Slow recovery with Pulsed noise into alfa

05apr08

Links:
04apr08: Pulsed noise into alfa Mixer
26aug08: Pulsed Noise into alfa horn shows problem is frontend.
11sep08: pulsed noise into alfa horn, modify bias voltage and currents.
16sep08: Remove bias circuit zener, amp3B, modify Vd,I.
18sep08: modify V,I for 3a and 3b.

INTRO: (top)

A pulsed noise source was built to simulate the FAA radar pulses that enter the alfa receiver. A signal generator was sent thru an rf switch that was pulsed with the same values as the FAA radar. The specs for the pulsed noise source are:

Pulsed Parameters
pulse width	9.8 usecs	FAA has 2* 5 usec pulses
ipp	317.5 Hz	FAA ipp about 2500 millisecond
BurstLen	100 millisec	FAA rotation take about 80 millisecond to sweep by the AO
Burst period	10 secs	The FAA Rotation period is 12 secs
AvgPwr reduction from duty cycle	-25 db -44 db	during the 100 Millisecond burst (31.710/1e5) Averaged over 10 seconds (31.710/10e6)

04apr08 : Pulsed noise into alfa Mixer (top)

Setup:

On 04apr08 the broadband noise plus a pulsed CW signal centered at 1340 MHz was injected into alfa Beam3a,b at the input of the alfa filter bank. The filter bank was set to wide band mode (no filter). The input level was measured using a spectrum analyzer in peak hold with rbw=3MHz at 1340 MHz. The alfa cover was placed on the front of the horn.
1 second spectral averages were taken with the wapps in 3 level 100 MHz bw , 4096 channels centered at 1375 MHz. The channel spacing was 20 kHz The frequency resolution was 40 KHz after hanning smoothing. Data was taken for 50 to 300 seconds at different input power levels.

Processing:

Input all the 1 second records for a scan. Hanning smooth. The channel spacing is 20 KHz and the resolution is 40 KHz.
For each 1 second record:

Find the peak spectral density at 1340 MHz (the pulsed signal frequency) in a 20 KHz channel. Search within 7 channels (140 KHz) of 1340MHz.
Compute the median spectral density between 1338-1342MHz (excluding the 7 channels around 1340 MHz). Use this to normalize the peak values to Tsys.
Compute the mean spectral density between 1345 and 1365 MHz. Use this as the off pulse data.

Plotting the data:

Page 1: A 1 second spectral average when the pulse was off. Black is pola, and red is polB. Each plot is a different beam.
Page 2: A 1 second spectral average when the pulse was one. Beam 3a,b was the only beam connected to the pulsed signal source.
Page 3: A 1 second spectral average blowup showing the pulsed signal profile. The channel steps are 20 KHz.
Page 4: Mean spectral density vs time (in units of Tsys) for 1345-1365 MHz (outside the pulsed signal frequency range). Any slow recovery times should be seen here. The vertical scale is +/- 1% of Tsys. The dotted green lines are where the pulse occurred.
Page 5: The peak spectral density at 1340MHz vs time. Only Beam 3 should have the pulses. The other beams will have pulsed data if there are radiation leaks..
-35 dbm input (.ps) (.pdf) :

Off 1340 time series:No pulses are seen in the off band data.
On 1340 Peak value: pulses seen in beam 3, 4,6all beams.

-15 dbm input (.ps) (.pdf) :

Off 1340 time series:No pulses are seen in the off band data.
On 1340 Peak value: pulses seen all beams. The mean level for beam 3ab is now 1.3 rather than 1. Looking at the off pulse spectra you see the 1340 power. This is caused by leakage of the input signal through the rf switch that is causing the pulsing.

-13 dbm input (.ps) (.pdf) :

off 1340 time series: No pulses.
On 1340 Peak values. Seen in all beams. beam 3 mean value has increased to 1.6 Tsys.

-10 dbm input (.ps) (.pdf) :

Off 1340 MHz time series: Pulsed power now seen in beam 3a every 10 seconds. The pulsed power is creating harmonics/intermods that are falling in the 1345-1365 band.
There may be a slight hint of a long recovery time on beam 3a (although it is very small).
On 1340 Peak value: Seen in all beams. Beam 3 mean value is now 2.3 Tsys.

