Rough pop/rock vocals

[ThatSoundsGood]

Technically, it's done the other way around than how you describe it.

Digital clipping is something I'm sure all mastering engineers avoid. What is technically happening is that the limiter "cuts" the transient peaks and creates a larger headroom, and that larger headroom makes it possible to increase the overall level of the music track. Most mastering engineers increase the level until the highest peaks reach -0.1 to -0.3 under digital clipping.

I don't see anything wrong with you calling the limiting "clipping", as it is a form of controlled clipping which can be made to sound "softer" than the digital clipping that would otherwise occur if the transient peaks were hitting digital zero.

With really good mastering engineers, the limiter often doesn't cut the transients (or cuts them very slightly or rarely). They are turned down but the integrity of the transients is maintained. This again is why I have an issue with calling it clipping. The term "clipping" is generally used in Pro Audio to signify that the signal has gone past the headroom of the circuit, especially when a digital circuit is clipped. A good combination of compressing and limiting will keep the integrity of the transient more often than not. This is what the mastering engineers have gotten so good at. They can squash the master but maintain the transients and the dynamic perception of the music. There are so many ways that they accomplish this.
One interesting thing is that there was a time when many of them were using some Burl A/D converters and they would clip the input of the circuit and it sounded amazing. This was actual clipping of an analog circuit (which sounds a lot better than clipping anything digital).
Another interesting note is that recording to tape actually makes transients more round because tape has natural compression. This whole topic is so much more complicated than what the cult of hifi understands.

Edit-----Another way to describe the end result of mastering and what you're seeing in the waveform is this:
The transient is still there but everything else is turned up louder after it. So the result is that there is less magnitude between the transient and everything else so the waveform looks different. But since the master waveform is a combination of many instruments projecting different frequencies, it looks square because everything is being turned up to the same level as the transient. But the transients are usually still there. It's just the other information is now louder.

 

[ThatSoundsGood]

Thanks a lot for this explanation! I am of course aware of later versions of the Eventide Harmonizer but have never experienced such phasy, flanger-like robotic voice coming out of one. But it absolutely makes sense and is from production point not that far from my initial suspicion that 2 different microphones in close proximity to the singer were used and mixed. It creates a surprisingly similar effect, but not as ´robotic´.

I'm guessing that this effect is an eventide, but I don't know for sure. I've used the old units and that's what it sounds like to me, but I could be wrong. You can also accomplish this through other means (delays, doubling vocals, overloading a reverb chamber or something else), but it really just sounds like a poorly behaved or poorly balanced Harmonizer. Which is similar to what 2 microphones could sound like on a vocalist so your hearing is excellent.

 

[Arindal]

it really just sounds like a poorly behaved or poorly balanced Harmonizer. Which is similar to what 2 microphones could sound like on a vocalist so your hearing is excellent.

Thanks! Had some experience with this effect as I was working with a professor for recording engineering who liked this double-mic trick to create more natural, minimally blurred phantom localization impression in jazz recordings as if they were made with a stereo main mic arrangements, but without getting the usual amount of reverb on tape which the latter would have caught.

If delay and panning was not properly adjusted, it sounded as phasey and flangy as you describe it, but not as ´robotic´, if that makes sense. The robotic effect I mainly know from vocoders, but they are also a known side-effect of pitch-shifters and autotune, so to me it sounds very plausible that you are absolutely right regarding Harmonizer!

 

[ThatSoundsGood]

Thanks! Had some experience with this effect as I was working with a professor for recording engineering who liked this double-mic trick to create more natural, minimally blurred phantom localization impression in jazz recordings as if they were made with a stereo main mic arrangements, but without getting the usual amount of reverb on tape which the latter would have caught.

If delay and panning was not properly adjusted, it sounded as phasey and flangy as you describe it, but not as ´robotic´, if that makes sense. The robotic effect I mainly know from vocoders, but they are also a known side-effect of pitch-shifters and autotune, so to me it sounds very plausible that you are absolutely right regarding Harmonizer!

