Small Oven, for 2x LT1236, (Python Reference) No1

[blackdog]

Hi,

And now, for something completely different... (Monty Python)

Lets call it the: Python Reference

This topic is about a small oven, it will house a voltage reference with two LT1236 IC's.
This wil not be a Ultra reference! just one to see what is posible.

The case comes from the dump store, it contained a beautiful GHz oscillator, which I demolished...

OK first picture, inside the box, removes all components.



Bottom of the box.



Made a small piece of thin circuit board, the reference will be build "Manhatthan Style".



The holes for the NTC resistors are drilled and the NTC are glued with WLK30.



Work in progress, unfortunately it never stays so neat...


uA723 on its back...



Firts stability test, bad, bad, bad....
This is the oven current on the display of the DMM.



It is working now, the large capacitor, it will end up with in a different place, just for testing on this spot.



NTC to measure the temperature in the box (5K connected to a 34461A)



Closed box.



Not made for this oven, but nice to have, a polystyrene box.



This is the current through the oven, there is no oscillation visible!



The power used, by the oven, LAB Temp ~22C



The temperature shift after power on.



It was difficult, to get the oven work properly...
I dit my best, to get sensor and the powersource (2xTIP3 close together to minimise the lagise the lag.
I used all kinds of compensation, nothing works...

This is how i started, the first schematic, do NOT build this one.



To keep it short, there was to much gain in the circuit.
I replaced the FET's for BJT's TIP31.
Repaced the resistor at the output of the uA723, made the 3.3 Ohm resistors 4.7 Ohm
Still to much gain, better behavior...
I had to lose some gain, but how?
Looking at the internals of the uA723, the only option was pin13, the AC compensation pin.
This is a Pull Up current source, and has a high impedance, a good place to ad a parallel resistor to power supply.
This lowers the gain that is available, BINGO!
After a couple of resistors to trie, the oven control work well.
Except for the power supply, there are no capacitors which determine the stability, they are not necessary.



This is the working circuit, the magic is in R6, a 330K resistor, controling the gain.
I have chosen the resistance value of R6, that the ovencontroller regulated, without overshoot.

A few points ....
Ovens must as voltage references, aging a little.
Ovens for voltage references, do not need to have a stability such as for Reference oscillators 0.05C is enough.
For a simple oven, you wil get between 25 and max 200 times reduction of the temperature change, 200x reduction, is already art   

There is much to tell about how your ovens build, but I have no more time today.
Nextime!

Kind regarts,
Blackdog

[SeanB]

Saw a lovely ocliiator this wekend, with an oven using a vacuum flask. I want to find another to take the crystal out of, or must look at the bits of Power Design reference I have. must dig out those NTC switches I ordered from RS a few years ago and use then as regulators.

[blackdog]

Hi,

Just a extra picture of the drift of the oven for about 4H



The dip at the end is opening the LAB door, the temperatuur dropt about 1C in the room :-)

From the bottom to the top on the picture is 0.002C.
This does not mean that the temperature suppression is nearly 1000x, to determine this en be more shure, i will have to do more testing.
I do not think, i'm going to do more measurements, because it's good enough for my application.

The next step is the circuit of the voltage reference, i do not have much space in the box.
A part of the circuit, will be outside the box, this is posible, because it is not necessary for the stability.

Back to the drawing board! 

Kind regarts,
Blackdog

[atferrari]

Hi,

Just a extra picture of the drift of the oven for about 4H



The dip at the end is opening the LAB door, the temperatuur dropt about 1C in the room :-)

From the bottom to the top on the picture is 0.002C.
This does not mean that the temperature suppression is nearly 1000x, to determine this en be more shure, i will have to do more testing.
I do not think, i'm going to do more measurements, because it's good enough for my application.

The next step is the circuit of the voltage reference, i do not have much space in the box.
A part of the circuit, will be outside the box, this is posible, because it is not necessary for the stability.

Back to the drawing board! 

Kind regarts,
Blackdog

Hola bd

Sorry for the question: what is on the vertical axis? milli what?

[blackdog]

Hi atferrari.

You can read it in the text => From the bottom to the top on the picture is 0.002C.
0.002C = 0.002 degrees Celsius

It is the drift from the starting point of the measurement.
assume that the temperature of the oven is set to 42 degrees Celsius.
The picture shows, how much the temperature setpoint in the oven is changed from 42 degrees celsius.

