[time-nuts] Tbolt temperature Control

WarrenS warrensjmail-one at yahoo.com
Thu Aug 27 02:53:01 UTC 2009


Neville

My responses to your well stated and mostly accurate 2c worth, 
even though much of it does not apply to this situation.

> I cannot remember the term in a PID equation that accounts for time delay.
Yea, pretty hard to remember, It called "Delay" on Smith predictor type of PID. 

Sounds like you also missed where I wrote "with a MODIFIED PID TYPE of S/W controller".

> Thermal control must face some nasty delay functions
Turns out for this application, a pretty basic type of PID works very well because the sensor 
has very high resolution and low noise, perfect for a large derivative term 
and the main temp TC is very slow and very dominate. 
No true delay going on just some slower faster Time constants that don't effect much.

 >It is the wrong algorithm for controlling almost any heater!
Probable get some disagreement from any OXCO manufacture on that.

> It is excellent for driving pens in pen recorders where mass,  ...
In fact a basic PID is not excellent for controlling much of anything, there are much better ways, 
BUT It can be adjusted to drive just about anything, which is one reason it is used so much.

>I would suggest controlling the air temperature around the case.
I agree that is the standard way and would work fine. 
However there are many advantages to heating  the case and using the internal sensor instead.
Some of which are Lower power, faster warm-up, less parts, it gets rid of that BIG delay TC you were referring to, 
does not need a fan, does not need a box, etc, etc.  
One of the things that makes the internal sensor control work 
so well is its better than 0.001 deg resolution.
  
>by  controlling the air temperature you will control the entire case of the TBOLT. 
Not necessary so unless until you add your 'local tornado of air" Don't need that with the case heater.

> (controlling air temp) will have much faster response
You may want to re-think the way you said that statement because it is VERY wrong.

> Air controls the case of the OCXO, voltage regulators and not just the temperature sensors.
In some situations controlling the internal temp sensor instead of  air or case temp will work better.
In this case it does not matter much, either are plenty good enough.
EXCEPT if the low resolution sensor is in the Tbolt like in all the new ones, 
in which case its better to keep the sensor constant so that 'Bad stuff' does nopt happen when it takes it giant steps.

> your best control algorithm uses the inverse transfer function of the thermal system to predict a somewhat tardy response.
If you want simple and 'good enough', a simple PD works fine here.

Bottom line
You should try it, It's easy top assemble and  works good. If you need any help or have any problems with it let me know.

>  There, I have had my 2c worth, cheers, Neville Michie
Thanks, and Now I have had my 2c also, and double cheers to you