So, the Eventide is a pitch-shifter. In general, it adds 2 extra voices to the main one. So you're hearing 3 of them in total. You have the original, then one that is slightly tuned up (normally about 9 or 10 cents) and one that is tuned down (about 9 or 10 cents.
What I'm guessing here is that the harder she sings, the more you hear the pitch shifter trying to work and the robotic effect is basically artifacts of it struggling to tune it.
The Waves Doubler plug-in does this sort of thing in case you want to experiment.

 

[kyuu]

Where did you get this track to find out it was clipping? Can you tell me your process here? Is it from Spotify, Apple Music?

I know exactly where he got the track, and his process, because I actually read his posts. I'm finding it weird that you seem to be unable to figure this out?

What you didn't answer was my question about what you've mastered. I would love to hear it.

Come off it. He clearly has no interest in responding to this, and has no obligation to. This seems to be related to your belief that no one who hasn't done mastering themselves has any right to complain about mastering, which is ludicrous. We can listen to the difference between a highly compressed "remaster" and an original release that has much better dynamic range and figure out for ourselves which sounds better. The general consensus is that overly compressed music sucks. We don't all need to start mastering our own music to be allowed to make that determination.

 

[ThatSoundsGood]

I know exactly where he got the track, and his process, because I actually read his posts. I'm finding it weird that you seem to be unable to figure this out?

I read his posts and his methodology is clearly flawed, which is why I continue to question it.

Come off it. He clearly has no interest in responding to this, and has no obligation to. This seems to be related to your belief that no one who hasn't done mastering themselves has any right to complain about mastering, which is ludicrous. We can listen to the difference between a highly compressed "remaster" and an original release that has much better dynamic range and figure out for ourselves which sounds better. The general consensus is that overly compressed music sucks. We don't all need to start mastering our own music to be allowed to make that determination.

So I should concede to someone who has a flawed methodology and is criticizing the mastering engineer with completely flawed information? On Audio Science Review? Are you saying that we should give in to modern consensus even though it clearly doesn't understand the process? This is a dangerous sentiment and clearly not what this website is about.
I never said that someone who hasn't mastered a record can't complain about it. Complain all you want. But when someone doesn't understand what they're talking about, I think it's important for us to provide better science to the conversation in order for us all to learn. I've been schooled many times on here by people who have far more knowledge in their pinky finger than I do and I learned from it. But this topic happens to be something I am knowledgable about and I get concerned when I continue to see that lack of knowledge on the topic lead to misinformed sentiment. You can disagree with me here, but can you present actual information in regards to it? Otherwise, maybe you should "come off it."
Whether more compressed masters sound better is a subjective thing and could only be measured by blind listening tests so that argument is opinion and futile for this discussion unless we add some actual studies in regards to it. You say that the general consensus is that overly compressed music sucks, but can you actually provide any studies on that? Are people actually listening to the same song with different mastering in a blind A/B test? What systems are we listening on? This is a far more complicated topic than the "general consensus" is willing to discuss and admit. It just became trendy in Hifi to think this because so many people repeated it and the "loudness wars" became a talking point. I hear all kinds of things from the hifi community that are just pure fantasy and they're repeating what other people have said. Can these people even hear the difference between a highly compressed master and a less compressed master? Of course some can, but we have to admit that there's a lot of snake oil gear and snake oil belief out there. It's funny to me that people who don't understand how compression works can actually have such a strong opinion about it. I have personally challenged people with blind listening tests and they almost always pick more compression (to a certain point). But again, this is mostly subjective unless we do some extensive studies on it.

What we can measure here and what is relevant is whether these masters are clipping. He said that they are. They clearly aren't. He thinks the mastering engineers don't know what they're doing but he clearly has an error in his methodology somewhere, but now you're defending that error by saying that his opinion is warranted. That's a bit wild to me, but have at it. Sure, your opinion and his are warranted but the data says that you're wrong. So when someone has such a basic and clear error in their methodology, I jump to question their opinion since uneducated, misinformed opinions can be quite damaging.
If my opinion says that Wayne Gretzky is the best BASKETBALL player to ever live, I'm entitled to that opinion, but I'm still clearly wrong.
Your tagline in your signature even talks about misinformation............now I'm really confused.

