advertisement


Room measurement, what to learn?

Tony L said:
He says in big bold letters "Attempting to get accurate SPL levels is the biggest mistake users of REW make!". I'm sorry but I don't agree. If the vertical db scale is out so will be the severity of the dips and peaks, i.e. something that gives the impression on the chart as being plus or minus 5db at an average level of 85db, but was in reality measured at 65 or 70db may only be a 2db swing. Inaccurate measurements are arguably a lot less useful than no measurements at all. I'd certainly hate to hear a system using electronic room correction arrived at with inaccurate measurement data. Science is worth doing right! I want to learn how to calibrate this thing properly so it is a real measurement tool and I can trust the result.
Click to expand...
Accurate SPL has nothing much to do with the dips and peaks, 20dB up or down is still exactly that, just that the scale on the left may read 60dB +/- 20 or 70dB +/-20, it doesn't matter so long as you aren't trying to get a dB/W figure
 
Tony L said:
He says in big bold letters "Attempting to get accurate SPL levels is the biggest mistake users of REW make!". I'm sorry but I don't agree. If the vertical db scale is out so will be the severity of the dips and peaks, i.e. something that gives the impression on the chart as being plus or minus 5db at an average level of 85db, but was in reality measured at 65 or 70db may only be a 2db swing. Inaccurate measurements are arguably a lot less useful than no measurements at all. I'd certainly hate to hear a system using electronic room correction arrived at with inaccurate measurement data. Science is worth doing right! I want to learn how to calibrate this thing properly so it is a real measurement tool and I can trust the result.
Click to expand...

I assumed the purpose of the calibration file is to EQ out the non-linearities of the particular mic and/or soundcard you are using? e.g. I use a Behringer ECM8000 which a test report shows to have a gently rising HF on-axis from 2kHz which peaks at 10kHz (+3.4dB) and rolls off thereafter, returning to flat at just over 20kHz. Its LF roll-off starts at 40Hz and is -2.4dB down at 20Hz.

My point is, if you only need to measure your speakers or room in relative terms (e.g. to identify before and after changes in positioning), and don't need an absolute accurate reference point, is calibration all that important?

If however calibration affects how sensitive your mic is to picking up +/- differences at a particular reference dB SPL, as Tony appears to suggest, then it would seem more important than I thought.

EDIT - David's post above addresses this issue, thanks.

FWIW - I've always run my sweeps at a pink noise reference level of 75dB at the listening position. I achieve this by placing a RadioShack SPL meter beside my ECM8000 turning my amp up until the SPL meter hits 75dB (C-weighting). I then make sure I have between -20dB and -10dB headroom on the REW meters when I run the 'Check Levels' test sweep, and if not I adjust the REW gain, my sound card's gain, and my amp's gain, until I reach this sweet spot. My resulting FR graphs typically end up in the 70dB range +/- 10dB so perhaps I need to crank it up a little more? I'm just concerned about frying my tweeters (I need to wear -30dB Peltor mufflers when sweeping at my current SPLs as it is already too uncomfortable on my ears!).
 
My Behringer DEQ2496 insists on quite ear splitting levels when doing room measurements, presumably so that any background noise effects are swamped, but its actually beyond the SPL capabilities of my kitchen speakers! However I found that placing the mic almost in contact with one speaker until it was happy with the level, then reducing the volume, and moving the mic to the listening position enabled me to get a correction. I then removed the attempts to remove the nulls, and slightly eased off the bass boost, and I was left with a couple of dips in the bass region, and a small hump in the midrange at about 3kHz, which is probably something I can fix in the electronic crossover, but for now, they sound ridiculously good for LS3/5a sized boxes, bunged on top of the kitchen cupboards!
 
Surely it's absolutely essential to calibrate your microphone, or the frequency response you measure will be down to the microphone as well as the rest.
Come to that also to use the best mike you can afford.

I tried 2 different (not inexpensive) microphones but neither were sufficiently sensitive as JH observes.

