^ That sounds like a fine idea. I'll do that now...
edit: hmm, I enabled PMs, and I'm clicking your name and then the link that says "send message," but nothing's happening. Maybe just try to send me a link and I'll let you know if/when I get it?
I believe the stated playrate and tempo resolution restriction is a myth based on that the display of them is rounded in the envelope lanes. I wouldn’t know how to test, but seeing that they are stored with 8 decimal places makes me think it would be weird to not use them all. If reaper doesn’t, then I’d think it’s a bug.
I rather think what makes your (super interesting btw) playrate experiment fail is resolution in the time domain, because if the changes are not 100% in sync with what the material does at all times it will introduce new fluttering at the spots where the playrate changes. Maybe you get more pleasing results if you don’t try to correct pitch fluctuations?
I rather think what makes your (super interesting btw) playrate experiment fail is resolution in the time domain, because if the changes are not 100% in sync with what the material does at all times it will introduce new fluttering at the spots where the playrate changes. Maybe you get more pleasing results if you don’t try to correct pitch fluctuations?
I couldn't agree more, TBH, and having heard some of the material, I don't think pitch/tempo fluctuation are even an issue. Pitch is an issue (or would be for me, because I would want to be able to add a VSTi or two), tempo is an issue (or would be for me, because I would want to be able to edit and use synchronised time modulation), but fluctuations are, at worst, a very minor factor here IMHO.
I believe the stated playrate and tempo resolution restriction is a myth if based on the display of them being rounded in the envelope lanes. I wouldn’t know how to test, but seeing that they are stored with 8 decimal places makes me think it would be weird to not use them all. If reaper doesn’t, then I’d think it’s a bug.
Could be the case. I also wondered why the data in file was not rounded, and also why the overall play rate can be adjusted to a finer level...
Oh, also, it's not just the value in the "envelope lanes" that shows only 2 decimals: If you try to set the value of a single point, it only allows 2 decimal places. Try putting in more and it will round the value. Maybe that's just for display purposes?? Who knows. I'd doubt it, you don't see that in other form fields...
Quote:
Originally Posted by gofer
I rather think what makes your (super interesting btw) playrate experiment fail is resolution in the time domain...
Yes, this crossed my mind after-the-fact. This is really hard to brain-wrap. The thing is, the playrate and the 'time domain' in this situation are or should be inextricably intertwined.
I've copied the tempo map timeline point values (the 'location' values) and used those as the playrate envelope 'location' values as-is. So those shouldn't change, right?.
And then, if you do what seems like pretty straight forward math - basically adjust the playrate of every item so that instead of the grid being adjusted the items get adjusted, to fit a constant tempo grid - well, it should work. And when I say "work" here I mean the down beat of each measure or whatever resolution you use for your tempo map/splitting should fall on the grid. And as I mentioned up there somewhere, my points were drifting off the grid... Whether the pitch miraculously gets fixed is sort of a separate question...
Maybe I'm still missing something. I could still be confused about what happens when I adjust the playrate of an item (I'll have to go check that when I finish this): Does the same segment of audio play, only faster - which means the item shrinks, or does the same segment of the timeline - i.e. the length of the 'item', play, but now more audio gets fitted within the item?
My experiment assumed the former, but now I'm thinking it could be the latter... If it's the latter then it wouldn't work. The items themselves have to shrink when the playrate speeds up (or lengthen when slowed down)...
Quote:
Originally Posted by gofer
, because if the changes are not 100% in sync with what the material does at all times it will introduce new fluttering at the spots where the playrate changes. Maybe you get more pleasing results if you don’t try to correct pitch fluctuations?
The original idea is that the tape speed fluctuated and that's what needed correction, or at least testing. And changing the playrate is supposed to compensate, which was supposed to also correct any pitch fluctuations that were due to speed fluctuations. So no active pitch correction should be needed. But I tried both. Both sounded bad, though in different ways.
I have to go check whether playrate change on the playrate envelope changes the length of items or changes the amount of audio played within an item...
^ OK, another duh moment: If I change the playrate of an item, say play it faster, it does not change the length of the item, it simply increases the amount of audio that can be played within that item. So regardless of decimal rounding or whatever other alleged shortcoming there might be, the method I tried simply wouldn't work - because audio items are handled differently than I thought they were...
This is exactly the case I was describing earlier - of me erroneously thinking of 'items' as discrete pieces of audio.
edit: OK, false alarm. If I change the playrate in the item properties window, yes, the item remains the same length and more audio (faster rate) or less audio (slower) will play within the same-length item space. But making a change to the playrate on the envelope does NOT do the same thing. If I change the rate with envelope points it simply slows down or speeds up the discrete section of audio that was within the original item length.
