| |
Repose
Registered: Oct 2010
Posts: 222 |
Ultimate reverse engineering
This person took 32 chips of every revision, connected them to a custom clock source and ran them independently, to read out the OSC3/ENV3 of every combination, plus analog recordings. He wrote an emulator which recreates every register setting and measures all the variations within and between chips. Its an amazing amount of work. He also reverse engineered CIA's as well, as those are needed for "real sid" type tunes, which require emulating the entire machine for music playback.
https://www.youtube.com/watch?v=pePq68HaI7M
Emulating the SID the HARD way.
plgDavid
Here is one of the "torture test" tunes which require very accurate emulation of a specific model, it sounds amazing:
https://www.youtube.com/watch?v=BVevORuyQ_8
Jammer - Club Stylier
Club Stylier
Forever 2010 - #1 |
|
| |
iAN CooG
Registered: May 2002
Posts: 3138 |
true dedication, very interesting |
| |
chatGPZ
Registered: Dec 2001
Posts: 11148 |
I wonder why you link to a video that plays the tune in a 10 year old VICE though :) I don't expect 8580 Tunes to benefit much from this research either - it's mostly relevant for the 6581 Filter distortion and the mixed Waveforms perhaps. We'll see when he posts the testrig :) |
| |
Repose
Registered: Oct 2010
Posts: 222 |
I agree it's not the best link, but its the one I could find. Please link a better if you find it :)
There are other videos where he plays some demos, and its true that he said the variations in the 6581 model of the SID are more dramatic. But, he did mention some tricky problems in ADSR bugs and combined waveforms and PCM techniques. |
| |
chatGPZ
Registered: Dec 2001
Posts: 11148 |
Yeah, most of it is old news though - he replicated a lot of what Alankila already did years ago - whats new is that he used analog recordings to get full 12 bit resolution for the combined waveforms.... however, given their totally unpredictable nature and variation on real chips, i really doubt it makes and audible difference :)
As said, we'll see when it's all posted - then we can run his stuff against reSID and fix it eventually :) |
| |
Repose
Registered: Oct 2010
Posts: 222 |
Not sure if you noticed, but it is posted here: https://plogue.com/products/chipsynth-c64.html |
| |
Repose
Registered: Oct 2010
Posts: 222 |
I've found the differences explained in the manual at https://s3.amazonaws.com/chipsynth/C64_manual.pdf
2092 Slightly more
resonant.
2792 Good average
8580.
4887 Has mellower
combined
waves. |
| |
acrouzet
Registered: May 2020
Posts: 80 |
There's some stuff emulated here that isn't in ReSID, such as oscillator leakage and the variation in combined waveforms between chips. Apparently the elusive R1 is emulated too. ReSID's 6581 filter also has a few bugs, so I'm wondering if this emulation is an improvement overall. |
| |
Repose
Registered: Oct 2010
Posts: 222 |
Here may be a better link to one of the test tunes that I mentioned.
https://deepsid.chordian.net/?file=/MUSICIANS/J/Jammer/Club_Sty.. |
| |
iAN CooG
Registered: May 2002
Posts: 3138 |
Quote: Here may be a better link to one of the test tunes that I mentioned.
https://deepsid.chordian.net/?file=/MUSICIANS/J/Jammer/Club_Sty..
You're just linking a site that emulates in the browser the sid tune contained in HVSC. Anyone is able to play a sid tune by himself with an emulator, which are knowingly "not exact".
A sound capture from an emulator has no example value exactly as one emitted directly by the emulator. We already know they're not sounding like the real thing, if we exclude 8580 chips which are more "digitally correct" than the mostly unpredictable 6581s.
What really counts as an example is an actual sound capture from a real C64 with a real SID chip. |
| |
Repose
Registered: Oct 2010
Posts: 222 |
I agree with you, however the complaint was that my first link was to a 10 year old copy of VICE, and the next link was improved by being a recent emulator.
I did say "may be better" :)
I still haven't found an authentic recording of this tune. |
| |
McMeatLoaf
Registered: Jan 2005
Posts: 106 |
Quote: I agree with you, however the complaint was that my first link was to a 10 year old copy of VICE, and the next link was improved by being a recent emulator.
I did say "may be better" :)
I still haven't found an authentic recording of this tune.
There is a recording of it in SOASC:
http://chuangtzu.ftp.acc.umu.se/mirror/media/Oakvalley/soasc/hv.. |
| |
4mat
Registered: May 2010
Posts: 63 |
The recordings in the QA sections of the videos are from real SID chips. |
| |
4mat
Registered: May 2010
Posts: 63 |
The author of this plug-in used to be a scener back in the day and also worked on the c64 demo we made for it:
Chipsynth C64 Demo
But they don't seem to be able to get an account here. |
| |
iAN CooG
Registered: May 2002
Posts: 3138 |
Scener in which group? on the c64 or other platform? |
| |
4mat
Registered: May 2010
Posts: 63 |
On C64, with the handle Sneiv in the group P.I.S
He just posted a recording of real SID vs the plug-in with Jammer's "Club Styler":
https://soundcloud.com/chipsounds/jammer-club-stylier-rea-8580-.. |
| |
iAN CooG
Registered: May 2002
Posts: 3138 |
ok, no idea about that group (maybe just a local group)
Plogue
also added is old handle, and added him as gfxian for the demo |
| |
4mat
Registered: May 2010
Posts: 63 |
Thanks. |
| |
iAN CooG
Registered: May 2002
Posts: 3138 |
That said, I could only add the scener entry.
