[time-nuts] Achievable temperature stability for Thunderboltenvironment? Answer found
dhooke at gmail.com
Sun Mar 31 22:20:29 EDT 2013
The answer is in the LH readme, and I completely overlooked it.
/k[?=#] - set temp control parameter '?'
> Hi All,
> I've setup a TBolt with a late model oscillator mounted externally to
> the case, and moved the C version temp sensor from the main PCB to the
> oscillator. With a little fan, LH is keeping the temperature within
> about 0.02dC. Amazing, given I have about 10dC ambient variations.
> How do I start LH with the PID parameters obtained during the autotune
> Which bit of which source file should I look at to figure this out?
> Thanks to Mark, John, Warren and others for LH!
> On 17/01/2011 9:49 AM, WarrenS wrote:
>> Here is a Plot of the results, (The post would not accept both
>> pictures at the same time)
>> This is the 'KISS' enclosure and driver I use with Lady Heather's
>> temperature controller that holds the temperature change to under
>> 0.01 deg.
>> I have the box just setting on top of a PC next to the a window.
>> 'KISS' = Keep It Simple and SMALL ...
>> I don't know if this low resolution picture will post correctly.
>> If anyone wants a high resolution view of the H/W,
>> or an expanded plot showing the results over a 2 week period,
>> I can provide them, if someone will send me the name of a site to
>> post them at.
>>> ----- Original Message -----
>>> From: "Mark Sims"
>>> <holrum-PkbjNfxxIARBDgjK7y7TUQ-XMD5yJDbdMReXY1tMh2IBg at public.gmane.org>
>>> Sent: Saturday, January 15, 2011 12:53 PM
>>> Subject: [time-nuts] Achievable temperature stability for
>>> Using Lady Heather's temperature controller (fan+cardboard box+solid
>>> relay+baffling and thermal mass to taste) I get around +/- 3
>>> temperature control when the AC/furnace is not running and +/- 20
>>> millidegrees with them cycling. Long term temperature average is
>>> down in the
>>> tens of microdegrees. Under ideal conditions, I have seen over an
>>> hour where
>>> the temperature sensor did not move a single microdegree! The active
>>> temperature control has a most definite positive effect on the device
>>> I place the power supply in the thermal enclosure to minimize its
>>> temperature coefficient. Yes, the thermal sensor is away from the
>>> module (and power supply) but in the semi-closed environment of the
>>> box, the
>>> thermal stability on one area is pretty much the same everywhere. I
>>> chosen my box so that if the fan stops (for whatever reason) the
>>> in the box still does not exceed 50C.
>>> For the ultra best uber performance you need to maximize the quality
>>> of all
>>> the input and environmental variables (temperature, power, antenna,
>>> location, disciplining parameters, etc) Choose your antenna mask
>>> angle and
>>> signal level threshold to minimize satellite constellation
>>> switching. With a
>>> little nutty attention to the details you can get parts per trillion
>>> performance out of the little beastie.
>>> As long as the Tbolt's Osc is being disciplined, I have found:
>>> Most Tbolt's with factory default settings (i.e TC = 100) will show
>>> effects with standard room temperature changes. No special protection
>>> A unit that is tuned a bit better (in a box and TC of 300 to 500), then
>>> temperature changes of less than 1 deg / hr will be OK.
>>> A better optimized setup with TC settings in the 500 to 1000 sec
>>> range, a
>>> max temperature rate of change of up to 0.1 deg C per hr will have
>>> If you want to go all out time nuts, (with "Special" TC setting
>>> above 1000)
>>> then best to hold the sensor temperature to within 0.02 deg total
>>> which can be done using Lady Heather's Temperature controller.
>>> How high you can go with the TC setting, depends on many things,
>>> such as how
>>> stable the Tbolt's Oscillator is. Each setup is different.
>>> A TC setting of 1000 sec is generally the max you should go. With this
>>> Tbolt, the best results can be obtained with a 'special' TC setting
>>> to 2000
>>> to 3000.
>>> Attached is the last 2 weeks of a Lady Heather plot, showing a
>>> tracking test I did to see how long it takes a Tbolt to learn a new
>>> This unit has a Poor antenna, 1 sec ADEV of 1e-12, Aging of 4 e-12 /
>>> Temp coeff of 2.5e-10 / deg.
>>> Have fun
More information about the time-nuts