What is "detail"?
May 9, 2019 at 9:16 AM Post #106 of 142
I was a long time ago, but when Inwas designing mixing desks my employer sent us on a training course at a sound recording collage, organised by the AES if I recall. There we were told and shown how classical recordings were edited and chopped between takes to get the best of a buch of takes. We were informed that they averaged two edits per bar generally. This may not be your experience, but it help explain to me why I found more modern classical recordings of the time sterile. How can you have musical thread building if you chop between 3-4 takes to get the best "note"?

To be honest, that is not my experience of classical recordings and it would be entirely impractical in most cases, especially with larger ensembles. Firstly, doing multiple "takes" obviously requires considerably more time than just doing one or two and the economics of classical recordings simply doesn't allow for paying say 50-100 orchestral musicians for days on end to do multiple takes. An entire classical album would usually be recorded within a couple of days, while a pop album would usually take several weeks to record (plus several more to edit and mix). Secondly, classical musicians (and classical music compositions) are acoustic musicians, they have to be able to perform those same pieces live, without sound systems, mixing or any sort of electronic interventions/manipulations and to audiences which are typically well informed and capable of acute critical judgement, unlike most pop gig audiences. Pop musicians are not acoustic musicians and even if they had the same acoustic performance experience/ability as classical musicians that still wouldn't help much, as pop music compositions/productions are not "acoustic" and cannot be performed live exactly the same as the recording. "Two edits per bar" (or more) is entirely usual for pop music genres but not for classical music, where it's more likely to be two edits per movement.

When I was recording on 24 track, we would "punch in" to fix pitchy vocals. This involved playing back the track and recording a patch over the bad part. I was working with excellent singers, so we didn't have to do that a lot. If we edited, we usually did it a bar at a time. It's a lot easier to edit with digital. Now we can easily assemble an optimal performance by editing together multiple takes. You can splice word by word or even parts of words imperceptibly. It's a lot better to just get a good take, but after the recording session is done and the talent is gone, it's just an editor working on the track, and you do what you have to do to make it work.

Of course, it varies by band, somewhat by genre, by song and by the desired outcome. Reportedly, one of the reasons that ABBA split is because they'd had enough of working so hard in the studio, not uncommonly doing a hundred or so takes for a single song, to achieve the quality of production they were after.

G
 
May 9, 2019 at 8:11 PM Post #107 of 142
The perfectionist thing has broken up quite a few bands...Rumours era Fleetwood Mac and the Eagles where both notable for having having a member with a perfectionist tilt that took the fun out of playing music.I personally prefer a looser sound from rock bands....less mechanical....the Stones,Faces,Zepplin,Black Crowes ect sound like they might be half cut on their best albums....might be that slack timing mentioned by jagwap earlier:)...the difference between capturing the spirit of an event and technical wizardry?
 
Last edited:
May 10, 2019 at 1:15 AM Post #108 of 142
I'm not a rock-n-roll guy at all, and I prefer Diana Krall rather than AC/DC. However, last week I prepared some DSP presets for rock-n-roll. I tried a few AC/DC hits and quickly realized that 350mW isn't really enough for my HFM HE400i to drive AC/DC pleasantly, that corresponds to about 120db SPL! Such music isn't required high-res at all, due to were recorded not really perfectly, but what such music is really need is the VOLUME, to be ROCK it has to be a lot louder vs vocal jazz or classic. Finally, I made a "maximizer" settings on the compressor cell(limiter with 1mS attack, 150mS release and pre-gain. Also I allow some light clipping on the kick drum with RMS detector instead of peak one) i.e. I get AC/DC virtually louder with the same physical power, and I got feel the drive as it suppose to be. Yeah, almost, I know but my ears are still safe because in a brute-force approach I have to go about painfull 130db or so )) Tonnes of modern music are compressed this way already(or processed by FM stations) i.e. maximized but not 70-80s rock music like AC/DC or Led Zeppelin, and physically it's not always possible to play that loud enough on an open-back planars. Also not a bad idea to add some light compression for applied only for about <-20dbfs, that's common for vinyl production, to compensate an ambient noise passing thru the open-back headphones, that's noticeably increasing the noise of "low-res" 16bits records but still perfect for 24bits yet.
PS: regarding more/less detailed performing audio stuff, B. Putzeys believes that the THD+N vs frequency plot can tell us about that more than usually expects. If distortions of DAC or Amp or whatever, aren't the same for mid and high frequencies(the's very common for many audio devices because nobody really cares), highs will sound smoother (like a bit rolled off with light LPF) than the same stuff but with equal THD across the audio. Interesting, the effect is still audible if THD+N vs Freq flat and .05% versus THD+N vs Freq isn't flat and THD@1kHz .005% THD@6.6kHz .05%(like the tpa6130 very popular HPA chip $1/pcs)! So it is not about the absolute value of distortions at all. If compared systems had .05% for 1kHz and 6.6kHz and another one .005%(or .0001%) for 1kHz and 6.6kHz then almost no one may surely hear the difference.
 
