Aug 11, 2015
CWTD Episode #73
... Accuracy & Stability in
the Ham Shack
Info, or go
directly down to the
CWTD VCXO Kit
episode is all about "Frequency", a fundamental element of our lives as
hams. In our typical fashion, we are addressing the topic with
techniques, devices and equipment to measure frequency with relatively
good accuracy. Being able to determine its stability.
Various time & frequency projects that are price-achievable and quite
usable in the shack. And "disciplining" approaches to improve
performance (such as phase noise). The nifty project this time is
the "CWTD Voltage-Controlled Crystal Oscillator", or VCXO ... a module
that will serve as the basis for ultimately producing our very own
GPS-Disciplined Oscillator for super stability and accuracy!
73, George N2APB and Joe N2CX
Listen to the Podcast
and download to your computer for best listening results)
Text discussion during the show ...
<20:38:32> "Ken - VA3KMD": a pity the NorCal FCC-2 is
no longer available
<20:45:14> "KD2E (Dave)":
<20:46:59> "Ken - VA3KMD": I think CHU has the same
stability and accuracy as WWV and has aa good signal on the east
<20:47:33> "Pete WB2QLL": Klockit is local to me and
they have an outlet store, and coupons and senior discounts are
<20:48:31> "Pete WB2QLL": I've bought the atomic
clock movement and it works fb
<20:51:24> "George N2APB": Yup! Ken, do you know of
any "CHU" receiver modules, such as we see for WWV? And are there
atomic clocks that also decode AM-encoded (or phase encoded) time
<20:51:59> "KD2E (Dave)": I've gutted about 10 normal
analog clocks, and retro-fitted with the WWVB (atomic) movement.
Pick the face thickness, and length of hands, second hand
<20:52:44> "Dave AA3UR":
<20:56:19> "KA7OEI": CHU uses AFSK - see the
<20:57:23> "Dan - KB9JLO": What about the HP
<20:57:31> "KA7OEI": The 100 Hz IRIG code on WWVH is
similar to WWVB - but the bits are in a different order. The former
is what the Heathkit clock used to use.
<20:58:48> "Joe N2CX": The HP unit is quite good as
well if available.
<20:59:39> "Dan - KB9JLO": I've been keeping one to
make a GPS-controlled_oscillator
<21:03:00> "KD2E (Dave)": Is the 60 cycles of line
voltage accurate these days to use as counter check....or is the
counter probably more accurate than the line?
<21:03:46> "Dan - KB9JLO": Remember that the 60 cycle
time signal is often "varied" to control generation on the grid.
Also loads on the grid affect frequency.
<21:03:50> "Larry - W2HHV": Video demo of
procedure for adjusting a secondary frequency source against a
primary frequency source: EEVblog #457 - Oscillator Calibration
<21:03:51> "Clint - KA7OEI": The line frequency is
"unregulated" these days: OK in the west, less good in the East and
can be terrible in Texas.
<21:04:16> "Joe N2CX": 60 Hz power on average is
stalbe but not nearly as good as a good xtal source. Besides, it is
only accurate "on average".
<21:05:00> "Dan - KB9JLO": It's not exactly
unregulated Clint. The utility companies use it to 'adjust' the load
on the grid and generation. I never realized what it was like till I
spent some time in that industry...
<21:05:38> "Dan - KB9JLO": Back when NY went dark
that one year (???) you should have seen the dip in frequency that
<21:06:57> "Clint - KA7OEI": When I was in the
industry WAPA (here in the west) held long-term pretty well. A few
years ago long-term stability was "relaxed" - whatever that is. On
the "Time Nuts" group a few years ago they talked about how
line-referenced clocks on the Texas grid were >10 minutes off in a
month: Here in Utah I seem them move 10-20 seconds/month.
<21:09:17> "Joe N2CX": For time reference, 1 ppm
equates to 2.59 seconds per month.
<21:14:09> "Clint - KA7OEI": Here is a "less
<21:18:14> "Dan - KB9JLO": I've got the PCB &
parts to do this one (someday)
<21:20:05> "George N2APB": I've observed the
time-nuts list discussing the HP OCXO mentioned here. Very nice.
(Do you have one?)
<21:22:09> "Steve, N0XC": I have the HP oven unit
inside the HP105B crystal standard, and it is very nice.
<21:23:12> "Pete WB2QLL":
History of Frequency Counting techniques ...
Did you know there are two types of frequency counter?
The method you'll probably think of first is to count the input
edges of the signal you want to measure and time a precise delay
e.g. 1s then the frequency you measure will be in Hz.
There is another method called reciprocal
counting where you count the edges of the internal clock instead
(used in commercial counters).
