DSP-10 Software Radio
Characterization and Operational Familiarity

back to n5bf/6 DSP-10 page

Status:  Operational at 6.4 watts

2005 July 16 2223 UTC

KG6IYN, Bruce, 58/59 from DM12 in the CQ VHF WW.  8 miles north of Mexico at 4900 feet.

Reports in all the contests indicate that nobody can tell this isn't a storebought rig.  They have no trouble copying, they either say nothing about the audio or say it is good.  They sometimes ask if I have "other bands on that radio."

2005 July 27  Forward upper driven element on 2 meter satellite beam found broken (by SWR LED), needs to come down.

2005 Aug 13  0045 UTC
KG6KUB on 144.200 USB 6 watts.  Fred on East Mountain, Catalina, hiking portable, listening for KH6 beacon.
Order an M^2  2M12, twelve element 12.8 dB 19-1/2 foot beam from Amateur Electronic Supply.
This is expected to improve my signal by one S-Unit.  Also lower receive noise on SSB.
(Can still use discone for vertical / FM.)

2005 Aug 17 New 2M12 beam arrives.

2005 Aug 21
Buy a 14 foot tree pruner at Orchard Supply Hardware.
Trim around the discone.
Trim the new turning radius for the yet-unassembled 20 foot long beam.

2005 Aug 23
Measured 12.09 VDC at the module at full power output.
Reworked the DC supply cabling to the amplifier.
Now get:
  Key Up: 13.20V with 4.95 bias
  Key Down:  12.67V with 4.93 bias
No appreciable change in RF output.
Went ahead and swapped out the module under the following rule:
  If it gets better, we're done.
  If it gets worse, swap back.
  If no change, done.
Retuned L1, L2, L4, and L5 (and rechecked L3) and got improvement back up to "not significantly different from old module."
Re-dipped L3 to extinguish SWR light at 148.5 MHz into dummy load.
Tried various L1:  0 turn (short), worse; as specified; and plus one turn, a few tenths of a watt better.
Bending coils together hurts, separating helps.  Left "plus one" in and peaked.
SWR light extinguished across the band on the dummy load.
The satellite beam is still broken, 5.0 forward, 1.9 reflected.
The discone loads like the dummy load.

2310 local.  This is the end of the diagnostic rework on the amplifier.
The 10C helped by 20%.
I no longer believe that using the wrong frequency slug means "mostly erroneous on reflected."
I now believe, "mostly erroneous, for indication only."
I either have two modules (probably from the same batch) that are slightly substandard (6.5 watts versus 7 watts spec.)
or I have an output circuit problem.  All of the toroids and L1 have been inspected and tweaked many times.
Note in the data sheets that 144 MHz is the low end of performance.

2005 August 27 Build and install 2M12 replacing 2" O.D. mast with 1-1/2".




2005 August 28 Remake the rotator clamp U bolts correctly. ... in a slight breeze.
2005 August 29 Order a KK7P DSPx Module and KDSP10 Interface Adapter from TAPR.  This is for the second DSP-10.

2005 August 30
0324Z 8/31, Work KG6LUL, Lyle, 30 miles north of San Diego as first contact on new beam.  S6 to me, S7 to him (6 watt USB 144.230).
Do the "Fast PTT" mod while listening to Hurricane Katrina Health and Welfare traffice on 3935.
Found the soldering iron still plugged in from last session.  Asked XYL (Viann, WD5EHM), "When was the last time I soldered something?"
She said, "Last Spring?"
(... No, more like last Tuesday 8/23, but still, I've got to remember to unplug that poor thing.)
Make appropriate notes in my schematic.
This is much better and doesn't hang up the DSP-10 when accidentally PTT in CW mode.

2005 August 31
Re-measured power usage, output, and beam performance.
Brickette efficiency is 32% at nominal output.
1 dB compression point is around 3.3 watts.
The Bird 10C and dummy load are the trusted measurements.
DSP-10 levels interpolated from Figure 12. power measurement circuit readings compared to extrapolated, fitted, Brickette values.
Inferred Brickette gain is 30.2 dB so the drive interpolations are low, if anything.

Here are the "final" charts:




2005 September 10
0400 UTC  Heard the Tehachapi Beacon on 144.294.5 sending 33 dits and "DE N6NB/B DM05sb"  Signal S-2, -140 dBm in -150 dBm noise.  Beam 340.
This is the first time I've ever heard anything to my north (and about 100 km north at that!) as I'm blocked by mountains that way.
Must be coming through the Grapevine (I-5 route north).
You can see the dit times and the ID times in the waterfall.


2005 Sept 10-12.  Operated the September ARRL VHF Test with the new beam and Fast PTT!
Made 33 QSOs in six grids:  DM03, DM04, DM12, DM13, DM14, and CM94 from DM04.  198 points.

Won the Section!

