Horrible Audio Clipping

[JonBrawn]

We are all aware that the audio on the VIC suffers a bit from clipping. I've made some graphs of this clipping in action - it's really quite bad!

The bit that surprised me was that you get clipping with just one channel active if you turn the volume up all the way!

[DarwinNE]

Interesting. What are you measuring, exactly? The ouput DC voltage? Where, exactly on the schematic? In this version, there is a CD blocking capacitor on the ouput, C11:

http://www.zimmers.net/anonftp/pub/cbm/ ... on%20E.pdf

Or is it an AC measurement? RMS, peak or peak-to-peak?

[JonBrawn]

Interesting. What are you measuring, exactly? The ouput DC voltage? Where, exactly on the schematic? In this version, there is a CD blocking capacitor on the ouput, C11:

I'm measuring the DC voltage at the pin of the VIC, whilst there's a TV connected to the AV DIN socket. This is so I can reproduce the horrible audio behaviour of the VIC in the FPGA I'm working on.

The graphs were produced using a limited number of data points, so I'm going to fill out the table (256 entries! Yippee!) and then graph it in different ways to see if I can refine my "acoustic model" to sound less like a kazoo being put through a meat grinder. I could use a simple lookup table, but that would use 32kb of FPGA memory, and I'm not certain I can afford to use that much.

Each voice has three output states:

• Disabled (doesn't contribute anything to the output signal)
  Low, which drives the output signal lower by a modest amount
  Hi, which drives the output signal higher by a less modest amount.

There's a steady DC offset of around 2.4V, and the volume control adds Vol/10 volts to the output regardless of what the voices are doing - even if they're all disabled. It the 0..15 value of the volume control is used as a multiplier of ((number of high outputs * high scale factor) - (number of low outputs * low scale factor)), so the final sound equation I WAS using was

2.4 + vol/10 + nHI * HIscale * vol - nLO * LOscale * vol

This sounds awful. Hence my desire to map the whole range, graph them all and scrutinize what I'm doing wrong (because I am doing this wrong - it makes sounds at the right frequency, but the clipping is badly wrong, so everything coming out of the speaker is clipped. I'm fairly certain that my fixed-point arithmetic in the FPGA is wrong, plus the scaling of "output voltage" to "value to poke into the DAC" might be another issue.

[DarwinNE]

Ah, now I see better what you mean.
In any case, if it can be useful, in the past I tried to explore a little the sound possibilities of the VIC20 for the game intro of Cavern Explorer. I noticed that everything was badly clipped if the volume was too high:

https://www.youtube.com/watch?v=JvhkRBfTAec

(sound starts at 1:30, the interesting bit is when the percussions start at around 1:50)

For this type of things, the maximum volume is about 5, everything above yields some clipping.

Did you check with an oscilloscope to observe the clipping in action? Do you need some measurements on a PAL device?