Audio filter mod

[Mick]

Ahh, I wondered what that circuit was all about. I already fitted my A600 audio circuit to A1200 2B spec so that means all of my tests had the same 34khz cutoff. What a waste of time that was then...

It's no wonder Commodore went bankrupt, so they were adding dozens of parts to their audio circuit for a feature that was barely even used, whilst the likes of Sega were actively sabotaging their own consoles sound quality to save as much money as possible.

Has anyone tried upgrading the op amp? or will the difference be pretty much zilch?

[Mick]

I'm tempted to try an OPA4134, thoughts? waste of money?

[pandy71]

Originally Posted by Mick

I'm tempted to try an OPA4134, thoughts? waste of money?

IMHO waste of money - if you searching for higher quality then you need faster OPAMP - Paula use PWM to control volume - slow OPAMP introduce distortions (technically audio signal produced by Paula is PAM like - Pulse Amplitude Modulation - OPAMP speed is quite important to not distort pulse).

[Damion]

Originally Posted by Mick

Has anyone tried upgrading the op amp? or will the difference be pretty much zilch?

I think it can make a change in sound characteristics for sure, though it's tough to gauge what's better in practice. For fun (not because I know what I'm doing) I tried a few different ones once when repairing an A1200 audio circuit (the op-amp had to come off). First was an MC33079, sounded very clear, but bass was very obviously lacking compared to the original LF347.

Next I tried an OPA1644, and left it in as it sounded really fantastic to my ears. Placebo effect said similar to the LF347, but a bit clearer and bass is more defined/slightly less "gritty" than the LF347. Again that's just subjective woo wording.

pandy71 might be able to speak from the techincal side as to whether or not that was actually a good idea..

[Mick]

Originally Posted by pandy71

IMHO waste of money - if you searching for higher quality then you need faster OPAMP - Paula use PWM to control volume - slow OPAMP introduce distortions (technically audio signal produced by Paula is PAM like - Pulse Amplitude Modulation - OPAMP speed is quite important to not distort pulse).

What op amp would you recommend as an upgrade?

How about OPA4992? I might just buy a few and compare.

[Mick]

I did a test with the 3khz filter forced on and can confirm it sounds dire.

https://ufile.io/f/wllgc

On the one hand it makes you wonder why Commodore were spending money on something optional that made the sound so bad but on the other you have to be impressed by how their audio circuit requires hardly any noise filtering at all and yet is so clean.

[pandy71]

Originally Posted by Mick

What op amp would you recommend as an upgrade?

I would search for OPAMP able to cope decently with PAM in Paula - large signal step response seem to be most important in I/V converter - there is few discrete circuits or using special video dedicated OPAMP in Current to Voltage converters (Paula seem to be current DAC type so some I/V converter is required)

Originally Posted by Mick

How about OPA4992? I might just buy a few and compare.

It looks well on datasheet - another cheap possibility is TLE2074 - Linear technology made few decent OPAMP's in past but since they was acquired by Analog Devices then their prices raised beyond sanity... (LT never was cheap but today...).To improve audio quality instead replacing standard OPAMP to "audio" OPAMP i would rather redesign Paula power circuitry (particularly bias/reference for DAC's - perhaps something around active virtual ground generator - something like TLE2426), properly decouple power lines and ground, use local linear regulator to filter power line and use ferrite beads to clean signal.Decent polyester/polypropylene capacitors, metal film resistors - usual audio (not high end audiophile) things.

[Mick]

TLE2074 looks to be 16pin : but is pin compatible so I guess I could cut the unused legs off and give it a try.

Do you mean using something like LM4040 2.5V rather than the resistor voltage divider? if you used a 5V voltage regulator to power Paula you would need to take power from the 12V rail right? like the CD32 circuit powering the LC78835M and then going on to VDAC/AVref?

[Mick]

Is it something like this you have in mind? I might be able to able to do a small board for my A600 if there's nothing too radical needs changing.

[pandy71]

Yes, generally Paula especially in DIL package but also PLCC suffer from insufficient decoupling analog from digital (let skip discussion about DAC's quality etc) - feeding clean power into Paula will be beneficial for audio quality more than changing OPAMP with distortions 0.003% to new one with distortions like 0.00006%.

[Mick]

I get the feeling it's going be totally underwhelming just going by human ear but I might give it a try out of curiosity, you never know. Is there anything you would add/change? pin 8 of Paula will have to stay connected to normal ground because I'd have to cut traces and I don't want to do that.

edit: I made a small pcb to solder in using castellated holes but after seeing the quote for something so tiny I think I'll just buy some throughhole parts instead.

[pandy71]

Originally Posted by Mick

I get the feeling it's going be totally underwhelming just going by human ear but I might give it a try out of curiosity, you never know. Is there anything you would add/change? pin 8 of Paula will have to stay connected to normal ground because I'd have to cut traces and I don't want to do that.edit: I made a small pcb to solder in using castellated holes but after seeing the quote for something so tiny I think I'll just buy some throughhole parts instead.

