# Questions regarding wide temp range OCXO

Discussion in 'General Electronics Chat' started by osx-addict, Jul 10, 2012.

Feb 9, 2012
122
9
Hi all.. I'm looking into yet another angle of my project -- this time spurred on by a comment in another thread about building your own Oven Controlled XTAL Oscillator (OCXO) -- namely pointing to The_RB's cool proof of concept here.

In his design he has purposefully set the circuit up to target ambient room temps and heats his OCXO to about 95F.. However, I'd like to see if his circuit/design can be tweaked to allow use in a much wider temp range -- specifically somewhere around 30F (or lower possibly) to say 115F.. To that end, I'm thinking I could effectively tweak his design to heat the xtal up to something around 130F. If I do that AND package it up into a larger thermal mass, do you see any other issues I might have in achieving my goal?

In my case this would be used in conjunction with a 32.768khz crystal to drive the RTCC module in a PIC32 processor to provide a, hopefully, rock-solid clock that can be used in warm or cold environments.

Also, if this device is mounted about an inch off-board from the rest of the PIC32 circuit is that asking for trouble since it's providing the clock for the RTCC?

TIA!

2. ### #12 Expert

Nov 30, 2010
18,076
9,676
The critical information seems to be, "15ma heats the assembly by 8C"
You must adjust the set point to 130F and make sure you can burn enough power to heat from your lower temperature, 30F.

5V @ 15 ma is 75 milliwatts per 8 C.
100F = 55.55 C
6.9 x 75 mw is your heater goal.
That's 104 ma.
Your maximum current capability must be at least 104 ma.

Got it?

Feb 9, 2012
122
9
Thanks.. I'll have to play around with making one of these -- I'll report back my findings if anyone else is interested.

By the way-- in your post one number is wrong but the math is correct- -- see revised calcs below along with a few missing conversions for those of us not so great at math -- myself included! Hope you don't get perturbed!:

1. 5V @ 15 ma is 75 milliwatts per 8C of temp.
2. 131F = 55C
3. If the temp is 0C (32F), we must raise the temp by 55C to achieve our goal -- so 55C / 8C = 6.875 (or 6.9 for simplicity sake)
4. 6.9 x 75 mw = 517.5mw -- this is the heater goal.
5. Plug .5175 into Ohms Law (I=P/E) and you get ~104ma.
Your maximum current capability must be at least 104 ma -- good thing I'll have 12V access!

Jul 17, 2007
22,198
1,797
5. ### Sensacell Moderator

Jun 19, 2012
2,143
684
The key to this project is not thermal mass, but thermal isolation.
Good isolation (insulation) will reduce the power required and improve the short term stability. "A fluffy ball with some leads sticking out"

6. ### crutschow Expert

Mar 14, 2008
20,878
5,930
Sensacell is correct. Thermal isolation is the key. Mount the crystal and heater inside a cube of Styrofoam at least an inch thick on all sides.

7. ### THE_RB AAC Fanatic!

Feb 11, 2008
5,430
1,311
Yeah if you are heating to 55'C the heat loss will be significant, so better insulation will reduce the power output and also make it more immune to room temp changes like drafts.

I found that styrofoam is a pretty poor insulator at low temp differentials, it conducts heat much better than an air gap. About the best small insulator I found was a loosely wadded up ball of tissue, which is mostly air with very fine paper layers. That's not a great permanent insulator but gives you an idea of a goal, lots of air layers held together by something with poor heat conduction and not letting air flow through.

As for a control circuit if you are after quality I would not use my circuit, that was stable enough for very low heat range (indoor) but for your case I would use a good voltage reference device, and a thermistor sensor and opamp or comparator.

8. ### ErnieM AAC Fanatic!

Apr 24, 2011
7,969
1,827
How accurate do you need this? The Maxim DS3231 has an accuracy out of the box of:

♦ Accuracy ±2ppm from 0°C to +40°C
♦ Accuracy ±3.5ppm from -40°C to +85°C (185 °F)

One problem with the PIC internal RTCC is keeping it running during power outages. The 3231 has a battery backup so it runs on it's own for several years. You read it (has an I2C interface) and load that a power up into the PICs RTCC. It also has a 32KHz output to drive the PIC.