But - and here's another conceptual lapse on my part - everything speeds-up, even the click track. So when I change the playrate via envelope points, the audio plays faster and the click track plays faster, i.e. the grid is being re-sized at the same time, so nothing ever 'falls off the grid'. The item stays the same length in 'Reaper-world', even though the item is playing faster and less time elapses in the real world...
I also see that "preserve pitch with playrate change..." box in item properties has no impact when changing playrate via playrate envelope...
Thanks Fex. You've done some interesting things with this. I sound like a real man, now!
I like that reverb-ish Fct you put on the vox, it fits rather nicely...
Looks like you used 'Melodyne'. Not something I've ever tried, but I keep running into it these days (I think I saw it cost like $800 or something like that, a little rich for me)...
I was thinking more about what I wrote maybe yesterday, about there being a 'fine line' between preserving the original vs. improving the 'audio engineering', and I realized that in reality it's less of a philosophical/aethetic question for me and more about feasibility: I have so much to learn that I simply can't do it all. I have to pick my battles. I'm already so sick of this song that there's no way I'll be able to actually 'mix' it when I'm done with all the grunt work, not without a good, long vacation from it.
I like that reverb-ish Fct you put on the vox, it fits rather nicely...
It's not something I put a great deal of effort into - I just pulled some FX chains from stock and auditioned them until I found one which sounded nice. I thought you might be interested to hear that.
That particular chain comprises Exciter, Stereo enhancer, EQ, Reverb and Delay. If I was mixing properly, I would perhaps want to tweak some or all of that stuff. Or maybe not. I'm lazy, and not great at mixing anyway.
Quote:
Originally Posted by eq1
Looks like you used 'Melodyne'. Not something I've ever tried, but I keep running into it these days (I think I saw it cost like $800 or something like that, a little rich for me)...
Melodyne 5 studio is €699, but I don't have access to it any more, or I'd have tweaked some formants - that one note I had to pitch down a semitone (after "fixing" the pitch centre) sounds a little off to me (not quite as perfect as I first thought). A more studious approach would have been to undo the pitch changes to that one note, and "correct" the pitch centre to <100%. A balance is to be found which improves the pitch without mangling the sound.
I used Melodyne 5 Essential, which is €99, but was much cheaper to me because I bought the license second hand. I'm cheap. I'll work for pizza.
Melodyne 5 assistant is €249, Melodyne 5 editor is €399. You pays your money and you takes your choice.
Edit:
In order to bring the entire project to concert pitch, I first wanted as little distraction as possible, and the vocal, while not at all bad, was distracting me; hence the pitch correction. I don't do that sort of thing as a matter of course. Having done it, I knew the song would better survive the vocal being routed through a BF delay, so started playing with the FX chains for shitz & giggles.
The first word in the snippet is has a horrible stray frequency, BTW. I should perhaps have done something about that.
Last edited by Fex; 08-09-2020 at 12:56 PM.
Reason: Afterthunk
I believe the stated playrate and tempo resolution restriction is a myth based on that the display of them is rounded in the envelope lanes. I wouldn’t know how to test, but seeing that they are stored with 8 decimal places makes me think it would be weird to not use them all. If reaper doesn’t, then I’d think it’s a bug.
So, I did a little test, one test, that goes like this:
-I started with a selection of one measure of audio, rendered it, and established a 'baseline' audio clip with known length
-2.796 sec based on Reaper reported length value, 2.7961678 based on total samples divided by 44.1k.
-I inserted two playrate envelope points at the ends of the selection and set their values to 1.025. Interestingly, the value field and the envelope labels rounded to 1.02. I tried inserting 1.029 and it rounded to 1.03...
-I then rendered the selection with playrate envelope active.
-If Reaper is rounding to only two decimal places, then the length of the rendered selection should be 2.7961678/1.03 (or 1.02 if it's really truncating or rounding down for some odd reason).
-At 1.02 playrate the length should be 2.741341; at 1.03 it should be 2.714726.
-If Reaper is using the actual playrate value typed in, 1.025, then the length of the rendered selection should be 2.7279686...
The result was 2.727959.
It looks like Reaper is NOT rounding for playrate envelope; it looks like Reaper is using the actual value typed into the 'set point value' field, at least to the 3 decimal places I tried.
And now I've tried 4 decimal places, typed in a playrate of 1.0254. I calculate the length should be 2.7269044, and the length that results is 2.7270068. Not exactly the same, but the rendered audio is obviously shorter than it was for a playrate of 1.025, so it also looks like reaper is doing playrate envelope to at least 4 decimal places...