Confirm - you need probably new PCB layout focused on decoupling and audio quality - from my perspective this will be more beneficial than simple OPAMP swap - there is on EAB thread with audio signal path modifications - for example adding ferrite bead substantially improved noise floor - so even small modification can give a lot to deliver cleaner audio signal.But adding local linear power regulator, creating way better filtered virtual ground are next steps to improve Paula environment.

[Mick]

Adding ferrite beads to where? the op amp power inputs instead of the resistors? do you have a link to the thread?

I was thinking something along the lines of this kind of mod but it's a bit risky to try for $75? when I ordered something of a similar size before it was like $25.

[pandy71]

Originally Posted by Mick

Adding ferrite beads to where? the op amp power inputs instead of the resistors? do you have a link to the thread?I was thinking something along the lines of this kind of mod but it's a bit risky to try for $75? when I ordered something of a similar size before it was like $25.

Check this thread https://eab.abime.net/showthread.php?t=109051 , particularly https://eab.abime.net/showpost.php?p...&postcount=22I doubt if without redoing whole PCB (Paula area) you get silent (low noise floor) audio path - ferrite beads shall "brake" continuity between digital and analog grounds (as a rule analog ground is connected to noisy digital ground in one, strictly controlled point trough some type of filter - usually ferrite bead) - for DIL package it is slightly easier for PLCC package or use dummy PLCC package and apply PLCC socket with custom PCB (also resolder Paula) as a piggyback or solder PLCC socket and use dedicated PLCC plug.

[Mick]

Cheers. The R309 resistor is replaced with the regulator whose input/output capacitors should filter a lot of the noise out and I'd assume that the 12V rail is much less noisy anyway. I've ordered some PCB to try with, YOLO.

[Mick]

Why did Commodore use resistors at R301/302/309 instead of ferrite beads? was it a cost thing or is there an advantage to it?

[pandy71]

Originally Posted by Mick

Why did Commodore use resistors at R301/302/309 instead of ferrite beads? was it a cost thing or is there an advantage to it?

My impression - perhaps wrong:

I bet lack of/insufficient knowledge - PCB layout obviously was not focused on analog signals integrity - high quality audio PCB layout was started with CD audio players and other modern audio sources using digital to analog converters etc. Something that today is very obvious, 30 yrs ago was uncommon...

But of course i can be wrong on this - doubt if there is significant cost advantage to use resistors but if you compare CDTV and CD32 circuit and PCB design then it begin to be quite obvious - also CD32 was ranked way higher than CDTV in audiophile magazines (like German Stereoplay)...

So from my perspective simply audio from Paula was never considered as deserving to be considered as requiring high quality approach.


Btw - Paula is very demanding DAC - 8 bit DAC is nothing special but way how volume control is implemented push demand on speed very high - you need at least 36MHz bandwidth (10 times 3.58MHz) to keep PWM pulse as much as possible free from time distortions (slew rate, large signal response are very important). Side to this isolate Paula as much as possible from digital power noise and ripple - hard to achieve on generic Amiga PCB layout - doubt if even this problem was addressed in modern Amiga PCB recreations - they probably don't care on this focusing on replicating primary PCB approach (similar way of thinking - audio is just audio - 8 bit PCM who cares...)

[Mick]

Very interesting.

Also, given what I've learned I've just had another look at this video:
[ Show youtube player ]

The conclusion he gives is that with 1.5K resistors in the first op amp stage the gain is set too high but the oscilloscope shows a 10V peak to peak (something around -2.25V-7.25V which with the 2.5V vref Commodore used is about right) which is well within the -12V-+12V rails. It seems like during the period of releasing A600 to A1200 2B Commodore tweaked the audio circuit and reduced the audio amplitude from 10V peak to peak to 4V peak to peak but isn't a higher amplitude actually better? could the distortion have been caused by the limitation of the op amps like you have suggested? so would fitting better op amps and restoring the 10V peak to peak (or even more) be preferable to the 4V peak to peak seen in the later A1200 2B?

[pandy71]

Higher voltage provide higher SNR but also it is not compatible with consumer levels - expected audio level in SCART is 5.65Vpp RMS (based on EN_50049-1) so probably they (CBM) realized too high level (another point to support my previous assumption about not considering audio as highly important topic) probably based on people complains and introduced changes to keep output level within EN_50049-1 specified range.

[Mick]

Ah right, bit of a waste then that they had 24V of swing but in the end only used 4V, they could have just used a single supply op amp like many others did and like they did in CD32. Which brings me to another question... what is the purpose of R320/R330 in the CD32? path to ground for DC noise? sorry for bombarding you with questions.