Why a 10MHz crystal?

Nick McClick

Just picked up a hydra (mostly for the book!). Why is the crystal 10MHz? Is it better to run a faster crystal & lower PLL multiplier?

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Concentrate on understanding the problem, not applying the tool

Mike Green

There's no magic to the choice of crystal for the Hydra. The Hydra was designed very early, before the work was put into characterizing the performance of the Propeller chip under different operating conditions. Based on what we know now, it would have been slightly better to have used a 5MHz crystal for 2 reasons: 1) Slightly less power use by the crystal oscillator at 5MHz vs. 10MHz. 2) Although the PLL works fine under normal operating conditions at 160MHz (10MHz x 16), testing has shown that not all Propellers will work at that speed at the extremes of the temperature range and supply voltage that the Propeller is supposed to work at.

AndreL

On the other hand, using a higher input frequency means less jitter in the scaled output, so the 10Mhz will give a cleaner signal ultimately since the slower 5Mhz has to be scaled so many times and as that input frequency drifts the drift on the pll drifts by a scaled amount. Also, the 10Mhz does allow you to over clock, but either way, that's why I put it in a socket, so you can change it and play around yourself, maybe put 8 Mhz in there if you like scaling to a power of 2, or maybe put 2xNTSC or 2xPAL in there so you can scale by integers to derive clock rates rather than make the PLL work so hard.

Andre'

Phil Pilgrim (PhiPi)

Andre',

My experience — at least with broadcast video — has been different. I get a much cleaner broadcast signal (Channel 3 - 60 MHz) at 5 MHz and pll16x than I could at 10 MHz and pll8x. In the latter case, there was quite a bit of hue drift through a given scan line that did not present at the lower crystal frequency, even though this was not apparent in baseband video. There may be factors in my setup that have thus far eluded analysis; but, so far, the crystal frequency appears to be the only variable. My initial conclusion was that the PLL syncs to a 5 MHz input more reliably and with less jitter than it can when being driven at 10 MHz.

-Phil

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Still some PropSTICK Kit bare PCBs left!

AndreL

Hmmm, I have different results, but the bottom line is this can be true for the PLL, its got a "sweet" spot, but in general you always want to start with a high frequency and multiple the least amount. But, obviously these are analog devices, and the xtal type is very important as well. If you are using a poor xtal or a xtal rated for series rather than parallel oscillation etc. then your results will be off. In some cases, using a lower f and scaling it is more accurate, but if your target is a close integral multiple of the fin then its better to start with a larger fin, so its give and take.

I think the take way is for anything analog, you have to test things, and cover the application area you are using and see what works best.

Andre'