Re: [NTLK] Implant Speed App

From: Humphreys, David (david.humphreys_at_honeywell.com)
Date: Thu Oct 21 2004 - 05:52:16 PDT


 

> -----Original Message-----
> From: newtontalk-bounce_at_newtontalk.net
> [mailto:newtontalk-bounce_at_newtontalk.net] On Behalf Of Sonny Hung
> Sent: Wednesday, October 20, 2004 9:42 PM
> To: newtontalk_at_newtontalk.net
> Subject: Re: [NTLK] Implant Speed App
 
> If anyone has done or tried this before I would appreciate a
> chime in on this thread. I know there was the photo of 275mHz
> that was not stable but I'm looking for the next stable jump up...

From the labs of PCBman:

I have seen this topic surface time and time again and every time I see it I
cringe.

Why? Well, this method of switching clock frequencies is very poor.
Unfortunately, while the method of changing the base crystal is sound (to a
point) switching it 'raw' is not.

Take the pixsolution board. There is a reason for the additional components.
It is important to ensure that signals are switched at the appropriate time
with no bounce.

I realize that a cheapo solution is appealing but it has many drawbacks.

The clocks for the system are all derived from the basic crystal frequency.
If you change this frequency you will change all the other clocks that are
derived from this. In an effort to over clock the uP this would seem to be a
good thing. But, where you run into problems is with the other devices that
are now clocked faster.

For example, the Newton was never originally designed to run at increased
speeds. The Dram chips will fail (or operate marginally) with increased
clock rate.

What irks me is that it is a crap shoot. Some of you will replace the
crystal, see it running 'normally' and then declare it to be working. This
is a situation I wish to guard against. It won't work for everyone and those
it seemingly does work for will experience data loss or worse sooner or
later. Don't bother to regale me with stories of your perfectly working
220MHz+ Newton. You have just been lucky so far.

I will try to explain why.

The base frequency is set by a crystal (3.6864MHz). This is multiplied by
various amounts to give the system clock, memory clock etc. etc.

Any slight change to the base clock will alter the others by a much greater
amount.

The memory chips (in fact all the chips) have a speed rating usually
expressed in nS. Trying to go faster than this will cause errors.

The base clock provides the cadence for the whole system. Switching crystal
freqencies on the fly requires waiting for the next clock cycle before that
action. If you switch at the wrong time there will be an incomplete clock
cycle and all manner of problems can arise.
They can range from a total reset to more subtle problems like an incomplete
write to memory. So, just because you may not have seen a problem doesn't
mean it isn't waiting in the wings.

A certain eBay seller offers this modification as a bonus on the Newtons he
sells.

It is NOT the bargain you might think.

I realize that the following might sound like a sales pitch but it is on
topic.

I have a solution for this. It is called 'Gearshift'. I have mentioned this
before. I have designed a circuit that has NO CRYSTAL and switches
glitchlessly (is that a word??) between normal and accelerated speeds.
Furthermore, it requires no user intervention. What this means is that the
Newton will run at accelerated speed until a function that requires normal
speed is invoked. It then throttles down until that function is completed
and switches back to turbo. Nice ehh? The benefits of an accelerated Newton
WITH serial port compatibility.

Now this is concept and prototype at the moment but it works thus far and I
just have to make a decision whether to make it stand alone or incorporate
it into a 'Newton mega board' incorporating other goodies.

Sonny, as far as the crystals go I will save you some time and tell you that
5MHz is the limit. Any faster and the memory will act flakey.

Regards,

PCBman

-- 
This is the NewtonTalk list - http://www.newtontalk.net/ for all inquiries
Official Newton FAQ: http://www.chuma.org/newton/faq/
WikiWikiNewt for all kinds of articles: http://tools.unna.org/wikiwikinewt/


This archive was generated by hypermail 2.1.5 : Sat Oct 23 2004 - 16:30:02 PDT