|
A gadget that I built a couple of years ago.?
The purpose is to compare an unknown freq to the output of a GPSDO, so that the unknown can be tweaked to match.
It's only useful when the unknown's freq is rather close to a known value.
The circuit does an AND of the GPSDO output with the unknown, thus "mixing" the two signals, then a small MCU measures the difference freq.
Pete
|
I use one source to trigger a scope. Other source goes into vertical. Tweak until sine wave quits moving.Using this technique I had a sine wave take all day to get 10 divisions.
Idea is from NBS or NIST nowdays.
|
On Tue, Aug 2, 2022 at 08:29 AM, saipan59 (Pete) wrote:
A gadget that I built a couple of years ago.?
The purpose is to compare an unknown freq to the output of a GPSDO, so that the unknown can be tweaked to match.
Is the output of the GPSDO just 10 MHz like mine, or is it programmable??
I recently wanted to adjust a 10 MHz OXCO? in a 40 GHz frequency counter. The approach I took was to set my highest frequency signal generator (20 GHz) to 20 GHz, with that locked to the GPS frequency reference. The 20 GHz was then fed into the frequency counter. The fact the signal generator was producing a frequency 2000 x higher than its reference standard meant that any difference between undisciplined OCXO and the GPSDO were magnified 2000 x. So I adjusted the OCXO until the frequency counter read 20 GHz.
I do have a few HP 5370B time-interval counters. One would start the counter with one oscillator and stop it with another. Then if the two signals are at the same frequency, the delay between them stays constant. However, this does not work too well for sine-wave inputs, as the triggering is not consistent enough. Of course, it's easy to convert a sine wave to a square wave .but not so easy to do it whilst not causing jitter of more than the 20 ps resolution.
|
On Tue, Aug 2, 2022 at 07:59 PM, Dr. David Kirkby, Kirkby Microwave Ltd wrote:
Is the output of the GPSDO just 10 MHz like mine, or is it programmable??
Programmable. The 4 buttons on the front panel talk to the MCU, which allows setting the GPSDO freq. As I recall, it is settable from something like 1 KHz to 30 MHz. The little "square wave" on the display is a representation of the difference freq - basically a simple o-scope display, with a sweep speed of 30 seconds. So, when the freqs are very close, only 1 or 2 cycles are displayed in 30 seconds. >I use one source to trigger a scope. Other source goes into vertical. Tweak until sine wave quits moving. Using this technique I had a sine wave take all day to get 10 divisions. Yes, like the classis Lissajous pattern. I have several o-scopes, and an HP 5345A counter, which is mostly a better way to go, but it's not much "fun", eh?. In part, I wanted to prove to myself that I could make an AND gate function as a Mixer, but with digital signals. It works fine! I'm thinking I should add a tiny speaker to the difference signal, so that there is audible feedback to make it easy to tune. Pete
|
Isn't classic mixer an XOR gate followed by an LPF?
Leo
|
For at-a-glance monitoring of
frequency/phase differences between my Rubidium, DOCXOs and GPSDOs
at 10 MHz and 100 MHz I use AD8302 log amp/phase comparator chips
to drive edge-reading centre-zero moving coil meters.? With a gain
control and 10 turn offset control and a precision op amp, I can
see the phase changes caused by a frequency difference of less
than 1 mHz very easily.? It's as sensitive as using a dual channel
scope with the gain turned up high to watch the phase difference,
but a lot more convenient and runs all the time, sitting in a 1U
rack case. It has DC outputs from the phase signal so I can watch
the levels on an external meter.? I suppose I could fit a
microcontroller with ADCs and e-ink display and maybe an ethernet
connection so I could monitor the phase outputs, but the simple
analogue meters and electronics are easy on the eye and the unit
doesn't cause any radio noise. If I want to monitor the outputs to
look for GPS phase glitches and diurnal/temperature changes, I
just connect my Fluke 8845A to the output and log the output from
that.
I sometimes use a similar approach to
Dave's method, using a receiver at? around 10 GHz to monitor a
harmonic from a 10 MHz source using a x4 multiplier and bandpass
filter, then a x3 multipler and filter to 120 MHz, then use that
to drive a snap-recovery diode comb generator. I combine that with
a harmonic from my HP E4433B which I lock to whatever reference I
want to compare the 10 MHz source with.? The beat note between the
two harmonics gives me a very fast way to adjust the frequency of
the 10 MHz source to within a few Hz, which represents a few mHz
at 10 MHz. That audible feedback means I don't need to watch a
meter, so it makes the process of trimming a little more
ergonomic.