May 10, 2019 at 2:00 AM Post #109 of 142
I'm not a rock-n-roll guy at all, and I prefer Diana Krall rather than AC/DC. However, last week I prepared some DSP presets for rock-n-roll. I tried a few AC/DC hits and quickly realized that 350mW isn't really enough for my HFM HE400i to drive AC/DC pleasantly, that corresponds to about 120db SPL! Such music isn't required high-res at all, due to were recorded not really perfectly, but what such music is really need is the VOLUME, to be ROCK it has to be a lot louder vs vocal jazz or classic. Finally, I made a "maximizer" settings on the compressor cell(limiter with 1mS attack, 150mS release and pre-gain. Also I allow some light clipping on the kick drum with RMS detector instead of peak one) i.e. I get AC/DC virtually louder with the same physical power, and I got feel the drive as it suppose to be. Yeah, almost, I know but my ears are still safe because in a brute-force approach I have to go about painfull 130db or so )) Tonnes of modern music are compressed this way already(or processed by FM stations) i.e. maximized but not 70-80s rock music like AC/DC or Led Zeppelin, and physically it's not always possible to play that loud enough on an open-back planars. Also not a bad idea to add some light compression for applied only for about <-20dbfs, that's common for vinyl production, to compensate an ambient noise passing thru the open-back headphones, that's noticeably increasing the noise of "low-res" 16bits records but still perfect for 24bits yet.

What is your source of 350mW? Also Tyll's measurments suggest 5Vrms is needed for 120dB on these, and that's 0.583mW. If it is on a phone and you meant 350mV, then yes you need more power as that is 96dB peak, 84dB average. You can hear it, but not rock out. May I suggest an LG phone with the impedance trick. That can give you an extra 12dB.

120dB is the peak. AC/DC are a loud band, but there best stuff comes from before the loudness wars, so the average is around 12dB lower. Even so that is loud. 108dB. Damaging your ears quickly.

By compressing the track you are bringing up the average level AND adding distortion. Both make things sound louder. Compressors distort.
 
May 10, 2019 at 2:28 AM Post #110 of 142
What is your source of 350mW? Also Tyll's measurments suggest 5Vrms is needed for 120dB on these, and that's 0.583mW. If it is on a phone and you meant 350mV, then yes you need more power as that is 96dB peak, 84dB average. You can hear it, but not rock out. May I suggest an LG phone with the impedance trick. That can give you an extra 12dB.

120dB is the peak. AC/DC are a loud band, but there best stuff comes from before the loudness wars, so the average is around 12dB lower. Even so that is loud. 108dB. Damaging your ears quickly.

By compressing the track you are bringing up the average level AND adding distortion. Both make things sound louder. Compressors distort.
agreed about 5Vrms for 120dB (based on Tyll's PDF), but 5²/45ohm= 0.56W not mW.
 
May 10, 2019 at 3:10 AM Post #112 of 142
I'm talking about so-called Fully Digital "Amplifier" or Power DAC(USB->digitally modulated PWM->72 LVC gates in parallel->passive LC demodulation so no analog amplification or feedback involved) which in my case powered from 5VDC USB rail, the highest power is available in case if power Y-splitter used(USB audio stream could be from your LG phone, and 5VDC power from 5V 2A charger). In that case, I can get nearly exact 3.5VRMS@1kHz@32ohm i.e. over 380mW. The official specs of HE400i claim 35ohm, hence 3.5VRMS^2/35 = 350mW(or over 600mW@16ohm), next, according to the same spec list 1mW produces 93db SPL, hence 350mW corresponds to 118.44068044350275635498477363868db SPL it is 1.559db less than 120db SPL, that's why in my text you can see "that corresponds to about 120db SPL!" :wink: Compressor does manipulate with original dynamic if you call that distortions - Ok but THD+N is not affected.
 