 

[tengiz]

...and to wrap up my initial issue: it turns out that by minimizing the impact of the seats (I simply folded the backs down and placed a couple of acoustic panels on them), the ML Ethos produced raw, un-EQ’d measurements that required no correction above the transition frequency, with a smooth roll-off in the frequency range I can still hear . The upper limit of the transition frequency and the actual slope match the calculated values for my room.

As a result, to take advantage of Audyssey’s bass management without altering the frequency response above the bass crossover, the "L/R Bypass" setting works best in my setup. The perception of exaggerated processing on certain vocals is gone—now it sounds consistent with my other setups.

So, it seems Audyssey was compensating for the measurement mic being close to the backs of the seats, in addition to averaging over the measurement area. This correction is not only unnecessary above the transition frequency, but it actually worsens things for the ML Ethos in my room.

The seats do, of course, affect the listening experience—there’s a subtle but noticeable shift in imaging when I lean forward (away from the seat back) compared to sitting normally. The soundstage moves downward from slightly above ear level. But beyond that, I’m not sure I hear any other effects. Somehow, the ears and brain are able to compensate just fine.

I appreciate all the help, advice, and insights. Thank you!

 

[Arindal]

it seems Audyssey was compensating for the measurement mic being close to the backs of the seats, in addition to averaging over the measurement area. This correction is not only unnecessary above the transition frequency, but it actually worsens things for the ML Ethos in my room.

Automatic room correction systems are kind of notoriously failing when it comes to correcting an in-room response curve with speakers that are a bit non-standard in terms of directivity. Particularly large horns, planar speakers, dipoles, line sources - I have not seen a single system delivering an even acceptable result, including the full Trinov toolkit. Presumably that has to do with the balance between direct and indirect sound in certain frequency bands not delivering the ratio that such an algorithm would expect, and there obviously is no algorithm smart enough to distinguish this.

Had a funny incident with such a setup once doing an automated correction and with the first track played after the procedure, literally everyone in the room was shocked thinking all of the tweeters got blown. Fortunately that was only a very kinked shelf filter the system has applied for unknown reason.

 

[Chris A]

Automatic room correction system are kind of notoriously failing when it comes to correcting an in-room response curve with speakers that are a bit non-standard in terms of directivity. Particularly large horns, planar speakers, dipoles, line sources - I have not seen a single system delivering an even acceptable result, including the full Trinnov toolkit. Presumably that has to do with the balance between direct and indirect sound in certain frequency bands not delivering the ratio that such an algorithm would expect, and there obviously is no algorithm smart enough to distinguish this.

I strongly agree. I've never gotten any "automatic EQ" facility to work in my listening room.

My setup includes fully horn-loaded loudspeakers in a 5.1 array with directivity control down to about the calculated room Schroeder frequency (100 Hz) based on front-of-room RT measurements.

One of the most egregious aspects that I've found with these automatic EQ packages: they apparently all require you to place the measurement microphone too far away from the direct arrivals of the individual loudspeakers (e.g., 1m) to determine loudspeaker in-room EQ. (I'm not talking about setting the individual 5.1 channel gains or the time delays--which turns out to be a completely different affair from setting up EQ for each loudspeaker/subwoofer.)

What I've found is that these apps simply cannot distinguish minimum phase direct arrivals from in-room reflections when measuring at the listening positions.

In one test, I carefully dialed-in all loudspeakers individually, then did the channel gains and delays for the listening positions, then turned on Dirac. It immediately tried to suppress the amplitude and perceived phase growth response around the transition region (about 70 Hz to 350 Hz)--when it should have done no correction at all. It consistently tries to do this no matter how I run the measurement sequence or placement of microphone.

I've never been able to get any of these packages to work (the EQ portion) such that I could listen to the results for more than a few minutes.

Chris

 

[tengiz]

Automatic room correction system are kind of notoriously failing when it comes to correcting an in-room response curve with speakers that are a bit non-standard in terms of directivity. Particularly large horns, planar speakers, dipoles, line sources - I have not seen a single system delivering an even acceptable result, including the full Trinov toolkit.

