SID filters - questions.

[Sebo]

Hi:

This week I received the Core and SID kits from Smash TV (Exelent quality).

I just finished the soldering and plan to do some test this weekend.

Last night I was reading the 6581 and 6582 data sheets, and the recommended

capacitors for 6581 filter are 2200 pF and the maximum frequency cutoff is

calculated with this formula:

FCmax = 0.000026/C

And the minimum cutoff is 9 octaves below.

With a 2200 pF the filter range is aprox 30 Hz to 12 KHz.

In the SID Module design this caps are 470 pF, applying the formula give me

a max cutoff of 55.3 KHz and an aprox minimum of 108 Hz.

I don't know if this choice for the 470 pF if for another reason.

I will try both values, with 2 different 6581 (a R2 and a R4).

Anybody tryed another cap values other than the 470 pF?

For 6582 (same as 8580) the data sheet caps recomendation is 6800 pF, so

no discrepance here.

Sebo

[stryd_one]

Post removed, forum glitched

[Nattgris]

FCmax = 0.000026/C

And the minimum cutoff is 9 octaves below.

With a 2200 pF the filter range is aprox 30 Hz to 12 KHz.

In the SID Module design this caps are 470 pF, applying the formula give me

a max cutoff of 55.3 KHz and an aprox minimum of 108 Hz.

I wouldn't trust that formula much. The 6581 data sheet was written before the circuit was completed and it turned out that the circuit didn't follow the specs very well, especially not the filter. A quote from a transcription of the data sheet at http://stud1.tuwien.ac.at/~e9426444/ :

FCmax = 2.6E-5 / C

Where C is the capacitor value. The range of the Filter extends 9 octaves below the maximum cutoff frequency.

[Transcriber's note: Allegedly, one of the designers should later have stated that the above formula is wrong].

470 pF is what's inside the Commodore C64 (at least the later models), I guess that's where the value comes from. (I still don't get why the 8580 cap was chosen different from what's inside a C64 using the 8580.)

stryd_one: Your last post is somewhat dangerous... I almost got stuck in an infinite loop. ;)

[Sebo]

Nattgris:

Thank you for the repply.

[stryd_one]

Sorry about that! For some reason, that post came up in a thread of it's own, that's so weird!!

[TK.]

470 pF is what's inside the Commodore C64 (at least the later models), I guess that's where the value comes from. (I still don't get why the 8580 cap was chosen different from what's inside a C64 using the 8580.)

The reason why the caps are not specified like in the original schematic is, that at the time where I created the MBHP_SID module (4 years ago) it was assumed that each SID needs different caps anyhow and the best specific values have to be found out by the user due to the variance between the revisions.

Nowadays we know, that this mainly is true for the 6581, but not for the 6582 and 8580.

Due to the fact, that at this time the 8580 filter was controlled with the same curve like the 6581, the 6.8nF value which can be found in the 6582 datasheet was more advantegous. After I changed the control curve based on the Razmo's great (!) suggestion (see his detailed description above), I never spent so much attention on testing different caps again.

Ok, to end this thread I've just tested my 8580 with 22nF caps. As mentioned several times before by other people, the cutoff frequency goes deeper, this has the advantage, that in between the typical frequency range where the resonance plays a role when bass notes are played, there is less stepiness than with 6.8nF caps. I've tested some sounds which are using bandpass and highpass as well, and don't hear any restrictions in the sounds

So - are 22nF caps really better, what are your oppinions?

For the 6581 I still believe, that the 470pF caps (taken from the original C64 schematic) are better.

For the 6582 I will test this now (thanks to Wilba for providing me two of these chips! :))

Best Regards, Thorsten.

[TK.]

Here the test results:

8580:

  - first impression was, that the sounds are more crispy with 6.8 nF caps - but this only depends on the filter envelope settings which are used. If the envelope depth is slightly increased when 22nF caps are stuffed, then you will get identical results - from the mathematical point of view, this behaviour makes sense.

  - I tried 100 nF caps, and the result is, that between 0..20 CutOff frequency (the "stepiness range") you won't hear any sound, and that within 21..127 the cutoff can be controlled with absolutely best resolution (-> no stepiness)

  - with 100 nF caps, it is not possible to set the cutoff frequency so high, that the filter has no effect... on the other hand: this is no problem if you want to play bass sounds. E.g., also the original TB303 doesn't allow to open the filter completely (if you want to take this old gear as reference... ;-))

6582:

  - I don't really hear a big difference to the 8580, considered that e.g. on a filter sweep via envelope, the depth has to be adapted to get the same "crispiness"

So: are 100 nF better for your purposes, or 47 nF, or 22 nF as a compromise (still filter stepiness in the low range, but filter can be completely opened) - or does it make sense to add a manual switch for different sets of caps? ;-)

Best Regards, Thorsten.

[Wilba]

After playing with 22nF on my 8580s and 6582s, I tend to like them better than 6.8nF, the only downside is I need to tweak the cutoff for the preset patches (and the depth of any envelope or LFO that modulates cutoff). I haven't tried anything higher though.

I put machine pin sockets in the filter capacitor holes so I can swap them around, and I've thought about a manual switch, and even contemplated a software controlled switch too... there's a couple of spare outputs on the shift register which could be used to drive a 4066.

As I said before, the only downside to switching over to 22nF is having to tweak all the existing patches to make them sound the same... which really isn't that much effort (especially if you have a four SID setup and can compare SIDs with different capacitors). It would be nice if 22nF became the new default (it's also the HardSID default, btw), perhaps as part of the MBSID v2, each patch can store what capacitors should be used (hint to user to flick a manual switch, or firmware can drive an electronic switch).

I'm going to run some tests with a 4066 and see if it works... :-)

[TK.]

