As a start, this is the original, https://www.youtube.com/watch?v=6H5we9iNGIA, by Thirsty Merc (yes, as in MB). NB: You are required to wear a certified DBT blindfold while watching, 
.
On a decent system this is very heavy going ... high intensity aggressive ...
. On a decent system this is very heavy going ... high intensity aggressive ...
Ouch ! It is inaudible.
A miraccle: Not a single instrument that still sound like something.
Even the pan on the waves at the begining is dud.
Like everything was compressed abominably (on purpose?) individually, it is irretrievable.
The musical landscape looks like a city after a bombing.
Last edited:
All meters, analog mechanical by definition or led electronic have some form of averaging which means in practice that short term and more especially very short term peaks are not truly represented.
Mechanical VU meters are the classic overload error causing level indicator in the hands of inexperienced operators, followed closely by PPM level indicators such as the BBC 7 (don't go over 6) scale.
Even sample peak metering is inaccurate, see diagram below...
bbc.co.uk/WHP202.pdf.
Peak_programme_meter
The only real and correct level indicator for digital audio is DAW waveform display, or level calibrated real time oscilloscope monitoring analog signal lines.
With all the averaging PPM metering standards in current use, it is no wonder that modern recording methodologies and final master results are in such a sorry mess, particularly in the hands of operators who do not understand visual metering standards behaviours and limitations.
In the the current 'loudness wars' environment is it to be expected that final playback DAC 0dBFS will be violated ?.
The real question is...How is it that over modulated/clipped recordings that should have at least some headroom make it to CD release ??.
Is it that nobody in final mastering actually takes a close look at DAW waveform display to determine real short term peak levels ??.
For example, how did the 'famously' clipped Norah Jones recordings make it to final release ??. AudioCompression.php
16 bit done correctly is good (enough), 24 bit can and ought to be better, and higher than 44k sampling rates takes things to a better level if and only if correct level practices become the norm as it was in the days of analog (tape/vinyl) in the hands of skilled and experienced operators.
The old days of 'white coats' audio engineering did produce great results that have stood the test of time, hoodies, t-shirts and tattered jeans audio engineering seems to be the norm nowadays and has not been an advance.
Another big problem in modern audio is SRC....48k or 96k source recording SRC to 44k is not a happy outcome but is the modern norm.
192k to 44k SRC is better but still not ideal/correct.
Max Headroom rant over....for now.
Dan.
Mechanical VU meters are the classic overload error causing level indicator in the hands of inexperienced operators, followed closely by PPM level indicators such as the BBC 7 (don't go over 6) scale.
Even sample peak metering is inaccurate, see diagram below...
bbc.co.uk/WHP202.pdf.
Peak_programme_meter
The only real and correct level indicator for digital audio is DAW waveform display, or level calibrated real time oscilloscope monitoring analog signal lines.
With all the averaging PPM metering standards in current use, it is no wonder that modern recording methodologies and final master results are in such a sorry mess, particularly in the hands of operators who do not understand visual metering standards behaviours and limitations.
In the the current 'loudness wars' environment is it to be expected that final playback DAC 0dBFS will be violated ?.
The real question is...How is it that over modulated/clipped recordings that should have at least some headroom make it to CD release ??.
Is it that nobody in final mastering actually takes a close look at DAW waveform display to determine real short term peak levels ??.
For example, how did the 'famously' clipped Norah Jones recordings make it to final release ??. AudioCompression.php
16 bit done correctly is good (enough), 24 bit can and ought to be better, and higher than 44k sampling rates takes things to a better level if and only if correct level practices become the norm as it was in the days of analog (tape/vinyl) in the hands of skilled and experienced operators.
The old days of 'white coats' audio engineering did produce great results that have stood the test of time, hoodies, t-shirts and tattered jeans audio engineering seems to be the norm nowadays and has not been an advance.
Another big problem in modern audio is SRC....48k or 96k source recording SRC to 44k is not a happy outcome but is the modern norm.
192k to 44k SRC is better but still not ideal/correct.
Max Headroom rant over....for now.