The Audyssey documentation—though somewhat vague—does acknowledge that automatic correction may not always work properly. For this reason, even the “consumer” version of Audyssey includes both the “L/R Bypass” mode and a basic option to limit the frequency range over which correction is applied.

The “advanced” version, which requires purchasing a license for each receiver, among other useful things including a much better/proper workflow, offers more precise control over how these limitations are implemented.

So, not surprisingly—but something I had to learn the hard way—it’s just a tool, not a panacea .

 

[Keith_W]

If you are still experiencing harshness when you listen to music, I wouldn't rule out distortion just yet. Distortion is a function of playback volume, and if the volume of the measurement is not controlled, it may look better or worse than it actually is.

Given that you have ESL's, I suggest you do a compression test. This is how:

1. Place your mic 1m on axis to the speaker
2. Using an SPL meter with C-weighting, play pink noise. Adjust the volume output of the system until it registers 76dB. If you have a UMIK1/2 these are factory calibrated so adjust the volume until it is 76dB. If you don't have an SPL meter and don't have a UMIK1/2, place your microphone 1m from a smoke alarm and press the TEST button. Adjust REW until this reads 76dB. Note, the pink noise should be played for about 1 min to comply with IEC standards for long term compression testing.
3. Take a sweep from 20Hz - 20kHz of one speaker only. Label this sweep "76dB"
4. Increase the volume by 10dB using REW (NOT using the SPL meter!). Take another sweep. Label this sweep "86dB".
5. Increase the volume by another 10dB and take a sweep. Label this sweep "96dB".
6. Zip and post the MDAT to ASR. Or if you like, you can continue with the following steps:

Now that you have all your sweeps, we will process them to obtain a compression graph.

7. Under "All SPL", align all the measurements to 0dB at 1kHz with an alignment span of 3 octaves
8. Trace Arithmetic A-B and perform the following operations: 76dB-76dB, 86dB-76dB, and 96dB-76dB.
9. Examine each graph for distortion.

The reason we do this test is because ESL's are dynamically limited in the first place - they don't like high SPL's. As ESL's get older (as mentioned in the other thread), the membrane degrades or gets dusty, stators lose performance, etc. and nonlinearities creep in. These changes may not show up in a routine freq response measurement. Although I suggested taking a sweep of one speaker only, both speakers may degrade at different rates so it may be worth doing the same test for the other speaker.

One final note: in-room frequency response measurements of dipoles is unreliable, at best. I certainly would not use a full range measurement of a dipole taken in-room as the basis for DSP. I would compare the measurement I obtained with known reliable measurements by overlaying my measurement and the reference measurement in an image editing program like GIMP or Photoshop. Otherwise, I would take the speakers outside to obtain a free-field measurement as basis for DSP.

 

[tengiz]

If you are still experiencing harshness when you listen to music, I wouldn't rule out distortion just yet. Distortion is a function of playback volume, and if the volume of the measurement is not controlled, it may look better or worse than it actually is.
...
The reason we do this test is because ESL's are dynamically limited in the first place - they don't like high SPL's. As ESL's get older (as mentioned in the other thread), the membrane degrades or gets dusty, stators lose performance, etc. and nonlinearities creep in. These changes may not show up in a routine freq response measurement. Although I suggested taking a sweep of one speaker only, both speakers may degrade at different rates so it may be worth doing the same test for the other speaker.

Thank you.

Unfortunately, I’ve already parted with those speakers - they showed signs of other potential electronic issues, so no new measurements

I’d previously had an experience with my older panels (Vistas) collecting dust to the point of arcing, though it took many years for them to reach that condition. The sound of arcing is, of course, unmistakable once you’ve heard it. As expected, vacuuming and blowing air resolved the issue. Those older panels still work flawlessly today.

The problematic pair had relatively new panels - I bought the replacements from the manufacturer in 2021.

I normally don’t listen at high SPLs, and the harshness I noticed didn’t seem volume-dependent at all. I’d also assume that membrane degradation, if nonlinear, would affect the entire frequency range of the panels. But I didn’t hear any signs of nonlinearity outside of certain vocals with a fundamental frequency around 350 Hz. I mean, I think I’d recognize it if it were there. But I would not be shocked, if I am wrong.