On 6/14/2010 7:45 AM, Invisible wrote:
>>> It's the 6510, which is similar to the 6502 that everything else used
>>> but with a few minor alterations. It's not a C64-specific chip though;
>>> just a different model in the range.
>>>
>>
>> But, see, it isn't an off-the-shelf component.
>
> How do you figure that?
>
> MOS Technology made the very successful 6502, so they designed an
> upgraded version, the 6510. It just so happens that Commodore purchased
> the 6510 for their Commodore 64 system. How is that not "off-the-shelf"?
>

If there are any customizations to it, then it is no longer just an 
off-the-shelf device. I'm not sure Commodore did anything custom with 
the 6510, though


>
> Fabricating a 40 nm Intel Core i7 might require high-resolution
> photo-etching such that the fab itself is expensive. If you're just
> throwing a few logic gates together, you won't be using a 40 nm fab. So
> the fabrication costs are quite small; it's setting up the production
> line that costs $$$. And, of course, the design costs. But I would
> imagine tooling up a fab is astronomically expensive. If you aren't
> making 80,000,000 units, it would seem prohibitively expensive.
>

Yeah, OK... but still... clean room. ;) it also depends a lot on the 
size of the die. A larger die is more expensive because the yield is 
actually lower.

>>> More like, they know their clients need a shaker that's guaranteed to
>>> work repeatably, so they can charge anything they like.
>>
>> I'm sure there's a point where the cost would be more than the client
>> will accept.
>
> You wanna bet?
>


consider alternatives.

>
> Well, the newest mass spectrometer we have was purchased about 5 years
> ago. (Second-hand, obviously. We can only afford to buy the old crap
> that other companies are throwing away.) Still, I gather GPIB is
> supposedly quite popular in scientific applications. (Could explain why


>


>

Hmm, I suppose it makes sense that it would use it. From what I was 
reading it's a standard bus for instrumentation.

FWIW, one of our older machines used something called ProfiBUS, the 
interface card for it is extremely expensive ($1000, iirc) You didn't 
dare screw with the interface card.

>> You still have to have the hardware to execute the commands of the
>> software, and depending on the application it could be much more
>> reliable to have it done in hardware.
>
> If all you're doing is cycling stepper motors, it would seem that all
> you need is some memory-mapped registers and the analogue electronics to
> map a binary digit to a step signal.
>

Well... dealing with multiple sensors where the number of sensors is 
more than the number of IO ports available on the controller or drive, 
then you need to do some kind of logic outside the program in order to 
get conditions from a combination of sensors. You can't easily or 
cheaply add more IO lines than the manufacturer supplies.

Most of what we have done are servo, which is generally analog or PWM.

>> Some things safety related are not controlled by software at all...
>> Take the Emergency Stop on industrial machines, for instance ...
>
> Then again, that's an extremely simple function.

right, but there may be a wired-or situation where if the e-stop for 
example is pressed or a sensor is activated (in one case on a large saw 
there's a rope around the perimeter of the saw, if someone were to get 
pulled into the machine, they'd hit the rope and trigger an e-stop.)

-- 
~Mike

Post a reply to this message