In a message dated 11/6/2007 16:29:16 Pacific Standard Time,
nebula@telus.net writes:

Hi Peter,

there was some discussion about this a couple of days ago.

All I can add in regards to the Fury GPSDO is: the EFC control  voltage
ground, and heater ground MUST be Kelvin-sensed.

Any ground loop error above about 10 - 50 microvolts would affect the
systems' accuracy on good double oven OCXO's.

Whoever conceived to put both the heater current and EFC ground return on
the same pin must not have been paying any attention in his/her engineering
classes.

bye,
Said

************************************** See what's new at http://www.aol.com

[from two separate messages]

I'm missing something.  MTI has been in this business for a while.  I'm
assuming they aren't dumb.  Why are they still making (otherwise) very high
quality parts with only a single ground pin?

I'm not picking on MTI.  They are just a handy example.  Many of the data
sheets I've looked at have only one ground pin.

My first guess is that it's a backwards compatibility mess.  Ages ago
somebody made a part with a single ground pin and somebody used it in some
telco gear and everybody has been using the same footprint ever since.  If
that's the case, I'd expect there would also be a variation with a second
ground pin to get better performance.

How much better would their part be if they had a second ground pin?

Are they running temperature tests with the EFC pin open so they don't see
any problem as the heater current changes?  (or something like that)

--
These are my opinions, not necessarily my employer's.  I hate spam.

Hal, Said, and others,

I need convincing that this single-ground-pin OCXO worry is
valid.

During the first few minutes of oven warm-up one can expect
large variations in heater current. But during this time no one
is concerned with EFC stability. So we ignore that case.

After that, oven current is relatively constant by comparison.
In a GPSDO the EFC is part of a closed loop. So even if there
is a small voltage offset in the EFC control or return path due to
heater current, it seems to me that it still has no effect on either
the accuracy or the stability of the GPSDO. So we ignore that
case too.

Is the worry, then, about the minor ambient temperature related
variations in steady-state oven current? Or something else?

/tvb

Hal Murray wrote:

Even the FTS1200 only has a single ground pin.
However the heater current is a function of ambient temperature, so it
is possible to correct for the variable ground voltage drop by sensing
either the ambient temperature or even better the supply current and
having the microprocessor "learn" how to correct the EFC voltage for
this effect.

Normal ground drop due to heater current is around 0.5mV max at room
temperature.
This corresponds to about a frequency off set of 1.5 -4.5 E-11 depending
on the model.
Since the heater current could be 5x this at low temperatures the
corresponding frequency offset will be around
7.5- 22.5E-11 depending on model (neglecting the decrease in the wiring
resistance as the temperature drops).

Since the temperature drift is 1-2E-9 the effect of the varying ground
voltage drop is 5 -10 x smaller, it doesn't really matter whether the
EFC pin is grounded or not during temperature testing.

Bruce

Tom Van Baak wrote:

Tom

Even without frequency disciplining it doesn't appear to be a
significant issue (if we accept Peter Schmelcher's data) for OCXOs like
the MTI 260 and 230 series as this effect is swamped by the frequency
tempco due to other effects such as the finite change in oven
temperature due to the limited thermal gain etc.

However for an OCXO using a high thermal gain oven controller (E1938A)
or a double oven (OSA B8607) such effects may be significant if a single
ground pin is used.

For example if the resistance of the shared ground wiring were say 0.1
ohms and the current varied by 300mA over the temperature range the EFC
voltage would change by 30mV with a corresponding frequency change of
3E-10  which is comparable to the max frequency change (2E-10) over the
rated temperature range for a standard 8607 and 10x that for some 8607
options.
With a star ground system wthin the OCXO it may be possible to reduce
the shared ground resistance by a factor of 10 or so.

The conclusion being that the smaller the oscillator tempco the more
significant such effect are.
For most single oven OCXOs this effect isnt likely to be significant.

Bruce

Bruce,

Thanks. Here's my math. An average 10 MHz OCXO has an
EFC range of, say, 1e-7 over 10 V. That's 1e-8/V; 1e-11/mV.

Now look at the short-term frequency error of a Fury:
http://www.leapsecond.com/pages/fury/log31837v.gif
( from http://www.leapsecond.com/pages/fury/ )

You see Fury frequency varies by at least 5e-11 over minutes,
as is normal for this class of OCXO. To make that much noise
with EFC effects alone you'd need on the order of 5000 uV
variation on EFC line. Another reason why I think the single
ground pin question is a non-issue.

The steady-state oven current variations are quite small. See
this plot, for example:
http://www.leapsecond.com/pages/fury/fury-oven.gif

Said, have you measured the ground return loop resistance on
a Fury? We can do the rest of the calculations based on that.