I'm considering adding an analogue
voltage to frequency chip and speaker/headphone socket to the
analogue phase comparator so I can get that same hands-free audio
feedback as a tone.? Using the same voltage that drives the meter,
I would get the advantage of the offset and gain controls so a
change of 10 mV per degree phase difference can be mapped to a
change of perhaps 100 Hz per degree of the tone. I can check for
DC drift in the comparators by using the same signal at both
inputs, although it's also amusing to use the same signal to each
input but with different cable lengths, then watch the wobbles and
shifts as you manipulate the cable or warm it up.
Definitely still a work in progress. I
must make a video about it one day before my Rubidium source wears
out and dies.
--
|
On Wed, Aug 3, 2022 at 03:13 AM, Leo Bodnar wrote:
Isn't classic mixer an XOR gate followed by an LPF?
I'm no expert on the theory, but "intuitively" I think both XOR and AND works. I can say for certain that AND works... Pete
|
Yes, indeed. Although, as Leo said, an XOR is a classical mixer,
pretty much anything that's not strictly linear and time-invariant
will perform mixing of some kind. And "some kind" is often good
enough. I've even used a CMOS inverter (74C04) as a mixer. It's not
a great mixer, but it mixes.
--Tom
--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
On 8/3/2022 06:38, saipan59 (Pete)
wrote:
toggle quoted message
Show quoted text
On Wed, Aug 3, 2022 at 03:13 AM, Leo Bodnar wrote:
Isn't classic mixer an XOR gate followed by an LPF?
I'm no expert on the theory, but "intuitively" I think both XOR
and AND works.
I can say for certain that AND works...
Pete
|
Here is my take on XOR versus Inverter as a mixer. The XOR, with two digital inputs (for example divided down from RF signals) will act as a phase detector. The output will be zero when two, 50% duty, digital inputs are exactly out of phase. Okay to be clear, an XNOR will be zero – a XOR will have flat positive output. As the two inputs deviated from that phase and/or frequency, it will generate pulses that can be averaged using and RC or integrator to give you an analog representation of the phase difference. I had seen these in PLLs for example, since that uses a phase detector. ? To use an inverter as a mixer, I am guessing you would do the trick where you bias it up as a linear gain block (two large series resistors from OUTPUT to INPUT, with maybe a bypass cap to ground at the miid-point). This creates a very high gain (analog) amplifier, so when you combine two input signals and apply to the input, it will bang rail to rail, basically it’s a nonlinear gain block, and you get that square law behavior that is the classical mixer: (SinA + SinB) ^2 which expands to the sums and differences (and some DC). ? I once measured a CD4049 inverter, biased as an amplifier, at 10.7MHz, and was amazed that it was >50dB, and I believe it consumed about 10mA. ? Dan
|
Exactly right, Dan, and very well written, if I may say. An XOR is
popular because it acts very much like an overdriven
Gilbert-type/Jones cell. Both make excellent quadrature phase
detectors.
The inverter is a surprisingly useful small-signal amplifier, but it
has pretty poor large-signal performance. It'll mix whether you want
it to or not. Moving the bias around helps to maximize the 2nd-order
nonlinearity, but that is a little bit like putting lipstick on a
pig. But sometimes that little extra is enough to make it useful for
non-critical applications.
-- Cheers,
Tom
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
toggle quoted message
Show quoted text
Here is my take on XOR versus Inverter as a
mixer.
The XOR, with two digital inputs (for
example divided down from RF signals) will act as a phase
detector. The output will be zero when two, 50% duty, digital
inputs are exactly out of phase. Okay to be clear, an XNOR
will be zero – a XOR will have flat positive output.
As the two inputs deviated from that phase
and/or frequency, it will generate pulses that can be averaged
using and RC or integrator to give you an analog
representation of the phase difference.
I had seen these in PLLs for example, since
that uses a phase detector.
?
To use an inverter as a mixer, I am
guessing you would do the trick where you bias it up as a
linear gain block (two large series resistors from OUTPUT to
INPUT, with maybe a bypass cap to ground at the miid-point).
This creates a very high gain (analog) amplifier, so when you
combine two input signals and apply to the input, it will bang
rail to rail, basically it’s a nonlinear gain block, and you
get that square law behavior that is the classical mixer:
(SinA + SinB) ^2 which expands to the sums and differences
(and some DC).
?
I once measured a CD4049 inverter, biased
as an amplifier, at 10.7MHz, and was amazed that it was
>50dB, and I believe it consumed about 10mA.
?
Dan
|
Another way I have heard about to align two 10 MHz oscillators it to use a VNA. Perform a 2-port calibration of the VNA, which must be fed from a GPSDO or similar. Set the VNA to measure S21, and feed the oscillator you wish to align into port 2. If?the phase of S21 remains constant, then the two oscillators are at the same frequency.?
?
