PIC18 at 60MHz?

Pret · Joined: 18 Jul 2006 Posts: 92 Location: Iasi, Romania

I have the PIC18F8722 development board from CCS. It has a soldered 20MHz Quartz. At some point i wanted to make it work at 40MHz throught PLL. But this PIC does not have any divisor... so i could not get it to 10MHz and use the H4 fuze to push it to 40MHz...

But, as curiosity, i let the H4 fuse on. Basically it should not work, or... work chaotically...

As started the FW, the application seems to be working, but the USARTs had an invalid baudrate (ofcourse). Then after, i was playing with "#use delay" value. First... i tried 80000000. FW was running, but still USARTs were sending junks. Then... i tried other values: 50M, 70M... until i used 60M. With this value, all seems to be working. My application includes PPP/IP/TCP stack, SD/FAT32 driver, USARTs, RTOS and other minor modules. All seems to be working. Dont know if this is the real clock or if it's even coherent.

Of course i did not test it for months, and i know this is risky and it might be a matter of time until it crashes, or it damages the mcu... but still... there is someone who knows something about this behaviour?

Ttelmah · Guest

It is well worth realising that specifications for things like speed are usually pretty flexible. I'd guess you were running at 5v?. If so, it is worth looking at Figure 28.3, which has a formula for the supported clock rate, versus frequency for these chips. The graphs is a nice straight line with voltage, but then just stops at 4.2v. If you extend the graph to 5v, you get just over 53MHz, as a predicted operating frequency. In fact there have been a couple of magazine articles about this, with some rather more exhaustive tests than yours, and in almost all cases, the chips will work reliably well over their specified frequency 'rating'.
Historically, some time ago, I needed to use a crystal that would be about 5% 'over' the data sheet specification. Because it was for a large batch, I contacted Microchip, and they actually issued a certificate, that over the voltage/temperature range we required, they were happy to certify that the chips cold operate at this extended frequency. So they are obviously aware that there is a significant margin in the normal operation.
The area that might well be unreliable is the PLL itself. Given that it appears to be latching onto a 4:3 subtone, it is quite likely that at some voltages/temperatures, it might switch to 2:1. If you changed the incoming oscillator to 15MHz, and ran some long term tests,I'd not be suprised to hear that it was completely reliable.

Best Wishes

Pret · Joined: 18 Jul 2006 Posts: 92 Location: Iasi, Romania

Anyway, personally i do not have the guts to deliver(sell) a product with specific configurations that are out of official graphs. Especially when you don't know the environment which client will use. Now i'm using this configuration when i need to do operation with SD card (like format).... or maybe graphic LCD testing.

I'm sure that this kind of 'overclocks' have their price...