Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Recording games with quieter music
#1
I've been making some rips from a stock VA6 Genesis, and in most cases, everything's groovy. With the console's volume slider at max, the noise floor of my recording setup hovers around -58db, and at that setting, the average game is even a little too loud, and I have to bring the console volume down a couple notches (which reduces the noise floor even more). Perfect.

However, a couple games I took a pass at today seem to have been mastered (for lack of a better term) to a WAY lower level. Cadash is a good example. Usually, I aim for the MAX peak level of the entire recording to be around -1db (for the occasional transient), but to get that level from Cadash, I need to turn the gain up on my Scarlett 2i2, and that brings the noise floor into unacceptable territory (like -53db... yuck). The noise is definitely coming from the console, rather than the hardware (bring the volume down on the Genesis, and there's pretty much no appreciable noise).

What to do here? I guess the dilemma is that the soundtrack, as programmed, isn't loud enough on the YM2612 to get a good recording level without bringing in excessive noise. Are games like this just unrecordable, for purposes of the project? Given that we want basically stock Model 1s, is there any way to bring the noise floor on the console down further? Or should we just deal with recordings that, while having an acceptable noise floor, are also recorded with the average signal hovering around -12db?

If it was just for my own personal use I'd just clean it up with some noise gating, but I suspect that would be frowned upon for our purposes here.
Reply
#2
This is a topic I thought I covered properly in the guidelines, but I think I'll need to expand it further.

Consumer output level is standardized (or, at least, there's a standard although lots of stuff even in hi-fi and hi-end world doesn't follow it) at -10dBv (or 2Vpp) and we must follow it.

The problem presents itself when we have audio interfaces with variable input gain because we need to find what is called the "unitary gain", that is, the interface doesn't add any kind of amplification of its own.

Now, bear in mind that the Mega Drive doesn't follow the standard consumer level either: it is capable of producing louder output than the standard and it's why sometimes we have to lower the volume on the Mega Drive itself to avoid clipping, but that's already covered in the guidelines. If a game is too loud, lower the volume on the MD itself.

Now comes the question: how can we find the unitary gain for our purposes?
I've talked to Artemio and he confirmed me that the audio test in the 240p test suite uses the highest amplitude fundamentals which is perfect for our purposes.

What's going to happen is that I'll produce a test file which is just a recording of the 240p test suite made with the Mega Drive Volume slider set at max and at unitary gain volume on the recording interface.

If you have an interface with a variable input level, you'll adjust it so that it matches the test file and you'll keep it for *all* the recordings and, if the game clips, just slide down the volume on the MD.

I'll update the guidelines with this shortly.
Vi veri veniversum vivus vici
Reply
#3
EDIT, just realized I didn't answer to the main question.

The answer is: leave it at that.
We're archiving music the way it's reproduced by the original hardware without any kind of modification because it may prove useful in the future as documentation.
So even if it is very quiet, to the point where the noise floor starts becoming noticeable, leave it be and don't make a big deal out of it.
And don't amplify the signal, I think I've already stated as much in the guidelines, but I think I'll rewrite completely that section to make it more clear.
Vi veri veniversum vivus vici
Reply
#4
(27-02-2020, 12:00 AM)Richard Mixin Wrote: What to do here? I guess the dilemma is that the soundtrack, as programmed, isn't loud enough on the YM2612 to get a good recording level without bringing in excessive noise. Are games like this just unrecordable, for purposes of the project? Given that we want basically stock Model 1s, is there any way to bring the noise floor on the console down further? Or should we just deal with recordings that, while having an acceptable noise floor, are also recorded with the average signal hovering around -12db?

If it was just for my own personal use I'd just clean it up with some noise gating, but I suspect that would be frowned upon for our purposes here.

Instead of fighting against the noise floor, you'd need to raise the pull-down resistors on the 2612 output as well as the PSG just for this game (assuming you want a cleaner rip).
Reply
#5
Bear in mind though that this would lead to hardware modification which is not accepted for the purpose of our own project, but it's absolutely valid solution to make a cleaner and proper rip without resorting to noise gating or post-processing.
Vi veri veniversum vivus vici
Reply
#6
Quote:This is a topic I thought I covered properly in the guidelines, but I think I'll need to expand it further.

I think you probably did cover it just fine; my confusion here, I think, is because there's some tension between the goals of the project (at least as I see it), between archiving and listening. If the true north here is the most accurate archival recordings possible, then it's probably best to just leave a LOT of headroom, enough to accommodate even the loudest games, and record them all at the same level. But IMO that makes playback less pleasant (the "audiophile" part). From a listening pleasure point of view, I personally want a volume level that won't force me to constantly fiddle with volume knobs (95% of my music listening is through a Sonos Play:1, at pretty low volume, as I write code all day), which implies recording as hot as possible without clipping anything.

