[time-nuts] Thunderbolt Fan mounting

[WarrenS]

Mark

Although Lady Heather's temperature controller can hold the temp at the TBolt's sensor constant to delta 0.000x deg over 72 hrs, 
this does not hold the temperature at the oscillator close enough to maintain an exact constant frequency because the TBolt's sensor location is not at the OCXO.
The change of the Oscillator's open loop freq vs. room temp can be seen when the room temp is also plotted and compared to the DAC voltage in an extended length LH plot.

What Lady Heater does is plenty good enough for most, but if you want to be even more nuts (and who wouldn't), 
there are still further improvements possible that can eliminate all of the TBolt's OCXO sensitivity to small room temperature changes.

1) One way to improve LH's temp control, which is not very practical and not recommended, is to reposition the TB's temp sensor.  :(

2) There are several ways that Lady Heather could be modified, to change the control loop set point a little as a function of room temp.  

3) The simple mechanical way that I use to minimize any remaining variation due to room temperature changes when using LH's temperature control loop
 is to adjust the position the LH controlled fan to compensate for the difference between the TB's sensor and it's oscillator.

Shown in the attachment,
I placed the stock Tbolt in a tight fitting foam lined small box where only the top case of the TBolt's is exposed. 
This minimizes temperature gradients on the case and causes the TB's temperature to rise a safe amount.  
I use a small 12v, 1W fan loosely mounted to the top to a 1/4 inch sheet of aluminum plate that sits on top of the TBolt. 
For course adjustment, I change the position of the aluminum plate on the TB. 
For fine adjustment, I change the location of the fan a little so that it is blowing air at a different spot on the plate.

Not shown in the attachment, is a small insulator / deflector that only allows fan air to blow on the aluminum plate, not on the TBolt's exposed upper case.
This insulator and the foam forces most all of TBolt cooling to go thru the 1/4 inch aluminum heat sink plate.
I can then adjust the fan position so that there is no visible effect on the TBolt's OCXO due to small daily room temperature changes.

The power supply I'm using is stable enough that I have not seen any TBolt changes when the PS is heated with a hair dryer or the line voltage is changed by large amounts with a variac.

ws

**********************

Mark posted:
I tried putting a Tbolt in a small (six-pack sized) Coleman cooler.   The temperature rose above the alarm temperature...  it does not take much insulation to cause problems.

Lady Heather's built in temperature PID controller works very well.   When properly set it up,  short term variations can be well under 20 millidegrees.   Long term RMS temperature error can be a few microdegrees or less.   Attached is a screen dump showing a 0 microdegree RMS temp error over 72 hours!

Besides the TT command to set the desired setpoint temperature,  there are some built in PID parameter commands (KW sets a slow pid,    KM sets a fast pid,  KA attempts a PID autotune...  you can also tweak the various PID parameters individually...  see the routine edit_pid_value() in heathui.cpp for some idea of the available parameters).   

For best performance,  it helps to have the power supply in the temperature controlled box.  This will minimize the effects of temperature variation on the supply,  which can be a third of the overall system temperature sensitivity.   It is also a good idea to not have the box so well sealed that the unit overheats if the computer/PID/fan shuts down.  You probably don't want a flammable box in case your cheapo Chinese power supply bursts into flame.

For best operation of a new Tbolt,  you should first run the 48 hour precision survey,  then execute the &a auto-tune command.  Autotune sets the oscillator parameters,  elevation mask,  and signal level mask to time-nutty values.  Before running &a,  first set the elevation mask to a low value and collect data for a couple of hours.   This lets the program find the satellite elevation angle where the signal starts to degrade.

Note that it can take several weeks for an old, unused oscillator to fully stabilize and age in.

            
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TBolt-Fan.jpg
Type: image/jpeg
Size: 62121 bytes
Desc: not available
URL: <http://www.febo.com/pipermail/time-nuts/attachments/20120712/9aa84ffa/attachment-0001.jpg>