Maker Pro
Maker Pro

Need help understanding HP10544 osc oven circuit.

J

Jim Flanagan

Jan 1, 1970
0
Hi -

Recently, I acquired an HP ovenized oscillator (HP10544A)which was
missing the proportional oven controller. I found the schematic for
the oscillator at the following link:
http://www.leapsecond.com/museum/10544/


After building the oven controller on a small PCB, I purposely disabled
the unijunction oscillator stage initially, in order to test the
controller. My thought was that the switching supply (if you want to
call it that) section was just to make the the oven controller more
efficient. With the oscillator disabled, I found that the oven would
oscillate itself. From a room temp start, the oscillator oven would get
to temp then shut itself off, then repeat at about a .1hz rate. It
wasn't until the oscillator section was enabled, that the oven would
work correctly. By correctly, I mean that as the oven approaches it's
set point temperature, the op amp section would go into its linear mode.
This is apparent as the heater current begins to decrease from about
500mA to about 180mA.

I would appreciate if someone would take a look at this circuit, in
terms of the oven controller, and help me understand exactly how the UJT
oscillator is functioning in this closed thermal system. Also, I would
appreciate some input as to how an simulate a thermal system like this
in spice. I simply don't understand how to simulate the thermal
feedback portion.

Your insight and expertise is appreciated... Thanks
Jim
WB5KYE
 
J

John Larkin

Jan 1, 1970
0
Hi -

Recently, I acquired an HP ovenized oscillator (HP10544A)which was
missing the proportional oven controller. I found the schematic for
the oscillator at the following link:
http://www.leapsecond.com/museum/10544/


After building the oven controller on a small PCB, I purposely disabled
the unijunction oscillator stage initially, in order to test the
controller. My thought was that the switching supply (if you want to
call it that) section was just to make the the oven controller more
efficient. With the oscillator disabled, I found that the oven would
oscillate itself. From a room temp start, the oscillator oven would get
to temp then shut itself off, then repeat at about a .1hz rate. It
wasn't until the oscillator section was enabled, that the oven would
work correctly. By correctly, I mean that as the oven approaches it's
set point temperature, the op amp section would go into its linear mode.
This is apparent as the heater current begins to decrease from about
500mA to about 180mA.

I would appreciate if someone would take a look at this circuit, in
terms of the oven controller, and help me understand exactly how the UJT
oscillator is functioning in this closed thermal system. Also, I would
appreciate some input as to how an simulate a thermal system like this
in spice. I simply don't understand how to simulate the thermal
feedback portion.


If the ujt is off, the loop gain is very high and the thing pretty
much works in bang-bang mode. A very small change in opamp output will
slam the heater full on or full off.

The Q1-Q2 differential pair compares the opamp output to the swatooth
created by the ujt. So the amp output has to span a roughly 7 volt
range to move the heater from full off to full on, which is
effectively a much lower gain.

One less obvious advantage of pwm, as compared to a linear system, it
that it makes heater power linear on error. A linear voltage or
current drive into a heater is a square function.

You can model the thermal stuff as a group of resistors (thermal
resistance) and capacitors (thermal masses). The approximate (within
5%) mapping is

1 ohm == 1 degc/watt

1 amp == 1 watt of heat

1 volt == 1 deg C

1 farad == 1 gram of aluminum

the catch being that the components tend to be distributed, not
lumped, so nasty diffusion math applies.

John
 
J

Jim Flanagan

Jan 1, 1970
0
John said:
If the ujt is off, the loop gain is very high and the thing pretty
much works in bang-bang mode. A very small change in opamp output will
slam the heater full on or full off.

The Q1-Q2 differential pair compares the opamp output to the swatooth
created by the ujt. So the amp output has to span a roughly 7 volt
range to move the heater from full off to full on, which is
effectively a much lower gain.

One less obvious advantage of pwm, as compared to a linear system, it
that it makes heater power linear on error. A linear voltage or
current drive into a heater is a square function.

