Home Theater Forum and Systems banner

1 - 7 of 7 Posts

·
Registered
Joined
·
261 Posts
Discussion Starter #1
I have a CSL EMM-6 (w/ sensitivity) and a Scarlett 2i2. I used this basic method to calibrate the SPL of REW. I have the mic calibration file loaded in REW as well as one for the audio interface.

I take a sweep at a given level using HDMI (ASIO4ALL) from the PC to receiver, say -30dB in REW with my receiver at -10dB and get something like this:



I'd say the SPL level of the subs is ~77dB.

However, using the sub cal signal at -30dB with my receiver at -10dB the SPL meter in REW will read 2.5-3.0dB less than the 77dB "average" in the sweep. Initially I thought it might be the crest factor of the sub cal tone which was clipping, but I decreased the signal level from REW and increased the master volume on my receiver and that didn't change anything.

What accounts for the difference? Which method should I use to set the level of the subwoofer correctly?
 

Attachments

·
Super Moderator
Joined
·
9,258 Posts
I take a sweep at a given level using HDMI (ASIO4ALL) from the PC to receiver, say -30dB in REW with my receiver at -10dB and get something like this:
However, using the sub cal signal at -30dB with my receiver at -10dB the SPL meter in REW will read 2.5-3.0dB less than the 77dB "average" in the sweep.
I can’t figure out what you’re asking. See the two quotes I've highlighted - it looks like you took one measurement with the sub cal signal and one with (sorry, not sure what exactly), and are surprised that they’re different?

Aside from that, a sweep isn’t a good method for determining sub levels. A SPL measurement with a broadband pink noise signal is the way to do that.

Regards,
Wayne
 

·
Registered
Joined
·
261 Posts
Discussion Starter #3
I can’t figure out what you’re asking. See the two quotes I've highlighted - it looks like you took one measurement with the sub cal signal and one with (sorry, not sure what exactly), and are surprised that they’re different?

Aside from that, a sweep isn’t a good method for determining sub levels. A SPL measurement with a broadband pink noise signal is the way to do that.
I thought it was pretty clear... A REW sweep from 10 to 200Hz at -30dB should give the same average SPL as the sub cal signal at -30dB. But they don't. The signal level is the same in both cases. Both test signals are measured in RMS, REW measures SPL using a RMS method. For a given master volume they should read the same.

AFAIK, band limited pink noise only works if the subwoofer has the same bandwidth as the signal and a flat frequency response. For example if the sub cal signal has energy between 10 & 150Hz, but your sub only reproduces 25-100Hz you're going to end up setting the subwoofer's SPL wrong because the effective RMS value of the reproducible signal isn't the same if rolled off at 25-100Hz as it is if reproduced fully from 10-150Hz.
 

·
Registered
Joined
·
261 Posts
Discussion Starter #4
AFAIK, band limited pink noise only works if the subwoofer has the same bandwidth as the signal and a flat frequency response. For example if the sub cal signal has energy between 10 & 150Hz, but your sub only reproduces 25-100Hz you're going to end up setting the subwoofer's SPL wrong because the effective RMS value of the reproducible signal isn't the same if rolled off at 25-100Hz as it is if reproduced fully from 10-150Hz.
So I decided to run the numbers on the specific scenario I outlined. A synthetic band limited pink noise sub cal (not the REW sub cal) signal with energy between 10-150Hz (2nd order on both end) if used on a sub with a flat frequency response and frequency rolloff on each end consisting of a 100Hz LPF (in the receiver / 4th order) and a low end acoustic response that mimics a 4th order HPF at 25Hz. You'd have an error of about 3dB.

AFAIK, the REW sub cal signal is band limited pink noise with 2nd order rolloff at 30 and 80Hz. Even just running that though a bass management crossover at 80Hz (4th order) will reduce the effective RMS value of the test signal by ~1dB.

However, in my particular case I'm not using the bass management to send the signal to the subwoofer. My subs have a 4th order roll off at 120Hz and something like a 6th order roll off at ~15Hz. That should only introduce an error of about .4dB with a band limited pink noise signal with 2nd order rolloff at 30 and 80Hz, so that doesn't explain the variance I see.
 

·
Registered
Joined
·
261 Posts
Discussion Starter #5 (Edited)
Just throwing out an idea here. Are REW sweeps displayed in peak SPL levels not RMS? REW sweeps do use sine waves. That would account for a 3dB delta since sine waves have a crest factor of 3dB.
 

·
Super Moderator
Joined
·
9,258 Posts
I thought it was pretty clear... A REW sweep from 10 to 200Hz at -30dB should give the same average SPL as the sub cal signal at -30dB. But they don't.
And they never will. This is an apples to Chevrolet comparison.

Putting aside the fact that a sweep is not a good way to determine SPL average, how exactly did you arrive at the 77 dB average figure for the sweep that you mentioned in your first post? I’m guessing it’s just an “eyeball” conclusion based on looking at the graph? Not terribly scientific, because a real average would require knowing the actual SPL of every frequency in the range in question, adding all those figures and dividing by the number of frequencies – you know, the basic formula for calculating an average.