If most people are thinking, that the 22 nF works better (for myself it doesn't really matter, I'm happy with both), I will change my own 8580 preset sounds accordingly.

perhaps as part of the MBSID v2, each patch can store what capacitors should be used (hint to user to flick a manual switch, or firmware can drive an electronic switch).

yes, I can consider this:

 * the information, which SID and caps have been used should be stored in the patch. 
SIDs: 0=6581, 1=6582, 2=8580, 3=FPGA; 
caps: 0=470pF, 1=1nF, 2=2.2nF, 3=4.7nF, 4=6.8nF, 5=10nF, 6=22nF, 7=47nF, 8=100nF[/code]
Best Regards, Thorsten.

[Razmo]

So basicaly what you do by changing the caps, is that you "squeeze" the frequency range of the analog filter. It sacrifice top end, but allows "better resolution" instead as the Filter Frequency range has been put under "magnifying glass" in the low end... so to speak.

Nonetheless, this does NOT change the anyway exponential curve of the digital control logic, and thus the special 8580 code I came up with would still be valid for a true linear response curve, though it sacrifice resolution the closer you get to the low end (gradual steppiness).

The good thing about "squeezing" the range though is, that as the high range of the frequency (thanks to the digital control logics) is of much greater resolution, so "squeezing" the range does not make a very drastic change in the new max. range value, but greatly does so in the low end (again, because of the digital control logic)... so yes; for bass patches this should not really be a problem.

Don't caps come in adjustable versions? like risistors does with potmeters? TK? ... if they do, can they not be used on the boards so that changing the cap values would be much more flexible?

Regards, Razmo

[Razmo]

... and also, changing the range like that, would probably ruin the tuning of the filter... maybe the cap values currently used was chosen because they go "hand-in-hand" with the frequency of the oscillators? ... in other words, the filter frequency is easier tunable in semitone steps... this feature is very important when making acid-sounds with high resonance... where the filter cutoff freq must follow the pitch as played on the keyboard.

Regards, Razmo

[stryd_one]

Hey Wilba don't spose you could make up some mp3's to give us an idea?

Razmo you make a very good point there about the tuning. Yet another thing for the lab...

[Wilba]

Trimmer caps are only in the very low pF range.

** edit ** dumb idea removed  ;)

Hey Wilba don't spose you could make up some mp3's to give us an idea?

An mp3 of what? My SIDs sound like TK's and Razmo's demos anyway... do you just want a non-steppy filter sweep using 100nF caps?

[stryd_one]

I was curious to hear examples of the different caps - maybe it was posted and I missed it  :-\

[ganchan]

Don't caps come in adjustable versions? like risistors does with potmeters? TK? ... if they do, can they not be used on the boards so that changing the cap values would be much more flexible?

Regards, Razmo

http://en.wikipedia.org/wiki/Variable_capacitor  ;D

[TK.]

Don't caps come in adjustable versions? like risistors does with potmeters? TK? ... if they do, can they not be used on the boards so that changing the cap values would be much more flexible?

Instead of adjustable caps, I just have mounted a pinhole header on my SID module which allows quick exchange:

Best Regards, Thorsten.

[TK.]

I was curious to hear examples of the different caps - maybe it was posted and I missed it  :-\

no, you haven't missed them - there are no MP3's yet - maybe somebody could add a nice documentation about the usage of different cap values into the Wiki? :)

Best Regards, Thorsten.

[TK.]

... and also, changing the range like that, would probably ruin the tuning of the filter... maybe the cap values currently used was chosen because they go "hand-in-hand" with the frequency of the oscillators? ... in other words, the filter frequency is easier tunable in semitone steps... this feature is very important when making acid-sounds with high resonance... where the filter cutoff freq must follow the pitch as played on the keyboard.

I think, that the 22nF cap values have been choosen, since they are the best compromise for covering the whole LP/HP/BP frequency range.

I will propably stuff two of my 8580 SID modules with 47nF caps in future, and integrate a switch into the firmware which does take this into account in 303 mode.

Currently only the first quarter of the cutoff frequency value is used in 303 mode, but in the alternative version, the firmware could access the upper half (or 3/4 of the upper range) instead, which results into the best utilisation of the available filter resolution.

Best Regards, Thorsten.

[ganchan]

Instead of adjustable caps, I just have mounted a pinhole header on my SID module which allows quick exchange:

Best Regards, Thorsten.

I've do exactly same thing on mine monoSID, with pinhole header solder oveturned :)

[Wilba]

Machine pin headers work well:

http://www.sparkfun.com/commerce/product_info.php?products_id=743

Or you can just cut an 8-pin machine pin IC socket in half.

[Jaicen]

I don't know if it will work in this particular application, but over on the DIY stompbox forum a solution has been devised which allows a psuedo variable capacitance for such things as treble boosters and fuzz face variants.

This works by stuffing the lowest desired capacitor in the pcb, and in parallel, wiring a 1M pot and a larger value cap  in series (usually more than twice the value of the first). As the pot is turned down (less resistance) the capacitance of the smaller cap is increased. This effectively gives you a variable capacitance with a fairly wide range. I don't think it works well with timing circuits, but for passing audio it usually works very nicely. I use it in all my Fuzz Face circuits to stop them being too mushy.

[stryd_one]

Hey that could just work..... Wonder if a digipot would be up to scratch ....  hmmmmmm

[ganchan]

Machine pin headers work well:

http://www.sparkfun.com/commerce/product_info.php?products_id=743

Or you can just cut an 8-pin machine pin IC socket in half.

yes i use exactly this one, but some time you have to sold some tin on the cap's terminals because this are too thin for the hole.. instead i prefer use this solution for 78x:

:)