Thanks Jay for saying I'd work at the 1/100th of a dB level! But "Holmanism?" Ug. -18 dBFSrms = 85 dBC was not a Holmanism, but was SMPTE RP200 for years--but you had to read it carefully because it was buried in the text.
The fact is that -18dBFSrms instead of -20 dBFS was not a misprint. It was there to solve a problem and to make it so that 0 VU made 85 dBCslow. Please follow the history:
Dolby originally used a wideband (100 kHz) average responding HP400FL voltmeter to set the level of noise for the hundreds of generators that they built and supplied to technicians who in turn used them to set thousands of theaters.
This had three problems: the bandwidth of the noise they measured was far wider than the audio band, an average detector under-measures random noise by about 0.8 dB, and the time constant of the measurement was not long enough to promote good accuracy (this was the analog era; fancy add- 'em-all-up-across-the-file-length stuff didn't exist).
The added bandwidth had the effect of lowering the level in the audio band when you then actually did band limit, and thus to get back to the same level you had to turn it up.
The effect of using an average-responding detector instead of an rms one was to turn it up.
The effect of using a fairly short time constant was to make the reading difficult and unpredictable.
In contrast at Skywalker I used a Radford Audio Noisemeter ANM5 which had 1. a defined bandwidth corresponding to the audio band (DIN Audio Band 22 Hz - 22 kHz -3 dB like an Audio Precision), 2. a true-rms detector (an Analog Devices chip), and 3. arbitrarily long time constant (just put a big, big cap on the Analog Devices chip smoothing port, so long as you are willing to wait for the answer). Then set it to Dolby internal bus level, 300 mVrms for the CN85 which today corresponds to -20 dBFSrms.
Now when the Wow! laser disc came out having pink noise at this level, we got a message from Roger Dressler at Dolby saying that our level was wrong. I checked the work others had done and found that our disc was correct to our method within 0.2 dB, and that the pink noise that Dolby had supplied for the Video Essentials laser disc was wrong in level and was not flat. I sent them a 5 page or so FAX regarding my history of measuring and specifying noise, dating to the time I was chief engineer at Advent in the '70's. I've got it somewhere; you'd find it amusing reading.
Time passed. We could never get agreement in SMPTE because practical people wanted 0 VU pink noise on the bus to be 85 dBC slow on the Radio Shack, and I wanted reference level (say +4 dBu) rms band-limited pink noise measured over a long interval to be 85 dBC slow.
The problem with this is that VU meters have a consistent error and are also inconsistent one to the next unless expensive, and many aren't. Radio Shack meters ain't bad, especially when calibrated, but the tolerances on their C weighting filter is pretty loose, leading to at least a +/-1 dB variation between them when they are set identical on tone.
All of the above went into specifying RP200, which once again, was written by Ioan, not me. The -18 dBFSrms noise is 85 dBCslow (meaning -20 dBFSrms is 83 dBCslow) because then 0 VU on the bus IS 85 dB on the Radio Shack! The two statements are the same thing. This is why you got good results with the Martinsound CheckMax, which we make, and yet the numbers sound different.
Now, RP200 is being revised. Dolby LA never agreed to it (get the irony), and they now use a band limiting filter, but still an average detector, and 1 minute+ averaging. So things are converging (over 15 years!). I recently took my pink noise card out there and as a courtesy to me they calibrated it to their standard, so we'd be alike, but note that their contemporary practice now uses most of the things I'd been complaining about years ago. So we are converging I think.
Former posts in this thread had some errors. Here are some replies:
1. We've been told to set the subwoofer level to +10 dB on a sound level meter. WRONG. Since the bandwidth is narrower than the whole spectrum, the actual level of the subwoofer measured on a Radio Shack is about +4 dB (exactly depends on its bandwidth). I wrote a whole Surround Pro article on this which should be archived.
2. There's a lot of stuff about the X curve. I disagree with Jay about moving the break frequency out to 4 kHz in a small room, and I believe that Ioan and I are in agreement to follow the standard for film material, which has an adjustment for room volume given as the number of seats. Small rooms such as 60 seat dub stages I set up typically flat to 2 kHz and then down about 1/2 of the X curve above there, thus -4 at 10 kHz and additional roll off beyond there. Also, the level of this smaller dub stage is set 2 dB below the level of the main 185,000 cu. ft. auditorium (Norris that many people in LA know from SMPTE spring seminars for years).
3. I agree you don't want an X curve for DVD-A material. In fact our Tesseract system includes a "film-video" switch which does the right thing; it switches in two different high frequency curves, one for X curve compatible film material and one for flat home monitoring compatiblity. Note that it is re-eq in Home THX that fixes the difference on Hollywood program material for the home; otherwise it is too bright.