Have fun,
ws

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

"Neville Michie" namichie at gmail.com

> Hi,
> your problem should be to adjust to physical reality, then you may
> be able to control you TBOLT temperature.
> I cannot remember the term in a PID equation that accounts for time delay.
> In fact there is not one! It is the wrong algorithm for controlling almost any heater!
> It is excellent for driving pens in pen recorders where mass,  velocity and position
> are the terms.
> Thermal control must face some nasty delay functions, like Gauss's error function.
> When the temperature error signal is delayed, as in the TBOLT, your best control algorithm
> uses the inverse transfer function of the thermal system to predict a somewhat tardy response.
> An alternative method that works very well is to create an isothermal  wall around the TBOLT
> which is held at constant temperature. I would suggest your light  bulb and a small
> computer fan to create a local tornado of air around the TBOLT and by  controlling the
> air temperature you will control the entire case of the TBOLT. This will have much
> faster response and may control the case of the OCXO, voltage regulators and not
> just the temperature sensors.
>  There, I have had my 2c worth,
> cheers, Neville Michie
> 
> 
> 
> 
> On 27/08/2009, at 5:56 AM, WarrenS wrote:
> 
>>
>> This can also be made to work on the TBolt units that have the 1/2  
>> deg course Temp sensor.
>> The control loop gets a bit trickier, but it can hold the Temp very  
>> close at one its course transitions points.
>> This would come in handy for those not wanting to change their  
>> sensor to the high resolution type.
>>
>> Another useful feature to add to a completely independent micro  
>> controller so that it needs no other inputs is to have its program  
>> smart enough to automatically and slowly readjust its set point  
>> control Temperature to just above the highest day to day Peak temp  
>> it sees. Needs no extra inputs, It just needs to monitor its own  
>> PID loop to see if it even comes out of control due to excess temp.
>>
>> Note you do not have to pass anything thru or output any RS232,  
>> just the ONE bit  if the micro is going to control the temp out bit.
>> It just needs to pick up the Tbolt output data in parallel and  
>> capture the Temp info.
>> If the micro is not doing the Temp control then there is no need to  
>> have the extra micro.
>>
>> ws
>> ********
>>> Don Latham djl at montana.com
>>> Wed Aug 26 17:38:21 UTC 2009
>>
>>> Heck, Warren, I'll put it on the line :-). Use a very simple and  
>>> cheap
>>> processor such as a Picaxe, pass the rs232 through it to Lady  
>>> Heather etc,
>>> and capture  the temperature, using it appropriately. Should work  
>>> nicely
>>> and keeps the temp control local. Even put a little LED on it to  
>>> indicate state.
>>> Don
>> **************
>>> TBolt Nuts
>>>
>>> To keep My TBolt's temperature constant so that the environment has
>>> minimal effect on it,
>>> I use an aquarium temperature controller (modified to have low  
>>> hysteresis)
>>> connected to a low wattage light bulb, placed in a box with the  
>>> TBolt.
>>> It works OK and keeps the TBolt's temperature constant to well  
>>> under 1 deg.
>>> To do it better and make it  more hi tech,
>>> I'd like to have the temperature control based on the TBolt's  
>>> internal sensor.
>>>
>>> What I have found works well is to use the TBolt's own RS232  
>>> temperature sensor output data,
>>> and with a modified PID type of S/W controller, turn an external  
>>> heater &/or cooler on off.
>>>
>>> The heater can be an appropriate power resistor or transistor  
>>> dissipating up to about 4 Watt,
>>> mounted to the TBolt case. What I use to cool the Tbolt up to 5  
>>> deg C, is a small fan blowing
>>> at a heatsink mounted on the top of the TBolt's case.
>>> Turning the fan on & off with a S/W driven switch, can be used to  
>>> keep the TBolt's
>>> internal temperature very constant over a limited external  
>>> temperature range.
>>> A standard PC chip fan & heatsink may be OK, if it does not add Phase
>>> noise due to it's vibration.
>>>
>>> Because of the long time constant and slow response of the internal
>>> temperature sensor,
>>> a single digital on-off bit, updated at a max rate of once per second
>>> works great for control,
>>> No analog needed. To keep the hardware and interface circuit simple,
>>> I'd like to be able to use one of the unused standard RS232 outputs,
>>> such as RTS, CTS, DSR, DTR  as the heater/cooler control bit(s).
>>> This is no problem when doing this in a DOS program or from an added
>>> microprocessor
>>> that monitors the Tbolt's  communications,  But the question is,
>>> can it be done in Windows in such a way that a modified existing  
>>> program such as
>>> Lady Heather or Tbolt monitor could control an already existing  
>>> readily available digital bit?
>>>
>>> Being a control person, Doing a software algorithm is the easy part.
>>> Making Windows do any kind of non standard I/O control, is way  
>>> above my capability.
>>> I'd like to get feedback from a Windows expert if there is a  
>>> simple way to control an existing Digital bit
>>> that would be available on a PC being used in a typical setup that  
>>> is used to monitor the Tbolt.
>>> One way I have heard suggested is to use the sound card output,
>>> but I'd like to keep it even simpler than that, Any suggestions?
>>>
>>> If anyone is interested in developing a program to make an  
>>> existing stand alone micro
>>> or basic stamp to include this function they can contact me off  
>>> line for some sugestions.
>>>
>>> Thanks,
>>> ws
>>> ******************
>>




More information about the time-nuts mailing list