Detecting the steep drop in oven heater current when the temperature regulator loop assumes control shouldn't require an extremely low tempco sensing resistor.

In principle PWM DACs can offer high resolution at relatively low cost, one tradeoff being the settling time of the low pass filter between the PWM output and the OCXO EFC input.

PWM DACs are inherently monotonic even when they have very high resolution.

Bruce

Tobias,

You have not shown the rest of your GPSDO, but presumably it includes a
phase detector and processor which implement a phase locked loop that
compares the oscillator phase with a 1 pps signal from the GPS.

For open loop VCXO testing, you will certainly need to be concerned with
the tempco and stability of the DAC and associated circuits. But for a
GPSDO, because the DAC and associated circuitry are inside the PLL, you do
not need to be excessively concerned with the temperature stability of this
part of the circuit.

The PLL will generally have a time constant of hundreds to a few thousand
seconds, so the DAC can be very slow. Other designs use a slow but high
resolution PWM DAC because it is inexpensive and guaranteed to be
monotonic. If you want to stick with a hardware DAC you might consider the
18 bit AD5680, which is a 16 bit DAC that interpolates 2 additional bits.

On Mon, Oct 7, 2019 at 6:02 PM Tobias Pluess tobias.pluess@xwmail.ch
wrote:

--

--Jim Harman

Hi Bruce,

oh yes I know, for sure no special resistor will be used for the oven heater!
By the way, many GPSDOs I have tested (e.g. Trimble Thunderbolt, STAR4) show info about the OCXO temperature when connected via Lady Heather. Do you know how they estimate the temperature? I checked almost every IC on the STAR4 PCB and was unable to find a temperature sensor. Is it possible to estimate the oven temperature from the current drawn? (for sure it is, but most likely it will be strongly dependent on the particular OCXO used, I think).

OK, so using a PWM DAC is more a matter of cost than something else. In this case I think I am fine with a DAC8501 or DAC8560, which are both based on a resistor string, and therefore are also monotonic by design.

Tobias

From: time-nuts [time-nuts-bounces@lists.febo.com] on behalf of Bruce Griffiths [bruce.griffiths@xtra.co.nz]
Sent: Tuesday, October 08, 2019 00:19
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] Question for my new GPSDO

Detecting the steep drop in oven heater current when the temperature regulator loop assumes control shouldn't require an extremely low tempco sensing resistor.

In principle PWM DACs can offer high resolution at relatively low cost, one tradeoff being the settling time of the low pass filter between the PWM output and the OCXO EFC input.

PWM DACs are inherently monotonic even when they have very high resolution.

Bruce

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.

Hi Jim,

oh yes, I have not shown yet the full schematics because I am not finished yet. But sure, it includes a phase detector. I plan to use a STM32 microcontroller with a timer being used in capture mode, and I plan to add a TDC7200 as external interpolator. (I am still thinking about the additional glue logic needed to interface with the TDC7200, because this IC can only measure up to 8ms, but with 55ps resolution).

So the OCXO will not operate in an open loop. However, I found that the tempco of the resistors I have is 100ppm/K. So, even if the OCXO is operated with a PLL, if the temperature in my lab changes (e.g. door opened, window opened, whatever), the resistors change so much that the resulting change in voltage comes close to 1..2 LSB of my DAC, i.e. the frequency will be slightly off and the PLL will need to correct for that. Sure the PLL will detect the frequency error and correct it, but since the time constant will be some 10min or more, it will take a while until the frequency settles to the new value.

So I wonder whether my assumption about the resistor tempco is correct, and the commercial GPSDOs use expensive low-tempco resistors, or whether I am exaggerating a bit.

BTW, thanks for the tip with the AD5680. I will plan this as an alternative, so that any of DAC8501/8560/AD5680 can be fitted.

Tobias

From: time-nuts [time-nuts-bounces@lists.febo.com] on behalf of Jim Harman [j99harman@gmail.com]
Sent: Tuesday, October 08, 2019 01:11
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] Question for my new GPSDO

Tobias,

You have not shown the rest of your GPSDO, but presumably it includes a
phase detector and processor which implement a phase locked loop that
compares the oscillator phase with a 1 pps signal from the GPS.

For open loop VCXO testing, you will certainly need to be concerned with
the tempco and stability of the DAC and associated circuits. But for a
GPSDO, because the DAC and associated circuitry are inside the PLL, you do
not need to be excessively concerned with the temperature stability of this
part of the circuit.