It's interesting: Obviously there's a major difference between something played back at 1.025 speed vs. 1.0. Clear pitch difference, and timing, of course. Between 1.025 vs. 1.0254? I can't tell any difference in pitch - but with both the 1.025 and the 1.0254 playing at same time there's an obvious phase difference, like a heavy 'flanger' effect.
I was wondering whether something else might be in play during this process, that might make the rendered audio length not exactly what you'd calculate. I rendered a 1.0255 playrate clip and I'm surprised I hear such a strong phase difference between the 1.0254 and 1.0255 - that's only 1/10 of a millisecond difference... Then I made a 1.02551 clip. I can't tell the difference between that and 1.02550, though Reaper is clearly using that extra decimal value - the clip is shorter.
Numbers are dangerous, y'know? It's only a matter of time before you realise that reality as you know it is only a simulation, and then where will you be?
If you can reduce anything you can't account for to a "reasonable question" (I don't know what that means, but I suggest that brevity might be a factor, and try to avoid overuse of the "R" word), you could try this website, which has left no reasonable question unanswered since 2009....
If someone feels like explaining the difference between the two following procedures, I'd be reading it attentively...
Stretch markers vs. changing item playrate
Say I have a fixed 86 BPM grid and an audio item. The audio item has a mean tempo of 85.5 BPM and I'd like to boost that to 86. What's the difference, in terms of actual processing going on, between placing a stretch marker at the beginning and end of the item and shrinking it, versus calculating the appropriate, faster play rate and changing that in 'item properties'?
Both seem to achieve the same thing, the same outcome, though I'm not sure if the actual processing is the same... I'm not sure why/when I'd use one or the other...
I don't know anything much about processing.... I think it involves maths. Shudder.
Stretch markers allow you to stretch parts of the audio within an item, while leaving other parts of the item intact. To achieve a similar result with item playrate, you'd obviously have to first of all split the item, which itself changes the audio (simply placing a stretch marker, before you actually stretch anything, does not).
Splitting an audio item would cause a popping noise at the split point (unless you were to split at the nearest zero crossing) - in Reaper, however, this is handled (by default) by placing a very short fade either side of the split. The fades are usually too brief to be audible.
For your purposes, I reckon you'll get away with splitting the audio if necessary. I think it would be easier for you than messing about with stretch markers.
If you just want to stretch the whole item en masse, just Alt & click and drag the edge where you want it. Unless it sounds bad, there's really no reason to take a more complex approach.
I don't know anything much about processing.... I think it involves maths. Shudder.
I get the impression you're not a maths fan...
Quote:
Originally Posted by Fex
Stretch markers allow you to stretch parts of the audio within an item, while leaving other parts of the item intact. To achieve a similar result with item playrate, you'd obviously have to first of all split the item, which itself changes the audio (simply placing a stretch marker, before you actually stretch anything, does not).
Yes, I think I'm understanding all this. It seems like using stretch markers could be a convenience factor if you have to do some tweaking in spots, or perhaps multiple tweaks across an item, split audio item, etc...
Quote:
Originally Posted by Fex
Splitting an audio item would cause a popping noise at the split point (unless you were to split at the nearest zero crossing) - in Reaper, however, this is handled (by default) by placing a very short fade either side of the split. The fades are usually too brief to be audible.
Wasn't aware that splitting in itself causes 'popping'. I've never noticed that...
Quote:
Originally Posted by Fex
If you just want to stretch the whole item en masse, just Alt & click and drag the edge where you want it. Unless it sounds bad, there's really no reason to take a more complex approach.
Now that's interesting. Makes my question about the difference between stretch markers and item playrate changes even more ponderous. If it's that easy to shrink or stretch an item, changing playrate in the process, I can't see why one would ever use stretch markers in such a case (one item, fixed grid, need to shrink to increase playrate so audio item fits grid).
Still, I'd like to understand what if any processing differences there are...
If the root cause is transport malfunction causing the speed to ramp up and down, fix that before touching anything else.
1st choice:
Do maintenance/repairs or replace the deck. Significant return vs the time to correct this after the fact.
last choice:
Repair the speed fluctuation digitally. The good news is linked veri-speed time/pitch adjustment is lossless in Reaper with its Elastique Pro algorithm.
If the speed ramps up and down, you can use stretch markers to match the ramping. This will be some labor of love to stare the whole thing down and adjust back to correct consistent speed/pitch. You can probably use your ear as a guide and follow pitch. Or if the original did in fact use a click, you could do it visually by that. When it's back in time, the pitch will follow.
Again, make sure you use linked pitch and time to veri-speed adjust! The DJ tricks with altering pitch or time separately would mutilate the recording. Along the same lines, don't even think of using quantization or tempo mapping and the like. The issue is the transport malfunction causing the speed to vary. Those kind of digital "DJ tricks" would just mutilate it further. A proper veri-speed correction would genuinely restore the recording with no artifacts or quality reduction.
