Experiments with Vref Ovens

[enut11]

There are lots of hits on the Forum if I search for 'ovens' or 'voltage Refs' or both and I will be spending some time catching up on the latest info.
Meanwhile I would like to post my recent experiments on a relatively cheap approach that I am using to improve my own Vrefs.

A friend once said, referring to the study of Voltage References, “temperature is everything”. Granted it is not that simple and other things can come into play but the thought has stayed with me ever since.
A couple of years ago, when the Aussie Cal Club was active, I built several LM399 based voltage references and left them powered up to condition the components. They did soak for over a year until my office got flooded from a leaking water pipe. All my gear had to be removed and the whole mess took over 3 months to repair. That was more than 12 months ago.

I recently dug out all my references from storage and re-powered them. Then, inspired to improve the outcome, I decided to experiment with a controlled temperature oven. The oven controller/heater design, also from that same friend, was a simple bridge circuit monitored by an opamp. One of the bridge legs is a thermistor. The opamp drives a power transistor to control the heating element, all standard stuff.

The controller pcb and Ref pcb are mounted inside a small metal box which in turn sits inside a larger insulated foam box.
The idea is to keep the set temperature as stable as possible to isolate the Ref from ambient variations.

This was most important for this project as my Ref pcb used garden variety components to reduce costs. The thought being that low tempco components would not be necessary if the temperature does not vary!
More to come…
enut11

[enut11]

The setup for measuring Ref oven temperature vs ambient consists of a Laptop PC connected to a HP 3456A DMM via USB-GPIB.
Ambient and oven temperature data is fed through a home made 2-input scanner to the 3456A in 2W/Th mode.

Because XL wants to 'join-the-dots', the temperature plot comes out as a green band.
The upper part of the green band is internal oven and the lower part is ambient temperature.

[enut11]

The first oven controller was built on veroboard and installed, along with the Ref pcb, in a 100x100x60mm diecast metal box. The diecast box was insulated from ambient inside a small polystyrene ‘broccoli box’ sourced from the local vege market.  I have since used several of these foam boxes as they are very easy to work with and cost nothing.

First 24hrs data from this prototype showed a nice flat profile for the oven temperature along with variable ambient so things were looking good.

[enut11]

I was curious about the stability of the oven so took a closer look.

Direct data over a shorter period of time showed an interesting oscillating temperature pattern.
The variation of around 0.15C peak to peak was thought to be caused by poor thermal coupling between the controller sensing thermistor and the heating element.
Persistent overshoot and undershoot of the set point was not desirable and had to be investigated.

[enut11]

Oven C was not suitable for the mods I had in mind so I built another oven in a smaller diecast box, about 1/3 the size, just big enough to hold the controller (now on a proper pcb) and the Vref.
In this build I located the controller sensing thermistor closer to the heating element.
A 24 hour plot of oven temperature vs ambient temperature again shows a nice flat upper area with a variable lower part. Even at this stage it looked better than the first run.
The gap in the plot was due to loss of power to the scanner for a short time.

[enut11]

A closer look at Oven A temperature over a shorter period showed no signs of oscillations. Indeed, the oven temperature profile was virtually flat with max variation of 0.01C.
Now that is not to say that it will stay that close to the set point over 24hrs but much better than the first oven.

[TimFox]

I'm unfamiliar with the term "broccoli box":  could you give the rough dimensions?

[enut11]

Good question @TimFox. I was about to show some more pix of the oven materials. The foam boxes measure 38cm long x 28cm wide and height can be 15-20cm. Wall thickness is 1.5cm.
It is easy to cut out holes for gauges, etc using a sharp razor knife.

For additional insulation I have used polystyrene, foam rubber, packing foam and roof/ceiling insulation bats. They all work. The only difference that I have noticed is the oven idle current varies depending on the effectiveness of the insulation.

For smaller projects, I have cut the boxes down and glued the parts together with white glue or epoxy glue. Some glues will melt foam so beware.

[enut11]

For the oven, I use the smallest diecast metal box that fits my Vref and controller boards. This box measures 11cm x 6cm x 3cm.
I don't use those bulky banana sockets inside the oven as they take up too much space. Instead I wire directly from the PCBs to the front panel sockets mounted outside the foam box.

The Ref pcb is 5x3cm and the controller 6x4cm and both fit comfortably inside the metal box along with a string of twelve 2W 100ohm resistors configured as a 33ohm heater plus some insulation to prevent circulating air currents.