Normal direct frequency counting
The obvious method is to feed the measurement
signal into a counter chip which is turned on for a set
period e.g. 1 second.
The counter value is actually the frequency measurement
since: f = events/time =
counter value/1 second = counter value in Hz.
The problem with this method is that the
resolution of the counter is 1Hz and the number of digits
displayed is dependent on the input frequency:
A 1Hz input gives 1 digit.
A 100Hz input gives 3 digits etc.
The resolution is in Hz (dependent on the
gate time -which must be a multiple (or sub-multiple) of a
second) - this is why the reciprocal counting method is
better as it gives a resolution in terms of the master clock
frequency inside the unit that is not dependent on gate time
and the number of digits is also not dependent on the input
Reciprocal frequency counting
The second method is reciprocal counting
where instead of counting the input signal you count periods
of a master clock and instead of counting the input signal
edges you let the input signal start and stop the counter.
For example the counter is started at the
rising edge of the input signal and stopped at the next
rising edge. Now the counter reading is actually the period
of the input signal in multiples of the master clock.
Its called reciprocal counting since you have
to work out:
f= 1.0 /
... to work out the frequency.
A simple reciprocal counting example.
Let's say your internal counter runs at 1 MHz
and you have an input signal that has a frequency of 1345Hz.
The number of 1MHz periods in 1345Hz is:
Tsig/Tmeas = (1/1345)/(1/1e6) or
1e6/1345 = 743.49
We lose the decimal point to get:
743 counts of our 1MHz counter.
So in the microcontroller we have a value of
743 and to show the frequency on a display we need to
calculate the following equation:
1/(743*(1.0/1e6)) = 1345.895 Hz
which can also be written as:
1e6/743 = 1345.895 Hz
... and the easiest way to do that is using a
floating point library.
slow, slow so
consider using fixed point maths.
The reason for using reciprocal counting
Reciprocal counting is more difficult as you
have to use floating point routines to work out the
frequency but because the counter counts edges of a master
clock the resolution is fixed in multiples of that master
clock and not dependent on the input signal i.e. the
frequency counter will show all digits regardless of the
frequency of the input.
The resolution of the measurement is better
since in a normal counter the count will be out by ±1
Hz count whereas the reciprocal counter will be out by ±1
master clock cycle.
Note: This does not mean that it is more
accurate - accuracy depends on the clock stability of the
Reciprocal counters are useful for period
measurements and you will find for commercial counters that
they usually have two inputs labeled A and B. You can set
up the frequency counter to just measure A frequency on A or
on B but you can also do signal period measurements e.g.
trigger on the rising edge of A to get the high period etc.
In addition you can do measurements between
different signals e.g. rising edge of A and only after that
has triggered stop the count on the rising edge of B - that
would be useful for measuring timing characteristics for a
peripheral e.g. RAM access time.
- See more at:
Theory of Operation
Count number of cycles of input signal for prescribed
Resolution depends on input gate time
seconds for 10Hz, 1 second for 1 Hz, 10 seconds for 0.1 Hz
Accuracy depends on good time base
Should be no worse than 10 times accuracy desired
- NPO caps, negative or positive temp coefficient
- EMRFD Section 7-12 - drift experiments with oscillator in a
box with light bulb
- Crystal Freq-Time
Curves ... flat spot at room temp or higher, points of
inflection where freq curve vs temp is flattest
Accuracy – how close to desired frequency (parts in 10E6, 10E7
Precision – how many digits?
Precision without accuracy is meaningless
Noise (Phase Noise)
... Here are some very cool
"frequency-related" projects that we've either built or studied before.
1) Simple example homebrew project from ei9gq ...
This simple frequency counter project was
in Homebrew, RadCom October 2006. The original design
uses a gate time of one second which gives a count
resolution of 1Hz. Maximum input frequency is about
50MHz. A VHF/UHF prescaler for this counter is described
in Homebrew, RadCom February 2008. U2 is a PIC16F628
(PIC16F628-20 or PIC16F628A). The PIC crystal frequency
is 10.240MHz. The firmware is easily modified so that a
10MHz crystal can be used instead. Homebrew for March
2008 shows how this counter can be used with a GPS or
Loran controlled gate.
2) EBay freq counter ...
a temperature compensated timebase ... "TCVXO"
See this real nifty frequency counter in use with our
CWTD "GPSDO" project lower on this page.
MANUAL, as constructed from information on
the eBay page, and our own experimentation.
3) HP-5328B Frequency Counter ...
Don't let this eBay price intimidate you. N2APB
got his for about $150 at a hamfest, and it works great!