May 10, 2019 at 3:31 AM Post #113 of 142
I'm talking about so-called Fully Digital "Amplifier" or Power DAC(USB->digitally modulated PWM->72 LVC gates in parallel->passive LC demodulation so no analog amplification or feedback involved) which in my case powered from 5VDC USB rail, the highest power is available in case if power Y-splitter used(USB audio stream could be from your LG phone, and 5VDC power from 5V 2A charger). In that case, I can get nearly exact 3.5VRMS@1kHz@32ohm i.e. over 380mW. The official specs of HE400i claim 35ohm, hence 3.5VRMS^2/35 = 350mW(or over 600mW@16ohm), next, according to the same spec list 1mW produces 93db SPL, hence 350mW corresponds to 118.44068044350275635498477363868db SPL it is 1.559db less than 120db SPL, that's why in my text you can see "that corresponds to about 120db SPL!" :wink: Compressor does manipulate with original dynamic if you call that distortions - Ok but THD+N is not affected.

Compressors alway add THD if you are in their operating range. It is unfortunately in there nature as they are non-linear. Michael Gerzon was working on the theorectical limit when he died. They became the Gerzon Limiter plug-ins.

If the above numbers are right, you are at 106dB average. That is heading to hearing damage quite fast. Add the compression and you are speeding it up. What is your loaded output impedance?
 
May 10, 2019 at 3:40 AM Post #114 of 142
I specify Zout as 1.5ohm@1kHz but in fact, it is a bit lower. And again, the compressor doesn't affect THD+N of static sine if release time slow enough, anyway, for the PowerDAC which is open loop system and its THD at level .006% is considered as a great result, I can not see compressor's distortions contributions.
 
May 10, 2019 at 9:08 AM Post #115 of 142
The perfectionist thing has broken up quite a few bands...Rumours era Fleetwood Mac and the Eagles where both notable for having having a member with a perfectionist tilt that took the fun out of playing music.I personally prefer a looser sound from rock bands....less mechanical....the Stones,Faces,Zepplin,Black Crowes ect

Yep, Brian Wilson (Beach Boys) is another famous example but there have been many. In fact, this was largely the reasoning behind the Punk movement in mid/late '70's, to get away from the highly polished/produced trend of the time and NOT incidentally, make a bucket load of profit on the basis that the recording/production cost was just a small fraction.

[1] Finally, I made a "maximizer" settings on the compressor cell(limiter with 1mS attack, 150mS release and pre-gain.
[2] I get AC/DC virtually louder with the same physical power, and I got feel the drive as it suppose to be.
[3] Yeah, almost, I know but my ears are still safe because in a brute-force approach I have to go about painfull 130db or so ))

1. That's not a "maximizer" setting. A "Maximizer" is a specific class/type of processor (the term is an abbreviation of "loudness maximizer") which is somewhat different to a compressor/limiter (which is effectively a "level maximizer"). The difference between the two (loudness and level) is human perception, certain regions of the audible freq spectrum sound significantly louder than others (at the same level). A "Maximizer" therefore always consists of a compressor/limiter, typically a "look-ahead" limiter (and therefore an attack time effectively of 0ms) and an "aural exciter" to add harmonics and/or EQ and/or multi-band compression to emphasise the regions to which we're more sensitive, thereby maximising the loudness rather than only the level.

2. It's not the same physical power, it maybe about the same peak power but it's higher RMS. And, it's not the feel/drive as it's supposed to be, the feel/drive it's supposed to be already exists on the recording/master, you are adding significantly more than "it's supposed to be", although maybe you prefer it that way, which is your choice of course.

3. No, your ears are NOT "still safe", you are FALSELY assuming the threshold of pain and the threshold of hearing damage are the same but they are NOT, the threshold of hearing damage is more than 100 times lower than the threshold of pain! Although, hearing damage depends on the duration of continuous exposure. With a continuous exposure to 106dB, damage is likely to start occurring after about just 5 minutes and at 130db, almost instantly. I strongly recommend you READ THIS.

[1] AC/DC are a loud band, but there best stuff comes from before the loudness wars ...
[2] By compressing the track you are bringing up the average level AND adding distortion. Both make things sound louder.