[enut11]

One of the things I had to decide on was the minimum temperature to run the oven at.

I mounted a powered voltage reference inside a foam box with insulation and monitored the temperature rise using a thermistor inside the metal box.
On a warm day here (27C inside my office) the metal box heated up to 39C just from heat generated by the voltage reference!
Hence, I now set the oven controller temperature to 40-41C. Not a big margin but it has been OK so far.

I monitor the controller heater current using an ammeter and can see if the oven shuts off by itself when ambient temperature rises.

[enut11]

I now prefer an analog meter to monitor heater current - easier to read at a glance. This pic shows the oven has settled at about 30mA (100mA FSD) with an ambient temperature of 20C.

The oven controller was set for 41C but the digital meter shows a slightly lower temperature because the sensor is outside the metal box in the oven.

Other front panel sockets give access to an internal 5K thermistor for data feed, regulated 15v input for Vref, Vz out and 10v out.

15v power supplies for vRef, oven controller and digital thermometer are all external so as not to interfere with the inner oven environment.

As an aside, I found the digital thermometer will also run happily from a usb 5v supply.

[enut11]

24hr oven internal temp vs ambient for a third Ref oven (Oven B) has just completed. Once again the oven temp is very flat compared to the significant changes in ambient temperature. Very happy with the end result so will continue to condition the Vref boards inside the 3 ovens.

The biggest challenge now is how to quantify the benefits of a stable Vref temperature environment but that is another story.
enut11

[dietert1]

When you have two ovens each one with its own voltage reference, it's more fun. Then you can measure the voltage difference (to 100 nV in the lowest 3456A range) as a function of the oven temperature while keeping the other temperature constant. You will need to use Low Thermal EMF connections to get that precision. Then you can determine TC of each reference as a function of temperature, find a good operating temperature where the curve is near flat, determine reference noise etc.

Regards, Dieter

[exe]

What thermistor do you use?

Do "typical" NTC thermistors drift over time?

[enut11]

When you have two ovens each one with its own voltage reference, it's more fun. Then you can measure the voltage difference (to 100 nV in the lowest 3456A range) as a function of the oven temperature while keeping the other temperature constant. You will need to use Low Thermal EMF connections to get that precision. Then you can determine TC of each reference as a function of temperature, find a good operating temperature where the curve is near flat, determine reference noise etc.

Regards, Dieter

@dietert1, this does sound interesting! More info please. This will be a new learning for me.

I am lucky enough to have 3 good meters (3456A, 34401A, 34461A) but they are all based on LM399 internal reference.
In effect, measuring an LM399 with an LM399 does not sound practical?
enut11

[TimFox]

If the two references are close to the same voltage, then measuring the difference between the two voltages (to the resolution of the lowest full-scale range) will have more resolution than measuring the two voltages individually.  The internal reference of the voltmeter will have almost no effect on the result, since it will proportionally affect the (small) difference voltage.

[enut11]

What thermistor do you use?

Do "typical" NTC thermistors drift over time?

@exe, you did not specify which thermistor so I assume the feedback in the controller bridge circuit.
This is a thin film 10k NTC available on eBay
Search for "NTC10K Film Thermistor Ultra Thin Temperature Sensor ". They are used in 3D printers.
I purchased the one with the orange film cover.

Do "typical" NTC thermistors drift over time?
I don't know. Maybe others on the Forum can advise on this.
enut11

[exe]

Search for "NTC10K Film Thermistor Ultra Thin Temperature Sensor ". They are used in 3D printers.
I purchased the one with the orange film cover.

That's what I meant

[enut11]

I use a squashed down lug connector. Insert the thermistor tip and set with thermal paste.

I have also tried bead thermistors but they are not as quick to respond to temperature changes.

[enut11]

When you have two ovens each one with its own voltage reference, it's more fun. Then you can measure the voltage difference (to 100 nV in the lowest 3456A range) as a function of the oven temperature while keeping the other temperature constant. You will need to use Low Thermal EMF connections to get that precision. Then you can determine TC of each reference as a function of temperature, find a good operating temperature where the curve is near flat, determine reference noise etc.

Regards, Dieter

@dietert1, this does sound interesting! More info please. This will be a new learning for me.

