The results for the RPD-1's are about what I expected: there's little
difference in slope between either a 50 ohm +47nF termination or a 47nF
termination.
The slopes are about 6.5x greater than the values given by Minicircuits.
(8mv/degree = 2.88V/cycle). I assume that they use 500 ohms connected
directly to ground not via a capacitor.
So there's something in NISTs claims of improved slope at least for the
RPD-1.
I suspect that NISTs original 50 ohm terminations were actually 50 ohms
direct to ground not via a capacitor.
Using a series capacitor increases the termination impedance at the beat
frequency substantially over that when the resistor is connected
directly to ground.
Since its is also claimed by NIST and others that reactive termination
reduces the noise, one also needs to measure the output noise spectral
density for the various IF port terminations.
Bruce
Bob Camp wrote:
Hi
Here's some data:
The setup is very simple:
Two oscillators at 10 MHz driving common base / 50 ohm output buffers. The
buffers ensure that the source impedance is really 50 ohms. One puts out 9.3
dbm the other 9.5 dbm. They can be tuned for a beat note in the 0 to 100 Hz
range.
The basic mixer termination filter is a pair cascaded / identical L networks. Both have
10 uh in the series leg and 0.047 uf to ground in the shunt leg. The "audio
end" of the filter hooks straight into a digitizing scope.
The termination options for the mixer are:
1) inductive - just running into the 10 uh of the first L network.
2) 50 ohms - drop a 57 ohm in series with 0.047 from the mixer output to ground
3) Capacitive - 0.047 uf to ground at the mixer output.
The data is computed from the time to cross the center 50% of the output waveform. If the
output is 1V p-p then the data would cover the range -.25 to +.25 volts. I've normalized
it to "volts / cycle". Divide by 2* pi if you want to get volts / radian.
mixer 50 ohms inductive capacitive
ZAD-3 3.51 2.96 9.98
RPD-1 #1 17.77 10.50 18.85
RPD-1 #2 17.40 10.058 18.53
10514A #1 5.796 4.396 10.31
10514A #2 5.826 4.406 10.33
10534A 5.402 4.078 10.88
ZP3-MH 8.06 5.81 11.28
ZAD-1H 7.73 5.93 9.38
Since not everybody has memorized mixer catalogs:
ZAD-3 typical minicircuits 7 to 10 dbm mixer
RPD-1 500 ohm output phase detector (50 ohms is the "wrong"
termination for it)
10514, 10534 HP products from a ways back
ZP3-MH a 13 dbm class mixer, 9 dbm should be under driving it
ZAD-1H a 17 dbm class mixer, should be 8 db under driven.
Bottom line - Capacitive termination helps some parts more than others. The
RPD-1 does not get a real big boost, but the ZAD-1 certainly does. There's no
real way to know what it's going to do without checking your mixer under your
conditions.
A few other notes:
1) The measurement technique slightly under states the slope for the 50 ohm
case. Since the beat note is approximately a sine wave in all cases, the true
slope at zero is a bit higher than this technique indicates.
2) The Inductive termination gives the widest linear region. The output is very nearly an
ideal triangle wave. It would make the best "wide range" phase detector.
3) The terminations are not precise, but they are identical in all cases. A
more purely inductive load could be constructed. The parts are just what I had
lying around.
4) No strange bumps or peaks were detected in the beat notes of any of the
mixers. Never seen one, regardless of what NIST says they have seen.
5) Eventually I'll go back and check the RPD's with 500 ohms. I stuck with 50 simply to
keep everything as "same same" as I could.
6) Sweeping the beat note from 100 Hz to 1 Hz showed no change in the output
amplitude.
7) Contrary to my previous post the peak-peak output voltages are within 10%
for all terminations. Slope and peak to peak are different things.....
8) All mixers are running into essentially an open circuit load at audio. The scope
input is> 1 M ohm and the capacitive reactances are>100 K ohms.
9) No attempt was made to set up directional couplers and figure out what the
"real" input to the mixers actually is. Ditto on playing with series resistors
to improve the match.
So there it is. Anybody else got some data to compare to.
Bob
On Feb 28, 2010, at 3:53 PM, Bruce Griffiths wrote:
My simulations indicate that terminating the Mixer IF port in an RF short (with
both RF and LO ports saturated) increases the beat frequency zero crossing
slope by more than a factor of 2 (exact value depends on mixer component
characteristics) but doesnt significantly increase the beat frequency amplitude
over that with a high value resistive termination. To achieve this the IF port
termination impedance needs to be high at the beat frequency and its
significant harmonics. The value above which the impedance is considered high
depends on mixer details such as transformer turns ratio, RF source impedance,
diode characteristics and RF input levels, etc.
Bruce
Bob Camp wrote:
Hi
Putting The C on the feedback R in a positive gain setup is only going to take the
"roll off" gain down to 1. Doing the same with an inverting amp or using a
series R / cap to ground will drop the gain a lot more in the roll off region.
I would worry about any resistor that's marked as 10K and reads 20K. It's
likely noisy.
A typical DBM has a loss of 5 to 7 db when not in compression. With a +7 to +10
dbm drive that should give you an output of 0 to 2 dbm . The mixer output should
be in the .6 to .8 V p-p range into 50 ohms. You should get about twice that on
the beat note running into a load> 500 ohms. A gain of 20 should be plenty.
