[time-nuts] V refs

[Richard Moore]

RE comments from Bruce --

> Message: 1
> Date: Sun, 30 Nov 2008 10:33:36 +1300
> From: Bruce Griffiths <bruce.griffiths at xtra.co.nz>
> Subject: Re: [time-nuts] V standards
>
> Richard Moore wrote:
>> Brooke, I read the patent with interest, and I cannot see how using
>> the LM399 and AD587 together represents an improvement over just
>> using a 399 and a chopper amp for the needed gain to get 10V (or
>> whatever) out. Geller's external adjustment R chain will suffer
>> tempco problems similar to a set of Rs used to gain-set the chopper,
>> and making the chopper an integrator with a long TC should help with
>> the 399's noise.
> You need a low pass filter with a well defined dc gain not an  
> integrator.
> You can also average the outputs of several 399s to reduce the noise.
> However this requires a lot of 399s if one wants to reduce the  
> noise by
> a large factor.
> The LTZ1000 is significantly quieter than an LM399.
> Its significantly cheaper to use a single LTZ1000 than an array of  
> 399s
> with the same noise (using the same low pas filter).
>>
> Bruce
> ------------------------------

Bruce, thanks for the clarification, and also for your earlier  
suggestion on the chopper amp. I think I've got a couple of LT1150 or  
1151s in a drawer. If I remember correctly, noise should improve  
roughly 3dB for every doubling of devices, which means that 8 399s  
should yield a 9dB improvement, which is nothing to sneeze at, and is  
still pretty cost-effective. Seems like using a chopper on each one  
and then paralleling those outputs into a chopper gain stage would  
make effective LP filtering relatively easy and would further average  
out noise and drift contributions from the various discrete parts,  
which means that using 1% MF resistors, for example, might be  
practical for sub-ppm results.

I'm beginning to like this, and I can see replacing the dual zeners  
in one of my 7-1/2 digit Datron 1082s  might be a lot of fun (the one  
with the bad VF display that's going to get 7-seg LEDs instead,  
'cause I HATE LCD displays that aren't backlit -- no wonder HP went  
to VF in the 3458 -- they sure must have got tired of the complaints  
about the 3457). That 1082 shifts its reading of my Fluke 732A's 10V  
output about 8-10 ppm/year or a bit more, when I can get everything  
at the same temp for repeating the measuremement...

Nevertheless, if I can get LTZ1000s for $60 a pop, that's a much  
simpler way to go mechanically and electrically. Are you sure that  
the price you stated is for singles? I wouldn't be averse to having a  
couple of them around.

I appreciate very much the depth of both your theoretical and  
practical experience and the insight you bring to everything you  
comment on here.

Dick Moore