|
Post by EmRR on Jul 10, 2024 8:11:03 GMT -6
I've noted these changes at times but hadn't done a real comparison on a hot WAV master in awhile. Seeing these results suggests -1.5dBFS is even better. Look at sample peak level and possible (oh for sure) clipped samples, and the amount they rise with conversion. Note RMS actually goes DOWN slightly. Sample rate conversion creates these level artifacts too, but not quite as dramatic. Peak levels bear little relationship with RMS or LUFS. Kind of shocking, and given streaming will turn a file like this down, it's pointless to have high peak levels. It's already clipped more heavily if the conversion is before the stream leveling, which I'd ASSume to be the case for storage purposes. I've heard this conversion clipping on some hot masters from my shop, and it bugs the shit out of me....no reason for a clean sounding acoustic record to have mastering induced clipping artifacts. I'm hoping this will get a revision with lower peak levels. The track is a fairly dynamic live performance, so it's also kinda dumb the RMS is pushed so hard. It's already plenty damn loud before mastering, loud enough for streaming to turn it down already. Hey, if pressing CD's knock yerself out with this hot peak level shit. Otherwise...you're fired if I have any say. mastered WAV: ![mastered WAV](https://live.staticflickr.com/65535/53848073014_e2eb9d1f94_o.png) 256K VBR 10Hz filtered smart conversion: ![256K mp3 of same](https://live.staticflickr.com/65535/53848151305_e6a6e0a37f_o.png) 128K VBR 10Hz filtered smart conversion: ![128K mp3 of same](https://live.staticflickr.com/65535/53847978358_2cb16ffef0_o.png)
|
|
|
Post by svart on Jul 10, 2024 8:31:32 GMT -6
I find -11dB RMS/LUFS to be around the right feel for me. Not crushed to the point of distortion, but also not so dynamic that it's going back and forth between hard-to-hear and ear-splitting.
I really don't know how folks can get it into the -8dB RMS or even the -6dB RMS range without it getting so far out of balance, not even taking the distortion stuff into account. I've had folks ask for LOUD and I've been yet unable to get into that territory without it seeming impossible to make sound right.
|
|
|
Post by Dan on Jul 10, 2024 8:47:35 GMT -6
The true peak level is even higher than that from the naive 4x oversampling required in the BS.1770-5 standard. We've heard how well the UAD plugs do not work on sharp peaks. The Gibbs phenomom makes the clipped peaks even higher. They're mostly using digital "limiters" that are incapable of peak suppression and are in essence random defective pumpy AGC devices that reduce the gain far in advance of the signal, misfiring fast compressors, and random PCM clippers. Now some of the current crop of shitty limiters over sample the clipper only so they're just as misbehaved, just different. The limiters they are using mostly cannot operate based on anything close to the signal, only the discrete pcm samples that represent the continuous signal, then operate on the signal from these to horrific effect. The users of these refuse to even read the manuals. Even the hilarious dysfunctional L1 (hey it was the best digital limiter they could make ITB in 1994) has a recommendation of -.3 sample peaks at the absolute maximum.
Even the most expensive records are the highest levels of the industry are butchered. If the recorders, mixers, and mastering guys don't butcher it, the idiot artists and producers will do it themselves after. No wonder the the music industry is in the toilet. Who would buy these records other than people who don't like music? This has been going on since the mid 90s and the results are horrific. Now anyone can buy functional digital processing software for a few hundred dollars without artifacts but they don't want to use it. The same reason they didn't buy Aphex gear in the late 90s and early 2000s. The same reason you see flavor of the month, usually fragile, usually cheaped color pres come and go while the clean rack mount remote pres and better board/interface pres are still used, some of them for probably 40 years now. Who and what are being recorded are mostly in no way ready or worthy of being recorded.
These loud records don't translate either. They cannot. Even if they sound okay from the dc filters, headroom above 0 dbfs or digital volume control, and crossovers on one system, there's no guarantee they won't sound like a random clipped mess on another system. I've heard pop music on supermarket intercoms where the clipping and poor reverb settings were blatantly audible. Lo-fi records from the 90s done on ADAT translate better than these.
|
|
|
Post by Dan on Jul 10, 2024 8:52:41 GMT -6
There are even records now sold to teenage girls where the kick hits the limiter and pumps down the vocal lower so that it's lower than the vocal fx send. The balance can be even worse than the L2 and clip everything era. The multiband limiters do so many stupid things.