The PLL will generally have a time constant of hundreds to a few thousand
seconds, so the DAC can be very slow. Other designs use a slow but high
resolution PWM DAC because it is inexpensive and guaranteed to be
monotonic. If you want to stick with a hardware DAC you might consider the
18 bit AD5680, which is a 16 bit DAC that interpolates 2 additional bits.

On Mon, Oct 7, 2019 at 6:02 PM Tobias Pluess tobias.pluess@xwmail.ch
wrote:

--

--Jim Harman

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.

Some OCXOs have built in temperature outputs, sometimes its just a pir of wires connected to the ends of an internal thermistor.

Bruce

Hi Tobias,

Very interesting project you have. Not just that it's a fresh GPSDO, but
interesting because you seem to be taking so much care in the design. A
couple of comments:

1. The need for ultra low tempco is more important if you expect your
   GPSDO to be in holdover often; less so if it is permanently locked to GPS.

One approach is to spend the money for low tempco devices and trust
them; another is to monitor the temperature and apply calibrated
corrections in s/w. I can't predict which is better, but don't discount
the latter approach.

Still, either method will have to be verified with actual experiment,
strip charts, correlation analysis, and ADEV plots. So I hope you have
the test equipment at hand to measure these subtle effects in a GPSDO.

2. I suspect one can spend a great deal of time picking and testing
   optimal parts in a GPSDO based on tempco specs. But at what point is it
   cheaper to just control the environment of the GPSDO board itself?

You mention opening lab doors and windows and such. Wouldn't it be
simpler to spent time to design a box that is wind-proof, or
fan-controlled; maybe even oven controlled? That way you can relax all
your worries about exotic passive and active components and just build a
controlled enclosure.

That could not only take care of temperature, but humidity and pressure
as well if they were found to be sources of instability. One advantage
of this approach is that the same box design could then be applied to
any other T&F projects like distribution amps or phase / frequency
counters that you design in the future.

3. About sensors.

about the OCXO temperature

The Trimble Thunderbolt (TBolt) uses an onboard DS1620 temperature
sensor. I don't recall it ever being called "OCXO temperature"; it's
ambient board temperature. Perhaps because the board is inside a larger
aluminum enclosure the board temperature is related to OCXO case
temperature. But as far as I know the reported temperature value from a
TBolt is definitely not the temperature of the crystal resonator itself.

/tvb

Hi

What is the temperature stability of the OCXO you plan to use? In the case of
open windows, what is the dynamic temperature stability of the OCXO?
What is the tuning range of the EFC on the OCXO?

The wider the EFC range on the OCXO, the more the tempco of the control system
impacts the output. With a small enough range, you may not need a very good
DAC at all. With a super big EFC range then the requirements go up.

As long as the EFC stability is better than the OCXO’s temperature stability, there
really is no benefit to improving it.

One example of all this is not an OCXO, but it’s the same sort of idea. A Telecom Rb
might have a tuning range of +/- 2 ppb. It also might have a tempco of 0.1 ppb over
a 50 degree range. That comes out to 0.002 ppb / K. It also comes out to 1,000 ppm/K
in terms of the tuning range. (2/0.002). 100 ppm resistors / Dac’s / references would
be overkill in this case.

None of that gets into the issue of dynamic change. Pretty much every component
you will find is rated for a gradual rather than fast change. The impact of a fast change
can be orders of magnitude worse than a slower change. ( so indeed, don’t open the
window :) )

=======

If you want to go a bit nuts (this being time nuts):

A change in the temperature changes the current through the OCXO heater. That
current flows through a ground pin. The ground pin has a resistance. On pretty much
all OCXO’s that ground pin is also in the EFC ( = it is the EFC return). Thus ground
current “tunes” your OCXO.

However you go about measuring the tempco of your OCXO, be careful that the ground
current is handled the same in your test as in your final circuit …..

Bob

Jim mentions the LM34 for sensitivity and LM34 or LM35 for cost compared
with the DS1620. But also look at the BME280 : it has digital measurement
of temperature with 0.01C resolution (50 times better than the DS1620),
costs £5 on digikey but less mounted on a breakout from ebay, and also
measures pressure. The very similar BME280 also measures humidity.

On Tue, Oct 8, 2019 at 5:01 PM Bob kb8tq kb8tq@n1k.org wrote: