High Speed Issues
pjv
Posts: 1,903
Hi All;
As promised in an earlier post, I did some testing on an SX28 operating at room temperature and 5 Volts to get some sense of time skew between the oscillator and the processor's operation.
Using a 50 MHz resonator as a fixed clock, or the SX-Key as a variable·frequency clock,·and toggeling·an LED·output·pin on port C, that pin then ALWAYS transitions (high or low) 12 nanoseconds after the rise of the corresponding clock edge on the OSC1 pin, regardless of clock frequency.
The rise time of the output is in the order of 8 nanoseconds, and·its fall time is about 12 nanoseconds, again, independent of clock frequency.
The risetime of the SX-Key's clock is a surprisingly fast 2 to 3 nanaseconds, also independent of clock frequency.
Running a very simple program (my small RTOS) toggeling·the·output pin high, then immediately low,·with no NOPs in between,·yielding a 1 instruction long pulse with a rep rate·of 1.25 uSec,·gave flawless operation to 70 MHz. Above that speed, a NOP needed to be inserted between the set/reset instructions, then yielding a 2 instruction long pulse. This remained the case all the way to 110 MHz.
The processor operated the pulsing loop reliably and in a stable manner·at all sppeds to the maximum of the SX-Key's oscillator frequency, 110 MHz.
The surprising thing was that even after 20 minutes at·a speed of 100 Mhz,·the SX·barely got warm. The current consumed for the SX-Key, SX28, some periperal chips·and a few leds varied from 100 mA at the slowest SX-Key setting of 400 KHz, to 200 mA at 100 Mhz. This is interesting because others as well as myself have observed much more heat at high speeds when operating more complex programs. I'm going to check into that further, including operation at lower voltages.
Cheers, and keep on SX-ing,
Peter (pjv)
·
As promised in an earlier post, I did some testing on an SX28 operating at room temperature and 5 Volts to get some sense of time skew between the oscillator and the processor's operation.
Using a 50 MHz resonator as a fixed clock, or the SX-Key as a variable·frequency clock,·and toggeling·an LED·output·pin on port C, that pin then ALWAYS transitions (high or low) 12 nanoseconds after the rise of the corresponding clock edge on the OSC1 pin, regardless of clock frequency.
The rise time of the output is in the order of 8 nanoseconds, and·its fall time is about 12 nanoseconds, again, independent of clock frequency.
The risetime of the SX-Key's clock is a surprisingly fast 2 to 3 nanaseconds, also independent of clock frequency.
Running a very simple program (my small RTOS) toggeling·the·output pin high, then immediately low,·with no NOPs in between,·yielding a 1 instruction long pulse with a rep rate·of 1.25 uSec,·gave flawless operation to 70 MHz. Above that speed, a NOP needed to be inserted between the set/reset instructions, then yielding a 2 instruction long pulse. This remained the case all the way to 110 MHz.
The processor operated the pulsing loop reliably and in a stable manner·at all sppeds to the maximum of the SX-Key's oscillator frequency, 110 MHz.
The surprising thing was that even after 20 minutes at·a speed of 100 Mhz,·the SX·barely got warm. The current consumed for the SX-Key, SX28, some periperal chips·and a few leds varied from 100 mA at the slowest SX-Key setting of 400 KHz, to 200 mA at 100 Mhz. This is interesting because others as well as myself have observed much more heat at high speeds when operating more complex programs. I'm going to check into that further, including operation at lower voltages.
Cheers, and keep on SX-ing,
Peter (pjv)
·