|
|
|
Post by EmRR on Jul 10, 2024 8:53:07 GMT -6
I need to look at the 1996 remix of The Stooges - Raw Power. Iggy pushed it well past the clipping level the mastering team was comfortable with, and it sounded different on every CD player because of the purposeful constant distortion. If you have a player with a level meter the 0dBFS red light stays on the whole time. That's cool if your purpose is a super violent sound, and they got it, but most of this stuff doesn't sound purposeful, just bad.
|
|
|
Post by EmRR on Jul 10, 2024 9:02:46 GMT -6
"Penetration" from the 1997 Raw Power remix CD (0dBFS 99.9999% of the time), converted to 192K mp3. I don't have the WAV on hand to see what it looks like. ![](https://live.staticflickr.com/65535/53847837921_6c7868589a_o.png)
|
|
|
Post by Dan on Jul 10, 2024 9:16:45 GMT -6
I need to look at the 1996 remix of The Stooges - Raw Power. Iggy pushed it well past the clipping level the mastering team was comfortable with, and it sounded different on every CD player because of the purposeful constant distortion. If you have a player with a level meter the 0dBFS red light stays on the whole time. That's cool if your purpose is a super violent sound, and they got it, but most of this stuff doesn't sound purposeful, just bad. I have that on my shelf. It sounds terrible and incredibly stupid like a very bad barely in signal radio station on many systems
The 2023 CD version with remasters of both the Bowie and Iggy mixes that are only slightly slammed sounds good. Of course if they didn't limit anything but stray peaks, it would sound even better.
|
|
|
Post by EmRR on Jul 10, 2024 9:30:32 GMT -6
I find -11dB RMS/LUFS to be around the right feel for me. Not crushed to the point of distortion, but also not so dynamic that it's going back and forth between hard-to-hear and ear-splitting. I really don't know how folks can get it into the -8dB RMS or even the -6dB RMS range without it getting so far out of balance, not even taking the distortion stuff into account. I've had folks ask for LOUD and I've been yet unable to get into that territory without it seeming impossible to make sound right. I like that -11dBFS to -12dBFS RMS range for a sense of dynamics. If something is really steady state without much, I can get to -10 no problem but it's clearly strained in comparison. I'm sure if we dug into mastering full time we'd start figuring out louder.
|
|
|
Post by nudwig on Jul 10, 2024 10:17:43 GMT -6
|
|
|
Post by Ward on Jul 10, 2024 10:18:33 GMT -6
It's all LUFS to me. Isn't the 2022-2024 standard -7 anyhow ?
|
|
|
Post by Dan on Jul 10, 2024 11:31:58 GMT -6
It's all LUFS to me. Isn't the 2022-2024 standard -7 anyhow ? The standard is to sound like shit.
|
|
|
Post by bossanova on Jul 10, 2024 14:33:41 GMT -6
If we're already setting levels based on true peaks, does this mean that something like -1dbfs on the master bus is still okay? Or are you suggesting -1.5dbfs even taking true peak into account? This also might sound like a weird question, but who isn't checking final levels based on true peak these days?
Also, what meter program are you using for those numbers?
|
|
|
Post by EmRR on Jul 10, 2024 19:09:56 GMT -6
If we're already setting levels based on true peaks, does this mean that something like -1dbfs on the master bus is still okay? Or are you suggesting -1.5dbfs even taking true peak into account? This also might sound like a weird question, but who isn't checking final levels based on true peak these days? Also, what meter program are you using for those numbers? Numbers from RX11. It’s the conversion to mp3 that’s of concern, we can see from this example that -1 would still end up well over 0. At 128k (more like free spotify/pandora/etc) even the sample peak is +1.4. If the orig is -1 versus -0.1, do you end up with less of a boost? I don’t know if that’s a linear or nonlinear relationship but safety says treat as linear. Who's not checking? A zillion “mastering” guys of unknown provenance.
|
|
|
Post by Blackdawg on Jul 11, 2024 0:01:07 GMT -6
This is why I used TP limiting at -1 was for conversion intersample peaks.
But TP limiting can sound...not as good. And a VAST amount of people seem to not care about it and just send it with a limiter set to -0.5.