That would give you .6 x 2 x 20 = 24 V p-p out of the amp.
If you "rf short" the output of the mixer you may double the beat note again
(total of 4X the 50 ohm value). Net would be a 2.4 to 3.2 V p-p beat note. Anything much
over a gain of 10 would be a problem then. This is one of the cases where 2 X 2 probably
does not = 4, so measurements are indeed in order.
Bob
On Feb 28, 2010, at 1:01 AM, Brian Kirby wrote:
The values in the schematics are wrong for the op amp gain. The drawing was
from an earlier drawing where I made a preamp to start checks on the mixers,
and I sent it to you (Bruce G). Thats when you determined I did not have
enough gain to get near the noise floor. The THAT1512/1646 ICs were ordered to
make a new preamp for the future measurements on the mixers.
When I use the scope and check the outputs of the IC, I have 20 volts peak to
peak, sine-wave. I know from previous readings I see about 500 mv p-p out of
the mixer.
I went down to the bench and the resistors I used were still there (I bought
several taped reels of Dale RN55D resistors when a local business went out). I
used 294 ohms and 14.9 kilo-ohms, for a gain of 50 (the power rails are +/- 15
volts). Also not shown on the schematic is a 0.47 uF cap around the 14.9
kilo-ohm resistor. I think I was trying to limit the bandwidth to around 15
hertz.
Also the resistor going between the op amp and the limiting diodes was marked
10K, its 20K. The diodes are 1N4148. Corrected drawing attached.
This is what happens to time nuts who can only play on the weekend and stay up
all night....and my employer just thinks I party too hard.....for Monday
mornings.
Brian KD4FM
Bruce Griffiths wrote:
The LT1037 is shown with a gain of ~1690x, if this amplifier is used to amplify
the beat frequency signal, it will saturate.
Opamp recovery from saturation is poorly documented and may be very slow.
It would be better to use some diodes in the amplifier feedback network to
limit the large signal gain to 5x (so that the LT1037 remains stable as it
isn't unity gain stable).
This will ensure a somewhat faster recovery from overload as the LT1037 then
avoids saturation and the opamp input stage remains in the linear region.
Bruce
Bob Camp wrote:
Hi
Assuming that the junction of the back to back diodes goes trough a chunk of
coax to get to the counter:
You are forming a low pass filter with the 10K resistor and the coax
capacitance. The LT1037 is quite happy driving a 600 ohm load. You could easily
drop the impedance at that point below 300 ohms. That should give you a faster
edge into the counter.
You also should check the slew rate performance of the 1037. You don't want the
op amp to be slew rate limited.
Bob
On Feb 27, 2010, at 12:41 PM, Brian Kirby wrote:
I am in the process of designing a DMTD system. As an experiment to do basic
measurements on the chosen mixer, I used a capacitor (0.01 uF) in series to
ground with a 47 ohm metal film resistor. Where the capacitor and resistor
meets, another resistor is attached (390 ohms) that goes to ground. The idea
is to provide a 50 ohm termination at 20 Mhz and a lighter termination at audio
frequencies. I seen this is a NBS note and I can say, its a starting point for
my experiments.
This (my) system is designed for 10 Mhz, using a 10 hertz beat. A schematic is
attached of what I am experimenting with at the moment. A HP5370B is the
recording instrument. The noise floor from 1 days observations show 2x10-11
at 0.1 seconds, 2x10-12 at 1 sec, 5x10-13 at 10 sec, 6x10-14 at 100 sec,
7x10-15 at 1000 sec, and 7x10-16 at 10,000 secs. It will be interesting when
the project is completed to see how much improvement there will be.
As I understand (or learning..) mixer performance is the key to the DMTD
system. It occurs to me that maybe a capacitor designed for 50 ohms at 20 mhz
may be a better termination (for the IF port) for this mixer. A 16 pF
capacitor is 50 ohms at 20 mhz, and for comparison at 10 hertz, it would be 100
meg-ohms, which would give maximum amplitude at 10 hertz. As I understand, a
capacitor terminated mixer will give a triangle wave output, which is very
beneficial to the design - as the end result is to get maximum slope out of the
mixer. I would say, unqualified as I am, the capacitor termination matches the
20 mhz signal, and helps attenuates the harmonics of the mixer, and has no , or
very little effect on the audio frequencies that we are interested in.
And saying/rambling on... that if maximum slope is needed, its needed on the 10
hertz beat signal - so maybe a capacitive termination on the 10 hertz signal
only and something resistive on the 20 mhz signal........another idea use the
16 pF direct off the mixer, then a series resistor for isolation and then a
large capacitor on the 10 hertz beat for maximum slope.
At the present, I am awaiting parts to build a low noise preamp base on the
THAT1512 so I can make better measurements on the mixer. Bruce has provided a
lot of good suggestions and helpful comments on my project and Ulrich has
provided me quite a bit of user support on his program, Plotter. Thanks to all.
Comments ? Brian KD4FM
<DMTD_Plans.pdf>_______________________________________________
time-nuts mailing list -- [email protected]
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- [email protected]
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- [email protected]
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
<DMTD_C_Plans.pdf>_______________________________________________
time-nuts mailing list -- [email protected]
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- [email protected]
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- [email protected]
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- [email protected]
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
_______________________________________________
time-nuts mailing list -- [email protected]
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