The best parts about this "Universal Counter" is that it
- Can measuring frequency down to .01 Hz, it can
- Can measure time difference between two pulses
- Can be clocked from an external source
(like a 10 MHz OCXO or Rubidium Standard ;-), and
- It has an HP-IB interface port on the back for
automated measurement taking.
This is a REAL DEAL in the realm of frequency
B) WWVB Receivers
We've talked about WWVB in a previous episode
"What Time Is It?", but here in this Frequency episode, the
value and meaning of time (and frequency) synchronization with a
known-good source like WWVB takes on whole new meaning.
1) PV Electronics "WWVB
60 kHz Time Receiver Module and Antenna"
Many in our homebrewing
field had been lamenting the "unobtanium" nature of an
older-yet-popular Digi-Key item (CMMR-6P-60), but we
recently found another one over in the UK, from PV
Electronics. I have several of these on order
2) N0BHQ "WWVB Receiver"
See full size schematic, description
and PIC decoding software at ...http://www.ringolake.com/pic_proj/WWVB/WWVB_ISO.PDF
PCBs available from OSH Park
Hans Summers Projects ...
This guy is amazing, and his web pages
have tons of great projects with fantastic (low) prices.
Of particular interest to us here on this topic is his
homebrew OCXO Design:
We will revisit this some more later on
2) N2APB & N2CX "OCXO" Project ... an
Ovenized Crystal Oscillator using a Trimble OCXO Module
"Trimble 34134-T" OCXO ... packaged neatly and perfect for the
Trimble 34134-T OCXO, ~$15 ...
3) LPRO-101 Rubidium Standard 10 MHz
Rubidium Frequency Standard on eBay is a simple deal (and a GOOD
one!) for about $175.
Having a frequency
standard that can achieve such high accuracy, and is stable
N2APB packaging of the LPR-101
Rubidium Frequency Standard "Primer" ...
Our good CWTD listener Clint, KA7OEI, has an
excellent primer on Rubidium Standards, and use of the
Efratom LPRO-101 ...
Measurement In Action]
1) Frequency Standard and Measurement Set-up ... at N2APB
2) Zero beat oscillator with WWV at 5 or 10 MHz
Tutorial #58 from Alan Wolke, W2AEW ... How
to zero-beat WWV to check or adjust a Frequency Counter's
Good at best to 1 part in 10E7 depending on skill and
Our CWTD Project
this time ... a
(Click photo for bigger view)
YouTube video of VCXO Breadboard being tuned up
... Check out how our Breadboard of
this circuit actually works ...
(Be sure to turn up your computer's audio volume for this one ... I
had a FUBAR moment during video recording! -- n2apb)
CWTD "GPS-DO" Project:
A GPS Clock, or GPS-Disciplined
Oscillator (GPS-DO) is a combination of a GPS receiver
and a high-quality, stable oscillator such as a quartz or rubidium
output is controlled to agree with the signals broadcast by GPS and GNSS
satellites. GPSDOs work well
as a source of timing because the satellite signals must be accurate in
order to provide positional accuracy for GPS in navigation. These
signals are accurate to nanoseconds and provide a good reference for
the course of this, and the following two episodes, we will be
homebrewing our very own GPS-DO. The three phases are ...
Phase 1: VCXO ... A Voltage Controlled
Crystal Oscillator ... Tonight's episode! (Aug 11)
Phase 2: OCXO ... A
temperature-controlled "oven" + a VCXO = an OCXO (Oven Controlled
Crystal Oscillator) ... Episode airs on September 8.
Phase 3: Disciplining ... Phase-Locking
the OCXO to a stable and accurate frequency ... Concluding the
series, this episode will air on October 12.
PHASE 1: VCXO
The "Crystalizer"... A Voltage Controlled Crystal
Here is "phase 1" of our overall GPSDO project, the VCXO
(Click image for larger view.)
Some time ago, we developed a 10 MHz voltage
controlled crystal oscillator, or VCXO, for use as a fun project at the
immensely popular Atlanticon QRP Forums. This VCXO is able to be
put to use in many ways around the shack. For example, once calibrated, this
VXO can serve as an accurate frequency standard for receiver alignment,
or as a PLL standard, or even as an LO for a transmitter.
view/download the PDF version of the schematic and all kit information
The CWTD “VCXO”
produces approximately 1-2V p-p at
when C3 is
in the top position (C3a on the pcb). This signal is quite ragged but
very suitable for driving a balanced mixer like an SA612 (e.g., in
mixing applications). When C3 is placed at the “C3b” position, a much
more sinusoid-like100-200 mV p-p signal is delivered at RF Out, which is
more suitable for use as a “standard” 10 MHz oscillator standard in the
Capacitors, crystals and semiconductors have
temperature coefficients that represent how the component values change
with changing temperature. When the component values change, the
oscillator frequency changes. This drift is not good if you want a very
stable and unchanging oscillator frequency.