It's around \$10 in single quantities but may be cheaper then building an oven.

Any way you go, there should not be any issues with running a 32KHz signal a few inches across your board.

Feb 9, 2012
122
9
Thanks guys.. Since this would be used to control clock drift in an RTCC, it's not going to kill me if it's not accurate down to 10ms or something.. As long as it would be way better than it's unheated xtal cousin would likely be in the same conditions, that's more or less what I'm after.. It'd be nice if it could keep keep the RTCC accurate to well within a minute per month..

I did read a note on another board indicating that without an OCXO, you can expect a drift of ~3-4 seconds per week..

I'm not sure exactly what to expect with an OCXO equivalent..

However, in further researching the accuracy of these clocks and their standalone SPI/I2C cousins, I stumbled across some RTC's that have built-in TCXO's already in the package.. I had no idea there were such beasties out there.. In doing more searching on the different products and how accurate they are, I give you (in the off chance this may be of interest to someone else out there) their limited specs below :

• Maxim DS3234 (and cousins), temp range 0C to +70C (+/-2ppm accuracy) or -40C to +85C (+/- 3.5ppm accuracy) -- kinda spendy in low qty at over \$8ea (mouser)
• RAMTron Corp's FRAM processor FM31T378, only available in the industrial temp range (-40C to +85C) and has +/-5ppm accuracy (~2.6 minutes per year) -- even more spendy at over \$10/ea in small qty at Mouser
• NXP's PCF2129A (and PCF2127A), +/-3ppm accuracy from -15C to +60C. Way more reasonably priced at about \$2.50/ea in small qty
With that said, I think my homework is done.. I'll just opt for using the NXP part and be done with it.. I'm not gonna cry over \$2 added to my project.. I can live with that.. And if +/-5ppm = ~2.6 minutes per year of accuracy, +/-3ppm ought to be about 2 minutes per year accurate.. Works for me!!

However, this discussion has not been in vain.. I'll keep this in the back of my head for the occasion when I need an accurate time source that is not related to an RTCC.. One nice benefit of going this route is that the NXP part anyway can also provide a nice output clock for the PIC processor as long as it's a power of 2 frequency (32k, 16k, 8k,etc).. I'll need to look into that further.. The others can likely do that as well but I didn't look.

Feb 9, 2012
122
9
Thanks for chiming in Ernie.. I was in the middle of writing my reply when you beat me to the punch.. I'll probably go with the NXP part because it's cheaper and does about the same thing.. I guess I'll be learning up on the I2C bus!

11. ### ErnieM AAC Fanatic!

Apr 24, 2011
7,969
1,827
Aww, sorry I didn't read your post sooner. But congratulations on thinking outside of the box and finding these things for yourself.

I was not aware of that good NPX part. It was clever of them to have both I2C and SPI on the same chip. Seems obvious once you see it but I don't think I've seen that yet (on the slave side).

So thank you for posting!

12. ### THE_RB AAC Fanatic!

Feb 11, 2008
5,430
1,311
You could also consider a GPS module, prices have come down a heap and they have a number of benefits including the fact they are linked to very expensive atomic clocks in the sky...

Feb 9, 2012
122
9
Yeah, I had thought about that but the prices were still higher than I was interested in -- most in the mid teens or higher per unit regardless of qty (more or less).. The nice side effect of that route is that you can just set the time periodically -- perhaps once daily and turn off the GPS unit the rest of the time.

Another option I was pondering was including a WIFI module that would allow the controller to set the time once daily by connecting to a local WIFI network but those prices are actually higher than the GPS units.. I've been thinking about this angle for quite some time..

I think I've settled on the NXP part with the TXCO built in -- price can't be beat at <\$2.50ea and the interface is super easy -- either I2c or SPI.. and I'll be using it to drive my PIC's external clock input.. and it won't eat more than 50ma of current max according to the data sheet..

Feb 9, 2012
122
9
Ok.. Insert foot in mouth..

I did a quick google search for something like cheap GPS module and found this press release from last October promising a sub \$3/ea GPS module that is ready to go (possibly sans external antenna).. Unfortunately the company website doesn't seem to call this part out.. I did pop them a note asking about availability but haven't heard back yet.. I'd be cool to have access to a \$3/ea GPS module with UART/SPI.. Nice!