Conclusions:

inpPwr dbm	pulsed pwr in 1345-1365	pulsed pwr at 1340 pk	mean pwr Bm3 1340Pk	avgPulsedPwr in 100ms (-25db)	Notes
-35	no	b3,b4,b6	1	-60 dbm
-15	no	all	1.3	-40 dbm
-13	no	all	1.6	-37 dbm
-10	yes	all	2.4	-35 dbm	Harmonics in 1345-1365 of pulsed power

The peak level for the 1340 signal does not increase as we go from -15 dbm to -10 dbm. This is probably due to the 3 level sampling and the signal being non Gaussian.
The isolation of the rf switch that pulses the signal starts to let some of the cw signal through at -15 dbm input. The -15 to -10 change of the mean peak value of 1.3 to 2.4 is (1.4/.3) = 6.7 db. This is close to the increase of the power from -15 to -10.
The pulsed signal is radiating into the other beams. The cw signal (before the rf switch) is not.
At -10 dbm we start to see harmonics/intermods of the 1340 signal created in the 1345-1365 MHz part of the band. We don't know if this is happening in the down converter box or farther downstream.
At -10 dbm there might be a small increase and then slow recovery.

processing: x101/080404/alfatest.pro

26aug08: Pulsed noise injected into alfa horn shows problem is frontend. (top)

On 26aug08 the pulsed noise source was injected into the alfa front end via the cal coupler.

The setup:

The front end of alfa had the cover put on.
Beam 6 B of alfa was dead.
The rf band was centered at 1400 MHz.
The pdev spectrometer was used with 1 millisecond sampling and two 172 MHz bands each with 512 channels (336 KHz channel widths).

Band 1 was centered at 1450 MHz (clean band with 1440 tone)
Band 2 was centered at 1350 MHz (band with 1300 MHz pulsed tone).

The pulse generator was set to 1300 MHz so that the pulse would be well outside the 100 MHz rf filter which is centered at 1440 MHz.
A 1440 MHz tone was also injected to measure gain variations. It's level was about 70 times Tsys in a 300 KHz rbw.
Data was taken at 3 power levels. The levels were measured at the pulsed frequency (1300 MHz) at the output of the dewar.

The output levels were [-18,-13,-10 dbm].
This correspond to siggen levels of [0,5,8 dbm].

Data was also taken with the 100 MHz rf filter (centered at 1440 MHz) in and out.
24 seconds of data were taken with the following configs:

100 MHz Filter	1440 tone	PulsedPower Dewar Output (dbm)
out	off	[-18,-13,-10]
out	on	[-18,-13,-10]
in	off	[-18,-13,-10]
in	on	[-18,-13,-10]

Data Reduction:

The data was input and averaged to .1 seconds.
The total power was computed across the 172 MHz bands and then normalized to the median power value.
The 1440 sine wave frequency channel was extracted and normalized to it's median value.
For the 100 MHz Rf filter in, 1440 MHz Tone data:

The data was limited to 20 seconds (two periods of the pulser).
the data was rotated so that the pulse started in the first time sample.

This mixed the last part of the 2 second pulse with the data prior to the first pulse. This assumes that whatever happened to the data was periodic with period 10 seconds.

A linear fit was done to the inner 8 seconds of each 10 second pulse to measure the slope of the recovery. This gave two fits per pol/beam for each 20 seconds of data.

The plots:

The plots show the slow recovery time (.ps) (.pdf):

Set 1: 24 seconds of power vs time, 1440 MHz Tone on, -10 Dbm pulsed output Pwr.

Black in polA, green is polB.
The dashed lines are +/- 1% of the plotted power.
Page 1: 100 MHz Filter out. Total power vs time.

The entire IF/LO chain sees the 1300 MHz pulse (up to the pdev filters).
You see a long recovery for beams 1,3,4,5.
There is a large jump before the first pulse of beams 1-5.

Page 2: 100 MHz Filter in. Total power vs time.

The pulsed signal is only seen by the dewar and the 100 MHz rf filter.
There are long recoveries in beams 1,3,4,5,6 at levels that are close to those with the filter out. This shows that the long recovery is in the dewar and not in the first mixer chassis.
There are on large jumps prior to the first pulse as seen in page 1.

Page 3: 100 MHz Filter in. 1440 MHz tone power vs time.

This is the same data set as page 2.
I've plotted the amplitude of the 1440 MHz tone channel vs time. It has been normalized to it's median value over the 24 seconds.
The sine wave was about 70 times Tsys in the 300KHz channel (except for beam 2a which was only 46*Tsys??).
There is a jump and long recovery similar to that seen in the total power.
This implies that the change is a change in gain and not a noise increase.

Set 2: 1440 MHz tone on, 100 MHz Filter in, [-18,-13,-10] db pulsed output Pwr.