You can model the thermal stuff as a group of resistors (thermal
resistance) and capacitors (thermal masses). The approximate (within
5%) mapping is

1 ohm == 1 degc/watt

1 amp == 1 watt of heat

1 volt == 1 deg C

1 farad == 1 gram of aluminum

the catch being that the components tend to be distributed, not
lumped, so nasty diffusion math applies.

John
Hi John -
Thanks for the explanation. Seems obvious now. Another case of not
seeing the forest through the trees.

One last thing, do you have a suggestion as to a literature reference or
tutorial in regards to the thermal modeling?

Again - Thanks a bunch...
Take care - Jim
 
J

John Larkin

Jan 1, 1970
0
Also R11 and Q4 seems to be drawn wrong.


Yeah. The PNP should point the other way. Weird.

Maybe Jim can check and see how it's actually built.

SRS sells some clones of the old HP ocxo boxes, with nice SC-cut
rocks. Their stability and phase noise are impressive, but their
thermal design is bizarre. They use TO-220 voltage regulators as
heaters, and it looks to me like whoever designed the loop didn't
really understand the dynamics, so they used a proportional-only loop
with fairly low gain, then added feedforward compensation from an
ambient temp sensor to improve temperature regulation.

John
 
J

Jim Thompson

Jan 1, 1970
0
Yeah. The PNP should point the other way. Weird.

Maybe Jim can check and see how it's actually built.

SRS sells some clones of the old HP ocxo boxes, with nice SC-cut
rocks. Their stability and phase noise are impressive, but their
thermal design is bizarre. They use TO-220 voltage regulators as
heaters, and it looks to me like whoever designed the loop didn't
really understand the dynamics, so they used a proportional-only loop
with fairly low gain, then added feedforward compensation from an
ambient temp sensor to improve temperature regulation.

John

I can't seem to retrieve this thread's references, but it sounds like
an inverted bipolar device used for AGC.

Can someone repost the schematic?

...Jim Thompson
 
J

John Larkin

Jan 1, 1970
0
I can't seem to retrieve this thread's references, but it sounds like
an inverted bipolar device used for AGC.

Can someone repost the schematic?

...Jim Thompson

http://www.leapsecond.com/museum/10544/

I actually meant that Jim Flanagan might check the actual oscillator
so see how the heater driver PNP works.

John
 
M

MooseFET

Jan 1, 1970
0
[.....]
SRS sells some clones of the old HP ocxo boxes, with nice SC-cut
rocks. Their stability and phase noise are impressive, but their
thermal design is bizarre. They use TO-220 voltage regulators as
heaters, and it looks to me like whoever designed the loop didn't
really understand the dynamics, so they used a proportional-only loop
with fairly low gain, then added feedforward compensation from an
ambient temp sensor to improve temperature regulation.

That is very strange. SRS seems to know what they are doing in the
other stuff they do. I can even see an argument for using a 3 pin
regulator as the heater. They give a huge current gain, are self
protecting and don't have their own tempco.

The components needed for a PID controller wouldn't add much to the
size of the circuit. These days op-amps with very low bias currents
can be obtained easily so the capacitors don't have to be huge.
 
J

Jim Flanagan

Jan 1, 1970
0
John said:
Yeah. The PNP should point the other way. Weird.

Maybe Jim can check and see how it's actually built.

I caught that also. My unit didn't come with a controller board.
It had been removed for some odd reason. That is why I built a replica
and the subsequent initial posting concerning the thermal feedback
question. I built my unit the way we believe it should be. All is
well now, functionally.
 
F

Fred Bloggs

Jan 1, 1970
0
John said:
If the ujt is off, the loop gain is very high and the thing pretty
much works in bang-bang mode. A very small change in opamp output
will slam the heater full on or full off.

Ummmm, looks like U1 is integrating the thermistor bridge error with a
0.001Hz time constant and ~40dB gain, so hard to visualize a "slam"
on/off action in that circuit.
The Q1-Q2 differential pair compares the opamp output to the swatooth
created by the ujt. So the amp output has to span a roughly 7 volt
range to move the heater from full off to full on, which is
effectively a much lower gain.