Kind regarts,
Blackdog

[blackdog]

Hi :-)

Next step, The first attempt of the circuit.
The reference chips will be 2x LT1236 LS8.
The 2x opamp will be a ADA-4077-1
The gain resistors wil be Vishay presicion film resistors.
I do not know yet if i will make a trim function, maybe just measure the value en note is down on the box.

Schematic V1.0


Remember, this wil not be an reference like a LTZ1000 or a good LM399.
My goal is to see what is possible with a 2x LT1236 in an oven.

Shoot at it! 

Kind regarts,
Blackdog

[blackdog]

Hi,

Nobody is liking this project, almost no reactions, stange...
Sorry no blinking L.E.D., no Arduino, no LTZ1000, not known is unpopular, or just bad design? 

OK, i wil go on.
I only had three wire fittings in the box and decided to drill two additional holes.
These holes are for the sense wires, this is a pair of a CAT5 cable.

First, the modified schematic, i have also included a sense wire to the negative side.
That is why, I needed extra holes into the box.
I have also made it clear with ferrite beads where is sitting HF decoupling.

Schematic V1.1



These are all components that should be in the box.
There is not much space left, one part is not in the picture, which is the opamp, ADA-4077-1
This wil take a lot braintime to fit it neatly and electronically corect in the box.


I dit some more measuring of the temperature stability.
1.2 degrees celsius change outside the oven, gives 0:01 degrees Celsius change in the oven.
A suppression of about 120 times, with a very good insulation, it can be even better.

Until next time.

Blackdog

[Andreas]

Shoot at it!

Hello,

did not see it before.

I would use >=10V as input for the LT1236 (due to bad PSRR below 10V).

Ageing of ADC17 (from day 200 on) shows what is possible after some run in time for a LT1236AILS8-5.
I am doing a mathematical temperature compensation when measuring the LTZ1000A references.
Over 1 year there is +/-2ppm drift of the reference. Of course with stressless (in hole) mounting to the pcb.
So with a oven I would expect similar performance after some run in time (if the resistors have same stability).

With best regards

Andreas

[codeboy2k]

Nobody is liking this project, almost no reactions, stange...

I like this project alot.  I have also wanted to know how a decent buried zener like the LT1236 can perform in a stable environment.  You say it's not a LTZ1000 or a 399... but I think that you can get almost as good performance, and I look forward to the results.

I really like how you've reused an RF box with the feed through capacitors for an oven! I might try that too.

[blackdog]

Hi Andreas,

Thank you for your reply :-)

I thought about the 8V regulator, to remove it, because there there wil be more space in the oven and its easier to built.
If i leave out the the 78L08, then I have to give the 12 to 15V for the reference section more attention.
At the moment, i cant make up my mind.

If the 78L08 is in the circuit wil only see verry smal variations on it power supply.
I measured the noise this afthernoon of a 78L08, it was <50uV RMS @ a load of 3mA and  20Hz tot 22Khz bandwith.
Do you see any problems with this specifications, the voltage wil i think not change more than 10mV in its live.

If i use a 78L08, i can use a LM317 voor the power supply of the reference.
If i leave out 78L08, i have to use a better regulator, maybe a uA 723 or a LT3080....

This afthernoon i did also some measurements on the filter properties of the schematic.
Because my Audio Precision stop at the low side at 9Hz, I have made the capacitors 100 times smaller in value.
Then it is easier to measure the -3dB point.

Test setup the filter properties of the reference amplifier.



10 to 3.15Khz (= 0.1 to 31.5Hz) marker at -3.32dB at 3.15Hz.

 

10 to 3.15Khz (= 0.1 to 31.5Hz) marker at 1Khz, -13,62dB, (10Hz).




10 to 3.15Khz (= 0.1 to 31.5Hz) marker at 1.5792Khz, -20dB, (15.792Hz).



The latest version of the schematic, i increased the current of the output stage.
I could not find a fet that can deliver a larger current, so I took three pieces of the J310, cheap and easy.

Extra is D2, D3 and D4.
D2 and D3 are for safety, D4 is for the battery back-up.


Of course, the schematic is a work in progress, there will be more changes. 