Dan.
Last edited:
On commercial FM it is/was even worse.
There are a couple of lower level vocal bits in the recording that sound reasonably natural.
To be fair this multi-band over compression is used as an effect, but ultimately is tiring.
If the compression/limiting was pulled back a bit the track could go much louder on PB systems and be more tuneful, within the limits of mega distorted/overdriven guitar of course.
This modern 'indie 'sound is the kind of audio that youngsters expect however.
I have front truss follow spot show call for 'The Prodigy' this Sunday...might be interesting.
BIG bass will be the order of the day.
Dan.
Last edited:
Okay ... sounds like it's worthwhile finding what I managed to get back, ... this will sound like the dog ate my homework excuse, but this decompression exercise was done about 3 years ago - and I've just been reminded what happened last time I tried to find the finished file - there was a stuff up some years ago when some folders were shifted, the computer went funny on me, and a whole bunch of music work files just vanished, . However, there is still the burnt CD about, so I will go a-hunting to locate it ... if the worst comes to the worst, I can recreate the steps, and spit it out again ... .
... this will sound like the dog ate my homework excuse, but this decompression exercise was done about 3 years ago - and I've just been reminded what happened last time I tried to find the finished file - there was a stuff up some years ago when some folders were shifted, the computer went funny on me, and a whole bunch of music work files just vanished, . However, there is still the burnt CD about, so I will go a-hunting to locate it ... if the worst comes to the worst, I can recreate the steps, and spit it out again ... .
Yeah, just a 900kB grab that you can attach directly might be good enough.
Interested minds want to know.
My on loan DEQ2496 should be coming back in the next week or so.
In the past I did have a play with expansion ratio/threshold settings and got some interestingly good results in 'revitalising' overly squashed recordings.
When I get it back I should revisit the experiment along the lines of your experimentation.
Dan.
Interested minds want to know
.My on loan DEQ2496 should be coming back in the next week or so.
In the past I did have a play with expansion ratio/threshold settings and got some interestingly good results in 'revitalising' overly squashed recordings.
When I get it back I should revisit the experiment along the lines of your experimentation.
Dan.
it is a very complex problem.
Vu meters are perfect when you use analog tapes. It gives you a good idea of the average level. (the feeling of loudness). The analog tape can deal nicely with peaks, as long you set your 0dB VU somewhere like +8dB over the 0dB of you tape. And add some king of nice limitation on very fast transients.
But you have to be aware of the content you are metering. Transients of drums that you have to under modulate, or an electric guitar with sustain that you can over modulate. I like to use them to meter the final mix.
BBC peak meters are perfect on separate tracks to ensure you do not saturate the peaks. To be used in correlation with the vu-meters.
With digital, you need a real peak meter with memory to ensure you'll never reach the highest bit.
Sometimes, you are obliged to saturate. I remember, on a movie track this rubbed match noise witch was in the red while very low in the ears. Happily, ears are not sensible to very short clippings if they are not repetitive.
To simplify, like when you measure the light in a very contrasted scene in photography, you have to know what you want to do, what is your measuring instrument, and how to interpret-it. And to lie on experience and your ears...
The thing is that, if you record strait in digital with close miking, and never reach the top, you'll have too low average level at the end. A sound a little inconsistent.
Last edited:
Ah hah! Haven't found that one yet, but there was another exercise I did: Californication, by the Red Hot Chilli Peppers - this was notorious for being one of the most extreme examples at one time, of the loudness disease - and this was put through my "unwringer", so to speak. What I ended up doing was to create a special track, which had a short segment of the original, compressed clip concatenated with the restoration, then back to the compressed, and restoration, ... back and forth a couple of times, to make it easy to pick what's been done.
I'll put part or whole of this special exercise up, for download - not tonight, close to my bedtime! - I'll get onto it in the morning,.
I'll put part or whole of this special exercise up, for download - not tonight, close to my bedtime! - I'll get onto it in the morning,
.
Last edited:
Yes, since 1942 the VU meter ballistics standard is indeed intended as a Volume Units ie subjective loudness indication, with no intention of peaks amplitude real indication.