/tvb

Hi all,

being more than 30 years in the Frequency Control business, I want to
comment on what Hal Murray wrote:

Look for example to the very popular "Eurocase" package CO 08, as used
in the MTI 260 (and many OCXO from other manufaturers - like the AXIOM40
& 45 series of AXTAL). This package originated from a TCXO package,
and was first used to make OCXO (as TCXO replacements) in the early
seventies. The connection scheme for the 5 pins was given and included
only one ground pin.
Next generation in OCXO miniaturization (in Europe) was the 20x20 mm
package CO15 (as AXIOM30 & 35), which has the similar legacy, i.e. the 5
pins were defined  for TCXO, i.e. with one ground pin, and later (in the
late 80's) used for OCXO.
The next smaller THD OCXO was in the 4-pin DIL14 package (AXIOM20 & 25),
and it's clear that the 4 pins do not allow 2 GND pins.

Any additional pin means more cost - and as such special versions would
run in much smaller quantities - the cost are even higher.
And - the accuracy and stability requirements for those small OCXO was
(and is) less stringent in most applications such as for telecomms.

The cost situation changes, when it comes to SMD packages (like the
22x25 mm package CO 28 - AXIOM10). For those packages you can indeed
find a few specs on the market, which have two separate ground terminals
for oscillator and heater circuit. Such specs are written by people who
know about the effects of ground loops.
However I do not know any "modern" catalog or standard parts on the
market with two ground terminals.
In fact the impact on stability is mostly neglectable under stationary
operation conditions with rather constant environmental temperature.
For applications where temperature changes are an issue and highest
stability is required, there is no other way than having two separate
grounds.

Indeed, many (if not most) OCXO manufacturers run OCXO temp tests (and
aging tests with the EFC pin grounded. The undesired ground loop effect
can be reduced if the connection is made very close to the pin, and if
the inside connection to the PCB ground has low impedance.

Regards

Bernd  DK1AG

AXTAL GmbH & Co. KG
Facility MOS
Wasemweg 5
D-74821 Mosbach / Germany
fon: +49 (6261) 939834
fax: +49 (6261) 939836
www.axtal.com

Fantastic work on that page as usual, Tom.  Your documentation really sets
the curve (no pun intended) for the rest of us amateur-metrologist types. :)

I've been toying with the idea of disciplining a 10811 with the output of my
Thunderbolt to cut down on the spurs from the latter, so your PN and ADEV
comparisons between the Z3801 and TBolt are especially interesting.

-- john, KE5FX

Tom Van Baak wrote:

Tom

For a single (or other) oven OCXO with finite thermal gain the change in
oven temperature due to changes in heater power are, for small power
fluctuations at least, proportional to OCXO heater current fluctuations.
Thus it is, in principle at least, possible to use the fluctuations in
OCXO power supply current to correct for both the effects of finite
thermal gain and ground voltage variations.

However its probably difficult to impossible to separate the two effects
from ambient temperature fluctuations and the associated OCXO current
and frequency fluctuations.

The real question is off course how large a temperature fluctuation is
required before such effects become significant?

If one makes a few crude approximations:
The MTI 260 series frequency varies by around 1E-9 over a -30 to 70C
temperature range corresponding to a tempco of around 1E-11/C.
Typically the ADEV at 1s is about 2E-11 or so, so such correction is
perhaps only useful for temperature fluctuations > 1 C or so.
Even then such correction is only useful if the temperature fluctuations
are slow enough for the OCXO to track them, whilst the temperature
fluctuations are too fast for the disciplining loop to track them.

Bruce

From: Bernd T-Online BNeubig@t-online.de
Subject: Re: [time-nuts] Fury Interface Board: 5MHz needed?
Date: Thu, 08 Nov 2007 00:04:19 +0100
Message-ID: 47324473.8050403@t-online.de

Hi Bernd,

If ground loops is indeed a concern, you can reduce the effect relating to the
PCB layout by putting a voltage sense point at the GND pin and run it over to
the EFC buffer op-amp and add that voltage. That way the buffer op-amp drive
the EFC to ground-pin voltage rather than the EFC to signal ground (whatever
that is at whatever point you have).

It isn't the same but if you think of it, the only real solution would be to
have a diffrential EFC input. But then again, you can't win them all, and the
relevance may be very low except for a few odd cases.

Cheers,
Magnus

...

I believe the problem is that the EFC closed loop has a significant time
constant, so any current variation that is faster than the EFC loop time
constant will induce an uncorrectable error (at least as far as the EFC loop
is concerned). With a separate ground pin, or a circuit designed to
compensate for the effects of ground pin current, there will be no such
error.