So the real fun challenge for the builder,
and the excitement with our resurfacing this project here on CWTD, is to
find ways to dynamically adjust, compensate or insulate the circuit
against temperature changes. The Adjust signal on connector P1 pin 1
is delivered by a simple potentiometer that presents a nominal 3.4 volts to
D1, which is a "voltage variable capacitor" commonly called a
But this only a static level and if the temperature changes, the pot
setting would need to be changed to compensate for that change . One
could instead dynamically adjust that control signal in accordance to
the ambient temperature. Or perhaps use of NP0 “negative
coefficient” capacitors in the circuit could compensate for drift
.. which is what we've done now with the CWTD version of the circuit.
But other techniques will come along in Phase 2 and Phase 3 for this
Construction of the
VCXO is straightforward – just use the
schematic as a guide for placement of the components at the silkscreened
locations on the board.
Resistors are mounted “on end” with the top lead
bent over and going into the hole next to the bottom lead. Be sure
to use your ohmmeter to double check your guess at the color
coding on these resistors. You may be fooled with the
47-ohm and 12-ohm resistors! ;-)
NP0 ("NP Zero")
capacitors are supplied in the "outer" kit bag. NP0 caps
have zero drift with temperature (or at least to +/- 30ppm/Deg-K
... which is pretty darn close to zero at the temperature ranges
we'll be ultimately operating the oven (in Part 2).
a 2.1mm coaxial power jack is also provided for convenience.
Just plug in your standard 12V power source on the bench to
power the whole project.
Place R11, the 500-ohm, 10-turn trim pot into the used pins 3, 4
and 5, as shown in the top view photo. Then add the wire
jumpers to the bottom of the board as shown in the bottom view.
(Alternatively, you could mount the trimpot off-board, as we did
in the Breadboard photo at the top here, but it's convenient to
have it located firmly right there on the board.) BTW,
note how I placed the ground jumper, as a bare wire "outboard"
of the board, able to serve as nice clipping point for attaching
your scope and/or frequency probe!
NEW ... In order
to have your VCXO ready to be combined with the Oven Control
circuit board coming in the next episode, please mount the
crystal about 1/4" off the bottom of the board, as shown
in the photo below. If you have already mounted the
crystal on top, just unsolder the crystal and reattach it on the
bottom. Don't worry, it's easy. Just gently pull one side and
then the other while heating each pad. Then re-attach to the
same pads on the bottom side. And if you don't have 1/4" of lead
length, just solder some bare wire in the pcb holes and then
attach the crystal to these.
What do you put at the L1 position on the VCXO pcb? The
answer isn't obvious! We provide a 10 pF NP0 capacitor for
use at the "L1" position on the
board. This spot was originally intended to be for an inductor,
which would tend to lower the operating frequency of the
oscillator. But as we built up many boards in the
prototyping phase, it was found that the crystal was running a
little "too low" for the adjustment range offered by the
voltage variable cap (diode D1), so we shall instead use a capacitor to allow adjustment up to, and beyond the 10 MHz operating
point. But as Dave AD7JT points out, and nicely
illustrates below, it's not quite as simple as that!
Figures above: Finding the right capacitor value to
place at the L1 position on the VCXO pcb to provide
adjustability above/below the target 10,000,000 Hz point.
Dave AD7JT writes: "I
had to do some cut-and-try stuff to get the frequency range to
include 10 MHz. The initial frequency range with "L1" set to 10
pF didn't come within 2 KHz of 10 MHz. Shorting L1 got it in
range but just barely (took about 4V on the Adjust line). I
wanted 10 MHz to be at about 1.5 V or midway on the DAC output
range for a dsPIC. I finally settled on a value of 68pF for
"L1". The attached shows my readings for all the values I
tried. In all cases the frequency range was about 700 - 750 Hz
(not enough to even see on a scope)."
just connect up the 8-12V power source (a 9V battery will work for a
short while, with a current draw of about 130 mA), and you'll be off to the races!
remember, you'll want to use a frequency counter to serve as your
adjustment guide ... Tweak the trimpot to bring the VCXO operating
frequency as close as possible to 10 MHz, just as I show in the YouTube
video (link above.)
you don't have a frequency counter?!! Well, get one of those
nifty 8-digit, blue-LED counters shown above (and on our Breadboard) on
order from ...
You just can't beat this
$12 wonder (+about $2 shipping from China!!)
Click either photo for a bigger view.