For the account, it's up to the mods |
| |
Repose
Registered: Oct 2010
Posts: 222 |
Quote: ok, no idea about that group (maybe just a local group)
Plogue
also added is old handle, and added him as gfxian for the demo
That's really cool, just what I wanted to compare. It does sound pretty good, the combined waveforms and filters are working great.
There is another difficult comparison I could suggest for him. This sound doesn't play correctly in the current reSID:
https://deepsid.chordian.net/?file=/MUSICIANS/D/Deetsay/Raveloo..
You can hear it properly here:
https://www.youtube.com/watch?v=FN6L9-TkHQ8 |
| |
chatGPZ
Registered: Dec 2001
Posts: 11148 |
How does it not play correctly? Sounds just fine for me in current vsid. You aren't mistaking deepsid for reSID are you? |
| |
iAN CooG
Registered: May 2002
Posts: 3138 |
Sounds correctly in
Sidplay/w 2.6 (ReSIDfp V0.0.2 Engine)
Sidplay 3 (ReSIDfp V2.4.0a Engine) (but at a bit lower volume)
VICE VSID (3.2 and 3.71 tested)
too bad doesn't sound correctly in ACID64 3.x+JsidDevice |
| |
Repose
Registered: Oct 2010
Posts: 222 |
Yes, I made the mistake of assuming that DeepSID was using the reSID engine.
DeepSID "WebSID Player" and VLC are the two which failed for me. |
| |
chatGPZ
Registered: Dec 2001
Posts: 11148 |
Quote:DeepSID "WebSID Player" and VLC are the two which failed for me.
Deepsid is really only "okish" for relatively "normal" tunes without samples or any other trickery. VLC uses a very old libplaysid iirc, like reSID from 20 years ago :) |
| |
Repose
Registered: Oct 2010
Posts: 222 |
I am using Sidplay 3 now, but its really really quiet. I'm wondering if I can compile this myself... |
| |
NoiseEHC
Registered: Feb 2005
Posts: 51 |
Just two fast little related questions:
What is considered the "best" SID emulation engine? Is it ReSIDfp?
I have seen "resonance fix" patched sidplay26w versions like a decade ago. Are those patches already merged into ReSIDfp mainline? Or were they pointless to begin with?
I ask these question because I will need something to play back sound in a sound editor for my player. Obviously either it should play on a real C64 or the best emulator. |
| |
chatGPZ
Registered: Dec 2001
Posts: 11148 |
Current reSID and reSIDfp are pretty much on par. The difference is mostly the way filter can be tweaked to your liking, and perhaps the defaults for them. Its mostly a matter of taste. All fixes have gone into both since Leandro is maintaining libresid(fp) |
| |
Grue
Registered: Dec 2001
Posts: 153 |
Think that resid is the most accurate emulation what comes into digital parts of the sid chip. Unfortunately it doesn't sound much like the actual 6581's.
Resid-fp sounds much more like it
I bought this new chipsynth c64 and it has the best sounding 6581 emu I've heard so far, but currently 8580 is lacking distortion (I already made a bug report and its being worked on)
Problem with the 6581 is that the manufacturing process was quite crap what came into producing analog parts of the filter circuitry, lots of variance there and every chip sounds different.
Some of you have noticed that there are 2 external capacitors connected to the sid chip. Datasheet states that those should be matched pair but commodore didn't care and used cheapest available option, instead of accuracy. Also I'm quite sure that the internal parts connected to the different cap sizes are not linear to each other either.
You can tune 6581 filter cutoff by changing capacitor values, and if those are not matched to the same value you can imagine how that affects with everything else in the equation.
I like the resid-fp and chipsynth64 approach to measure the output instead of going to ideal solution from the engineering point of view. Both of those emulators have plenty of different filter curves for us to enjoy.
Which is more accurate then? Its up to your own ears and likings. And if you're not obsessed with sid sounds you shouldn't even care which emulator is the best, just enjoy the tunes. |
| |
chatGPZ
Registered: Dec 2001
Posts: 11148 |
Commodore even used totally wrong caps on some boards :) (eg in all SX64, but also some regular boards) |
| |
McMeatLoaf
Registered: Jan 2005
Posts: 106 |
There's a demo of it here you can download and run:
https://www.plogue.com/downloads.html
(10 min per session and you can't save, otherwise it seems to be the same as the full version)
The emulated filter IMHO sounds quite different from reSID-fp or Sidplay3 and clearly closer to real chips. Too bad a few of the emulated chips seem to glitch on some David Dunn tunes, e.g. Elite (both subtunes). |
| |
chatGPZ
Registered: Dec 2001
Posts: 11148 |
Quote:The emulated filter IMHO sounds quite different from reSID-fp or Sidplay3 and clearly closer to real chips.