Once you have the audio corrected, then start production work.
...If you'd prefer an informed opinion (ie, the facts), I refer you to my earlier post, #51.
Thanks. It may come to that...
Quote:
Originally Posted by serr
If the root cause is transport malfunction causing the speed to ramp up and down, fix that before touching anything else.
1st choice: Do maintenance/repairs or replace the deck. Significant return vs the time to correct this after the fact.
Overall the digitization doesn't sound overtly bad in terms of speed/pitch (to my ears), I think Fex agrees with that for the section of the song he listened to. There's various pitch-related problems, not related to the speed fluctuations, such as guitar a bit out of tune with everything else, and everything else not being in 'concert pitch' (because the synthesizer was detuned, too). But it wasn't until I did a careful tempo map that I saw how tempo fluctuated, from a low of about 83 BPM to a high of about 88 BPM (there's a scatter plot of the fluctuations a dozen or so posts up). Maybe in a couple places of the song the tempo fluctuation is actually noticeable, to me...
The point is, this is mostly academic, educational, for me. I mean, this song is going no where when it's 'done'. It's not winning any grammys! But I would like to make decent transfers of my 4-track tape stuff, and if there were modest-level of difficulty, DIY things I could do to the deck to improve those transfers I would likely find time to do them. I've looked around a bit and at the manual, and I haven't seen anything I can do to it to make it as good as it can be...
It's a Yamaha MT120, BTW, from around 1990. If you know of any good sources that touch on optimizing tape decks I'd be grateful if you shared them... I clean the head and pinch rollers, that's about it. I wouldn't buy another machine, the quality of the transfers thus far aren't bad enough to make that outlay worth it to me, at this point at least (it seems like I get deeper and deeper with every hour I put into the project, though; this is probably the 3rd or 4th time I've made 'transfers' of 4-track stuff, each time realizing it should/could have been better)...
Quote:
Originally Posted by serr
last choice:
Repair the speed fluctuation digitally...
Most of what you write here has been my thinking on the issue. But I'm confused when you say "don't even think of using quantization or tempo mapping and the like."
For example, you write: "if the original did in fact use a click, you could do it visually by that."
Isn't that a form of tempo mapping or quantization - using the 'click' visually to find beats, match beats to grid, or vice versa? And then I assume stretching or shrinking between the beats to correct timing, where pitch will (should) follow?
I've done this a few times now and it hasn't worked, just makes it worse. I'm not sure why. I think that's been the whole premise of this thread - that if I could correct speed fluctuations pitch would naturally follow. But thus far, no good. I started out thinking there must be a way this could be automated, that it should be automated. But it's not looking like that's the case...
Quote:
Originally Posted by serr
Once you have the audio corrected, then start production work.
Ha, yeah, I think that's one of the biggest takeaways from this work. All this 'correction' stuff should be one of the first things completed. But as I mentioned initially, I'd done a bunch of stuff already and have/had been hoping I could make corrections and 'plug' audio back into the projects with, say, envelope tweaks and stuff like that. I still might be able to do something without re-doing everything, but most likely I'll simply be recreating 'stuff' on top of new, somewhat corrected audio source material...
Overall the digitization doesn't sound overtly bad in terms of speed/pitch (to my ears), I think Fex agrees with that for the section of the song he listened to.
Fex absolutely agrees! It's not exactly Abbey Road DDD, but there's nothing wrong with it that a cheap fix won't make worse.
Quote:
Originally Posted by eq1
But it wasn't until I did a careful tempo map that I saw how tempo fluctuated, from a low of about 83 BPM to a high of about 88 BPM (there's a scatter plot of the fluctuations a dozen or so posts up).
You know why blind people get work as piano tuners, right? What the audio looks like is less important than what the food sounds like.
BTW, I'm not sure if the project you let me look at was more in the way of experimentation than a serious attempt to mix, but FWIW, I'd advise you to be a little more conservative in your EQ boosts. It is better to cut than to boost.
^ Those were 'serious' eq curves at one time, though applied to source audio at a different level, so they won't quite work in that project. But, I thought they were in the ballpark... If I wanted the shape to stay the same, is there a 'sonic' advantage to dropping the gain at each band equally, such as for the 'Korg' EQ curve, so that there's more cutting than boosting, rather than just using the EQ overall level fader?
* * *
OK, I have to report major major break through here. I can't believe it's taken me so long to understand what, in the end, is SO simple! I'm sure this is at least close to what y'all have been telling me to do all along... I guess some of the initial work takes some time simply to sort out what the problems are, tempo and pitch-wise, and to figure out what the correct tempo is. But once you figure that out, or if you know what it is -- simple simple simple.
All I had to do, or did do, was this:
Resize full song/audio item to fit length of song at the target tempo.
The actual mean tempo was about 85.7 and it should have been 86 BPM. So, set BPM to 86, grid resizes, alt + click and drag media item right edge to snap to where the song's supposed to end at a 86 BPM tempo. Without "preserve pitch..." checked, the media item resizes and pitch adjusts accordingly via item playrate. My item playrate for instance ends up at about 1.002.
At this point the song length fits an 86 BPM tempo, but some of it drifts a little bit in the middle. Now I need to adjust the middle, shrink or stretch the audio at points in the song so the beat falls on the grid.
The idea I started with was to adjust in some manner, maybe stretch markers, maybe splitting and resizing items - I thought every 16 bars might do it. But once I got started I realized that it could or should be done sequentially, one adjustment at a time - using stretch markers. SOOO simple...
I placed a stretch marker at the beat that was supposed to be at bar 17 and had to shift that point to the grid (left) just a little (maybe 20 ms?), so everything to the left of the marker gets squeezed a bit, and everything to the right gets stretched a bit.
I played audio with metronome ON over the next 16 bars, and by bar 33 it was off a little bit, too. So I placed a stretch marker at that beat and did the same thing - shifted a bit to the left.
I repeated this process, moving through the song, and in the end I only had to place 4 stretch markers to make timing fit the fixed tempo grid.
That is it! Everything sounds fine, I can't even tell the pitch differences, not in any specific way. Overall things sound better, I think, a bit tighter, more normal...
So now I have 'tempo-pitch corrected' items and can render those as new audio files to be used as source audio for the 'production' stage...
Those were 'serious' eq curves at one time, though applied to source audio at a different level, so they won't quite work in that project. But, I thought they were in the ballpark... If I wanted the shape to stay the same, is there a 'sonic' advantage to dropping the gain at each band equally, such as for the 'Korg' EQ curve, so that there's more cutting than boosting, rather than just using the EQ overall level fader?
"Dropping the gain at each band equally" doesn't mean much without defining the number of bands, and the frequency and bandwidth of each.
A gain control is just a gain control.
The EQ on the Korg track isn't really illustrative of the principle - you have High pass and Low pass filters (the track's carrying the bass, so I'd have gone with a Low shelf rather than a High pass, but whatever) and the bandwidth of the High pass on the right channel is such that it creates a boost at around 56Hz, which I think is fine (or would be, except that it's a little unconventional to have that boost only in the right channel, which is equivalent to panning the bass).
The problem with boosts is clearer in the higher frequencies - on the vocal, particularly, the boost serves to make the track more silibant, and the silibance gets a bit wild when you start messing around with FX (as I did).
The rules of thumb are, rather than boost any frequency, see if you can achieve the same effect by cutting nearby frequencies either side. This produces fewer audible artifacts, making for a much smoother sound.
Artifacts might not be so significant on a bass synth. What's it going to do, after all - make the bass sound synthetic? It's a synth!
A case in point:
No eq boosts allowed, if legend is to be believed. How smooth is that sax? How smooth are those cymbals?
Quote:
Originally Posted by eq1
OK, I have to report major major break through here. I can't believe it's taken me so long to understand what, in the end, is SO simple! I'm sure this is at least close to what y'all have been telling me to do all along...
YES! Hexagram 4 applies.
It has taken you an age to realise this. On the plus side, I think you might be a Time Lord now.
"Dropping the gain at each band equally" doesn't mean much without defining the number of bands, and the frequency and bandwidth of each. The EQ on the Korg track isn't really illustrative of the principle - you have High pass and Low pass filters (the track's carrying the bass, so I'd have gone with a Low shelf rather than a High pass, but whatever)...
Was referencing the EQ on stereo receive track of the left and right Korg tracks (see bottom of post): If I dropped each band equally, there'd be more cutting than boosting, right?
The EQ on the left and right channels aren't used, I don't know what I was doing with those. There's only a low shelf on the stereo receive track, no high pass, for instance...
I'm aware of the 'cut-first' idea. I always do that. I generally try to cut, boost if I think I have to. Though I can't say I've ever noticed a difference if I don't think about any of that and just do 'whatever'. At least, working with digital equipment. Seems like I used to have issues working with analog in the old days...
But this idea, that if I cut each of the frequency bands in the stereo EQ curve for the Korg receive track, in project I sent you - that doesn't make a difference, right? The curves would be cutting more, but then I'd just have to raise the output fader more. Seems like you might be saying there's something fundamentally wrong with the shape of the curve??
