Re: [NTLK] Knowledgenavigator

From: PCBman Service <pcbmanservice_at_ix.netcom.com>
Date: Sun Sep 07 2008 - 13:16:43 EDT

On Sep 7, , at 9:03 AM, James Wages wrote:

>> Be that as it may, if indeed you really do get a noticeable speed
>> boost
> without any sacrifices (e.g., a simple flip of a switch turns off the
> booster gets serial functions back), why doesn't every single sensible
> Newton user have one of these? Or are there other tradeoffs that
> were not
> mentioned about this accelerator in your recent dialog?
>

As a principle opponent, I could write a ton of stuff about Ferdi but
I will take the higher ground and refrain from hashing old grievances.

The simple method of switching an alternate crystal in and out of the
circuit is extremely poor design.

Here's why:

The action of switching connections from one crystal to another is
essentially a random event. One simply doesn't know what the state of
the clock is when you switch and the mechanical action of switching is
not repeatable either.

The base frequency of the crystal is multiplied within the ARM chip to
get the 162MHz clock using phase-locked loop techniques.

Simply put, each transition of the base clock yields 44 transitions of
the master clock.

When the switch is changed, one doesn't know when the contact with the
other crystal will happen, where you are on the current clock
transition or how long the crystal is going to take to start
oscillating.

These events, short as they are in our macro world are significant
durations at the micro level. If, for example, a DRAM refresh gets
delayed or shortened, data corruption is quite possible.

Messing with the base clock will cause problems eventually for ALL of
you. Just because you may not have experienced it catastrophically yet
doesn't mean it won't happen.

It may have already flipped some data bit from 0 to 1 or vice versa
and you don't know it yet as you haven't accessed it for a long while.

In principle, changing the clock frequency will work but it has to be
a little more sophisticated than an extra crystal and a switch. You
need some logic to isolate and de-bounce the switch and wait for a
clock transition so that you are sure an instruction is not taking
place.

Quite apart from all this is 'pushing the limit' with crystals faster
than 5MHz. As you increase the frequency, you decrease the amount of
time that the memory has to refresh. Due to manufacturing tolerances,
some people may be 'lucky' and have devices with wider margins -
others not so lucky. But, amongst other reasons, the 5MHz was chosen
to give an acceptable operating margin and ensure all the products
that went out the door worked.

By upping the base frequency you are moving into no man's land for
reliability.

Damage to your device, while unlikely, is always a possibility.

Also annoying, is the fact that the serial port is tied into the base
frequency too. What a pain to have to switch each time you want to use
the serial ports. Again, logic would be the answer. Monitor the power
line to the serial port. It is asserted each time the port is used. If
you do this then the speed would automatically ramped down when the
port is in use. Not only that, but the sound output would be unchanged
as it too asserts this line.

I could go on but...

PCBman

  

====================================================================
The NewtonTalk Mailing List - http://www.newtontalk.net/
The Official Newton FAQ - http://www.splorp.com/newton/faq/
The Newton Glossary - http://www.splorp.com/newton/glossary/
WikiWikiNewt - http://tools.unna.org/wikiwikinewt/
====================================================================
Received on Sun Sep 7 13:17:15 2008

This archive was generated by hypermail 2.1.8 : Mon Sep 08 2008 - 13:30:00 EDT