Hissing and whistling in the STe audio output

[sandord]

What is on FCLK ?

As far as I can tell a 250 KHz block wave. I think the filter is by default configured to a rate of 100/1 so that makes 2500 Hz... :o :shock:

[sandord]

I found a much better (and more scientific) FFT program: http://www.sillanumsoft.org/download.htm

This is what I measured on the MF4 at Filter In (pin 8):

before-filter.png

And Filter Out (pin 5):

after-filter.png

First I thought that the spikes around 3 KHz on pin 8 somehow translated to the 2.6 KHz spike on pin 5 but frequencies don't just move around now do they? So they must be unrelated, the 3 KHz spikes are probably filtered away and the 2.6 KHz spike is introduced from the clock.

The datasheet says that the filter can run in self-clocking mode. Perhaps we can try if that would lead to acceptable results. But this is where my knowledge is rather lacking and I don't want to start ripping my STE apart :roll:

[sandord]

What is on FCLK ?

Here's the scope reading of CLK IN (I figure that's what you meant when you said FCLK).

DS1Z_QuickPrint1.png

[sandord]

What is on FCLK ?

Oh cr*p, I looked at the schematics again and saw FCLK going to pin 2 (CLK R).
Here's the scope shot, although the signal is not so different from pin 1 (CLK IN).

DS1Z_QuickPrint2.png

[exxos]

ahhh, well you could try just lifting the output resistor of U505 (looks like 10K R503) lift the left side disconnecting the filter and see what you get on the main audio out then (RCA outputs )...

[exxos]

Think I am gonna have to look at that pdf when I am actually awake.. I mean pin 1 on the STE is left floating when its got a clock in, you just don't leave gates floating... I guess you could try a 10K to 5V or 0V on pin 1 and see if that changes anything.. need to look at it more though..

I'd also be tempted just to lift the 10K resistor off the output and wire link it to the input of the MF4, basically bypassing it to see what happens.. I don't know why they need that part, I mean you can make a simple LPF with a LC network anyway.

[sandord]

Think I am gonna have to look at that pdf when I am actually awake.. I mean pin 1 on the STE is left floating when its got a clock in, you just don't leave gates floating... I guess you could try a 10K to 5V or 0V on pin 1 and see if that changes anything.. need to look at it more though..

I'd also be tempted just to lift the 10K resistor off the output and wire link it to the input of the MF4, basically bypassing it to see what happens.. I don't know why they need that part, I mean you can make a simple LPF with a LC network anyway.

Apparently, the MF4s are used as anti-aliasing filters. The external clock is changed according to the sampling frequency the STE DMA sound is currently playing at. This could mean that the 2600 KHz spike could change to another frequency when playing DMA sound at a different frequency (the STE only supports a handful of different sampling frequencies, up to 50 KHz). This would be be an interesting effect to observe as we've only tested with the STE sitting idle.

Lots of further details can be found here: https://info-coach.fr/atari/hardware/STE-HW.php#fig2

[exxos]

From the datasheet (from what I remember last night) the clock was to set the roll off frequency.. So the clock is probably fixed.. So if you have 250KHz and we had /100 used, then thats 25KHz roll off which makes sense.

[sandord]

From the datasheet (from what I remember last night) the clock was to set the roll off frequency.. So the clock is probably fixed.. So if you have 250KHz and we had /100 used, then thats 25KHz roll off which makes sense.

Well, the page I linked to, mentions:

Finally the filters carry out an anti-aliasing (removal of the staircases on the sound curve) with the assistance of a low pass filter whose frequency is 40% of that of sampling that is 2503, 5007, 10013, and 20026 Hz.

[exxos]

Well, the page I linked to, mentions:

Finally the filters carry out an anti-aliasing (removal of the staircases on the sound curve) with the assistance of a low pass filter whose frequency is 40% of that of sampling that is 2503, 5007, 10013, and 20026 Hz.

The "stair step" isn't relating to the filter clock exactly.. the stair step steps will be smaller the higher the frequency is.. I don't know how it would clean that up exactly, some inherent smoothing somewhere. Sample playback speed is the stair step part, the clock to the filter sets the roll off frequency... different things.. Really the roll off is to remove the HF content which can burn up op-amps or even speakers, add bad harmonics etc, so its filtered out.

I would still suggest you bypass the MF4 and see how it affects things.. we have to start somewhere..