Here's image, I don't see anything wrong with this, but maybe you know better. This is the same as what's in project I sent to you, but I cut each band by 2db:
Quote:
Originally Posted by Fex
How smooth is that sax? How smooth are those cymbals?
Eiks -- Sade's voice in this medium does not sound like I remembered it. Sounds like she's got a sock in her throat, much more masculine than my recollection. I have this tape laying around, somewhere... Was in to Sade...
I don't have anything looking at all like that in the project you sent!
Ah, I see, I took that receive track out, to simplify... Sorry 'bout that.
Quote:
Originally Posted by Fex
On the vocal track, there's a 9.3dB boost at 9340.9Hz, with a bandwidth of 0.97. That seems a bit extreme. Just saying.
hmm, well, I don't think there's much useful data above that point, coming from a 4-track tape, so I don't see it as a narrow boost; rather, it's more of a 'gradual' boost starting around 4k and ending at the point beyond which it's just fuzz... I don't think there's a problem with that, I haven't noticed any, but I could be wrong...
Quote:
Originally Posted by Fex
I'm not an expert on EQ, by any means, and this might, perhaps, be a fascinating subject for a whole nother thread.
Probably. A little tangent is OK, but I can't think about timing/pitch AND EQ at the same time. Oh, actually, I'm done with timing/pitch; I'm moving on to part editing, trying to move stuff around a bit, while keeping things aligned.
hmm, well, I don't think there's much useful data above that point, coming from a 4-track tape
In which case there's little advantage in boosting, is there?
Quote:
Originally Posted by eq1
so I don't see it as a narrow boost; rather, it's more of a 'gradual' boost starting around 4k and ending at the point beyond which it's just fuzz... I don't think there's a problem with that, I haven't noticed any, but I could be wrong...
I was listening with headphones which are a little bright. That was a choice - I tend to mix a little bright, and I need to break that habit. They are rather good headphones, though....
I didn't notice any problem with the vocal until I applied that FX chain, at which point the main body of the vocal sounds really rather good, but there's a hard, sharp harmonic on the word "Do." I probably should have switched your EQ off first. Whatever else, I don't think that boost helps particularly in any way. YMMV.
EDIT: What I'm really saying is, for the purpose of preserving the integrity of the original tape, you should perhaps take a very conservative approach to EQ. It's like salt - easy to add, but a bugger to remove.
OTOH, when it comes to radically remixing the thing, knock yourself out.
GAH.... listened again, and I should retract bits of that, and qualify other bits....
That boost isn't causing the problem. I don't think it's doing anything very much. The frequency I don't like is substantially lower down, and would benefit from the judicious application of a Notch filter.
It is unfortunate, as well, that the first "Do" in the clip also coincides with a similarly pitched blip from the percussion, which exacerbates the effect. I had totally failed to notice that.
*Slaps self*
But anyway, my point stands - EQ - less is more. The clue's in your name.
Yeah, I know what you're talking about, with the harsh "Do...", etc. There's a few of those around the song. And sibilance. I'll have to take a listen around at some point, re-evaluate. With the "Do..." harshness, though, I think I was just leaving that there. I'll have to re-evaluate that, too.
Re. Boosting up there, there'd be no point in boosting above that point, but there must've been a reason I was boosting up to that point. I was probably trying to add some detail to the vox - maybe there isn't any there to boost, to begin with, or enough to warrant what might be a trade-off with artifacts. Not sure. I'll be going over it all at some time in the future...
Well, when you do, bear in mind that Professor Fex says that if you do need to bring out the highs in the vox, it would be better achieved with a cut or two below the target frequency rather than boosting very much up there.
A case in point:
[Sade youtube vid]
No eq boosts allowed, if legend is to be believed. How smooth is that sax? How smooth are those cymbals?
Quote:
Originally Posted by eq1
Eiks -- Sade's voice in this medium does not sound like I remembered it...
Here's a funny thing: You post that Sade vid, and it just dawned on me -- this song I'm working on, that you've heard the 2nd verse of? -- the first line is a spin on a couple lines from a Sade song, "Fear" off their second.
The first line of my song is, "Blue... Is the colour that I feel inside."
The first line of that Sade song is, "Blue... Is the colour of the red sky..." The 3rd is "Blue... Is the colour that she feels inside..." (I actually had to look this stuff up recently, before that I had just vaguely remembered that I got that 1st line from Sade). The melody of that line is also more or less a knock-off, not just the lyric.
I wonder if subconsciously you tapped-in to that 'vibe' when you posted that Sade vid as an illustration of a totally unrelated concept... I mean, there's a lot of examples you could have posted, right?...
