Log inRegister an accountBrowse CSDbHelp & documentationFacts & StatisticsThe forumsAvailable RSS-feeds on CSDbSupport CSDb Commodore 64 Scene Database
You are not logged in - nap
CSDb User Forums

Forums > C64 Composing > 6581 filter curves
2014-10-03 18:43

Registered: Jul 2007
Posts: 369
6581 filter curves

So I stumbled over this graph by Antti Lankila:

On the X axis we have the value that is written into the cutoff register (0-2047), and on the Y axis we have (if I'm not completely mistaken) the actual cutoff frequency within some working range.

Is it really this bad? I mean, I know that 6581s are supposed to have individual differences, but this is ridiculous. It means that if I write e.g. $80 into $d416 to get a 50% cutoff, I could get just about ANY cutoff frequency from like 10% to 90% of the total range.

How do y'all deal with this when composing for 6581? Do you check on multiple machines? Does that even make a difference given the absolutely ludicrous variation? Or do I misunderstand the graph?
2014-10-03 19:11

Registered: Apr 2008
Posts: 417
I believe Antti Lankila et. al. were looking for the variety rather than the most common to map the filter.

Personally, I have 5 randomly acquired C-64's with 6581 and they all are close to the average c-64 filter. My first one(RIP) had filter that had the strong curve.
2014-10-03 19:12

Registered: Nov 2002
Posts: 1288
2014-10-03 21:35

Registered: Jan 2005
Posts: 103
Some time ago zerozillion recorded a set of tunes on several different 6581s, one of which had a ridiculously light filter, the cutoff value at X=0 must have been close to 1000 Hz.

This tune seems to have been made on a chip with a really low filter curve:
2014-10-03 21:54

Registered: Aug 2010
Posts: 61
I often wondered how much does the values (and tolerances) of the filter capacitors connected to the SID-chip affect the result? Are the values in that graph a sum of SID+filter caps-combination?
2014-10-03 22:51

Registered: Sep 2006
Posts: 273
@Jammer EXACTLY! :)
2014-10-04 06:58

Registered: Apr 2004
Posts: 933
I do check my tunes on some (vice) 6581 variations and on 8580, lately. The Follin curve is the most annoying, it's not only open, it's also loud, but I doubt anybody uses that to listen to Sids. My own c64 comes closest to Nata's curve. I've also had one before that fits the strong filter.

All in all, for me setting filters for all Sids is dealing with high and low values and the values in between can (only) be used to interpolate. Using middle values as key filter values to create cool sounds will only lead to dissapointment when listening to them on other Sids, it makes Batman angry.

Still, most MoN tunes sound great on all Sids. It's interesting to notice there's not much difference between how I think they sounded in 89/90 and how they sound now with any Sid filter setting. Most likely also because of the well defined compositions and themes.
2014-10-04 07:32

Registered: Nov 2002
Posts: 47
yes it is really this bad! And compo composing on the 6581 just sucks, you can be sure your tune will sound muffled or too bright on compo machine.

We went with zrx thru recordings of all his chips to find a particular good one for hubbard songs (which i later on used to record project hubbard, with custom output amplifier - https://soundcloud.com/zerozillion/sets/project-hubbard) and of out of about 40 chips maybe 2 or 3 do sound the same. strong reference is for me the hubbard song "nemesis the warlock", it includes a bell sound in the intro that has a wide sweep, when this bell sounds nice and sharp, the curve is more to the left (trurl red), when the attack is slow-ish or the bell inaudible, its the curves more to the right. this bell sound depends a lot on the filter, it is not a sample, so it is a good indicator as it is right at the intro of the song and only takes a second to listen to (to my judgement, the correct type of hubbard filter would be near the red trurl curve, as this one plays the bell sound most accurate - as i think from the composition it must be a striking bell, not a slow attack bell). of course other oldskool tunes need other filter curves; and for modern tunes we gladly have filter models like jeff's one. though that only solves it partially on both ends.

i once came up with the idea of measuring at least the bottom and high point of the lowpass by ear, which gives somewhat reproducable results: SID Audit (test 3 & 4). the plan was to run this test on compo machines and index them, then to run the tests in vice to have matching models to each compo machine. as nobody caught interest in that (hi there!), i let it be and now i use 8580 as it saves me a lot of nerves, a pity for the filter distortion though.

one way to use the 6581 filter though is to use it only statically and move it a bit upwards or downwards, preferrably at spots where all curves meet somewhat, so no drastic fast moves or jumps, and no dedicated hour long filter editing. this ensures at least that bass sounds are not totally swallowed or other sounds are killing the tweeters in the PA system.

filter caps - i swapped them several times in machines, i see the tolerance f**ups inside the chips filter section are much wider than what could be off by the cap tolerance.
2014-10-05 01:56

Registered: May 2012
Posts: 28
Quoting onebitman
and for modern tunes we gladly have filter models like jeff's one

What does this refer to?
2014-10-05 07:17

Registered: Nov 2002
Posts: 47
Filter distortion emulation in sidplay and vice, you can select which curve is being used.
RefreshSubscribe to this thread:

You need to be logged in to post in the forum.

Search the forum:
Search   for   in  
All times are CET.
Search CSDb
Users Online
Guests online: 97
Top Demos
1 Next Level  (9.8)
2 Mojo  (9.7)
3 The Ghost  (9.7)
4 Coma Light 13  (9.7)
5 Edge of Disgrace  (9.6)
6 No Bounds  (9.6)
7 Uncensored  (9.6)
8 Comaland 100%  (9.6)
9 No Sprites  (9.6)
10 Wonderland XIV  (9.6)
Top onefile Demos
1 Party Elk 2  (9.7)
2 50 Shades of Zardax  (9.6)
3 Cubic Dream  (9.6)
4 Rainbow Connection  (9.6)
5 Copper Booze  (9.5)
6 TRSAC, Gabber & Pebe..  (9.5)
7 Onscreen 5k  (9.5)
8 Dawnfall V1.1  (9.5)
9 Daah, Those Acid Pil..  (9.5)
10 Quadrants  (9.5)
Top Groups
1 Nostalgia  (9.3)
2 Booze Design  (9.3)
3 Performers  (9.3)
4 Oxyron  (9.3)
5 Crest  (9.3)
Top Graphicians
1 Mirage  (9.8)
2 Lobo  (9.7)
3 Mikael  (9.7)
4 Archmage  (9.7)
5 Facet  (9.7)

Home - Disclaimer
Copyright © No Name 2001-2024
Page generated in: 0.051 sec.