Kind regarts,
Blackdog

[codeboy2k]

If the 78L08 is in the circuit wil only see verry smal variations on it power supply.
I measured the noise this afthernoon of a 78L08, it was <50uV RMS @ a load of 3mA and  20Hz tot 22Khz bandwith.
Do you see any problems with this specifications, the voltage wil i think not change more than 10mV in its live.

The LT1236-5 PSRR @ 8V is approx 93dB . 10^(-93/20) = 22.4E-06 . So that 10 mV of change in the regulator output will be 22E-06 times smaller at the reference output.   10 mV x 22.4E-06 = 224 nV at the reference output.

So if the 78L08 is able to keep the output voltage regulation within +/- 10 mV as you think, then this will equate to 224 nV in output change at the LT1236-5 reference output.  224 nV / 5V ~= 0.5 ppm.  That seems OK to me for this application.

[dom0]

Probably better to use a steeper low pass, tchebyshev would be ideal (passband ripple is a "don't care" here), if your intent is to reject 50 Hz + it's two-hundred harmonics.

[blackdog]

Hi dom0,


Its not about 50/100hz rejection.
Its al about noise below 10 Hz, and -3dB @ 0.1 Hz would be nice 
But, no room in the box (oven) in this project, i have to do it with the filter i made.
I have already 2 capacitors of 6,8uF in this small oven, no more room for bigger or more capacitors.

This is a reference to say measure multimeters and then you use the highest number of power line cycles to get the best reading.
More PLC  = higher rection of low frequentie noise.

Kind regarts,
Blackdog

[nadona]

blackdog: Thank you for the sharing. 7805's input and output is reversed(typo).

dom0: Typing mistake.

Probably better to use a steeper low pass, chebyshev would be ideal (passband ripple is a "don't care" here), if your intent is to reject 50 Hz + it's two-hundred harmonics.

Regards,

[dom0]

Yeah, Chebyshev is a bit hard for me to remember how it's correctly written. We Germans usually write it Tschebyscheff ...or one of the English variants

[JohnnyBerg]

Nobody is liking this project, almost no reactions, stange...

Not strange at all 
This is very specialized stuff, not for the average Arduino hobbyist 

I like to read it, as I have made I nice reference and working on the B revision.

[blackdog]

Hi,


I did some testing on the outputstage of the voltage reference, wy do i use so manny components you may ask...
This is why:
Verry low "Dropout Voltage", i tested it with several opamps, like the LT1012, LT1077 and the OP177.

With a load of 1 mA, the drop-out voltage of the reference circuit was 10,35V.
At a load of 12mA (a bit above my starting point), the drop-out voltage was 11,3V.
I've looked at the output voltage of the test circuit and then when the output voltage 1 PPM fell, i noted the voltage of the power supply.

By 1 mA load, the varied voltage on the opamp output is between 7.5 and 7,9V, so the opamp output is 7,9V @ the drop-out voltage of 10.6V.
I give so much attention to this circuit, to make maximum use of the energy in the batteries.

Another way to make it even more stable it is, a switched mode power supply, so that the reference circuit sees always say 15V, even at low battery voltages.
I find a "normal" switching power supply is not an option, too much noise.

I'm trying to develop a small sine wave power inverter, between few 1kHz an say 20Khz, low power, to power up reference circuits like this one.
But that wil be another project, it wil not come in this reference.

My starting point was actually, just to keep the reference under power.
So the oven will not cool down with a power outage or transport for a few hours...

This is how i tested is, i know, its not a buildop for a high quality reference, bus just enough to do this measuremt.
The reference is here loaded with the 10K resistor, the ref is a TI 5050, opamp a OP177.



Top left is the opamp voltage at dropout point with a 1mA load (REF505, OP177)
Bottom left is the power supply voltage at dropout.
Top right the 1PPM difference at dropout at the reference output.
Bottom right has nothing to do with this measurements, this multimeter is measuring a different 10V reference.


Of course, the one PPM differences com from a big part of the power supply and offset voltage shift to the power voltage variation between 10 and 15V.
That why a stable powersupply is necessary if you want good performance, if used om batteries.
I tested with a 9V batterie for the TI REF5050.

I stil do not know, if i wil use The 78L08...
When I go to use the reference on batteries, this 78L08 is needed like Andreas pointed out.
Decisions, decisions, decisions 

Kind regarts,
Blackdog

PS
Turnt the 78L08 around, stupid mistake 

[Andreas]

Hello,

and you should be patient....
What are 1000 hours more or less with precision stuff.
Sometimes you have to wait another 1000 hours until you can state a result.

With best regards

Andreas

[branadic]

I was thinking of a temperature controlled reference with LT1236LS8 too and had this in mind. The ref-part is on a slotted part, only connected at the corners with the rest of the pcb. The crystal heater QH40A is glued on the backside of this pcb section and keeps the reference at 40°C.

[JohnnyBerg]

I was thinking of a temperature controlled reference with LT1236LS8 too and had this in mind. The ref-part is on a slotted part, only connected at the corners with the rest of the pcb. The crystal heater QH40A is glued on the backside of this pcb section and keeps the reference at 40°C.

Interesting idea!

Perhaps it is even possible to create 2 planes on either side and connect them through some via's and use that is a heater?
How are you going to measure the temperature?
 

[dom0]

The QH40A seems to be a complete precision heater: http://shop.kuhne-electronic.de/kuhne/de/shop/professionell/prof-zubehoer/QH+40+A+Praezisionsquarzheizer+40%C2%B0+QH40A/?card=260

[blackdog]

Hi dom0

bin there done that 

Look at the link below, you wil vind a link to a Dutch site "Circuitsonline.net" with a TL431A project of mine.
Just for fun i researcht what was posible with a TL431A, a good reference was cheaper like a LT1021, LT1236 ect.
But this was fun and good specs to.
http://www.circuitsonline.net/forum/view/116156/1/tl431

Kind regarts,
Blackdog

[branadic]

Hi,

found you reverse engineering the circuit of the QH40A. I found the following parts to the smd markings:

358 --> LM358 (1459 in your circuit)
50K --> ZMR500
210 --> KTY82/210
1Ls --> BC848C (? in your circuit)
AS3 --> BST52

As the LTC6655 is available as LS8, has much lower noise, but high thermal hysteresis of 30ppm (–40°C to 85°C) maybe it's worth putting multiple of them into a temperature controlled oven, running 24/7, together with a A=2 LTC2057 amplifier with LT5400 as gain setting resistor. Sure, longterm stability would be worse the frist 2000h, but after it has aged, well...

[dom0]

Masturbation is healthy. I think that's independent of teh kind of masturbation.

[Kalvin]

Stupid question: Why not add the styrofoam insulation inside the metal enclosure? If one wants to have more "thermal mass", just place a smaller metal enclosure inside the insulated enclosure?

[daqq]

Very nice! Impressive that you got such a small reaction to outside temperature.

Perhaps it is even possible to create 2 planes on either side and connect them through some via's and use that is a heater?

Using PCBs as heaters is done (seen it in 3D printers), but the problem is that they have very low resistances - as such you need a low voltage with a lot of current.

Stupid question: Why not add the styrofoam insulation inside the metal enclosure? If one wants to have more "thermal mass", just place a smaller metal enclosure inside the insulated enclosure?

I'd say it would be better to put the metal enclosure into an insolation - that way you get a good thermal connection to the heated device and a lousy thermal connection to the outside.

[Alex Nikitin]

together with a A=2 LTC2057 amplifier with LT5400 as gain setting resistor. Sure, longterm stability would be worse the frist 2000h, but after it has aged, well...

You can get a pretty decent x2 precision amplifier using one INA133 inside the enclosure.

Cheers

Alex

[Kalvin]

Stupid question: Why not add the styrofoam insulation inside the metal enclosure? If one wants to have more "thermal mass", just place a smaller metal enclosure inside the insulated enclosure?

I'd say it would be better to put the metal enclosure into an insolation - that way you get a good thermal connection to the heated device and a lousy thermal connection to the outside.

Total brainfart from my side I thought placing the heating transistors with a heat sink inside the insulated metallic enclosure would give some benefit. Like you said, insulating the heated enclosure is the Solution.

[branadic]

Any news or results on that experiment?

-branadic-

[branadic]

I finally managed to power up LT1236LS8 assembled on a PI flex pcb, attached with a thermal adhesive pad to a QH40A crystal heater. The assembly is insulated and inside a glass flask. I still have a bunch of those flex boards left, so if someone is interested in let me know. Can ship them via a letter if you pay for the shipment.

-branadic-