You could try changing out the crystal on the Prop BOE to get a faster system clock. Typically 6MHz, 6.25MHz, or 6.5MHz crystals would give a system clock of 96MHz, 100MHz, or 104MHz. Although all of these have been used, that doesn't mean they will always work. The higher the frequency, the higher the temperature of the Propeller chip, and the lower the actual voltage of the 3.3V supply, the more likely you are to have problems. 80MHz is what is guaranteed to work. See the Propeller datasheet for details ... there's a nice graph.
Once again, Thanks Mike. I was not aware that the crystal was removeable. I doubt I will ever have a need to icrease the speed, I was just wondering if the crystal that was on it was only capable of 80 MHz.
I doesn't look the PropBOE's crystal is removeable. You'd need to use a soldering iron to remove it.
Lots of other Propeller boards do have removeable crystals.
The PropBOE's 5MHz crystal will allow 5, 10, 20, 40 and 80 MHz operation depending on the PLL setting. You can also not use the crystal at all and use the internal oscillator's RCFAST and RCSLOW settings.
If a system is running with a 6.25 MHz crystal (100 MHz) and the processor runs cool, what other sort of problems might it encounter.
Does it make the processor more susceptible to noise or put higher demands on the power supplies ?
Higher speed means higher power demand. That's true of CMOS in general.
The issue is that the chip is designed to run at 80MHz tops. The design is very robust (conservative), so it will normally run at higher speeds. Semiconductors don't work as well at higher temperatures, so the Propeller will run reliably at higher speeds if the chip temperature is kept low. The graphs in the datasheet give some idea of maximum speed vs. chip temperature. Semiconductors also work better at higher supply voltages ... up to the point where the structures on the chip (gates particularly) get destroyed from the voltage (oxide punch-through). Again, the datasheet gives some test information. Generally, the Propeller will work in normal room air at nominally 3.3V supply voltage at 100MHz and I've not heard of problems at 104MHz. If you package the chip in a box with poor heat dissipation so it runs hot or it's mid-summer and the air conditioner doesn't work or run it with a low supply voltage, it will start to become unreliable at the higher system clock speeds. If you want guaranteed performance over the industrial temperature range and more, stick with 80MHz.
Thanks, I don't believe in loading down the I/O's heavily, probably 2 mA max on any of them. We just use transistors on a separate power supply
for the heavy loads, this also isolates the load from the processor and improves noise immunity.
Mike, Is there really a noticeable difference when running at the higher speed? 20 MHz does not seem like it would make enough difference to make it worth messing with and chancing burning it up.
I would say you arent at any risk of actually frying your chip. I have run the turbulence demo (this thing is crazy demanding) on all my prop chips at 112mhz and they stay cool to the touch. The biggest issue in overclocking failures with the prop come from an unstable PLL and noise in the power supply. Basically if you want to use anything faster than 100/104mhz your going to need really good board decoupling and a clean 3.3v supply. If you take your clock to high, your program will just run buggy or completely crash. I would say the only way you would burn a chip up is if you were feeding it more than 3.3v or source/syncing more current than recommended.
As far as the usefulness of overclocking that is a completely different issue. 100mhz always sounds better than 80mhz but in reality, an abnormal clock rate causes problems with a lot of objects. One example is VGA/NTSC, running at 100mhz will cause most of the video objects to fail because they arent in sync anymore. I would say overclocking is really only useful when you have a specific application that needs it. One example is the propalyzer logic analyzer, I run my propalyzer board at 112mhz so i can sample higher frequencies, another reason I see people over clocking is when they need to generate specific frequencies for RF experiments and that kind of stuff. Maybe you need to read from another chip at the fastest speed possible, and overclocking will help you out. One huge reason I see to actually OC would be to run Phil's color image capture stuff.
I would recommend leaving the BOE alone, unless you really have a need to OC it. If your want to play with clocking a chip to see if your code benefits from the higher speeds, the best thing to do would be to just breadboard/protoboard a dip chip and play with the crystal and pll settings, testing it out with the turblence demo.
I would recommend leaving the BOE alone, unless you really have a need to OC it. If your want to play with clocking a chip to see if your code benefits from the higher speeds, the best thing to do would be to just breadboard/protoboard a dip chip and play with the crystal and pll settings, testing it out with the turblence demo.
Yea, That's actually a good idea. Or, I could always use my QS board as it has a removeable crystal. Even though the Prop BOE has a Lifetime Warranty, I would rather not mess it up.
NWCCTV
Posts: 3,629
Comments
I doesn't look the PropBOE's crystal is removeable. You'd need to use a soldering iron to remove it.
Lots of other Propeller boards do have removeable crystals.
The PropBOE's 5MHz crystal will allow 5, 10, 20, 40 and 80 MHz operation depending on the PLL setting. You can also not use the crystal at all and use the internal oscillator's RCFAST and RCSLOW settings.
If a system is running with a 6.25 MHz crystal (100 MHz) and the processor runs cool, what other sort of problems might it encounter.
Does it make the processor more susceptible to noise or put higher demands on the power supplies ?
The issue is that the chip is designed to run at 80MHz tops. The design is very robust (conservative), so it will normally run at higher speeds. Semiconductors don't work as well at higher temperatures, so the Propeller will run reliably at higher speeds if the chip temperature is kept low. The graphs in the datasheet give some idea of maximum speed vs. chip temperature. Semiconductors also work better at higher supply voltages ... up to the point where the structures on the chip (gates particularly) get destroyed from the voltage (oxide punch-through). Again, the datasheet gives some test information. Generally, the Propeller will work in normal room air at nominally 3.3V supply voltage at 100MHz and I've not heard of problems at 104MHz. If you package the chip in a box with poor heat dissipation so it runs hot or it's mid-summer and the air conditioner doesn't work or run it with a low supply voltage, it will start to become unreliable at the higher system clock speeds. If you want guaranteed performance over the industrial temperature range and more, stick with 80MHz.
Thanks, I don't believe in loading down the I/O's heavily, probably 2 mA max on any of them. We just use transistors on a separate power supply
for the heavy loads, this also isolates the load from the processor and improves noise immunity.
As far as the usefulness of overclocking that is a completely different issue. 100mhz always sounds better than 80mhz but in reality, an abnormal clock rate causes problems with a lot of objects. One example is VGA/NTSC, running at 100mhz will cause most of the video objects to fail because they arent in sync anymore. I would say overclocking is really only useful when you have a specific application that needs it. One example is the propalyzer logic analyzer, I run my propalyzer board at 112mhz so i can sample higher frequencies, another reason I see people over clocking is when they need to generate specific frequencies for RF experiments and that kind of stuff. Maybe you need to read from another chip at the fastest speed possible, and overclocking will help you out. One huge reason I see to actually OC would be to run Phil's color image capture stuff.
I would recommend leaving the BOE alone, unless you really have a need to OC it. If your want to play with clocking a chip to see if your code benefits from the higher speeds, the best thing to do would be to just breadboard/protoboard a dip chip and play with the crystal and pll settings, testing it out with the turblence demo.
Yea, That's actually a good idea. Or, I could always use my QS board as it has a removeable crystal. Even though the Prop BOE has a Lifetime Warranty, I would rather not mess it up.