The 3 power levels with 100 MHz filter in, tone on was used to measure how the slope of the recovery changed with dewar output power.
The 3 datasets were each truncated to 20 seconds and then rotated to align the first pulse with the start of the data.
Linear fits were done to total power vs time for each 10 second pulse period (ignoring 1 second about each edge).
Page 4: PolA total power vs time with fits

Black is -18 db, red is -13 dbm, green is -10 dbm. The straight lines are the fits.

Page 5: PolB total power vs time with fits.
Page 6: Slope of recovery vs pulsed power level at dewar output.

Top is polA, bottom is polB
The vertical units are Tsys/Second.
The two fits for each 20 seconds of data is shown
The largest slopes are:

beam1A, beam4A, beam3a,b

Conclusions:

The long recovery times are coming from in the dewar.
With a -10Dbm pulse out of the dewar the recovery can be 1% of Tsys over 10 seconds.l
The recovery is a change in gain and not a change in noise since the sine wave tone amplitude also shows the long recovery.
Beams 1a,4a, and 3ab are the worst.
What wasn't resolved.

The recovery lasted for the full 10 seconds.
When the next pulse came, the system had not completely recovered.
The amplitude of the jump was less than what you would get from an unperturbed system.
The slopes of the recovery are probably independent of the 10 second pulse period.

processing: x101/080826/alfatst.pro

11sep08: pulsed noise into alfa horn, modify bias voltage/current. (top)

On 11sep08 the pulse generator (described here) was used to see how the recovery time varied as we modified the bias voltages and currents. The setups was:

The noise generator was set to 1300 MHz. The pulse height out of the dewar was adjusted to -8dbm.
The 100 MHz RF filter (1390-1490) filter was in.
The alfa horn was covered.
Data was taken with the pdev spectrometer at .1 seconds integration with a 172 MHz band centered at 1450 MHz.
Beam 3 polb had its bias voltage and currents modified. The other beams were not changed.
We started taking data with the pulsed noise off. After about 10 seconds we turned on the pulsed noise. It repeated about every 14 seconds. For the tests run in the afternoon we turned off the pulsed generator after about 4 periods and continued to record data so we could see how long it took for the system to return to the initial gain level.
Beam 1 polA was not available (it was used for monitoring upstairs), Beam 6B was dead.
The following tests were done:

Bias set at normal values
beam 3b stage 3 I [19->15 ma]
beam 3B stage 1 I [7->5ma]
beam 3b stage 2 I [10->8ma]
beam 3b stage 3 I [19->22ma]
beam normal settings
beam 3b normal settings, 1415 tubular filter replaces 100 MHz filter.
Afternoon: normal settings.
beam 3b dc block before (regular) filter
beam 3b dc block and circulator before filter
beam 3b stage 3. I [19->15ma] (dc block and circulator still in)
beam 3b stage 3. I[19->22ma] (dc block and circulator still in)
normal settings, back to default filters, circulator/dc block out.

The first set of plots shows the total power versus time for all the beams for each test (.ps) (.pdf) :

each page is a separate test
Black in polA, red is polB. Only beam 3 polB was modified.
The 7 beams are shown top to bottom.
The vertical scale goes .98 to 1.07 of Tsys.

The 2nd set of plots show the total power vs time for beam 3b for each test (.ps) (.pdf):

There are 4 tests per page (except the last)
black is polA, red is polb.
the vertical scale goes .99 to 1.07 Tsys.

Results:

The beams that were not modified had responses that varied:

2b, 3a had the largest variations

After stopping the pulsar after 180 seconds it was sometimes not fully recovered.
Decreasing the stage 3 current 19->15 ma. sped up the recovery. Increasing the current changed the shape of the recovery but did not bring it back to the original gain.
The recovery is most sensitive to the current in the stage 3 amplifier. stages 1 and 2 and smaller effect.

processing: x101/080911/alfatst.pro

16sep08: Remove bias circuit zener 3B,amp3, modify Vd,I. (top)

On 16sep08 the zener diode for amp3 in the bias board for alfa 3b was removed. This allowed us to modify the bias voltage and current. The pulsed noise source was used to measure the long recovery times (described here) as a function of bias voltage and current. The setup was:

The noise generator was set to 1300 MHz. The pulse height out of the dewar was adjusted to -9 dbm.
The 100 MHz RF filter (1390-1490) filter was in. Alfa was centered at 1440Mh
The alfa horn was covered.
Data was taken with the pdev spectrometer at .1 seconds integration with a 172 MHz band centered at 1490 MHz.
The zener diode in the bias card circuitry was removed for beam 3b and the voltage ref was changed to allow for greater than 5 volts.
Beam 3 polb had its bias voltage and currents modified. The other beams were not changed.
For each test we took data for 240 seconds. The rf pulsing was turned off prior to starting the test to let the system recover. After starting data taking, the pulsar was turned on about 15 seconds after the start (to give a baseline). The pulsing repeated every 14 seconds.
Beam 1 polA was not available (it was used for monitoring upstairs), Beam 6B was dead.