In the end I forked out for a UMIK-1, which is not only calibrated, but integrates automatically with the latest version of REW. They're not cheap though.
 
Yes, calibrated microphone absolutely essential. A microphone calibration facility is not cheap, and so a microphone calibrated to 1dB for £75 seems cheap to me ... but perhaps I am in a minority.
 
I never bother calibrating absolute SPL levels - you will get a different reading when you move the mic closer or further from the speaker anyway.

Setting the absolute SPL level is different than calibrating the frequency response of the mic. The inaccuracies tend to be in the high frequency range though. So if you are doing low frequency room correction it really doesn't matter, particularly added to the huge variations the room introduces.

Where high frequency accuracy can matter is working on speaker design where you need to get the build right; is that blip at 16KHz from the microphone or tweeter faceplate? Even so, I would always listen and adjust general tonal balance to taste so really who cares? In the end probably only a lab with responsibility to publish accurate results.

Regards garden measurement I'd be careful - the more sources of reflection you remove, the more obvious the remaining few become. Particularly the ground! I've ended up lifting speakers a couple of meters in the air in the middle of fields to know what was really going on in the 100Hz-300Hz range. Even most anechoic chambers are not accurate below a few hundred Hz.
 
I would suggest getting very close will involve the microphone in the measurements. I would suggest 1m is as close as you need to get. All you want to do is make sure the measurement from the driver is 10dB above any room sound.

Try moving the microphone away and towards the driver. This will check for standing waves between the microphone and driver. Try off centre as well you might find that interesting.

If you want to look at the dip then use a single frequency. It's easy to downloads some tones, Tone (sorry).

Position the microphone vertically. Seems daft, but that will reduce errors due to direction. When listening your detectors are at 90 to the sound source (sort of!!).

Fascinating how you can worry yourself about the lack of a flat sensitivity to your drivers. Most music (except in C by terry riley) has varying frequency and volume.

Your driver and room will vary with the the volume. Try some response curves with differing volumes.

In the end don't worry.
 
chrisletts said:
Surely it's absolutely essential to calibrate your microphone, or the frequency response you measure will be down to the microphone as well as the rest.
Come to that also to use the best mike you can afford.

I tried 2 different (not inexpensive) microphones but neither were sufficiently sensitive as JH observes.

In the end I forked out for a UMIK-1, which is not only calibrated, but integrates automatically with the latest version of REW. They're not cheap though.
Click to expand...

I think the UMIK-1 is very reasonably priced and a nice hassle free solution.

Order from Keith at Purite and support pfm traders :)
 
Given you can bypass the need for a dedicated audio interface I think the umik is very cheap. I paid more than twice that for my measurement mic and it needs me to have a mic pre-amp too. Even that is not pricey - look and B&K stuff!
 
I'm not sure how accurate REW, or the set up I have to measure in, really is.

For example, if I make a measurement, then without adjusting anything, remake that measurement, the results will be slightly different. This could be ambient noise, internal machine noise - I don't know.

For setting up a speaker system I don't think absolute accuracy is necessary. It may be scientifically desirable but not practical or necessary. For instance, as you change the smoothing on the frequency response charts, peaks and nulls increase. Does that mean with 'no smoothing' you get an absolute figure? I don't know. If you do can you actually achieve your perfect response using a no smoothing setting? I doubt it. I usually finish up by using test tones and my ears, and of course music.
 
Bear in mind you are measuring the speaker/room response at seated position..only the sweet spot matters

Level of sweep tones via rew should be more or less the max volume you listen at, too high stimulates resonances you would most likely not hear in general music listening, too low wont drive the room enough and not stimulate boom etc you would hear at high levels

Play test tones within 5-10 hz of each other till round 2-300hz and track down rattles and buzzes and extraneous resonances ..damp them

apples with apples , dont do a measurement with a window open one day and closed the next .