I am willing to bet: In a proper blind test, most ppl will still not be able to tell what is what :) |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Also: what are "real chips" in the 6581 realm anyway? I feel 6581 will be lost in translation because everybody had a different one and even remembers it differently. There's a reason everyone uses 8580 nowadays.
When I read threads like this I'm quite happy that I got my first C64 so "late" and it's normal for me to listen to 6581 tunes on 8580.
Bought me a couple of 6581s in ~200X – afair they all sounded different. Regardless of "revision". |
| |
chatGPZ
Registered: Dec 2001
Posts: 11148 |
Indeed a LOT of this "doesn't sound like a real chip" can be attributed to "doesn't sound like i remember it from my c64". Which is exactly why many ppl will not be able to tell what is what in a proper blind test (containing various recordings from emulators and different chips).
Reminds me how years ago i uploaded two recordings, both from reSID and only different in volume, and people were convinced one of them CLEARLY is the broken emulator :) |
| |
Frantic
Registered: Mar 2003
Posts: 1630 |
Quote: There's some stuff emulated here that isn't in ReSID, such as oscillator leakage and the variation in combined waveforms between chips. Apparently the elusive R1 is emulated too. ReSID's 6581 filter also has a few bugs, so I'm wondering if this emulation is an improvement overall.
What's "oscillator leakage"? |
| |
acrouzet
Registered: May 2020
Posts: 80 |
Quote: What's "oscillator leakage"?
An oscillator's volume level being 0 still lets some of the oscillator's signal through. You can hear this in many demo recordings on YouTube. When a tune has faded out, you can still hear the SID voices very faintly holding the last sound register values. |
| |
acrouzet
Registered: May 2020
Posts: 80 |
Quote: Also: what are "real chips" in the 6581 realm anyway? I feel 6581 will be lost in translation because everybody had a different one and even remembers it differently. There's a reason everyone uses 8580 nowadays.
When I read threads like this I'm quite happy that I got my first C64 so "late" and it's normal for me to listen to 6581 tunes on 8580.
Bought me a couple of 6581s in ~200X – afair they all sounded different. Regardless of "revision".
6581 has some analog artifacts that emulators still don't get exactly right. They vary quite a bit between chips, but sometimes an emulator will do something that is rarely heard on a real chip, if at all. |
| |
F7sus4
Registered: Apr 2013
Posts: 112 |
Quoting acrouzetAn oscillator's volume level being 0 still lets some of the oscillator's signal through. You can hear this in many demo recordings on YouTube. When a tune has faded out, you can still hear the SID voices very faintly holding the last sound register values.
This is true. By the way, it is still possible to modulate sound with registers in that state and the result will lead to a lot of glitch-like interferences otherwise impossible to achieve. From what I see, this behavior is still yet to be researched (and possibly utilized) or at least I didn't find anyone else doing it to a larger extent before.
Quoting GroepazIndeed a LOT of this "doesn't sound like a real chip" can be attributed to "doesn't sound like i remember it from my c64". Which is exactly why many ppl will not be able to tell what is what in a proper blind test (containing various recordings from emulators and different chips).
It really depends on individual sensory receptivity and ear-training. No matter how precise the emulation will attempt to be, reSID will still remain distinguishable from the real chip in many circumstances, especially because of how organic the SID's lowpass filters sound at lower filter values. (Not mentioning SwinSID and its hardware alikes as they fall into the more obvious "I'm not a real SID sound" category.) |
| |
chatGPZ
Registered: Dec 2001
Posts: 11148 |
Quote:the variation in combined waveforms between chips.
This btw is - just like variations in the filters - something that is simply different, everywhere, and can not be attributed to some chip revision (for example). The difference between reSID and this new emulation (and also reSID-fp) regarding those thing is that reSID does not try to sound like a specific (real) chip. It sounds like "a" SID chip though :)
But yes, the emulators (also this new one, btw) do not simulate certain analog effects related to the mixed waveforms, but instead use tables for them. It will still take a while until this can be synthesized in real time (on a random pc anyway) :) |
| |
acrouzet
Registered: May 2020
Posts: 80 |
Quote: There's a demo of it here you can download and run:
https://www.plogue.com/downloads.html
(10 min per session and you can't save, otherwise it seems to be the same as the full version)
The emulated filter IMHO sounds quite different from reSID-fp or Sidplay3 and clearly closer to real chips. Too bad a few of the emulated chips seem to glitch on some David Dunn tunes, e.g. Elite (both subtunes).
If you're referring to the octave drops, its actually a thing some real 6581s do, its just rare. Like you said, only a few out of the many 6581s in the emulator do this. |
| |
Jammer
Registered: Nov 2002
Posts: 1293 |
'Club Stylier' is one of my old 'dead end' techniques back when I handled multi sounds with Attack set to 2 which is not the best and most stable thing to do with ADSR at such speed rates. Surely not doing it anymore ;)
Here's the leakage, btw:
https://youtu.be/YGxAWQzTzVo?t=75
Played with dual 8580 so glitch manifests on each chip separately with different delay (thus you can hear that left/right thing). Cutting sound with testbit or zero wave is the easiest way to reproduce. |
