Audio filter mod

[jarp]

Hi all,

I've been wanting to modify hardware audio filter from my Amiga 600 due to lacking highs when playing music. Amiga 1200 does not have this problem due to higher cut off frequency and I was planning to mod Amiga 600 and rise cutoff frequency to Amiga 1200 levels. I am not electronics engineer so perhaps somebody could kindly verify that I am doing the right thing?

So I've been studying schematics of Amiga 500, 600 and 1200. They all are implemented equally, it's just a simple low pass filter implemented by using opamp, resistor and capacitor per channel.



So cutoff frequency can be calculated using formula 1/(2*3,14*R*C). According to schematics different Amiga models uses following values:

A500: 0.1 uF, 360 Ohm -> 4,4 kHz
A600: 3900 pF, 1.5k Ohm -> 27 kHz
A1200 r1: 3900pF, 1.5kOhm -> 27 kHz
A1200 r2: 6800pF, 680 Ohm -> 34 kHz

So actually Amiga 600 should already have quite nice cutoff frequency - however schematics are wrong. I have three Amiga 600 from r1 to r2 and all have 360 Ohm resistors and according to RMAA measurements cutoff frequency is ~4kHz so I suspect that filter in Amiga 600 is equivalent to Amiga 500. Somebody else also said on other thread that schematics are wrong.

So to change cutoff frequency to match Amiga 1200, one can change resistors and capacitors to match Amiga 1200, but that would alter gain also, no? So alternative is just to replace capacitors and calculate suitable value using that formula 1/(2*3,14*R*C). That should keep original gain but alter cutoff, did I got it right? I calculated that when resistor is 360 ohm using 15 nF capacitor will produce 29 kHz cutoff frequency and using 22 nF capacitor will produce 20 kHz cutoff frequency. 20 kHz should be plenty, but one could use 29 kHz also since A1200 r2 uses 34 kHz, no?

Am I on right track?

Thanks a lot!

-J

[jarp]

I just had to test this already. I replaced filter caps with 15nF caps (only problem was that my 0805 caps were a tad too small to replace original caps, had to place them extremely carefully to avoid shorts) which should produce that 29 kHz cutoff frequency.

Here frequency response before mod (oops seems my input levels were way off, I calibrated levels using VU meter on my preamp but that probably did not take such agressive low pass filter into account):



And here's after mod (be aware of the scale, it's different on both pictures):



I call this success. Frequency response is actually better than on many 1990s sound cards Please notice that channel imbalance is probably due to my pre-amp which has a bit... sensitive pots. It seems, though, that audio will still be cut off at about 10...15 kHz, must be limit of Paula chip or something? So using 22 nF even 27 nF capacitor would be just fine...

Ps. RMAA says that noise level is 83 dB and dynamic range is 77 dB. I think this is amazing when taking Paula's age into account? Distortion is high though.

[ross]

Originally Posted by jarp

Ps. RMAA says that noise level is 83 dB and dynamic range is 77 dB. I think this is amazing when taking Paula's age into account? Distortion is high though.

Hi jarp, interesting results.

This can confirm feasibility for my thoughts:
http://eab.abime.net/showpost.php?p=1146073&postcount=6

We've log[2](4856)=12,25bit non-linear resolution (then more linearized), so a DR>73,72dB.
We cover 77dB with 12,79 Q-bit linear <- 20*log[10](2^Q)

There is somewhere on board a long discussion about real dynamics obtainable from 14bit Paula player.
Someone think 10, someone 11, 12, 13, full 14 Q-bit DR or better...

I narrow the field, my bet is somewhere betwheen 12,25 and 12,80 Q-bit
(which is absolutely an excellent result for a chip of the early 80s!)

Bye,
ross

[Amiga1992]

I achieved the "A1200 sound" just by removing the two filters underneath the board, C331 and C321. I was thinking about adding a controllable cutoff filter on those points, but I never did. Maybe a project for one day!
So the A600 for whatever reason has those extra filters on top of the lowpass filter it already has (the LED one all Amigas have). Why? No idea.

More discussion on this subject here: http://eab.abime.net/showthread.php?t=57506

[jarp]

Yeah those C332 and C321 caps are those low pass filter caps from schematics above. By removing them I believe you will remove totally disabled low pass filter which is not exactly like A1200. AFAIK that low pass filter is for removing aliasing but don't know if it really matters in real life.

