8701 replacement dot clock issues

If the source frequency is stable, then the Dotclock should also be stable, as they are tied to each other through a phase comparator/PLL circuit (that's what the two counters on the design are for).

Before we dig deeper into this, please list the accessories that are connected to your computer. It's a PLL, so there is a voltage-controlled part to it, and if the source voltage is fluctuating, the phase comparator might fail to sync under certain circumstances.

Also, I'd be interested in the method of frequency measurement that you're using. After all, it's a crystal that is used as a source, and the fundamental frequency is 17.734475 MHz and that should not deviate more than about 900Hz.

I have no idea if the Rigol is really a frequency *counter*, or if it just measures the time period between slopes and calculates from there. My Agilent scope is doing exactly that, with similar deviations that you're showing. If I use an actual frequency counter (HP 5334B), I can see much more accurate readings. The second symbol from the top left indicates that your scope is also only taking the two slopes and calculates from a single cycle - oterwise the "1/<->" wouldn't make sense. So no, your scope is not a counter, but it can give you a rough indication of frequency within an error margin that you have to be aware of.

Dot clock is generated using an actual PLL, which is different from the 8701 chip, which uses hard delays for generating the "divide by 18, multiply by 8" function. Since a PLL has a VCO, it is certainly depending on a good power supply, and I have commented on the Ray Carlsen solutions a number of times - with the result that his fanboys have attacked me viciously. A lot of the points I've been making against the bad Amiga PSUs out there also apply to Ray's creations, with one big difference: His PSUs haven't caused damage so far, and they usually also don't cause (intermittent) malfunctions. Still, they are not in line with Commodore's specifications, so I'm still taking them with a grain of salt.

Back to the measurements: 17.7337MHz is within 100ppm of the target frequency, which is 17.734475MHz, so no problem there. I'm sure you have perfect colour with this.

If the VCO overshoots on one cycle, it will undershoot on the next. Correction pulses are given at a rate of about 2MHz. This is expected behaviour of a phase comparator-controlled VCO. The deviations you get with a PLL are probably even smaller than the deviations that the original 8701 generator produces.

So in all honesty, I have no idea what you're asking for here.

Right now, the original "brick type" is the best you can get. It's rated 1.7A on the 5V rail and provides sufficiently-low ripple. We are planning a similar approach that we've made with the CA-PSU for the Amiga, so you can safely power the old computers, and increase efficiency to a level that today's laws require.