1. Not really. The loudness wars started in the 1950's with juke boxes and then radio play, "metal" bands in the 1980's pushed the loudness war as far as it was possible to go and even beyond "what it was possible to go", into what could be considered poor/amateurish/bad taste levels of compression/distortion. So much so, that by the early 1990's some/many mastering engineers were collectively and openly warning and complaining about the amount of compression/limiting they were being asked/required to apply. Then in the mid 1990's with the advent of software "look-ahead" limiters, it got even worse (!) and some time/years after that, the audiophile press finally noticed and turned it into a more public/consumer issue.

2. Two points here:
Firstly: We must be careful not to just state "compressors" in general. Early analogue compressors and limiters added noise and all sorts of distortions (IMD, THD and others), by the late 1960's they were sometimes being deliberately driven into that distortion (as a musical effect) and by the early 1980's they were being thoroughly "abused"! Over-driven into severe distortion by bands/engineers looking for something a little more "over the top". This culminated in the mid 1990's with a unit by Empirical Labs called the "Distressor" (derived from Distortion + Compressor), ownership of which was de rigueur for just about every cool music studio on the planet. It provided the over-driven distortion/compression characteristics of several of the classic, vintage compressors flexibly in one box and did it extremely well, a modern classic. It even had a "nuke" setting if I remember correctly! On the other hand, late analogue and then digital technology allowed compressors and limiters with an increasingly linear response and little or no unintentional distortion. These "clean" compressors and limiters were/are used by film/TV engineers, while the "coloured" compressors are used/prized by the music engineers (although classical music engineers tend towards cleaner compressors). An original, good condition vintage compressor can go for as much as $50k, although these days many/most use plugins which model/emulate them (even the Distressor is now available as a plugin). Many/Most of the compressor and limiter plugins for use in DAWs are designed for music production and are therefore usually at least somewhat "coloured". So if we're talking about compressor distortion we need to be careful about what sort of compressor we're talking about.

Secondly: The risk of causing hearing damage increases with sound intensity, not necessarily by what sounds louder. This brings me back to point #1 above in my response to E1DA and the difference between level and loudness. However, this doesn't detract from your warnings of hearing damage (with which I completely agree), as E1DA is maximising level rather than loudness.

G
 
Last edited:
May 10, 2019 at 9:33 AM Post #116 of 142
gregorio, quite accurate corrections, and I agree. Only maybe about level vs volume maximizing difference doesn't look clear, and about the power, I said that 350mW isn't enough for AC/DC drive that means peak ability(700mW), of course, AVG power I can get 2x more :wink: BTW, do you think it is bad idea to add classical compressor cell into the DSP chain of HPA(where are already parametric EQs, Fletcher-Munson loudness curves, bass/treble sliders)?
 
May 12, 2019 at 7:02 AM Post #117 of 142
[1] gregorio, quite accurate corrections, and I agree. Only maybe about level vs volume maximizing difference doesn't look clear, and about the power, I said that 350mW isn't enough for AC/DC drive that means peak ability(700mW), of course, AVG power I can get 2x more :wink:
[2] BTW, do you think it is bad idea to add classical compressor cell into the DSP chain of HPA(where are already parametric EQs, Fletcher-Munson loudness curves, bass/treble sliders)?

1. Yes, I agree that I wasn't clear about "physical power" and used it to mean the physical, acoustic SPL energy rather than electrical power. That's because electrical power is effectively arbitrary relative to SPL and it's SPL that we hear and that can cause hearing damage, which you obviously already know. For example 350mW maybe plenty to provide high SPLs with some IEMs but under other conditions (say a large cinema for example) even 10,000W may not be enough to provide the same SPLs. So I was going on the stated peak and RMS SPL energy rather than the wattage of the signal being fed to your transducers. I also agree that I wasn't particularly clear about level vs volume, which is because it's far more complex than it appears. The conscious process of determining volume/loudness appears trivially easy, it requires pretty much no conscious effort or thought, we just automatically know that something is louder or quieter than something else. The fact that all humans (without severely impaired hearing) can achieve loudness determination so easily implies that volume/loudness is a simple, objective property but actually that's not the case, it's actually a "perception" rather than a property and a rather complex process. You're obviously aware there's another variable at play beyond just differences in SPL, as you've mentioned F/M loudness contours, however there's still more variables at play beyond these two. For example, without exception (as far as I'm aware), a louder version of the same naturally occurring sound not only has a higher SPL but also a higher number and balance of harmonics. A shout has both a higher level and more harmonics than normal speaking, the same for a drum hit harder or an acoustic instrument played louder and our brains/perception uses both of these properties in it's determination of volume/loudness. So, if we take say a drum hit (or just about any other sound), artificially add some harmonics to it, adjust the level so it has the same SPL energy as the original hit and then compare the two, the one with the added harmonics will sound louder or, if we take a shout and reduce it's level to the same as normal speech, it still sounds louder. There's also the variable of context; a lower SPL sound can be perceived as louder than the same sound with a higher level depending on what precedes the sound. Additionally, there's the the variable of acoustic size/distance; even compensating for the loss of SPL due to distance, sound with the same SPL in a larger room will be perceived as quieter than the same sound (at the same SPL) in a smaller room. There's also the variable of duration; a short duration sound will be perceived as quieter than the same sound, at the same SPL but with a longer duration. And finally, all these variables interact, so adding more of one of these variables might not increase the perception of loudness depending on say the context. Maximising loudness is therefore significantly different to just maximising level, as maximising loudness will take all (or at least more) of these other variables into account. Incidentally, this is also why I have some objections to the characterisation of the "loudness wars", which isn't just an issue of compression/limiting but also of these other variables.

