overclocking 11.0592 in H4 mode 44.2368 for 0% error@115.2kb

[treitmey]

Has anyone overclocked 11.0592 in H4 mode 44.2368 for 0% error@115.2k baud?

Spec sheet says 10MhZ x4 ie:40MhZ is max but if figured there should
be a margin of error.

This is going into production,.. but I figured that if they made it through
24hr burn in I'd be in the clear.

Has anyone tried this?
Does it seem production worthy?
How many boards(volume) is in the field? outside??

[PCM programmer]

Why not use a 9.216 MHz crystal, which gives 36.864 MHz in PLL mode ?
Then it gives 0% error at 115200 baud and you don't have to overclock it.

[treitmey]

Yes, of course your right.

Just wanted to know if this is ever done.

What volumes are in production.

Please respond if you do this.

Thanks

[PCM programmer]

I never ship any over-clocked product.

[SherpaDoug]

I would be scared to ship anything that goes beyond spec unless I had written info from Microchip that it was OK. For hobby stuff sure... but not for paying customers.
On the other hand my products all cost over $1000 US. If the paying customer is paying for a $10 teddy bear things might be different.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[treitmey]

Yes,
And the thing is, I've never seen an over-clocking failure.
And I figure if it was to fail, I'd see it on burn in. I don't think it would start off OK in burn in, then fail latter on. Just my take.
Seems not too many people are doing this.
I'll do a 20 board run and see how they go.
(my boards are $300 and have a 4 year cross shipping warranty)
(if ANYTHING goes wrong they have a new board in less than 24 hr)

Anyone else have an opinion?

[bwhiten]

I personally would not ship it to a customer if there is an in-spec alternative. What's to gain by going out of spec just because you can? Chances are you will eventually run into a lot of components that are much closer to the listed spec and all hell will break loose with you customer base.
How mission critical is this product to the customer?
Have you performed "burn-in" as you call it, over the full temperature and humidity range the product can be exposed?
How about with elevated and degraded power rails?
If you still consider this an option I would do some 4-corner testing of the product, not just one or two but a representative quantity of boards that will show if you can live with the failure rate.
Is there another reason to overclock the PIC other than the baud rate?
These are all questions I would ask an engineer working for me.

[ckielstra]

Exceeding the specifications by more than 10% seems a large risk to me. A burn in test doesn't guarantee the next batch of processors will meet the same specifications. I would ask Microchip for a written confirmation that with your voltage and temperature ranges the chip will work.

You didn't mention the processor model you are using, but would it be possible to use one of the newer models? There are PIC18 models with a maximum of 42MHz, 48MHz and even 64MHz.

[PCM programmer]

If you care about staying in business you won't ship out-of-spec designs.

[treitmey]

AND This is why I ask questions.
Thinking about the corner cases, and that what works today may
not work with next silicon,

thus I have decided to stay within the written specification.

Thanks for the input.

[meereck]

crystal 22,1184MHZ works as well

[markleman]

The problem is you could make hundreds of boards, burn them in on the bench and it still would not help you know if it will be reliable in the field. It is quite likely that on the bench at room temp with a stable supply rail and no external noise all will be well, but raise the temp or lower the supply voltage and you may encounter problems. Unless you can control the environment your end users put you boards into your will never sleep well with an over clocked part. And then Microchip may change something (a die shrink, change fab or tweek the manufacturing process) and your next batch will fail, not on the bench but in the field where the cost of replacement and loss of reputation is high.

All of which is why the overwhelming consensus in this thread has been 'don't do it'

We have to test some of our designs in an environmental chamber over -40C -> +125C, whilst in a rotating pressure vessel at 10bar pressure, half full of liquid to prove they will work in the field . I went over the data sheet in detail to make sure we did not exceed anything and got Microchip to clarify a few grey areas. Which is why I sleep well, safe in the knowledge that 10,000+ PIC are out there reliably doing the job we intended them to do.

The bottom line is "Do you *really need* 11.0592MHz crystal, won't the design work with a 9.216 MHz crystal?"

[treitmey]

Yup, got it

[Ttelmah]

As a 'follow up' comment, there are occasions, when it may be necessary to 'overclock' (but rare). I had one quite a few years ago, where the interface timing, just wouldn't work, without a tiny increase in clock speed (just over 2%). Now I realised, that data sheets, are an _entirety_. The key is that things like clock speeds are quoted for a specified temperature range, and supply range, and will almost invariably be 'stretchable', _provided you are prepared to sacrifice something else in the data sheet_. Hence I contacted the manufacturer, and asked if they could provide data of what would need to be sacrificed. They could. I received graphs of maximum clock rates versus temperature, for a range of supply voltages. Because the temperature specification needed for the product fell several degrees below the range for the chip, I was able to 'overclock', but support the decision to do so, with graphs showing what the chip manufacturer said.
Getting data like this from the manufacturer, is the _only_ way that I'd consider 'overclocking' for a commercial product. Otherwise you are just asking for 'come backs', when failures do happen, even if they are not a result of the clock rate.....
So, don't rule it completely out, but 'use with wisdom'...

Best Wishes

[ckielstra]

Hi Ttelmah,
Thanks for putting things into perspective!