And they can get away with it because the only time it "matters" is if you do an ADM file and have to certify that it doesn't do any intersample peaks after Apples AAC conversion. Which is not checked by any means. You just have to say it's okay and then they get the stamp of approval. So it's kind of a joke.
|
|
|
Post by nudwig on Jul 11, 2024 7:21:29 GMT -6
Numbers from RX11. It’s the conversion to mp3 that’s of concern, we can see from this example that -1 would still end up well over 0. At 128k (more like free spotify/pandora/etc) even the sample peak is +1.4. If the orig is -1 versus -0.1, do you end up with less of a boost? I don’t know if that’s a linear or nonlinear relationship but safety says treat as linear. Who's not checking? A zillion “mastering” guys of unknown provenance. That is concerning you're getting that high of a jump when converting, I've never had anything over a 0.1 jump. Is RX11 handling your conversions? I have a few other rate/format converters if you need help invesitaging.
|
|
|
Post by Dan on Jul 11, 2024 9:03:31 GMT -6
This is why I used TP limiting at -1 was for conversion intersample peaks. But TP limiting can sound...not as good. And a VAST amount of people seem to not care about it and just send it with a limiter set to -0.5. And they can get away with it because the only time it "matters" is if you do an ADM file and have to certify that it doesn't do any intersample peaks after Apples AAC conversion. Which is not checked by any means. You just have to say it's okay and then they get the stamp of approval. So it's kind of a joke. Most true peak limiters are just over sampled clippers tacked onto normal crappy limiters. Just l2 with another clipper or oversampling the clipper portion like fabfilter pro l or limitless. Or are another limiter on the backend like Oxford and tdr. Lowering gain or backing off a limiter is better than using them or clipper usually. Or it can be just a better limiter. The tdr “limiter” portion js a better true peak limiter (as in fast compressor) than most supposedly brick wall true peak limiters. Then again so is kotelnikov on ultra fast attacks on insane What’s really stupid if you don’t limit the crap out of it with something nasty, they will say it sounds dull. Because acoustic limiters are functional generally until they go so much faster than the tonal content that they horribly distort it or clip it (for ones with clippers for brief peaks). The aliased digital garbage sample limiters just misapply the fast volume modulation and the clipper and crap up the highs with aliasing in a way that nothing acoustic or better programmed for modern cpus can. They crap regions impossible to harmonically distort in non aliased processors or even to the human ear in analog or sound waves propagating in air because the first harmonic of any fundamental above 10 kHz or so will be above 20 kHz. Add in most older or cheaper equipment like api, behringer, dbx, and rme and the like often just rolled off or distorting from older and cheaper opamps and they will just get duller usually until the stupid phase shift from crappy converters messes up cymbals. Then even a lot of better stuff from the 80s onward can have a bit of a fog to the air and now since 2000s we have insane opamps, killer remote Pres, and very good canned parts from THAT corp. you used to not really be able to get clean highs or really any sort of detail in the highs unless you had a lot of money. Now for a couplei thousand bucks, you can get clean stuff, and with digital we can bypass a lot of the stuff that had to be perfect like tape and mixer line level circuitry to not dull things without having to patching it around and hope their wasn’t noise and get a very clean high end that reveals all flaws of bad mics, filters, circuits, transformers, voices, and amps. But still even then, they do not have the massive distortion. It has to be added in from insane and idiotic aggressive and dysfunctional digital dynamics processing. The early functional digital compressors did not have this. They smoothed the attack and usually were dark or mid forward like Waves Renaissance Compressor, Sony / Sonnox Oxford Dynamics, and the current Fir Comp which takes pointers from them. The modern stuff capable of compressing highs and brief transients cleanly doesn’t have the insane high frequency distortion. The pcm limiters and awful aliased analog emulations add it. Whatever nice analog distortion or digital artifact free slight digital distortion you add pretty much doesn’t matter versus the horrible distorting and pumping of the pcm limiters. This is one of the reasons for analog gear becoming mostly stupid sounding toys instead of tools. Otherwise you can barely hear it through the digital bullshit. The same with the most popular digital distortions being insanely aggressive like decapitator or vintage warmer or incredibly stupid sounding like avid lofi. I’m trying Weiss DS1 (more distortion!) and then if that doesn’t work, simply inserting Oxford limiter with enhance cranked on a mix instead of brand spanking new but dulling MDWDRC2 or boring the glue or glowing Molot GE that fed the very pretty U-he Satin and just sending it in. Previous complaint was too clean then after a bunch of distortion added… too dull. Well time to kick this up a notch with a something that sounds like a synth envelope follower added to a clean comp and if that doesn’t work, the SSL bus comp meets L2 meets a high gain crack pipe amp that is the Oxford Limiter. The more modern Elevate is your typical multiband garbage that remixes tracks and undoes your moves. Meanwhile Waves makes everything crinkly like aluminum foil and Fabfilter Pro L polycarbonate microplastic faced pop star sheen. They all suck but at least the single band stuff doesn’t totally undo your mix moves. Dan