The following tests were done:

File	Volts	Current (mA)
000	5.00	19
100	5.25	19
200	5.50	19
300	5.58	19
400	5.58	18
500	5.58	17
600	5.58	16
700	5.58	15
800	5.00	19

The first set of plots shows the total power versus time for all the beams for each test (.ps) (.pdf) :

each page is a separate test
Black in polA, red is polB. Only beam 3 polB was modified.
The 7 beams are shown top to bottom.
The vertical scale goes .98 to 1.07 of Tsys.

The 2nd set of plots show the total power vs time for beam 3b for each test (.ps) (.pdf):

There are 4 tests per page (except the last)
black is polA, red is polb.
the vertical scale goes .99 to 1.07 Tsys.

Results:

The beams that were not modified had responses that varied:

2b, 3a had the largest variations

After stopping the pulsar, we took data for about 150 seconds. For some configurations it still had not recovered.
When changing the current the system took a long time to recover. This is seen in the tests where the final value ends up less than 1.0. During the test the system was still recovering from the current change.
Increasing the voltage made the system recover faster. Decreasing the current at 5.58 v increased the recovery time even more (except for the 16 mA test).
In all case there was always a large jump up of 3-6% Tsys. For v-5.58V and I=17mA (the quickest recovery) it still took about 1 second to get back close to the original value. So playing with the V,I can change the recovery but it is not changing the original jump.

processing: x101/080916/alfatst.pro

18sep08: Bm3a,3b Vd->5.92,I. (top)

On 18sep08 the zener diode for amp3 in the bias board for alfa 3A and 3b was removed. This allowed us to modify the bias voltage and current. The pulsed noise source was used to measure the long recovery times (described here) as a function of bias voltage and current. The setup was:

The noise generator was set to 1300 MHz. The pulse height out of the dewar was adjusted to -9 dbm.
The 100 MHz RF filter (1390-1490) filter was in. Alfa rf was centered at 1400 Mhz.
The alfa horn was covered.
Data was taken with the pdev spectrometer at .1 seconds integration with a 172 MHz band centered at 1450 MHz.
The zener diode in the bias card circuitry was removed for beam 3a,3b and the voltage ref was changed to allow for greater than 5 volts.
Beam 3 pola and polB had its bias voltage and currents modified. The other beams were not changed.
For each test we took data for 240 seconds. The rf pulsing was turned off prior to starting the test to let the system recover. After starting data taking, the pulsar was turned on about 15 seconds after the start (to give a baseline). The pulsing repeated every 14 seconds.
Beam 1 polA was not available (it was used for monitoring upstairs), Beam 6B was dead.

The following tests were done:

File	Volts	Current (mA)	misc
000	5.00	19	rfPulse=-10Dbm
100	5.0	19	rfpulse=-11Dbm
200	5.58	19	rfPuse=-10Dbm
300	5.92	19
400	5.92	18
500	5.92	17
600	5.92	16.5	16mA oscillated
700	5.58	17
800	5.75	17
900	5.92	17
1000	5.92	17
Turn off amps for two minutes
1100	5.92	17

The first set of plots shows the total power versus time for all the beams for each test (.ps) (.pdf) :

each page is a separate test
Black in polA, red is polB. Only beam 3 polB was modified.
The 7 beams are shown top to bottom.
The vertical scale goes .99 to 1.04 of Tsys.

The 2nd set of plots show the total power vs time for beam 3b for each test (.ps) (.pdf):

There are 4 tests per page (except the last)
black is polA, red is polb.
the vertical scale goes .99 to 1.04 Tsys.

Results:

In all case there was always a large jump up of 3% Tsys. For V=5.92 and I=17mA (the quickest recovery) it still took about 1 second to get back close to the original value.
We redid the V=5.58, I=17 from 16sep08 and the recovery was not as good as measured on that day.So playing with the V,I can change the recovery but it is not changing the original jump.

processing: x101/080918/alfatst.pro

page up
home_~phil