[time-nuts] Temperature sensors and bridge amps

Richard Karlquist richard at karlquist.com
Sun Nov 14 05:03:28 UTC 2010


 I understand what you are describing, however, the E1938A
 (I typed in the wrong number on the original posting) is
 a true single oven.  The "double integrator" is just a
 mathematical function in addition to the proportional,
 integral, and differential functions that are normally used.
 The double integrator improves transient response to
 the extent that a typical environmental chamber ramping
 as fast as it can go is not enough to get a significant
 temperature error.  Credit goes to Len Cutler for this.
 Without the double integrator, the oven still has a thermal
 gain in the many 100,000's in production.  (Hand built
 ones can exceed 1,000,000).    This breakthrough occurred
 one night when I was working late and hit upon the idea
 of adding an additional heater winding around the rim
 of the "hockey puck".  The first try was worth over an order
 of magnitude, and we later got another.  I agree that we are 
 mainly after stability.  However, another interesting 
 feature of the E1938A is that we carefully set  each oven
 at the exact turnover.  Unlike 10811 crystals, all E1938A
 crystals have turnovers.  So accuracy matters to some 
 extent.

 You mentioned insulation.  You cannot highly insulate
 an oven oscillator because you have to get rid of the
 overhead heat.  In any event, the key is getting rid of
 heater induced thermal gradients.  These are not caused
 by insufficient insulation.  See my 1997 FCS papers for
 additional details.

 The E1938A is a proof of concept that you can get 
 ridiculously good thermal performance without using
 any extreme measures like thermos bottles, or large
 thermal "flywheels", etc that are sometimes proposed for some
 "ultimate" oven design.  Whether they would work or not,
 they are not necessary.

 Rick Karlquist
 N6RK

 On Sat 13/11/10 7:41 PM , "Bill Hawkins"  wrote:

 Rick,

 Is the double integrator actually a cascade of two controllers,
 where the primary controls the crystal temperature and its
 output sets the setpoint for a heater temperature controller?

 That's how industrial control handles the lag between a 5000
 gallon reactor and its steam-heated jacket. There's a sensor and
 primary PID for the reactor contents (stirred) and a sensor and
 PID for the jacket temperature.

 When I said "ambient" I was thinking of the range that a military
 receiver like the R-390 was designed for.

 Seems to me that if the only disturbance to the crystal temperature
 is the ambient temperature, then you can lag ambient with enough
 insulation so that the internal controller can be slow enough to
 have adequate gain.

 There are disturbances within the internal loop that would have
 to be dealt with, such as the heater supply voltage and the offset
 voltage of the first error amplifier.

 We are going for stability, not accuracy, right?

 Bill Hawkins

 P.S. Really good story to wind up the "loosing things" thread.

 -----Original Message-----
 From: Richard Karlquist
 Sent: Saturday, November 13, 2010 3:24 PM

 Yes, E1938A.  I was operating on limited sleep when I posted that.

 Rick

 On Sat 13/11/10 8:35 AM , Magnus Danielson wrote:

 On 11/13/2010 04:48 PM, Richard Karlquist wrote:
 > The HP E9183A achieved 1 millidegree over -55 to +85C
 > in a single oven. The time lag was dealt with by adding
 > a double integrator to PID.

 I'd assume you would intend to write HP E1938A, right?

 Cheers,
 Magnus

 > Rick Karlquist N6RK
 >
 > On Fri 12/11/10 8:26 AM , "Bill Hawkins" wrote:
 >
 > in the heater control loop. Of course, you can't get to a
 > millidegree from ambient with just one oven. And you can't
 > eliminate time lags if you have any thermal mass.
 >

 


More information about the Time-nuts_lists.febo.com mailing list