That said, I think my goals for making hardware recordings are a little different from the 16bap, so the clarification helps. I don't mind recording the rips at whatever signal level you guys want, and then remastering them for my own purposes. I just figured that since I was going to be making a bunch of 96-24 recordings anyway, I may as well help the project out (and do as good a job as possible for your purposes).

Quote:What's going to happen is that I'll produce a test file which is just a recording of the 240p test suite made with the Mega Drive Volume slider set at max and at unitary gain volume on the recording interface.

If you have an interface with a variable input level, you'll adjust it so that it matches the test file and you'll keep it for *all* the recordings and, if the game clips, just slide down the volume on the MD.

That would be useful, I think. I have a VA6, which I've been using for recording, and a VA3, which is both a bit noisier, and at least 12db quieter at the same volume settings than the VA6. Which brings up the question of what the right "baseline" signal level is. Given that each individual piece of hardware varies so much, it probably makes sense to establish a reference recording level. I haven't played with the 240p Test Suite, but it we have something that can output a stable test signal, then we could establish that the reference recording level is the test signal at -6db or whatever. Might be worth testing some of the louder games to see if those clip, but it would be a good start.

Quote:Instead of fighting against the noise floor, you'd need to raise the pull-down resistors on the 2612 output as well as the PSG just for this game (assuming you want a cleaner rip).

Good advice - I might see about fiddling with my VA3 to clean it up a bit (but again, not for making 16bap recordings.

Thanks!
Reply
#7
Also FWIW, I just double-checked all of the rips I've submitted so far, by double-checking the "live" signal levels from each soundtrack vs the recorded levels, and confirmed they were all recorded with 0 gain on the audio interface (with gain attenuated by the volume pot on the console), so this isn't a concern with any of those. I'll re-record After Burner and Cadash today, which were the only two I've encountered so far with really low signals out of the Genesis.
Reply
#8
(27-02-2020, 05:42 PM)Richard Mixin Wrote: I think you probably did cover it just fine; my confusion here, I think, is because there's some tension between the goals of the project (at least as I see it), between archiving and listening. If the true north here is the most accurate archival recordings possible, then it's probably best to just leave a LOT of headroom, enough to accommodate even the loudest games, and record them all at the same level. But IMO that makes playback less pleasant (the "audiophile" part).
You have nailed perfectly the crux of our project: our two goals are in clear contrast between one and the other, I think this was discussed already on this site/forum, but for the sake of conversation I'll just point out that your observation is right in the money.

Our goal is archiving with the highest accuracy possible, but at the same time we're doing our best to do so with the best equipment and techniques with today's available technology.

This said, we had several iterations (hence you see "Remasters" around as well) because our target is not a static one and has only very recently become more stable, by giving the accuracy part the top priority. This has enabled us to give proper direction on what to do and how, and thus we have created the guidelines and let people help us out in our quest.
Ours is an imperfect job by definition, because to give a true "audiophile" result we'd need to employ techniques such as noise gating, EQing or even do hardware modification to improve the original circuit (which, to be honest, leaves much to be desired. FBX has done quite some work which you might want to check out: https://twitter.com/FBXGargoyle/status/1232912605800325121 ), but that goes against our goal of maintaining the source signal as unaltered as possible.

It's a lose/lose situation where we had to make a choice to make the best out of it and in the end we gave the accuracy part the highest priority so that our project could be used as "documentation" for those who're interested in studying how music sounded like in the Mega Drive and the differences between the various soundtracks.


If you've been around for a while, you might have noticed that our FAQs, Facebook and Twitter pages have recently modified their mission which is now not to offer "the best quality rips possible", but "archiving music with the best quality possible" which is quite a change.
I hope to have time to do a proper write up in the guidelines shortly, along with the test file.

EDIT: I've just read your latest post. Thanks, that would be immensely appreciated, although you might want to wait a bit until I do proper tests and give you the volume reference file.
It's really interesting that different revisions have different output levels. I own several revisions and never noticed much difference volume-wise between them.

EDIT 2: were both VA3 and VA6 at their max volume? Because if they were not, it might have been down to something as simple as a difference in the volume slider's taper.
Vi veri veniversum vivus vici
Reply
#9
And sure enough, Richard was right about difference in volume in different revisions.

Based on the MDFourier recordings done by Artemio, the Jap VA3 and VA6 are the same, but Jap VA1 is 4dB quieter.

This is really interesting as I'm not aware of differences in the audio amplification circuits of VA1 and VA3/6 besides the notorious fix.
Vi veri veniversum vivus vici
Reply
#10
Just out of random curiosity, would there be an objection if an audio rip was recorded on a Sega Genesis (launch edition) that had brand new Panasonic capacitors (in place of the aged capacitors)?
Reply


Forum Jump: