a 1Khz centered band limited pink noise level the way the processor/reciever internal tones are used to get the channel levels matched to 75dB spl?


I'm trying to see how closely all my channels match each other through out the entire frequency range for tonal accuracy.I want to use high shelving filters in the DEQ for broad trend adjustments on a couple of channels to get the overall frequency trend between them better matched.I'm using my previous saved plots with REW to compare and to predict the end result. I'm interested in knowing if the frequency plots are being compared to each other in a precise enough way that I can see the deviated trend correctly and use the EQ filters function in REW to model shelving filters to make a better overall trend match between channels and get an accurate end result.
 
#2 ·
I depends on how you use it. If you just use the SPL meter function, it'll work like any SPL meter, in which case it'll depend entirely on the signal you feed it. (click the generator button to see your options).
Alternatively you could overlay two sweeps, one of each channel. Issue there is any bias towards one channel or the other in your setup (mic placement, mic orientation, etc) will skew the result.

How accurately balanced are you looking to achieve and why?
 
#3 ·
I'm using the saved sweep measurements for each channel and overlaying them to see the 1/3 octave smoothed plots to see the overall trend from about 1Khz on up.All measurments were taken from the listening position in the sweet spot and the mic was never moved for the measurements.

My rears and left channel responses have a more tipped up trend from 1khz to 20Khz than the right and center which are more flat.I want to shelve down those 3 channels to match the right and center more closely.

I just want to make sure that when the sweep plots are compared that they are giving me an accurate comparison in level match between the channels to know the correct amount of dB I need to adjust the shelving down.
 
#4 ·
Well, like I said, any bias of the mic towards one channel or the other with skew you're results...
How different are they?
I'd personally be more interested in examining WHY they're different... such as acoustics, etc...
 
#5 ·
Well,my 2 rear speakers match each other in overall trend,but they tip up about 3dB from 1khz to 20khz and that I believe isdue to less acoustic treatment in the rear.My left main follows a similar trend.I have first reflection points for the LCR on the sidewalls and ceiling treated,but there is an opening on the right of the room were part of the sidewall reflection point of the right speaker would be.To help compensate, the left part of the wall parallel to the right side opening is treated in addition to the left speaker reflection point.The center and right channel match each other in overall trend and show as the flattest responses.

I can hear the tonal differences with panned sounds and vocals.The rears have a brighter tone to them and the same seems to happen with the left main,but to a lesser degree.
 
#6 ·

I assume you made valid full-range measurements using the calibrated ECM8000 (or similar) mic?

The easiest way to accomplish what you’re trying to do is to take all the measurements of all the speakers at the same time (i.e., one right after another) on the same day. First of course, you would level-match the speakers using the receiver’s sequential test tones, or whatever method you use.

Measure each speaker with the mic pointing at the speaker (make sure you’re using a horizontal-orientation calibration file for the mic). Once all the speakers have been measured, you can use the “Measured” tab to overlay all the measurements (smoothed for maybe 1/6- or 1/3-octave). Now you’ll be able to see where the response from the right and center deviates from the other three.

Regards,
Wayne
 
#7 ·
Yeah,I used the ECM8000.The measurements I saved were the finalized responses and were done one after the other at the same time period.

Ok,so you would have to level match them with pink noise before doing the sweeps.I thought the levels check before each sweep was supposed to set the the levels correctly.That's why I was unsure if what the overlaid sweeps showed were matched in level enough to get accurate results.

I thought you were supposed to point the mic at the ceiling for measurements?
 
#9 ·

Ok,so you would have to level match them with pink noise before doing the sweeps.I thought the levels check before each sweep was supposed to set the the levels correctly.
Either method is fine.


That's why I was unsure if what the overlaid sweeps showed were matched in level enough to get accurate results.
You’re interested in comparing frequency response, so “accuracy” isn’t related to absolutely perfect level matching. If one trace shows that a speaker is a bit too loud or too soft, you can just re-align the trace using the Trace Offset function (“Traces” icon, left side of the screen).


I thought you were supposed to point the mic at the ceiling for measurements?
Only if have a vertical-incident calibration file. But horizontal orientation w/ a horizontal file gets the best results, IMO. Even with a vertical calibration file, vertical orientation is going to pick up a lot more high frequency reflections from the ceiling than you would get with horizontal orientation. The direct signal from the speakers is what you primarily hear (unless you can’t close your eyes and point to where your speakers are – in which case you have massive acoustical issues :) ). As such, you want your response readings to minimize those extraneous reflections. Bottom line, IMO horizontal readings are more accurate.

Regards,
Wayne
 
#10 ·
Can you hear th difference between the same pink noise in each speaker? Sometimes pans are supposed to change timbre as they pan. Are the speakers matched? Are the speakers matched? How close are the rears to boundaries?
 
#11 ·
These are the measurements(1/3 smoothed) from 200Hz on up for all 5 main channels that I saved and overlaid on the measured tab.The area circled in red was were I wanted to match closer.I'll have to try taking measurements with the mic horizontally orientated at the speakers at a later time period to get more accurate measurements.

