PokeMon wrote:Well - it depends on the quality of TV/monitor you use to display.
A 46" Sony LCD TV for me. And yes, you mileage may vary depending on your TV.
I was not happy with the transistor solutions and regardless of the speed of transistor and filter you do - this is not noise from the power supply this is voltage drop inside the ULA power lanes which is measurable and viewable at output pin 16.
Agreed about the ULA signal noise, and this is exactly what I wrote by the way...

Note that the transistor speed doesn't change a thing about the noise, but if you use a "generic" (BC547 & Co) transistor, its Ton/Toff, and most important, Ts are simply too large (up to 500ns for Ts with a BC547 against 20ns for a ZTX313 or 2N2369A), which will cause trailing effects on white to black transitions and may also degrade the sync.
Keep in mind that in the video signal 700mV is the difference between white and black and already 5-10% noise (35-70mV) can be seen on a white background.
The noise is not seen *at all* for me on the white: like I explained, the signal is in fact
a little (*) too strong (over 1V peak-to-peak) and the white level saturates the video input of the TV, so whether there's 70mV more noise added doesn't change a thing (it's just clipped by the TV circuitry together with the excessive white level).
You can reduce this with individual contrast/brightness settings but when you connect it to a 42 inch plasma TV as I do, you can surely see those thin line pattern in the background and it is true that it is depending on the CPU usage during video display and you can see when the data is delivered for a character, when it is latched and so on. These actions do slightly voltage drops on the ULA output which is not as steady as it should be for a good analog signal.
Like I explained (and shown in the photo I posted), this is only seen on the black wide areas, as dark grey vertical lines, but they are barely visible even when using an excessively high brightness setting. For the best results, you'll want to push the contrast setting to the max and the brightness to the minimum (given the saturation, even with a minimum (0 !) brightness the ZX81 white screen stays very bright for me).
And of course I am not happy to change the default/custom settings of brightness and contrast on my home TV for just the ZX81 display.

My TV remembers a different setting for each input, and the AV1 input I'm using the ZX81 onto was unused, so not a problem for me.
(*) EDIT: quite stronger, actually: 2.2 Vp-p, measured on a scope, with black/backporch level at 1.1V and white level at 2.2V... For a result closer to the norm, I could probably replace D9 with two diodes in series(**), so to lower the signal by 700mV...
(**) EDIT 2: I tested the two diodes in series trick: with two 1N4148s, I get 1.6 Vp-p and I indeed can see the noise/vertical strips in the white background, but I also get a much deeper black. With one 1N4148 and one BAT85 (Schottky diode) in serial, I get 1.8V peak-peak, the white is saturated and noiseless again, and the black is still very deep (much better than with just one silicium diode); I adopted this latter solution, which definitely gives excellent results with my TV set (but your mileage may vary, depending on the AGC characteristics on the composite video input of your TV...).