By comparison, a sound level meter doesn’t “average” anything, generally-speaking, and it certainly can’t generate an average dB figure across a frequency range like you’re trying to accomplish. A sound level meter merely registers peak noise levels. When you generate the sub cal tone, the dB-SPL figure you get reflects the highest dB peak of whatever frequency is the loudest. For instance, with REW’s 30-80 Hz sub cal tone as an example, if your sub is generating a modal peak at 75 Hz, the dB figure the meter shows is based on that 75 Hz peak and nothing else. So you see it’s impossible to get an average for a frequency range from a sound level meter – at least not with a broad- or limited-band calibration tone.

Legitimate SPL averaging is a time-lapse operation, where the average figure is based on data accumulated over a period of time. Using the example above as a reference, it would average the variations in level of that 75 Hz peak that occurred between the designated start and stop times. In other words, a figure roughly half-way between the lowest and highest dB figures that registered.

Furthermore, sine waves are steady-state, while pink noise is a random signal. That means a pink noise signal jumps around, somewhat like a music signal would, while sine waves don’t. So how can anyone expect to reasonably compare “averages” between the two?

Again, this is a futile comparison. I don’t see how it has any effect on anything REW was designed to do.

Regards,
Wayne
 

·
Registered
Joined
·
261 Posts
Discussion Starter #7
And they never will. This is an apples to Chevrolet comparison.
Sorry, but it's not. On an ideal system with a flat frequency response a -20dB (RMS) signal generates the same SPL regardless of what is in the signal. SPL is properly measured with an RMS meter integrating for some time constant. You can determine the RMS level of any test signal in a file. Correlating the two isn't hard and is exactly what I'm trying to do.

Putting aside the fact that a sweep is not a good way to determine SPL average, how exactly did you arrive at the 77 dB average figure for the sweep that you mentioned in your first post? I’m guessing it’s just an “eyeball” conclusion based on looking at the graph? Not terribly scientific, because a real average would require knowing the actual SPL of every frequency in the range in question , adding all those figures and dividing by the number of frequencies – you know, the basic formula for calculating an average.
Well, for starters I did put average in quotes, and said about 77dB. Yes, it is a visual estimate. The average SPL from 30-80Hz of that graph sure isn't 74-74.5dB (the result of measuring the REW sub cal signal with the same RMS level). The entire trace of the sweep is above that SPL level. Basic math tells us the average can't be lower than any of the values.

By comparison, a sound level meter doesn’t “average” anything, generally-speaking, and it certainly can’t generate an average dB figure across a frequency range like you’re trying to accomplish.
Huh? A good SPL meter is an RMS type meter. It's required by IEC61672. (FWIW, Radio Shack SPL meters are not RMS meters.) They provide an average SPL level over some integration time. This is the time average Sound Pressure Level (La). The formula for this is defined by IEC61672.

A sound level meter merely registers peak noise levels. When you generate the sub cal tone, the dB-SPL figure you get reflects the highest dB peak of whatever frequency is the loudest. For instance, with REW’s 30-80 Hz sub cal tone as an example, if your sub is generating a modal peak at 75 Hz, the dB figure the meter shows is based on that 75 Hz peak and nothing else. So you see it’s impossible to get an average for a frequency range from a sound level meter – at least not with a broad- or limited-band calibration tone.
I'm sorry, but that's just not how it works. A proper SPL meter does not return peak values unless in a special "peak" mode. They must be RMS meters (per IEC 61672) and they return time windowed averages of total pressure. The frequency content of the signal is irrelevant. Fast is generally an integration time of 1/8th of a second, slow is generally an integration time of 1 second. Again, IEC61672 defines the formula to compute La (Time weighted sound level).

Legitimate SPL averaging is a time-lapse operation, where the average figure is based on data accumulated over a period of time.
Basically correct. A series of instantaneous SPL readings are integrated per a given formula for some period of time to arrive at La.

Using the example above as a reference, it would average the variations in level of that 75 Hz peak that occurred between the designated start and stop times. In other words, a figure roughly half-way between the lowest and highest dB figures that registered.
That's not correct. A SPL meter measures total sound pressure. It's not the pressure at the loudest frequency. It doesn't do a FFT and return the amplitude of the highest amplitude "bucket".

Furthermore, sine waves are steady-state, while pink noise is a random signal. That means a pink noise signal jumps around, somewhat like a music signal would, while sine waves don’t. So how can anyone expect to reasonably compare “averages” between the two?
Because they're both SPL readings (La) for an input signal of equivalent energy. A pure sine wave, pink noise, white noise, or any other signals that all have the RMS value will have the same time weighted SPL (La) [assuming the integration time is long enough to smooth out momentary variation]. A time weighted SPL (La) reading over 1 second doesn't care what the signal is. It's simply the result of the integral.

Again, this is a futile comparison. I don’t see how it has any effect on anything REW was designed to do.
It's not futile at all. It is a fundamental question about how REW reports SPL in two different sections of the program and why there is an apparent discrepancy.
 
1 - 7 of 7 Posts
Top