|
|
|
Post by Dan on Jul 11, 2024 9:09:21 GMT -6
Numbers from RX11. It’s the conversion to mp3 that’s of concern, we can see from this example that -1 would still end up well over 0. At 128k (more like free spotify/pandora/etc) even the sample peak is +1.4. If the orig is -1 versus -0.1, do you end up with less of a boost? I don’t know if that’s a linear or nonlinear relationship but safety says treat as linear. Who's not checking? A zillion “mastering” guys of unknown provenance. That is concerning you're getting that high of a jump when converting, I've never had anything over a 0.1 jump. Is RX11 handling your conversions? I have a few other rate/format converters if you need help invesitaging. Hot masters can do that, codecs clip differently, you have no control over the codec used, and all true peak meters read out differently. Only way to monitor them in real time is Sonnox Fraunhofer Pro Codec but even then, the mp3 guys will be using LAME and video codecs are different. www.sonnox.com/plugin/fraunhofer-pro-codecThe standards are poor. I mean the official standards are rather inaccurate. The stuff truly meant for radio broadcast like Orban and Thimeo has much better true peak metering to not overload stuff as does the TDR Limiter 6 GE because TDR does things the right way to the point I’ll even use it sometimes just for metering. Free Orban loudness meter mostly better than what comes with typical scrappy audio software: www.orban.com/meter
|
|
|
Post by EmRR on Jul 11, 2024 9:44:31 GMT -6
That is concerning you're getting that high of a jump when converting, I've never had anything over a 0.1 jump. Is RX11 handling your conversions? I have a few other rate/format converters if you need help investigating. Above is Apple's conversion from Music. VBR, highest, smart encoding adjustments, filter below 10Hz. I clarified in the OP. Some other variations, sticking with roughly 128 range. I looked at a couple other 256K range, and across all types, every variation and conversion software gives a different result. Apple 128K ![Apple 128K](https://live.staticflickr.com/65535/53849921066_7806ee41e2_o.png) Apple 128K VBR 10Hz filtered smart mp3 - same as above ![Apple 128K VBR 10Hz filtered smart](https://live.staticflickr.com/65535/53847978358_2cb16ffef0_o.png) Apple 128K VBR smart - lose the 10Hz filtering, slightly lower peak ![Apple 128K VBR smart](https://live.staticflickr.com/65535/53850358965_8e25dc7acd_o.png) RX11 128 - highest RX but lower clipped than RX VBR, true peak most similar to Apple 128K VBR smart ![RX 128](https://live.staticflickr.com/65535/53849921076_b2c8cf8608_o.png) RX11 115 VBR - barely lower but much higher possible clipped samples ![RX11 115 VBR](https://live.staticflickr.com/65535/53850182788_77478cd04f_o.png) RX11 115 VBR limit - this was about a 1 minute render on a 6:30 file. Lower limited levels all around. ![RX 115 VBR limit](https://live.staticflickr.com/65535/53849031882_bc018bfe38_o.png) XLD LAME 128 - highest true peak level, sample level, possible clipped count of any tried ![XLD LAME 128](https://live.staticflickr.com/65535/53850358960_ecc3f48ee7_o.png)
|
|
|
Post by gravesnumber9 on Jul 11, 2024 10:26:31 GMT -6
That is concerning you're getting that high of a jump when converting, I've never had anything over a 0.1 jump. Is RX11 handling your conversions? I have a few other rate/format converters if you need help investigating. Above is Apple's conversion from Music. VBR, highest, smart encoding adjustments, filter below 10Hz. I clarified in the OP. Some other variations, sticking with roughly 128 range. I looked at a couple other 256K range, and across all types, every variation and conversion software gives a different result. Apple 128K ![Apple 128K](https://live.staticflickr.com/65535/53849921066_7806ee41e2_o.png) Apple 128K VBR 10Hz filtered smart mp3 - same as above ![Apple 128K VBR 10Hz filtered smart](https://live.staticflickr.com/65535/53847978358_2cb16ffef0_o.png) Apple 128K VBR smart - lose the 10Hz filtering, slightly lower peak ![Apple 128K VBR smart](https://live.staticflickr.com/65535/53850358965_8e25dc7acd_o.png) RX11 128 - highest RX but lower clipped than RX VBR, true peak most similar to Apple 128K VBR smart ![RX 128](https://live.staticflickr.com/65535/53849921076_b2c8cf8608_o.png) RX11 115 VBR - barely lower but much higher possible clipped samples ![RX11 115 VBR](https://live.staticflickr.com/65535/53850182788_77478cd04f_o.png) RX11 115 VBR limit - this was about a 1 minute render on a 6:30 file. Lower limited levels all around. ![RX 115 VBR limit](https://live.staticflickr.com/65535/53849031882_bc018bfe38_o.png) XLD LAME 128 - highest true peak level, sample level, possible clipped count of any tried ![XLD LAME 128](https://live.staticflickr.com/65535/53850358960_ecc3f48ee7_o.png) Interesting. So what's the solution?