Ummm, since the Q4 current gain is so high, U1 is required to inject
just a few uA into Q1 base for full-on of the Darlington. There is not
going to be much of a "span" about this Q1 base voltage threshold.
One less obvious advantage of pwm, as compared to a linear system, it
that it makes heater power linear on error. A linear voltage or
current drive into a heater is a square function.

The UJT will be oscillating at something around 4KHz, and its purpose is
to chop the Q4 drive to minimize power dissipation, nothing exotic there
in the way of control loop processing.

The funny labeling on the heater wires and the diode in Q4 lead me to
suspect the heater drive may be AC...but then the color coding makes me
think it should not be.
 
M

MooseFET

Jan 1, 1970
0
[....]
Ummmm, looks like U1 is integrating the thermistor bridge error with a
0.001Hz time constant and ~40dB gain, so hard to visualize a "slam"
on/off action in that circuit.

How do you get 0.001Hz?

51.1K * 2 uF -> 1.55Hz

5.6M * 2 uF -> 0.014Hz

The latter being the zero in the PI controller, I'd expect the gain
cross over to be near that point.


[....]
Ummm, since the Q4 current gain is so high, U1 is required to inject
just a few uA into Q1 base for full-on of the Darlington. There is not
going to be much of a "span" about this Q1 base voltage threshold.

It is the voltage on the base of Q1 that matters here.
Q2 is fed with a ramp from C1.
 
J

John Larkin

Jan 1, 1970
0
Ummmm, looks like U1 is integrating the thermistor bridge error with a
0.001Hz time constant and ~40dB gain, so hard to visualize a "slam"
on/off action in that circuit.

Then maybe you should give it another try.
Ummm, since the Q4 current gain is so high, U1 is required to inject
just a few uA into Q1 base for full-on of the Darlington. There is not
going to be much of a "span" about this Q1 base voltage threshold.

Roughly 7 volts.

The UJT will be oscillating at something around 4KHz, and its purpose is
to chop the Q4 drive to minimize power dissipation, nothing exotic there
in the way of control loop processing.

Turning on the ujt cuts loop gain by a factor several hundred, and
changes it from super-nonlinear to mostly linear. Do you think that
might affect loop dynamics?

John
 
F

Fred Bloggs

Jan 1, 1970
0
John said:
Turning on the ujt cuts loop gain by a factor several hundred, and
changes it from super-nonlinear to mostly linear. Do you think that
might affect loop dynamics?

Looks like someone needs to review their basic arithmetic. In the
simplest case of linear PWM, the change in output power per unit of
error voltage drive into the UJT modulator remains the same over the
span (7V) of modulator input voltage. This is not a gain reduction of
several hundred. The circuit is a simple dominant pole regulator with
temperature->voltage->power->temperature forming the loop states.
 
J

John Larkin

Jan 1, 1970
0
Looks like someone needs to review their basic arithmetic. In the
simplest case of linear PWM, the change in output power per unit of
error voltage drive into the UJT modulator remains the same over the
span (7V) of modulator input voltage.

No, the sawtooth is not linear, but let's assume it's close enough.
This is not a gain reduction of
several hundred. The circuit is a simple dominant pole regulator with
temperature->voltage->power->temperature forming the loop states.

The gain reduction is relative to when the ujt oscillator is stopped.
In that case, the system becomes continuous/nonlinear with very high
gain, and becomes unstable as the op noted.

John
 
F

Fred Bloggs

Jan 1, 1970
0
John said:
No, the sawtooth is not linear, but let's assume it's close enough.




The gain reduction is relative to when the ujt oscillator is stopped.
In that case, the system becomes continuous/nonlinear with very high
gain, and becomes unstable as the op noted.

The gain reduction is relative to a completely different circuit
topology when the ujt is stopped, so the comparison is not meaningful.
Also, I'm not sure it even makes sense to use the word "gain" in that
context because the loop becomes a simple two-state in each of which the
incremental gain is zero. The experiment may be used to infer the
thermal time constant of the heater system and that's about it.
 
Top