What volume should be used for Room Correction reading/calibration?

[Dacapalooza]

Do different Room Correction technologies expect different volumes or should all readings for all RC tech be done at the same volume?
I have a Trinnov and have been using -18db for reading, although my listening volume is -25db.
I was told to increase volume for testing, but unsure if that is true?
Based on how finicky this forum is about volume tech, I thought I'd try calibrating at -25db yesterday so calibration & listening are one & the same.
Subjectively, I thought it sounded better with -18db reading. So now I am unsure.

Also, my room is a library. With the -18db reading I hear noises (resonances?) coming from the shelves during readings, while the -25db I don't hear noises. Should I take that into consideration?

 

[starfly]

Last time I measured with REW, I measured at around 95db (based on whatever the UMIK was reading). If you're talking about built-in measurement capabilities, doesn't the software automatically set that?

I've learned that you should take measurements about 40-50db over ambient noise levels to minimize the possibility of it affecting the measurement.

 

[Daverz]

I have the same problem. REW sweeps cause the glass doors of my fireplace to make a sympathetic buzzing noise.

 

[QMuse]

Do different Room Correction technologies expect different volumes or should all readings for all RC tech be done at the same volume?

You should use your "normal" listening volume (app 80dB of average SPL would do just fine).

 

[amirm]

The last two automated systems I used set the volume themselves. In general you want them loud enough so that the signal to noise ratio is good. REW has a level check where it warns you if the level is too low or too high if you are doing things manually.

 

[Dacapalooza]

Thanks!

I have used 2 softwares, both expected the user to change volume. I used the Audyssey standalone & Trinnov. Audyssey, I higher the volume in the connected PC.

Trinnov even uses the locked AES signal. If a source is not locked, Trinnov won't calibrate. Onetime I had it locked on 96khz and it gave warning if I was sure. I normally listen to 48khz so never gave it second thought. Now I wonder why Trinnov cares about the source and how it influences the calibration? I wonder if that means if Trinnov downrez 96 to 48khz? I'll know soon enough. I am getting DAC next week that has bit rate displayed. Or I could ask Trinnov I suppose.

 

[RayDunzl]

In general you want them loud enough so that the signal to noise ratio is good.

Example:

In my room, with the Krell and MartinLogans, long ago, 2dB incremental SPL increase:

1: Below this SPL level noise affects the frequency response measurement.

2: But, the lowest THD measure occurs at this level. THD decreases (better SNR) up to this level, then increases, as the actual THD rises out of the noise floor.

Frquency sweeps:

It can be seen that too low an SPL level will likely upset correction at low frequencies, and even be different for each time you create a correction.



---

For distortion:

The lowest sweep SPL is at the top, decreasing THD as level increases (better SNR), bunching at the bottom and increasing THD as the actual harmonic distortion begins to rise out of the noise floor.

Total Harmonic distortion using the above sweeps.

So, for frequency response, I measure at 70~80dB. Not so loud as to require hearing protection, but loud enough that the noise levels don't upset the measurement.

For an "accurate" distortion measure, depends on the actual distortion of the speakers vs the noise floor in the room.

 

[Dacapalooza]

In general you want them loud enough so that the signal to noise ratio is good.

Confused. I would imagine distortion/SNR plays a small part in RC software readings? We have -15db dips to contend with sometimes. I would think SNR is small factor in comparison? Are we saying that better THD will result in better readings. So if we simply throw a few AHB2's at this issue we can take readings on normal volume and not worry about optimizing the volume that produces the least SNR? Or we are talking about distortion of the speaker? Even still, question stands about SNR should be small factor?

 

[DonH56]

In my room I set the level to -30 dB to calibrate my SDP-75 (re-badged/JBL-tweaked Altitude). Most AVRs set the level themselves, but Dirac Live and Trinnov require you to set it high enough to read with a certain noise floor and/or crest factor for calibration. Usually in the 75~80 dB range, louder than I usually listen. I VNC'd to the SDP-75 and ran several calibrations with different settings to try (listen to) later.

 

[DonH56]

Confused. I would imagine distortion/SNR plays a small part in RC software readings? We have -15db dips to contend with sometimes. I would think SNR is small factor in comparison? Are we saying that better THD will result in better readings. So if we simply throw a few AHB2's at this issue we can take readings on normal volume and not worry about optimizing the volume that produces the least SNR?

It is a little tricky. You do not need a lot of SNR, but since speakers and room response can easily vary by 20~30 dB or more (mainly due to room modes), you need to get the signal level well above any noise in the room that could corrupt the readings, like HVAC noise or speaker hiss. Low-frequency rumbles from the furnace or air conditioner that we barely hear (or ignore) are typical culprits. Calibration is usually performed around 80 dB SPL because (or so I have been told) that is around the level our hearing is flattest so is a good place to set a "flat" measurement point. That way it is calibrated around the level they expect most users to be listening critically.