|
|
|
Post by nudwig on Jul 11, 2024 10:38:09 GMT -6
Thanks Doug and Dan for all the info, I'm going to go dig through and experiment with more codecs on the hottest masters I've received as well as some blown out audio I was sent to master (as a side note it's great to see Thimeo mentioned in here as that's a lifesaver when mixers insist on sending sausages and you need to make some space and bring some life back to it).
|
|
|
Post by bossanova on Jul 11, 2024 14:16:49 GMT -6
I use regular peak limiting and I check true peak on output. If I'm crossing TP -1 on final output, I lower the volume prior to hitting the master bus or make adjustments elsewhere in the chain. It seems like a really straightforward fix, but maybe it's because I've spent years thinking about stuff like this instead of finishing albums 😛
|
|
karlo
Full Member
![*](//storage.proboards.com/forum/images/stars/star_green.png)
Posts: 27
|
Post by karlo on Jul 11, 2024 17:02:41 GMT -6
I love crushing the bejeezuz out of a mix. Multiband too. 8 to 7 lufz all day. If the playing is dynamic you’ll still hear the dynamics in the timbre without the silly volume fluctuations - forcing the listener to have to turn it up to hear the soft parts - lame.
|
|
|
Post by Pueblo Audio on Jul 11, 2024 21:00:28 GMT -6
I’ll offer some perspectives for consideration.
If you are publishing files with levels over -15dBFS(rms) then you are over the line. You have already accepted distortion. Any kind of limiter you apply to it will just exchange one type of distortion for another. If the levels are anywhere near “commercial” level then the distortion will be unavoidably significant.
Being concerned about the potential distortion from that last dB of “true” peak measurement (which are in fact unknown, predicted peaks) is myopic. Kind of like telling someone their fly is open in the midst of a nudest convention.
Coming from decades of classical acoustic recording I appreciate the rejection of such distortion. Clean is peace and love.
But I have also spent the last few decades at a chart topping mastering studio. What can I say? Show me the records on the charts that remain below -1dB true peak. Even just one? If you’re swimming with sharks you gotta know how to handle yourself. Craft that distortion for best possible outcome.
Fixating on that last dB may help some, but it’s not the real, tragic flaw.
|
|
|
Post by robo on Jul 12, 2024 8:28:40 GMT -6
Interesting topic. I remember when folks started talking about leaving -0.3 db headroom on the master, then -0.5, and recently -1. Now some are doing more?
I can’t say I’ve ever heard the problem demonstrated, or circled back to test my work post publishing beyond casual streaming. Does this manifest as harshness? Cloudiness? Dullness? What is the aesthetic problem we’re avoiding, since as Pueblo pointed out there’s distortion happening regardless?
|
|
|
Post by Dan on Jul 12, 2024 9:52:51 GMT -6
Interesting topic. I remember when folks started talking about leaving -0.3 db headroom on the master, then -0.5, and recently -1. Now some are doing more? I can’t say I’ve ever heard the problem demonstrated, or circled back to test my work post publishing beyond casual streaming. Does this manifest as harshness? Cloudiness? Dullness? What is the aesthetic problem we’re avoiding, since as Pueblo pointed out there’s distortion happening regardless? You are avoiding digital clipping from the DA chip's processing and subsequent analog clipping. You can clip both in a converter. So infinite harmonics of infinite harmonics and the peak level gets louder from the Gibbs phenomenon. When this is on tonal content like many modern slammed productions, it sounds awful on many systems, especially cheaper ones or pro ones not designed to accomodate butchered signals where the user cannot simply digitally lower the volume.
|
|