I am lucky enough to have 3 good meters (3456A, 34401A, 34461A) but they are all based on LM399 internal reference.
In effect, measuring an LM399 with an LM399 does not sound practical?
enut11

I forgot to mention that I also have a CROPICO 1.00000v ESC1 Standard Cell so this may be useful in some way to verify my ovenised 399 references.
However, all my gear is un-calibrated.

[doktor pyta]

these thermistors are most likely manufactured by this company:
https://tewa-sensors.com/products/

[antintedo]

Do "typical" NTC thermistors drift over time?
I don't know. Maybe others on the Forum can advise on this.

Yes, that's why glass encapsulated thermistors exist. Manufacturers rarely publish this information, especially for cheap stuff.
Some info I found while selecting sensors for a TEC oven:
http://www.farnell.com/datasheets/169207.pdf
https://ieeexplore.ieee.org/document/7281460
https://www.eevblog.com/forum/metrology/long-term-stability-of-temperature-sensors/
Ysi 55031 glass beads are available from Farnell - that's what I eventually got.

these thermistors are most likely manufactured by this company:
https://tewa-sensors.com/products/

Unlikely since it was purchased on ebay from Chinese seller. Plenty of local companies produce identical looking sensors.

[essele]

I've been experimenting with ovens for a few months now, this is my latest incarnation...



It uses a PTC (pair) and an ADS1247 ADC and then generates an analogue heater control voltage using two of the current DAC's in the PSoC which then drive 6 different transistor/resistor heaters around the board.

It's intended to support a daughter board with a large number of parallel REF102's (as per the paper that's floating around that I can never find when I need it), so the rest of the components are power supplies and some options for buffering and filtering to experiment with.

It's all in a small(ish) die-cast enclosure, but I've had to increase the size of it to cope with the (future) power dissipation of the REF102's. I had some good results with air. I tried adding a circulation fan, but that made things incredibly noisy (electrically), and now I'm doing some experiments with mineral oil.

Without crazy tuning I'm typically seeing temperatures within about 20mK (of 44C) measured on an external DMM and the thermistor you can see in the picture, this is with overnight ambient variations of 5 to 6 degrees. However I am starting to doubt the accuracy of that as a validation mechanism -- I'm fairly sure my DMM isn't great with temperature changes, and I'll definitely have some thermal emf in the test setup. I'm able to keep within four or five mK if you believe the ADC results.

I gave up on PID as a control method quite some time ago, I now have a focus on the first and second derivatives of temperature and am getting much better results. At the moment I'm specifically working on trying to improve recovery from upsets ... which can be quite slow in oil!

Lots of things still to do:

- I've built a TMP117 module for use as an external sensor to pre-warn of ambient changes to get a head start before the ADC detects it.
- The ADC is noisy and I have no hardware filtering, so I need to do something about that. (Software filtering has helped a lot though)
- My test rig isn't particularly well sealed because of the cables, I'll probably add another layer of insulation.
- The heaters presumably heat the PCB and the PTC's are directly on the PCB so I think that's skewing results, ideally I'll put the PTC's on the daughterboard, but that needs a bigger redesign. May just add some more slots for the next revision.
- There are a few other board tidies to do (as you can see I missed the Vref cap and connections completely on the ADS1247!)

Very keen to get any input ... and happy to share schematics/code if anyone is interested (neither of which are tidy at the moment!)

EDIT: I've added a photo of the daughterboard if anyone is interested. It's double sided and can take 120 x REF102's. It needs a redesign and I need to be comfortable everything is looking good before I shell out hundreds of pounds on REF102's!

[enut11]

@essele, quite an impressive setup and much more sophisticated than mine. Where is the oven heater?
Also, is the pcb with a large number of REF102s designed to condition the chips or for another purpose?
enut11

[essele]

Hi enut11,

The heaters are Q1 to Q6 and the pairs of resistors near them. I went for 6 distributed around the board, it meant they didn't need to be too chunky and I assumed being distributed would help even things about a bit.

The REF102 board is intended as a massively parallel set of REF102's to reduce noise and improve stability. I was inspired by this...

https://www.researchgate.net/publication/305800682_Eleven_years_of_monitoring_an_ultra-stable_10_V_zener-based_voltage_standard

... and I know it's not really a cost effective thing to do (120 REF102's don't come cheap), but they did seem to get quite impressive results and it felt look a good mad hobby project ;-)

I've added another layer of insulation now and it's actually reduced the heater power needed by quite a bit, a good test for my algorithm, but actually a problem as I'm not going to have much headroom over the heat generation of the REF102's!