?
|
"The world's largest phase detector"
:)
Cheers
Tom
--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
350 Jane Stanford Way
Stanford University
Stanford, CA 94305-4070
On 8/3/2022 23:22, Dr. David Kirkby,
Kirkby Microwave Ltd wrote:
toggle quoted message
Show quoted text
Another way I have heard about to align two 10 MHz oscillators
it to use a VNA. Perform a 2-port calibration of the VNA, which
must be fed from a GPSDO or similar. Set the VNA to measure S21,
and feed the oscillator you wish to align into port 2. If?the
phase of S21 remains constant, then the two oscillators are at
the same frequency.?
?
?
|
This is the method used for the VNWA frequency meter. Also for generation Allan deviations on the fly. Kind regards Kurt ? ? Another way I have heard about to align two 10 MHz oscillators it to use a VNA. Perform a 2-port calibration of the VNA, which must be fed from a GPSDO or similar. Set the VNA to measure S21, and feed the oscillator you wish to align into port 2. If?the phase of S21 remains constant, then the two oscillators are at the same frequency.? ? ?
|
Since I don't have a VNA and constructing one would take more htan a couple of weeks, I'd probably take the following approach using components in my current inventory:
1
|
Since I don;t have a VNA and constructing one in less than a couple of months is a major undertaking, I would probably take the following approach, using components in my junk box.
1) Construct 2 input channels, each with a MAX913 comparator, 2 74HC40102 dividers and a 74HC series gate (for truing up the counter output to be synchronous with the 10 MHz input). This will pruduce a 10 Hz train of pulses (1 kHz if only one 74HC40102 is used).
2) One channel is the start input of a Time-to Amplitude Converter (TAC). The other is the stop input.
3) The TAC outputs (at a 10 Hz or a 1 kHz rate) can be fed to a MCA or examined in some other way to determine the frequency difference.
An Ortec 566 TAC can achieve 15 pS resolution (100 nS range) or 10 pS (50 nS range). Jitter in the MAX913 and the 74HC logic would add some more uncertainty (currently being evaluated for another project). Since the 566 TAC is a general purpuse instrument, a dedicated TAC, built specifically for this purpose might achieve the same or better resolution.
Measurement resolution (15 pS/10S) in the 10-12 area might be possible (15 uHz) and 100uHz might be readily in reach.
Stephen Menasian
|
Oops, It's too early in the morning - Upon proofreading of my posted message, I find that I have mistakenly substituted 10 Hz for 0.1Hz throughout.
Sorry,
Stephen Menasian
|
Oops again. I meant 74HC4059 (/10000 capable), not 74HC40102(/100 capable).
Also note that the Kb control inputs on the counters can be used with a small amount of logic to initialize the counters for each cycle.
Stephen Menasian
|
OOps again - I meant 74HC4059 (/10000 capable), not 74HC40102 (/100
capable). Also note that it is easy to force the counters to a known
starting state for each measurement with the Kb mode input and a small
amount of logic.
Stephen Menasian
|
I have the following:
?
HP 3575A gain-phase meter
HP K34-59991A phase comparator
Tracor 527A & 527E frequency difference meters
?
The 527’s are handy for fast convergence of a significantly off-frequency signal.? Then for fine adjustment both the HP meters are called into use.? The analog output of the 3575A can be fed into a PC for long-term analysis and is handy for setting the crossover point of quartz crystal based standards.
?
Greg
|
This happens to be exactly what a circuit I am currently working on (should) do. My circuit is not as simple; the chip count is 17. If it
works and meets my design objective, frequency comparison will be better than 1e-11. If all goes perfectly, 1e-12 will be achieved. If that happens, I will substitute 3 74AHC chips for 74HC to get into the 1e-13 area. That might be a bit better than what you are looking for but part of the core circuitry is a Time-to-Amplitude converter (TAC). Commercial TACs resolve 10 pS for arbitrary time differences; I am restricting myself to 10 MHz and might do a bit better. I divide the 2 input frequencies by a factor of 1e8 to get a 0.1 Hz output, which is trued up (made synchronous with the input) with a gate. Both channels are identical. The first one to reach the end point starts the TAC, the purpose of which is to interpolate the time difference between the end points of the 2 channels. The astute reader will note that the start points will not be identical; there are several ways of taking care of this - the most straightforward of which is to use the TAC at the start points as well. My circuit (if it works) will get around this with some analog magic. The output is bipolar (the polarity of the analog output depende upun which frequency is higher.
I am not done yet; but the project is currently active.
I hope I've given you a good idea of how it works. Anyway, check to see of a TAC can be used to accomlish what you need. Full documentation will take a while.
Stephen Menasian
|
saipan59 (Pete)
Dr. David Kirkby, Kirkby Microwave Ltd
Kurt Poulsen