I repeat crevasses DON'T matter , you dont hear them..peaks do.. forget stressing over severe dips with high Q's (spread of octaves the dip encompasses , the higher the q the more like a spire)

Umik -1 is cheap as chips and is the de facto std mic used for most home measurements

And finally .. its illuminating to see that compared to 10 years ago that room/speaker acoustics and treatments etc are now considered as the prime component of how stuff will sound and that there is robust discussion
And as a postnote: it's all subjective anyway
 
I made my own measurement microphone using a Panasonic electret capsule at the end of a 30cm wand of the same diameter into a PP3 powered preamplifier, Those electrets are very good, if mounted correctly. Ordinary microphones don't work well because of reflections.
I used to have one of those expensive B&K sets for telephone measurement and we found the electrets were dead flat up to 10kHz
 
Hipper said:
I'm not sure how accurate REW, or the set up I have to measure in, really is.

For example, if I make a measurement, then without adjusting anything, remake that measurement, the results will be slightly different. This could be ambient noise, internal machine noise - I don't know.

For setting up a speaker system I don't think absolute accuracy is necessary. It may be scientifically desirable but not practical or necessary. For instance, as you change the smoothing on the frequency response charts, peaks and nulls increase. Does that mean with 'no smoothing' you get an absolute figure? I don't know. If you do can you actually achieve your perfect response using a no smoothing setting? I doubt it. I usually finish up by using test tones and my ears, and of course music.
Click to expand...

The point of smoothing is to de-emphasize narrow peaks/troughs and thus to make general, broad trends more easily recognisable. Since it is (arguably) these general trends that have more impact on SQ, or conversely since very narrow troughs aren’t obnoxious, smoothing is helpful.
 
Tony L said:
Can you just clarify that even if the SPL scale at the left of the graph is, for point of argument, 10db out, i.e. suggests the median level is 85db when it is in fact 75db, is an indicated 5db deviation on the chart still exactly 5db? It is this aspect I've been struggling to get my head around!
Click to expand...

Yes. It's a ratio.
 
Tenson said:
I never bother calibrating absolute SPL levels - you will get a different reading when you move the mic closer or further from the speaker anyway.

Setting the absolute SPL level is different than calibrating the frequency response of the mic. The inaccuracies tend to be in the high frequency range though. So if you are doing low frequency room correction it really doesn't matter, particularly added to the huge variations the room introduces.

Where high frequency accuracy can matter is working on speaker design where you need to get the build right; is that blip at 16KHz from the microphone or tweeter faceplate? Even so, I would always listen and adjust general tonal balance to taste so really who cares? In the end probably only a lab with responsibility to publish accurate results.

Regards garden measurement I'd be careful - the more sources of reflection you remove, the more obvious the remaining few become. Particularly the ground! I've ended up lifting speakers a couple of meters in the air in the middle of fields to know what was really going on in the 100Hz-300Hz range. Even most anechoic chambers are not accurate below a few hundred Hz.
Click to expand...

Can you just clarify that even if the SPL scale at the left of the graph is, for point of argument, 10db out, i.e. suggests the median level is 85db when it is in fact 75db, is an indicated 5db deviation on the chart still exactly 5db? It is this aspect I've been struggling to get my head around!

Robert said:
I think the UMIK-1 is very reasonably priced and a nice hassle free solution.

Order from Keith at Purite and support pfm traders :)
Click to expand...

This what I'm using and it seems very good. I'm just trying to make sure I'm using it correctly!

PS I've just cocked up a post merge on my iPad so cooky's reply is now actually above my question!
 
Tony, 5dB is 5dB nice and simple. Add 5dB or subtract 5dB it's all the same regardless of absolute level. That is the purpose of using the dB scale rather than a linear scale like air pressure in millibar.

What I expect confuses you is that the specific power increase / decrease that 5dB represents will vary depending on the absolute level. Adding +5dB at 90dB mean level represents more power than adding 5dB at 40dB. But it's all nicely proportional using dB, so you can just forget about the specific pressure / power / voltage levels.
 
You must log in or register to reply here.


advertisement


Back
Top