As I recall I never really liked the way that song ("Fear") went from that first line/verse into this sort of 'matador/Spanish' thing with screechy saxes. But I liked the vibe of the first verse... It seems strange that I would have liked Sade enough in my youth to have purchased like 3 or 4 of her/their albums at the time. I hardly bought anything, not really 'my style'...
I wonder if subconsciously you tapped-in to that 'vibe' when you posted that Sade vid as an illustration of a totally unrelated concept... I mean, there's a lot of examples you could have posted, right?...
Well, as you were explaining that, it did seem somewhat familiar.... but, no, I know why I chose that example, and that wasn't it....
I had the pleasure of an account of the Sade sessions from someone who was there. His recollections of the experience largely revolve around EQ cuts and cocaine. Nuff said.
Overall the digitization doesn't sound overtly bad in terms of speed/pitch (to my ears), I think Fex agrees with that for the section of the song he listened to. There's various pitch-related problems, not related to the speed fluctuations, such as guitar a bit out of tune with everything else, and everything else not being in 'concert pitch' (because the synthesizer was detuned, too). But it wasn't until I did a careful tempo map that I saw how tempo fluctuated, from a low of about 83 BPM to a high of about 88 BPM (there's a scatter plot of the fluctuations a dozen or so posts up). Maybe in a couple places of the song the tempo fluctuation is actually noticeable, to me...
I'd only repair anything that crossed the line. Either from pitch or a tempo artifact. Obviously the pitch and speed move together with tape. (Or if it wasn't obvious... As the tape slows down, the pitch lowers and vice verse. This is why you need linked speed/pitch veri-speed type adjustment to correct for this. Reaper has the lossless Elastique Pro. Lossless in veri-speed mode that is. In Reaper, select all the items you will be correcting. Open item properties. Untick the box for 'preserve pitch when altering tempo'. It's worded a little weird and yes, the default is separate pitch/time DJ tricks mode! If you don't change this first you will destroy all your work. Note all selected items together respond to the change.)
Quote:
Originally Posted by eq1
It's a Yamaha MT120, BTW, from around 1990. If you know of any good sources that touch on optimizing tape decks I'd be grateful if you shared them... I clean the head and pinch rollers, that's about it. I wouldn't buy another machine, the quality of the transfers thus far aren't bad enough to make that outlay worth it to me, at this point at least (it seems like I get deeper and deeper with every hour I put into the project, though; this is probably the 3rd or 4th time I've made 'transfers' of 4-track stuff, each time realizing it should/could have been better)...
Yep, clean everything. Adjust the speed to the tape. Adjust the head azimuth to the tape. (Quick and dirty: listen in headphones for most high end on all tracks.)
Quote:
Originally Posted by eq1
Most of what you write here has been my thinking on the issue. But I'm confused when you say "don't even think of using quantization or tempo mapping and the like."
Again, when tape slows down, the pitch lowers with it and vice verse. You need veri-speed correction. There are digital tools now that let you alter either the pitch or the speed separately. (Obviously this can never be lossless. It's a glaring alteration! There's also only so far you can go before hearing crazy artifacts.) If you were to accidentally use a "DJ trick" style separated speed or pitch move, you would seriously mutilate the recording beyond any hope of repair! The quantize features use the separated speed from pitch and alter only the pitch. This is a classic "DJ trick" and would just mutilate the recording.
Quote:
Originally Posted by eq1
For example, you write: "if the original did in fact use a click, you could do it visually by that."
Sure. Use the tempo for a visual telltale so you don't have to rely on your pitch perception. Might make it quick and easy. Since you know it was recorded to a click originally, when you hit the original speed with a veri-speed adjustment you should hear the pitch also fall into place.
Quote:
Originally Posted by eq1
Isn't that a form of tempo mapping or quantization - using the 'click' visually to find beats, match beats to grid, or vice versa? And then I assume stretching or shrinking between the beats to correct timing, where pitch will (should) follow?
The point is not to use anything that alters time alone. The quantization tools alter time separately. You can only use linked pitch/time "veri-speed" style adjustment to correct for analog tape transport fluctuation.
Quote:
Originally Posted by eq1
I've done this a few times now and it hasn't worked, just makes it worse. I'm not sure why. I think that's been the whole premise of this thread - that if I could correct speed fluctuations pitch would naturally follow. But thus far, no good. I started out thinking there must be a way this could be automated, that it should be automated. But it's not looking like that's the case...
Make sure you set for veri-speed mode in item properties and for every item you are adjusting. Also note that if you need to ramp the pitch/speed up/down over a section, stretch markers are the tool to let you work that. Again, you need to be in veri-speed mode with the item properties setting.