Purple:center

Green:left

Blue:right

yellow:left rear

light blue:right rear

Text Line Green Pattern Plot
 
#12 · (Edited)
Are you sure that you want to equalize the high treble region? At one point I was investigating how Audyssey determined the level on my speakers. I noticed that REW's speaker pink noise for the mains uses a band limited generator with the 3db rolloff points at 500Hz and 2000Hz, and 12dB/octave slopes. (If you bring up the Signal Generator dialog in REW, it shows you this in a pop-up help when you bring the mouse over Speaker Cal.) When I used RTA to look at the test tones of my Denon receiver, they appeared similar, 500Hz to somewhere less than 2000Hz, and 12db/octave slopes outside there. So apparently industry practice is to focus on this normal voice region when level setting.

Here is the spectrum I got from the Denon test tones. Front left and right are in red and green; blue is the sub test tone, but it is biased by the response curve of the sub itself. Someday I will go back and redo the measurements at the preouts, instead of in the room, to record the test tone spectra themselves.

Text Line Plot Slope Pattern


Bill
 
#13 ·

Not sure that I get what you’re saying, Bill. The test tones REW uses for measuring and plotting frequency response are full-range (if measuring the main channels). What do the limited-bandwidth tones used for level-setting have to do with that? :scratch: Drudge would have used a full-range Log Sweep tone for his individual speaker measurements, not the signal for level setting.


Regards,
Wayne
 
#14 ·
I expect you're right as regards the REW test tones. I may have overstated. The REW generator does offer speaker calibration Pink Noise test tones, with a default range for the mains of 500-2000Hz and for the sub of 30-80Hz. These seem to match the spectrum the Denon uses for pink noise test tones for speaker level setting. And, as I recall, Chris has indicated something similar when writing about the range Audyssey looks at when calculating its trims for each channel, but I may be remembering something he wrote when discussing the sub channel alone.

So, if Drudge is looking at the high end to provide more accurate leveling because the curves are easier to compare, that may not give the hoped for results. Looking back at the earlier comments on filtering to improve the tonal match, I can see where justifying the curves at the high end to determine what filtering to apply in the midrange might be a reasonable approach.

Bill
 
#15 ·
The band limited pink noise is centered at 1kHz for spl level matching in receivers,processors,and test discs because the ear is most sensitive to that range and spl level differences in that area would be most noticeable.If that range has a dip in it or a peak and you raise or lower your levels for that channel to match level with the other channels then the other ends of the frequency spectrum would be raised or lowered as well.That would cause everything below or above that range to higher or lower in level than the other channels.

That brings me back to why I asked if REW uses that same band limited pink noise to level match for the sweep measurements.If that's the case and you have dips or peaks in that area it would cause the measurements when overlaying different channels for comparison on the other frequency bands to be off by whatever amount you had to adjust to get the narrow band pink noise area to match in level.

I had this problem when I was setting my sub for 75dB with pink noise.the pink noise is centered right were I have a large dip in my response and was causing me to raise the level 3-5dB in level.When I used REW to look at the sweep after everything was pink noise level matched,the sub level was overall 3-5dB higher than my mains as compared to a flat response.

From what I've read if you use broadband correction on the higher >300Hz area (such as using an adjustable shelving filter)it would be just like using the treble control on your receiver/processor(which typically are hinged at 1Khz)to adjust a broad area up or down in level over a larger slope.There should not be any really negative affects as long as what you've measured gives you accurate enough data to make a proper and accurate correction if your trying to match the overall frequency trend of the other channels.

I don't know if the area I'm looking at would make that much difference if I adjusted it though.:scratch:
 
#16 ·
... That brings me back to why I asked if REW uses that same band limited pink noise to level match for the sweep measurements.If that's the case and you have dips or peaks in that area it would cause the measurements when overlaying different channels for comparison on the other frequency bands to be off by whatever amount you had to adjust to get the narrow band pink noise area to match in level. ...
Looking back at the Calibrate dialog in REW, it appears to give a choice of the subwoofer cal signal or the speaker cal signal. So I think my original remarks above about the band limited pink noise spectrum for the mains applies, that REW uses band limited pink noise to calibrate levels.

To obtain the most comparable measures, I would follow Wayne's suggestion above, to calibrate once and take all the measures in one session.

When I take measurements, I've adopted the procedure to note the REW SPL level of the band limited pink noise with each measurement, so I can move them up or down if needed to compare them across sessions. But this is still only as accurate as the initial calibration to the RS meter at the start of each session.

Bill
 
#17 ·
It seems to perform a proper tonal match or frequency response trend adjustment with eq can become a real slippery slope.:rubeyes::bigsmile:

I'll have to give Wayne's suggestion a try whenever I get the time to do it.It's a real time consuming process for me to setup and do tests on all five channels without distractions,noise etc...:laugh:


Thanks for all the input Bill,Wayne,glaufmen:T
 
#19 ·
I was going to use the paragraphic eq function of the DEQ2496(it mimics a parametric with the graphic eq in this mode) which allows you to use an adjustable high shelving filter.I have 3 DEQ's in my system that I use on all channels for room correction and the delay function that allows for more precise delay times than my processor and player.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top