XI and XO lines, Overclocking thoughts, are they any good?
rwgast_logicdesign
Posts: 1,464
Ok so when I first started out I refused to buy a premade propeller board, I studied stuff for a month or two as this is the first electronics design Id ever done. One thing I wanted was the ability to OC as much as possible. With alot of reading about decoupling, and OC threads on here by cluso and sephia I came up with a design which works well it may be overkill though. I run two 3.3v regulators, one is specially for the prop only, the other is for eeprom and everything else. I layed that LDO out as the closest one to the prop. I then put .1uf and 10uf ceramic caps on both of the props power input lines, and a 33uf tant on the crystal side. It works fairly well ive been able to use a 14mhz xtal on some prop chips.
Well ive been thinking about laying out a new board, this started becuase i dont like the system im using to run 3/5v supplys to plugin board i make, I just threw two screw terminals on and run wire for lets say the vga module to the 5v screw terminal on the board. I would like a system kind of like the spin board used.
Anyways thinking about building another board had brought up two questions id had in my mind for a long time and they have to do with clocking the chip. The first being is if I used a switching regulator to power the propeller chip instead of an LDO could that help bring transients down and smooth out the power even more? Would it be worth it if so? Switching power supplys arent an easy task and prebuilt module setups are aa whole lot more than an LDO. Let get this out there worth it to me is bumping from 112mhz to 120!
Secondly I had been contemplating how cool it would be to change the propellers clock on the fly. For this i would need some kind of variable frequency generator to feed the XO and XI lines. But what i dont really know is if the clock lines need square waves or AC waves. I was under the impression a crystal oscillated an AC wave, but when I stuck a logic analyzer on the xo,xi pins I got a square wave that looked like 5mhz. The next thing is what would be the best way to get started researching this. I want to adjust from 1 to 20mhz, I have a PLL generation chip on the way not for this project, just goofing with. From what I understand the pll generators can put out any frequency but require huge amount of power and get really hot the higher you go, im not sure if im mistaken or not. Im also looking to get clock accuracy doing this if possible, if I could somehow make the thing generate signals as tight as a 32khz crystal that would be awesome, if not at least as accurate as the crystal we all use regularly is a must. I have a feeling the AD9951 http://www.analog.com/en/rfif-components/direct-digital-synthesis-dds/ad9951/products/product.html may be what I need, but im just not sure. The last thing that I dont really understand is when you use a crystal you need one with an 18pf load capacitance how would that translate into a circuit like this?
Im not saying im going to run out and build this tomorow Im just trying to figure out where to start researching a project like this, what parts I made need so I can do a little research on what they are and the things I would need to learn to build this. Im guessing if it were esy it would be on other OCable propeller boards instead of an xtal socket. I always lose my xtals swaping them around all the time.
Well ive been thinking about laying out a new board, this started becuase i dont like the system im using to run 3/5v supplys to plugin board i make, I just threw two screw terminals on and run wire for lets say the vga module to the 5v screw terminal on the board. I would like a system kind of like the spin board used.
Anyways thinking about building another board had brought up two questions id had in my mind for a long time and they have to do with clocking the chip. The first being is if I used a switching regulator to power the propeller chip instead of an LDO could that help bring transients down and smooth out the power even more? Would it be worth it if so? Switching power supplys arent an easy task and prebuilt module setups are aa whole lot more than an LDO. Let get this out there worth it to me is bumping from 112mhz to 120!
Secondly I had been contemplating how cool it would be to change the propellers clock on the fly. For this i would need some kind of variable frequency generator to feed the XO and XI lines. But what i dont really know is if the clock lines need square waves or AC waves. I was under the impression a crystal oscillated an AC wave, but when I stuck a logic analyzer on the xo,xi pins I got a square wave that looked like 5mhz. The next thing is what would be the best way to get started researching this. I want to adjust from 1 to 20mhz, I have a PLL generation chip on the way not for this project, just goofing with. From what I understand the pll generators can put out any frequency but require huge amount of power and get really hot the higher you go, im not sure if im mistaken or not. Im also looking to get clock accuracy doing this if possible, if I could somehow make the thing generate signals as tight as a 32khz crystal that would be awesome, if not at least as accurate as the crystal we all use regularly is a must. I have a feeling the AD9951 http://www.analog.com/en/rfif-components/direct-digital-synthesis-dds/ad9951/products/product.html may be what I need, but im just not sure. The last thing that I dont really understand is when you use a crystal you need one with an 18pf load capacitance how would that translate into a circuit like this?
Im not saying im going to run out and build this tomorow Im just trying to figure out where to start researching a project like this, what parts I made need so I can do a little research on what they are and the things I would need to learn to build this. Im guessing if it were esy it would be on other OCable propeller boards instead of an xtal socket. I always lose my xtals swaping them around all the time.
Comments
http://www.silabs.com/support%20documents/technicaldocs/si5351.pdf
This can generate from 160MHz to 10KHz, using two PLLs and 3 dividers, to 3 pins.
You can buy a Eval board, but not that cheap as Eval boards go... SI535X-TMSTK
Edit: Another alternative, if you want cheap bench use, ahead of size and precision, is to use one of the lower cost uC eval Boards, as most 32 bit uC now have reasonably complex PLLs.
I see the ST STM32F4xx claims a PLL VCO from 64..432MHz, and the LM4F120 from TI seems to have 5 PLL control regs, but always targets a 400MHz VCO, and gives no Limits - which is a little unusual.
also if i wanted the reference design is fairly simple its just a crystal ground and outputs! ive never done smt though if i were to attempt to etch a breakout board on a piece of copper clad which chip package that has 8 outputs would be the easiest to hand soldier? here is the mouser list
http://www.mouser.com/new/silabs/silabssi5351/
Gull wing packages are easier to solder, and you can sometimes find break out boards.
The 0ppm is more typical - it uses fractional dividers to give more precise averages, but you can see the exact number in their. software & can disable the fractional, if jitter matters more than tolerance.
My only beef, is SiLabs are very slow in allowing any crystal value in their software (but they do allow a ratio, which means the kludge of external maths )