Quote:
Originally Posted by eq1
Ha, yeah, I think that's one of the biggest takeaways from this work. All this 'correction' stuff should be one of the first things completed. But as I mentioned initially, I'd done a bunch of stuff already and have/had been hoping I could make corrections and 'plug' audio back into the projects with, say, envelope tweaks and stuff like that. I still might be able to do something without re-doing everything, but most likely I'll simply be recreating 'stuff' on top of new, somewhat corrected audio source material...
100% has to be speed/pitch corrected first!
Save your previous work.
Start over and do the speed correction first.
Then cut/paste from your previous work to save anything still relevant. Reaper makes it slick to cut/paste between project tabs.
Using the "sonic" method maybe somebody might want to create an automatic tool for the task the OP describes, and with that replace high priced alternatives .
...Overall the digitization doesn't sound overtly bad in terms of speed/pitch...There's various pitch-related problems, not related to the tape speed fluctuations... But I would like to make decent transfers of my 4-track tape stuff, and if there were modest-level of difficulty, DIY things I could do to the deck to improve those transfers I would likely find time to do them. I've looked around a bit and at the manual, and I haven't seen anything I can do to it to make it as good as it can be...
It's a Yamaha MT120, BTW, from around 1990... I clean the head and pinch rollers, that's about it. I wouldn't buy another machine, the quality of the transfers thus far aren't bad enough to make that outlay worth it to me, at this point at least (it seems like I get deeper and deeper with every hour I put into the project, though; this is probably the 3rd or 4th time I've made 'transfers' of 4-track stuff, each time realizing it should/could have been better)...
I came back here to review how I ended up successfully 'tempo mapping and quantizing' my digitized tape audio - cuz I've forgotten most of it already - and I came across this passage above, realized I have some new info to add, you know, for all those people who are transferring 4-track tape material from a Yamaha MT120 recorder...
I basically just finished transferring pretty much all my 4-track stuff. In the process I dug-in a little into the recorder itself and found that there's some very simple things you can do to make the transfers much better. I'm kind of kicking myself for not doing it earlier.
First of all, I've attached a link to the service manual, which will help you get the thing open and show you which adjustments are possible. There's a host of things you can test and adjust, but I didn't have a scope or the proper test-tone tapes, so just kind of winged-it with a few things. Here's that link:
Here's three things that were easy and made/make a big difference:
1) I'm using the tape outs, was getting crackle, I cleaned the tape out RCA jacks with alcohol and a cotton swab and haven't heard crackle since.
2) Mess around with all the switches and knobs, make sure nothing unneeded is 'ON' or turned-up. I found that there was horrendous hum/noise with the meter selection switch set to '4-track' rather than 'stereo' -- so when digitizing I made sure that was set to 'stereo'.
3) ADJUST AZIMUTH for every digitized piece. There's a tiny screw on the head that can be accessed when a tape is playing. Turn that screw left or right, tiny amounts, to adjust the alignment of the tape head. Listen for the loudest, clearest high frequency signals, such as cymbals. Listen for phasing and try to get rid of it.
Unfortunately, I learned this late in the process - that this can make a big difference - AND it can change for each playback! I did back-to-back digitizations, dialed-in azimuth, recorded but got pops from some external source so had to do it again. The second recording was noticeably worse than the first. I had to re-adjust azimuth and do it a 3rd time...
Here's a 4th thing: In general, turn-up your headphone amp or whatever you're listening through so you can clearly hear all the noise sources - and try to get rid of them. I'm pretty sure that, in the end, you shouldn't hear anything but white noise - fuzz - with your headphones cranked...
That's pretty much it. I had tested the Dbx noise reduction circuits and those measured fine. I also messed with the playback level by track, but I probably just made things worse with that, you need a proper test tone for that... You can also adjust 'bias' by track. This stuff is in that service manual.
I opened-up the unit and tried to clean everything with a vacuum cleaner and can of compressed air, 99% alcohol swabbing in some places. Can't say what this accomplished. It wasn't terribly dirty, at least not the parts related to the tape-out circuits. The mixer section/faders, etc. were pretty filthy.
The internals are separated into a bottom and top half, and those halves are connected via 3 weird old-school 'ribbon' cables. It seems like you should be able to 'un-plug' these things and separate the halves when working, but they don't seem to 'un-plug' very easily. I'm not sure they're supposed to be. I tried to unplug one - the connection is actually tinned wires pressed into 'press-fit' 'friction-fit' 'interference-fit' type terminals, they're weird, hard to loosen and then don't seem to hold they're gripping power if you do loosen them... Be careful with those, probably best to try to work around them, leave them connected. Interestingly, one of the narrower cables has a wire that deals with tape speed, and it can easily come loose and tape speed will go all over the place.
Linear Mode
