Very nice !!

Great track, Dave!

By golly, it's hot! When reconstructed, it has peaks that go nearly 1dB over FS.

Thanks Bos. Odd that db reading. this was professionally mastered. I can't imagine my mastering engineer allowing anything like that. It's definitely a loud record. I wonder if it's just certain frequency ranges that are louder than others. :unsure:

Im curious Dave if you did anything to the studio with regard to the electrical wiring, like isolated grounding, or isolation xformer, things of that nature for "clean power"?

Hi Kyle. Other than making certain of my circuits and replacing all the receptacles and grounding everything properly I really didnt have a lot to do with this room. The main live room has a separate set of circuits and the control room is the dining room which is a 20 amp. I don't need much as I'm not running a console. I do have a transformer for isolation but it's not being used right now. I got zero buzzes.

It was an addition on the house back in the 60's so the main live room is slab-on-grade with 9' ceilings. Mineral board locking tiles already in place since they built it as the family room. It's right at 18.5' X 17'. Separated from the dining room by a slider. Dining room is my control room. One of the bedrooms is a small remote dead room although it's storage for guitars and the drum kit when it's not being used. I used to use it as a guitar amp room and have recorded a kit in there also.

Excellent, those 9' ceilings are a treat in a home enviornment. Its refreshing to hear that standard electrical hasn't been any issue. Thanks for the reply, cheers!

Davedog, post: 465421, member: 4495 wrote: Thanks Bos. Odd that db reading. this was professionally mastered. I can't imagine my mastering engineer allowing anything like that. It's definitely a loud record. I wonder if it's just certain frequency ranges that are louder than others. :unsure:

I'm not sure it has anything to do with frequency ranges. When I was involved in a loudness analysis contract some years ago, one of the things I was tasked with was estimating reconstructed waveforms that had been limited at the final stage, knowing how users' DAC reconstruction filters would perform. Having seen what mastering engineers tried to get away with, I got sensitised to it. Your track is not in that league and sounds just fine, but there is evidence of analogue limiting to allow the average level to be set high, and reconstruction filters take a few of the peaks over the nominal max level. I haven't done a LUFS measurement on it!

Boswell, post: 465431, member: 29034 wrote: I'm not sure it has anything to do with frequency ranges. When I was involved in a loudness analysis contract some years ago, one of the things I was tasked with was estimating reconstructed waveforms that had been limited at the final stage, knowing how users' DAC reconstruction filters would perform. Having seen what mastering engineers tried to get away with, I got sensitised to it. Your track is not in that league and sounds just fine, but there is evidence of analogue limiting to allow the average level to be set high, and reconstruction filters take a few of the peaks over the nominal max level. I haven't done a LUFS measurement on it!

Hey Boz, did you do this analysis from the mp3 Dave posted?

kmetal, post: 465432, member: 37533 wrote: Hey Boz, did you do this analysis from the mp3 Dave posted?

Yes, partly, as that is currently all the information to go on. You have to take into account that the levels on the original CD track are likely to have been modified by going through an MP3 encoder, but this track has a feel of the MP3 being generated as part of the mastering process. Maybe Dave could comment on that.

Level overs are not easy to see using digital means, as they do not present as overs until they are put through a DAC and reconstruction filter. As part of the LUFS project I mentioned earlier, I built a DAC -> filter -> ADC rig that captured average levels and output overs up to 150% of nominal FS (about 4dB over). This track of Dave's has peaks of less than 1dB over, which is minor compared with some of the really hot rock tracks I was monitoring in the lead up to LUFS formulation.

I think I have posted here before that I was also active in this area at the time of the first CDs coming on to the market, although the equipment I had in those days was very crude compared with what can be done today. What I do know is that my work helped to weasel out the many designers of CD players who apparently had no idea that the analogue output of a reconstructed waveform could go higher than the reference voltage applied to the DACs, as quite a few players were presenting with occasional flat-top waveforms, even when reproducing classical CDs. In addition, many players had standard output filters that did not implement the sinc (sin(x)/x) characteristic necessary for a true reconstruction filter. I have always thought that these shortcomings were part of the "digital sound" of the early CDs, but, to be fair, there were several other factors contributing to this.

Interesting discussion and observations my friends. This is what I know about this track. It has been encoded to MP3 through iTunes because the site won't let me load it in M4a. I'm not sure there's any difference in overall volume because of this. I hear certain bandwidths being chopped a bit and thats just on the computer speakers.

This track was taken from the released mastered CD. I had to convert it to MP3 to post it. I would be interested to know what this does to the tracks and why.

Davedog, post: 465435, member: 4495 wrote:

This track was taken from the released mastered CD. I had to convert it to MP3 to post it. I would be interested to know what this does to the tracks and why.

missing information (the dropping of least significant bits). i heard an MP3 of a Doors song where the guitar sounded like a banjo. depending on the rate the effect is exacerbated or reduced.

Thanks Kurt. I am well aware of the sonic differences I was thinking more about Boswell's analysis through his equipment he has developed for such a thing and whether or not he can map the frequencies missing.

The techniques I used were pan-spectrum, that is, they considered all the frequencies in the audio range. It would be perfectly possible to make moving spectrograms of the audio before and after compression, and this could show, for example, that certain instruments were contributing unduly to any spectral peaks.

A case in point might be the way you had mixed yourself playing the bass guitar in the "Move" track. The bass comes across as well placed in the mix, and it does not show exceptional levels in a moving spectrogram. However, inspection of the temporal waveforms shows that it's the bass that contributes most to the excess levels.

This is not unusual, as the Fletcher-Munson loudness curves show that at 60dBA intensity it takes something like a 20dB greater amplitude of 60Hz to sound as loud as 1KHz. The shape of the curves can also go some way to explaining why you may feel that some frequencies may be "missing", particularly at different loudness levels.

It also brings to mind the RIAA frequency pre-compensations in LP pressings, as it is used to reduce the excessive groove excursion in passages that have a strong bass component. I have toyed in the past with applying a similar frequency-dependent amplitude reduction in digital recordings (CDs and others), with corresponding expansion on replay. I argued that the technique could open up a greater dynamic range in 16-bit recordings (CDs) without the side-effects of Dolby and similar noise-reduction methods. I did not continue with the idea after I got little interest from the digital media profession, who thought that the mass market was descending to lower and lower sound quality and that 24-bit SACDs would be the rescue for the purists.

Very interesting stuff.
Not sure how far back you developed this, but with the SACD mention I can kinda figure it. It's a shame the "digital media profession" didn't take hold of your discoveries at an earlier time. Perhaps we could have missed out on the digital "hash" we experienced when the media changed.

Now it is, as you said, a shrinking delivery system with very little regard for fidelity at the end-user. Though I will say, a better made recording still sounds better on any delivery system it's put through. It's the budgets that suffered the most and disposable music took over the airwaves. It's in most genres now and that's really a shame. I know I'm not alone in wanting things to sound as pristine and connected as I can and the budget be damned in some cases.

It took me a long time to learn how to "fill a track" in the digital media in a way that got the best fidelity out of the instrument without asking for a bunch of plugs to repair or steady what the mix called for. When I started really getting into stems, aux subs, and VCA masters it started to make more sense. Some mixes I see from time to time have a huge amount of plug-ins and a huge amount of tracks. And while I may indeed work 30-50 tracks on a song it's really only 8-16 that matter and are the basis for the song and it's structure.

Thanks for sharing this Bos!

Interesting stuff. So to clarify the mp3 encoding doesn't effect the peaks/gain?

I have heard that mp3 encoding can cause overs that aren't otherwise there. I don't have any direct references, and am not sure if that was specific to intersample peaking.

kmetal, post: 465452, member: 37533 wrote: Interesting stuff. So to clarify the mp3 encoding doesn't effect the peaks/gain? I have heard that mp3 encoding can cause overs that aren't otherwise there. I don't have any direct references, and am not sure if that was specific to intersample peaking.

Yes, I have heard these reports, but since my analyses have usually been on .wav files, I've never investigated them. I do know that some MP3 encoders have a separate gain element that can be applied at the encoding stage, allowing a conservative .wav file to become a hot MP3, but I have no information about how widespread its use is. The gain element could be a simple normalisation, or it could be a user-entered gain factor that could take peaks over scale.