It is more about the room, its noise and response, as well as the speakers than the amplifiers.

FWIWFM - Don

 

[RayDunzl]

Confused.

Look again at my post above.

With low Speaker SPL, ambient noise can dominate the readings.

My room is reasonably quiet, suburbs, no busy road, no dogs barking, etc.

 

[Dacapalooza]

I VNC'd to the SDP-75 and ran several calibrations with different settings to try (listen to) later.

No longer need VNC. Trinnov changed the WEB GUI and you can access the Calibration screens from there. Much better, because the web GUI never times out.
I have no ambient noise at all I think. My room is sound proof-ish and my house is far in from street. Providing sirens don't go by there is no noise. I Turn off the split unit and even my rack fans during calibration. I even just had my kid just drop a pin and I can hear it from other side of room.

Look again at my post above.
With low Speaker SPL, ambient noise can dominate the readings.

You first graph represents ambient noise and second distortion? Since you mentioned your room is "reasonably quiet" what accounts for the ambient noise in graph? Your HVAC? Question: If we are sure that there is no ambient noise, we can calibrate at listening level?

It can be seen that too low an SPL level will likely upset correction at low frequencies, and even be different for each time you create a correction.

Perfect you should say that. In my experience many times Both Audyssey & Trinnov create different correction for same readings minutes later. I can scientifically prove it because my ammeter uses more/less power for same scene. Tron is my best test post calibration as it goes up an extra 15 amps from mute with no correction - so minor diff produces big power diff. With correction scene varies from 5-8.5. If less < 7 amps, I always re-calibrate. If < 7a, the scene just does not sound right. I just assume the reading software hiccups due to running in PC. Some other background process or something. If I was to bet, probably the PC culprit produces far more inaccurate readings than SNR. When I have patience, I only accept readings if after 3 times the graphs & Tron's ammeter look the same. Very rarely this occurs on first 3 tries.

 

[RayDunzl]

You first graph represents ambient noise and second distortion?

The first graph is 24 test sweeps of frequency response, and irregularities between traces shows the effect of ambient/measurement noise on the reading, particularly in the bass region and low SPL test level.

Noise level here in this room, as seen by the measuring microphone, looks like this right now, 4:45pm local coronavirus time. It is always "about the same" here unless the garbage man drives by or the golf course maintenance man is mowing grass.

Ambient/Measurement Noise:

For frequency response swept sine measurement, a minimum level of about 75dB will be needed here, so that the "noise" in the bass region is swamped by the tone level.

You can see that in the irregularity in the SPL traces for the bass region for the sweeps at increasing levels that was posted before. Measurement below the #1 line are obviously affected by ambient/measurement noise.

The bass measurement doesn't smooth out till the SPL of the sweep is around 75dB SPL (ignore the 48Hz dip)

For distortion, in my case, the noise level in the room is higher than the distortion produced by the speakers until an SPL level of about 90dB SPL is produced by the speakers.

Since you mentioned your room is "reasonably quiet" what accounts for the ambient noise in graph? Your HVAC?

Low frequency noise, though you can't always consciously hear it, it travels long distances and passes through our rooms. Aircraft, helicopter rotors, highway traffic, railroad,

When my AC comes on, there's a little spike at 18Hz (inaudible), and I can't hear the compressor from here (it's 80 feet away through five walls.

Fortunately, the AC compressors for both adjoining neighbors are on the other sides of their houses.

Question: If we are sure that there is no ambient noise, we can calibrate at listening level?

There is always ambient noise. Microphones make their own noise, too.

Listening level for me is 75dB or more, so, it works out that that is a reasonable level for frequency response measurement.

Higher SPL is needed if trying to actually see speaker distortion with low distortion speakers.

 

[dasdoing]

The first graph is 24 test sweeps of frequency response, and irregularities between traces shows the effect of ambient/measurement noise on the reading, particularly in the bass region and low SPL test level.

Noise level here in this room, as seen by the measuring microphone, looks like this right now, 4:45pm local coronavirus time. It is always "about the same" here unless the garbage man drives by or the golf course maintenance man is mowing grass.

Ambient/Measurement Noise:

View attachment 63047

For frequency response swept sine measurement, a minimum level of about 75dB will be needed here, so that the "noise" in the bass region is swamped by the tone level.

You can see that in the irregularity in the SPL traces for the bass region for the sweeps at increasing levels that was posted before. Measurement below the #1 line are obviously affected by ambient/measurement noise.