In any case beats me why Commodre had set it to 4.4 kHz. That is totally too low and cripples Paula's capabilities badly. Who cares if there are no aliasing when you cannot output frequencies higher than that!

[Amiga1992]

I'll do some recordings to compare, I never bothered, since it sounds really good to me.

[xArtx]

I’d have assumed it was simply because the Amigas after A600 are faster and the highs would sound like crap with the slower models.
Are you able to produce a high audio sine and hear that on it’s own?

[ross]

I do not think it has to do with machine speed but with the available Chip memory.
In OCS/ECS Amiga normally we have 0.5/1MB, on AGA 2MB.
Paula do not have a fixed period output, so more high-frequency sampled audio (better quality) more Chip RAM is required.
In old models i'm constrained to use low-frequency samples and then a filterer output is a better compromise...

[Amiga1992]

Originally Posted by jarp

Yeah those C332 and C321 caps are those low pass filter caps from schematics above. By removing them I believe you will remove totally disabled low pass filter which is not exactly like A1200. AFAIK that low pass filter is for removing aliasing but don't know if it really matters in real life.

Can I bump this thread? I cannot find where this discussion happened, but someone mentioned that removing these capacitors might introduce some distortion on the high end. Whoever you are, sorry I cannot credit you for it but chip in if it was you.

I've been using my A600s for years without these two capacitors, and guess what, I started to notice that distortion. It doesn't always happens, but when it does, it's nasty.

So hopefully an idea can come up here on what to replace those caps with, because it seems leaving them empty can be problematic. In al inked above thread, 8bitbubsy said they changed components with those found in an A1200, and it was an improvement. Would that be the solution?

[Jope]

I reckon installing the A1200 values in there are a good idea.

[pandy71]

There is also problem with OPAMP's used in Amiga audio path - they are simply to slow to correctly convert current to voltage and do lowpass filtering - at least in first stage.
passive LC filter should precede OPAMP's - this would help a lot with slow OPAMP (or replace OPAMP to something significantly faster with decent GBW - capable to deal with few V square wave 3.5MHz), additional issue is with virtual analog ground and power supply quality also affect overall audio quality.

[Amiga1992]

Originally Posted by pandy71

There is also problem with OPAMP's used in Amiga audio path - they are simply to slow to correctly convert current to voltage and do lowpass filtering - at least in first stage.
passive LC filter should precede OPAMP's - this would help a lot with slow OPAMP (or replace OPAMP to something significantly faster with decent GBW - capable to deal with few V square wave 3.5MHz), additional issue is with virtual analog ground and power supply quality also affect overall audio quality.

OK all cool but how to solve? let's be practical and fix Amiga audio posting actual procedure on what to change and for what.
I also notice that the output level of every Amiga I have is really low, doesn't seem like proper line level.

[pandy71]

Originally Posted by Amiga1992

OK all cool but how to solve? let's be practical and fix Amiga audio posting actual procedure on what to change and for what.
I also notice that the output level of every Amiga I have is really low, doesn't seem like proper line level.

There was at least one thread with such modifications ( http://eab.abime.net/showthread.php?t=109051 ) but some problems would require different approach so new PCB (at least some piggyback/daughterboard for Paula, filtered virtual ground, low noise power supply, faster OPAMP and passive lowpass 1.79MHz filter).
It is quite important to understand that Amiga was designed in first half of 80's past century and from this point we have significantly more knowledge about let say digital audio. Paula can sound better and also 14 bit mode can be done more accurate but this require lot of changes when compared to generic Commodore.

[8bitbubsy]

Yes, the schematics are wrong. All A600's are ~4.4kHz (just like A500), and I think all A1200's are ~34kHz. Not sure about the very early boards.

[Mick]

Looking at the schematics both the A600 and A1200 have the same 3khz 2nd order Sallen-Key active low pass filter in the 2nd stage, the only difference is in the 1st stage. It's the two 10K resistors in series, the 6800pf in parallel with the op amp and 3900pf to ground. I'm not exactly an expert but I've been toying with trying to improve audio on a MD2 for what seems like ages now.

Use this calculator if you want to modify the cut-off frequency:
http://sim.okawa-denshi.jp/en/OPseikiLowkeisan.htm

I just looked at the A500 schematics as well and they too have the same 3khz filter on the final stage.

CD32 has it too for the audio coming out of Paula, A4000 as well... seems fairly consistent, probably all of them. If you want higher frequencies than 3khz to be output you'll need to modify that 2nd order filter.