2. Firstly, I'm not sure what you mean by "classical compressor", do you mean a classic "coloured" compressor or the sort of clean compressor usually preferred in classical music production? If it's the former, then personally I would consider it a "bad idea" as you are "colouring" a recording which already has exactly the amount of that type of colouration desired by the artists (musicians and engineers). In other words, by using a coloured "classic" compressor you are not only changing the fidelity of the intended relative levels (RMS, Crest Factor, difference between quieter and louder sections) but also changing the fidelity of the colouration (intended frequency content, distortion, etc.). However, ultimately it's a matter of personal preference, some people prefer the colouration of using tubes for example. Secondly, are you saying that you are applying Fletcher/Munson loudness curves (or the inverse, to compensate for them)? If so, that's a mistake IMO. Mixes/Masters are created by engineers who are human, who, along with all other humans, have FM loudness contours themselves and therefore, the FM loudness contours are already built-into the released recordings. And, I'm not just talking about a single contour. Typically during mastering, a song will be played back at different levels and the amount of bass and treble set at a compromise level, which would usually err towards having a bit too much bass at higher playback levels. So, unless you're playing back music at especially low levels (in which case adding some extra bass might be a good idea) then I personally wouldn't apply any compensation for FM curves as you're effectively compensating what has already been compensated! Thirdly, it also depends where in the chain you insert the compressor: The input signal to the compressor will be different if the compressor is inserted into the chain before say the EQ than if it's inserted after the EQ and therefore the characteristics of the compression will be different. Where in the chain to place the compressor is a common question from student engineers but there's no right or wrong answer, it depends on the signal/song (and the desired outcome), what might be best for one signal/song might be worse for another and so the best advice we can give to students is to try both ways around and build experience of what difference it makes.

Again, personally I think it's all a "bad idea"! I personally am after the highest fidelity I can get and therefore the only EQ or any other sort of DSP I would ever add is to compensate for some freq response weakness of my particular transducers/listening room acoustics and even then, only in relatively small amounts. If I couldn't get the power required to drive my transducers to the SPL I desired, I would get a more powerful amp rather than effectively mangling the recording to get more RMS into the amp (and volume out) but with the caveat that the "SPL I desire" does not present a significant risk of damaging my hearing. Given the choice though, many audiophiles (and other consumers) would choose their own subjective preferences over high fidelity, that's their choice and not such a "bad idea" (for them), providing of course they're not risking hearing damage.

G
 
Last edited:
May 12, 2019 at 7:42 AM Post #118 of 142
I using quite normal compressor cell with no artificial coloration or so, if you wish you can take a look my Android App here https://drive.google.com/open?id=1GNMPSEoxISauSIg4YKfEDFLSfvknuu3D It can work in demo mode without HW. Only one thing there a but unusual is Energy slider which allows making the transition between Peak to RMS detection. Fletcher-Munson curves implemented in the loudness(and only low-frequency part of it, Ti did that algorithm, not me) for Volume slider, that's it.
 
May 12, 2019 at 9:35 AM Post #119 of 142
gregorio, well know your precise&slow style of reply, let me ask you one more question right now.. What's your opinion about the statement of Bruno Putzeys regarding that the flat THD+N vs freq plot is the factor(one of) of sound neutrality?
 

Users who are viewing this thread

Back
Top