Didier KO4BB

Magnus Danielson wrote:

The HP10543A OCXO EFC connections were isolated from the rest of the
circuit which had a single ground pin.
The anode and cathode of the diode were connected via resistors to 2
isolated pins.

Bruce

From: Bruce Griffiths bruce.griffiths@xtra.co.nz
Subject: Re: [time-nuts] Fury Interface Board: 5MHz needed?
Date: Thu, 08 Nov 2007 14:06:50 +1300
Message-ID: 4732612A.9060703@xtra.co.nz

Bruce,

Such an arrangement would allow the diode to be put into a feedback loop of an
op-amp in virtual ground arrangement if a designer feel it is a better drive
mode.

Cheers,
Magnus

For others on the list, John is referring to the PN plots at:
http://www.leapsecond.com/pages/fury/phase.htm

Let me know when you have the 10811 running from the
Thunderbolt and I'll add that to the set of results. Note that
the Trimble GUI can be used to adjust the EFC parameters
as necessary.

/tvb

Thanks for that explanation.

Note thermal effects have a time constant as well. Depending
on the OCXO, or the enclosure used, it may be shorter; more
likely it's actually longer than the GPSDO time constant. This
is especially true for diurnal temperature changes where the
thermal TC is 100x to 1000x slower than the GPSDO TC!

I understand what you're saying, though. However, in your
scenario of a more rapidly changing temperature, I suspect the
frequency changes due to the OCXO tempco itself (internal
resonator temperature, thermal lag(s), gradients, etc.) will far
exceed any changes in external EFC due to changes in oven
heater current.

You might check this yourself by shorting the EFC pin(s) and
then run some rate dependent temperature tests.

Another check someone could run is to deliberately place a
1R or 10R or 100R carbon resistor in series with the EFC and
see what it takes to make this effect rise above the noise.
With these measured points in hand, you could then extrapolate
down to 0.01R or 0.001R (PCB trace) to see how far below
the noise you are in real life.

/tvb

An illustration of what I meant would be what happens when someone opens the door to the lab where the GPSDO is located. The room temperature may change by a few degrees over a few minutes time, causing the oven to kick in (or out). That kind of change is too fast for the GPS to compensate for, and the increased (or decreased) oven current will drive a frequency change.

Slow diurnal changes would be compensated for, but as Said commented, the loop can only compensate for an error it sees. If there is no error to begin with, there is no error to compensate for.

Didier

---- Tom Van Baak tvb@LeapSecond.com wrote:

This is interesting.

I would like to setup a data logger on a 10811 (with EFC grounded) and record the oven current as well as the output from a 5370 comparing the 10811 to my Thunderbolt and the temperature in my shack. Temperature regulation in the shack is very poor, as it is at the end of the house and upstairs. Large temperature swings are routine, so we should be able to estimate the potential effect of a common ground in the EFC return as temperature changes and see how significant it is compared to the natural OCXO temperature sensitivity.

Only problem, the Thunderbolt is in the same room at the moment, and I am not sure how it's OCXO's temperature stability compares to the 10811. I probably need to put the Thunderbolt into an isolated box.

Currently, I have a 10811 that has been running for about 6 months from a fairly well regulated 18V supply, so all I need is to put a current sense resistor and send that to one of the microprocessor board I have which is already configured to measure temperatures  through standard 10k NTC thermistors (I have 4 inputs available, so I could record the ambient temperature around the 10811 and also around the Thunderbolt). My Visual Basic GPIB data logger can take inputs from the microprocessor simultaneously through a serial port, so I could record current, temperature and 5370 data every second or so.

I see another project coming for this week-end...

Didier

---- Tom Van Baak tvb@LeapSecond.com wrote:

Didier Juges wrote:

Ideally you should use a 4 terminal current sensing resistor and a high
input impedance instrumentation amplifier.
A Thomas style 4 terminal resistor in a temperature controlled oil bath
would be nice, but failing that a low tempco 4 terminal resistor will
suffice.

Bruce

Bruce,

I have some 4 terminal current sense resistors at work, but they are very
low values (0.1 ohm and lower), so what I may gain at the sensor I may loose
in the amplifier. Right now, the 1 ohm resistor and OP-27 give a voltage
that is way out of the noise, as the curve shows. If I can find more
temperature stable resistors for the instrument amplifier, then I could use
one of those 4 terminal resistors. I will see what I can get to improve this
setup. As it is, it may not be extremely accurate in absolute terms, but it
is precise and probably sufficiently relatively accurate for what I want to
do.

What do you think of the 2mA/degree current sensitivity? Is it in the
ballpark? I need to run the test much longer, and over greater temperature
variations, but I am still tinkering with it... The engineer's curse :-)

Didier