This is where skill and experience come into play in order to interpret/correlate what such (VU) metering is indicating according to the nature of the programme content.
Mechanical VU metering was/is indeed quite valid for tape systems and 'old timey'/natural music content, but perhaps not so well suited for modern rock/pop music playing into digital recording systems with defined clipping overload behaviour as opposed to the soft overload characteristics of analog/tape captures.
The beauty of modern digital 24 bit capture is that average levels can be set relatively low in order to capture peaks/transients undistorted without incurring significant system noise as is normal/expected with tape systems.
This can then allow post track recording processing/mixing/mixdown without signal degradation provided that operators understand correct digital gain structuring
Dan..
This is where skill and experience come into play in order to interpret/correlate what such (VU) metering is indicating according to the nature of the programme content.
Mechanical VU metering was/is indeed quite valid for tape systems and 'old timey'/natural music content, but perhaps not so well suited for modern rock/pop music playing into digital recording systems with defined clipping overload behaviour as opposed to the soft overload characteristics of analog/tape captures.
The beauty of modern digital 24 bit capture is that average levels can be set relatively low in order to capture peaks/transients undistorted without incurring significant system noise as is normal/expected with tape systems.
This can then allow post track recording processing/mixing/mixdown without signal degradation provided that operators understand correct digital gain structuring
Dan..
Like this,
OTOH this is a newish sound card from ASUS that claims 124dB SNR and seems to meet the "standard".
I was thinking last night of a point possibly missed, professional recording gear has to work in general at real time hence the sometimes massive amount of DSP chips, cost, etc. With enough computational load a home user might find the need for offline processing. Though it is amazing how much you can do. Numerically (resolution wise) there is nothing an Intel Pentium can't do that a DSP can. Remember also a Pentium costs 10-100 X a DSP chip and has a large programming overhead. For instance there's nothing the basic miniDSP does that a decent sound card and readily available free software can do. But at only $100 it can be put in standalone mode and do it by itself.
In the end it is programming skill and knowledge of the algorithms, I don't think the state of digital audio has anything to do with the quality or sophistication of the currently available analog or digital IC's.
Sorry, I'm just trying to point out that a single stake in the ground number for SNR is about as useful as THD at 1kHz as a measure of amplifier sound quality. In any case with some real analog engineering ASUS seems to have a sound card they spec at -124dB and one user here measured -118dB in his setup with very low spurs too.
And yes I did read your link as I already said archiving historical recordings at no further loss is an admirable goal. If you can point me to a single master tape with -118dB SNR I would be interested in the data point.
Last edited:
What the purpose ?
In the best recording studios i can imagine, between the ambient noise, the hisses from mike preamps, the noise out of the electric guitar amps etc... there is really no need for any dithering, even at 16 bits ;-)
96dB ? Oh Lord !
Thanks God for the masking effect.
When you measure digital SNR you do it broadband, in practice weighting is very important which is why it is used for acoustical measurements from SPL to NC ratings.
There really isn't much issue that human hearing covers a greater range particularly when you look at critical bands and masking.
Now in my typical use a sports announcer may turn to the side and look at his statistician or director, lower their voice as most humans do face to face while still announcing. This provides an input level of around 65 dBa at a distance of 24" off axis (15 dB of pattern control.) They will then get excited and scream (95 dBa) into the microphone at a distance of 1/8". Now to get any intelligibility there needs to be at least 30 dB of modulation. So there is 30 dB of signal increase, 15 dB pattern increase, 46 dB of distance decrease and 30 dB dynamic range required for a worst case announcer. So a 121 dB range converter would seem to meet the need except that requires perfect adjustment of the gain trim.
Now for fun most stadia use an analog microphone control box that doesn't even have this range!
Now there are actual solutions. The best is to replace the announcer with a competent one, sometimes happens. My favorite of course is a head vise with an amplifier driven shock collar, but that isn't going to happen.
ES
That's the point, there is nothing to archive at that level. Safety margin OK maybe.
- Status
- Not open for further replies.