Hi-stability oscillators use proportional oven control techniques such 
as a copper container that has a heater coil and sensing. Then the ckt 
is inside with some insulation. Very high quality variations of this do 
two containers with separate heaters and sensing – one inside the other 
with insulator between the containers. The idea is to have high thermal 
conductivity containers separated by insulation to reduce gradients and 
thermal transients. A cheaper variation would be to use aluminum pipe 
rather than copper – not as good in the thermal mass dept., but still 
pretty good thermal conductivity. There are lots of examples of these in 
the amateur radio community – see issues of QEX for several examples.

A very simple way to reduce air current variation is to put your circuit 
in a small, open-top box and pour in clean, dry, sand. I like #30 silica 
sand (the fine, white sand you see in hotel lobby ash-tray cans), about 
$4 for a 50 lb. sack at Home Depot. Also helps with microphonics – if 
you have them. Just pour out the sand to work on the circuit.

Regards,
Charles Patton

ChrisAtUpw@....... wrote:

> 
>In a message dated 10/03/2005 16:11:50 GMT Standard Time,  jpopelish@........ 
>writes:
>
>Jack  Ivey wrote:
>  
>
>>Bret Nordgren wrote:
>>    
>>
>>>Another factor that you  may want to consider is thermal variation.  At 
>>>      
>>>
>very
>  
>
>>>low  frequencies, below 1Hz, the effects of micro-variations in the  device
>>>temperature can add additional "noise". 
>>>      
>>>
>
>  
>
>>I've  seen this effect with thermocouple amplifiers, where moving your hand
>> near the circuit would move the air enough to create low-frequency  noise.
>>It can be almost eliminated by pressing the circuit board  between pieces
>>of foam rubber.
>>    
>>
>
>It also helps a lot to keep the  internal temperature rise of the front
>end opamp to a minimum.   Reducing the opamp supply voltage as much as
>possible without degrading the  performance of the amp helps keep the
>chip cool and reduce the thermal  effect of changes in air currents.
>
>For this reason, if two amp choices  have similar noise specs, but one
>may be operated at lower supply voltage  or draws less supply current,
>its lower self heating may allow it to out  perform its hotter
>competition in the low frequency  realm.
>
>
>
>Hi John,
> 
>    Assuming that you are using a 16 bit ADC with a  range of +/-10V, one 
>count is 305 micro volts. Normal amplifier gains can result  in very significant 
>count drifts with temperature unless great care is taken in  the design and 
>construction.
> 
>        There are two different  factors operating here. One is the 
>temperature sensitivity of the opamp input  circuit in micro V / C Deg. Remember that 
>this relates to temperatures  on the IC chip itself, so it is effected by the 
>chip dissipation.
> 
>    The CAZ type opamps have very greatly reduced  thermal input drifts and 
>1/f noise.
> 
>    The other is the signals derived from external  thermo electric junctions 
>and are rarely less than a few micro V / C Deg.  These can be between the 
>chip header and the socket or the wiring,  or between cables and the input 
>clamps, or even between different cables or  connections. You will see differences 
>across the circuit board, if there is a  thermal gradient across it.
> 
>    Some resistors, like the metal oxide types,  generate high EMFs if there 
>is a temperature difference between the two  ends. Don't even try to use 
>carbon resistors, either composition or  film. 
> 
>    It can be an advantage to stick a strip of soft Al  or Cu to the top, or 
>even to both sides, of the input amplifier chip and bolt  this onto the outer 
>Al Screening Case. Another alternative is to use double  sided circuit board. 
>This greatly reduces temperature variations across the  board. You can bolt a 
>Cu chip cover strip onto the board. This is  preferable to trying to reduce 
>the dissipation by reducing the supply voltage.  Having said this, it may be 
>desirable to use separate IC regulators for the  input opamp supply, to give low 
>noise and drift and high AC supply  rejection. The first amplifier does need 
>very good supply noise decoupling. 
> 
>    Seismometer amplifiers often have two distinct  gain stages, with a high 
>pass filter set to maybe 20 to 30 sec in between.  This will greatly reduce 
>thermal error signals and 1/f noise at the output.  For geophone circuits, the 
>filter maybe set to 1/10 the resonant  frequency.
> 
>    The seismometer amplifier case is preferably made  of metal and earthed. 
>It should be kept dry, screened from drafts and any  temperature variations 
>should be minimised. It can be an advantage to fill  the case with glass wool to 
>inhibit convection.
> 
>    You might include the LF412 for  second amplifiers. They have quite low 
>drift.
> 
>    The INA118 is very useful as a low  noise true differential input opamp. 
> 
>    For information, noise calculation and  selection of your photo diodes 
>See 
>_http://usa.hamamatsu.com/assets/applications/SSD/photodiode_technical_information.pdf_ 
>(http://usa.hamamatsu.com/assets/applications/SSD/photodiode_technical_information.pdf) 
> 
>    Regards,
> 
>    Chris Chapman
>
>  
>

__________________________________________________________

Public Seismic Network Mailing List (PSN-L)