[time-nuts] Oscillator temperature compensation

[Magnus Danielson]

On 06/22/2013 11:10 PM, Hal Murray wrote:
>
> magnus at rubidium.dyndns.org said:
>> On 06/22/2013 05:27 PM, Didier Juges wrote:
>>> A real treat would be to do the GPS receiver with tubes ;)
>> The the correlation channel(s) would be possible to do in tubes. The rest of
>> the processing is "problematic".
>
> The IBM 709 was tubes.  (Well, mostly, they used transistors on the front end
> of the memory.)  Memory was 32K 36 bit words.  Call it 128K bytes.  That
> might be enough.
>
> It had a cycle time of 12 microseconds.  12 ns would be 1000x as fast.
> That's 80 MHz, a reasonable speed for an ARM.  So the CPU in today's GPS
> systems is 300x to 1000x faster than the 709.
>
> Anybody know what fraction of an ARM it takes to do the GPS calculations?
>
> Do you have to keep up, or can you do the calculations for every N-th second?

I have articles desribing how they looked at the feasability of using 
RCA 1802 and 6502 as processors. Some of the early 2-channels receivers 
used a pair of similar 8-bitters if I recall correctly. You don't need 
much, but it's still a fair bit more tubes than just the bandpass 
processing would require which is mostly shift-registers, counters and 
some integrators.

Cheers,
Magnus