The bass measurement doesn't smooth out till the SPL of the sweep is around 75dB SPL (ignore the 48Hz dip)


View attachment 63048


For distortion, in my case, the noise level in the room is higher than the distortion produced by the speakers until an SPL level of about 90dB SPL is produced by the speakers.




Low frequency noise, though you can't always consciously hear it, it travels long distances and passes through our rooms. Aircraft, helicopter rotors, highway traffic, railroad,

When my AC comes on, there's a little spike at 18Hz (inaudible), and I can't hear the compressor from here (it's 80 feet away through five walls.

Fortunately, the AC compressors for both adjoining neighbors are on the other sides of their houses.




There is always ambient noise. Microphones make their own noise, too.

Listening level for me is 75dB or more, so, it works out that that is a reasonable level for frequency response measurement.

Higher SPL is needed if trying to actually see speaker distortion with low distortion speakers.

this is very intresting.
we could probably answer OPs question like this:
since he wants to avoid his shelfs rattling,
start low, make new measurements increasing volume until you have 2 paralel measurments

 

[North_Sky]

I was going to say 75 dB.


Reference Level and Calibration

A reference level is the PEAK level (in terms of output SPL in decibels) when the digital audio sample is also at its PEAK (ie '0 dB'). You can use any level of your choosing but generally when people talk about 'reference level' they mean Dolby Reference Level which has a well-defined meaning (from Dolby) and that is 105 dB (115 dB for LFE but the receiver automatically adds the +10 dB boost for the LFE channel).

So how do you get reference level? By the process of calibration and that is where the receiver's test tones (or an audio calibration disc's test tones) come in and where it is important to understand the prior discussion of digital audio levels.

The receiver's test tones are broadband pink noise and normally are at a level of -30 dB, which is a THX specification (and as mentioned above is the average level of a typical movie soundtrack). Some setup discs use a level of -20 dB, which is a Dolby specification. The level of the tones you use is only relevant when it comes to which number you want to see on an SPL meter when performing the calibration.

So...to get Dolby Reference Level, you run the test tones and adjust the channel trims on the receiver up or down as necessary until each channel reads roughly 75 dB on the SPL meter. Because the average level of the test tones is -30 dB, doing so yields 30 + 75 = 105 dB PEAKs when the digital audio sample peaks.

In other words, the majority of the time the SPL you hear will be 30 dB below the peak of 105 dB but when any sample is at a peak (again '0 dB'), you'll get blown out of your chair as the SPL instantly rises to 105 dB.

●

https://community.roonlabs.com/t/a-guide-how-to-do-room-correction-and-use-it-in-roon/23800

 

[QMuse]

For an "accurate" distortion measure, depends on the actual distortion of the speakers vs the noise floor in the room.

As @JohnPM mentioned in some thread here, for in-room distortion measurement stepped sine is a more acurate method than sweep.

 

[dasdoing]

why would you need to meassure at reference level? if you get a clean meassurement at lower level it will be paralel to the reference level meassurement. unless you hit limiters at reference level, which would mean you need new speakers

 

[Dacapalooza]

why would you need to meassure at reference level? if you get a clean meassurement at lower level it will be paralel to the reference level meassurement. unless you hit limiters at reference level, which would mean you need new speakers

Logically: Ideally readings should be the same as listening level. Why bring in other factors?
Scientifically: Trinnov does give different readings at different levels. Even if it is external factors, the correction/filter is now wrong. Why I thought to ask.

In general you want them loud enough so that the signal to noise ratio is good.

Example:..

Not getting parallel graphs with db vol changes. Trinnov is giving different readings/filters with every 3 db diff I would say: dips/bumps are being read as +-2db. Sometimes more. (Actually I think every 1db, but didn't save graphs)
So what you are saying is go ahead and use wrong correction/filter. Increase to volume you will never listen at because the SNR is great!
I guess you are saying that the SNR/ambient noise interferes with readings more so than all other factors?

since he wants to avoid his shelfs rattling

My shelves are stuffed with books weighing 100lbs+. Only happens when I increased to <= -18db. My left speaker has 15db bump <50hz. So -3db + sweep noise. I'm convinced it is resonance. I tried in other rooms with -18db and I hear noises coming out of walls even.

 

[QMuse]

Logically: Ideally readings should be the same as listening level. Why bring in other factors?
Scientifically: Trinnov does give different readings at different levels. Even if it is external factors, the correction/filter is now wrong. Why I thought to ask.

EQ made at say 85dB should be valid at all SPLs, barring LF compression which happens at high levels.

 

[Dacapalooza]

EQ made at say 85dB should be valid at all SPLs, barring LF compression which happens at high levels.

As an odd coincident, precisely 85db happens to be the threshold of permanent damage according to Google.