See:  ARRL September VHF QSO Party Log.  (This beats the 2-meter subset 30 QSOs in 4 grids from June test.)
Time limited, I decided not to operate other band/modes except for DSP-10 on 2 meters linear (SSB).
CM94 is tough from here, not so much for the hundred mile distance due to blockage from hills.
The new 2M12 beam made a noticeable difference over prior antennas.
W7PUA is correct, this is not particularly a contest rig.  It overloads easily, and this would be worse in a better location or with a preamp.
It is, however, clean, quiet, and sensitive and the filter options are great.  The "Fast PTT" was fun too.
And, I can copy the N6NB beacon in DM05 (see above) although I didn't hear anyone operating the contest from up there.

2005 September
When my Astron RS-50M burped, I realized I had no backup 12 V supply in the shack.
Acquired an MFJ-4035MV variable supply for bench testing and backup.

2005 September 19 around 0400 Z heard local lightning on the DSP-10.  Unplugged everything!

2005 October 11
Used the new MFJ supply to make this plot.
  - Vs measured with BK Test Bench 390
  - Vcc2 measured with BK Test Bench 388A
  - Po measured with Bird 10C
The OV led comes on at 14.75 on the way up and goes off at 14.48 on the way down.
MFJ supply peaks out at 14.83 VDC with OV led on, but OV goes off key down.



2005 November 22 Acquired a Mac PowerBook G4 1.67 Ghz, OSX 10.4.3 and Virtual PC 7.0.  Began moving in.

2005 December 3  Determined that all amateur radio soundcard and serial applications would work on the Mac under Virtual PC with a Keyspan USA-19HS USB to serial adapter, except for the DSP-10 loader and uhfa.exe.  Looks like serial timing problems.

2005 December 17  Upgraded to UHFA / UHF3 version 3.50, put in a PL of 131.8 and had a QSO with W4EF on WR6JPL on 147.750 / 147.150.

2006 January 21-22 Worked 20 stations in 6 grid squares (DM03, DM04, DM13, DM14, CM94, DM12) in the ARRL VHF Sweepstakes using the same 2M12 beam and 6 watt brickette.  Didn't try as hard as before, have other things going on.  Rig works well for this, but did overload and go non-linear on one local.  Tried using TRLOG on the Mac through Virtual PC.  This was a disaster in half a dozen ways (doesn't set time right, doesn't keep time right, doesn't do Cabrillo right).  TRLOG is now retired.

2006 January 22-23  Big windstorm, gusts 50 - 80.  JPL was closed for the day Monday, some damage of JPL ARC antennas.  Worried about my 2M12 but it stayed up!

Through the winter did lots of 8 hour captures of screen at a very low waterfall rate (1200) to see what kind of activity there was day to day on various frequencies.

2006 May 10  Reading through the operators manual systematically, I came upon EME2 and got sidetracked for six weeks but, that's what it's all about!  Read all about it here:

EME2 Without Hardware Upgrades

In the process, discovered two things about the power output problems above.  1)  I had a two foot cable with 1.6 dB loss.  No longer.  2)  There is ripple in the output power versus frequency.  Tuning around just a few KHz to get away from birdies during EME tests could make a difference of a watt or more, on the order of 1 dB.  So, when choosing a lucky frequency:

Status:  Operational at 7.2 - 7.5 watts

2006 June 10  Operated the June ARRL contest.  Took a break to capture a moonrise pass (EME) during the test.  Was prevented from operating on Sunday but still did as well as last year.  What was great was that, as I worked on EME2 post processing software, listening to the calling frequency Monday and Tuesday evening after the test there was considerable chatter among the locals about the weekend test.  They were all selling each other rigs, amplifiers, and antenna, shifting all the equipment into new hands for "next time."  Later, there was a two hour QSO between a couple of guys who had been in the test.  One was commuting home across the LA basin, the other was nearby and when he got home went into the house and got on his rig there to continue the QSO.  The local contesting gossip was great.  This is the way I remember ham radio being!  Maybe it still is the "good ol' days" on some frequencies.

2006 June 19  Reading Chapter 12.5 in Experimental Methods in RF Design about DSP-10 weak signal work and EME2 in particular.  Confirmed some of what I had figured out and learned some new things as well.

2006 August 18  Upgraded to UHFA / UHF3 version 3.80.  Printed some new cheat sheets.

2006 September 9-10  Did the September ARRL VHF QSO Party.  See contest.

2006 September 23  Plotted up antenna pattern data from last summer.

On July 1, shuttle audio was being transmitted by W6VIO/R on 147.15 about four miles away.  Beam heading to the repeater would be 99 true, but that direction is blocked by a hill.  Apparently I see mostly a reflection off of hills to our mutual south.  During quiet periods (carrier only) I took signal level measurements at beam headings in 15 degree increments.  Since my beam is misaligned, (true - indicated = +17 degrees), the plots are calibrated in degrees true running from 2 around to 347.

The left plot is in micro-volts and looks a lot like a beam antenna pattern with maybe a little multipath to the left.  The outer circle is 100 micro-volts.  The right plot is in dBm, a log function of the left plot.  The outer circle is -68 dBm.  I don't know the power output of W6VIO/R or what the path losses, including the reflection, should be.  Actually, these are signal levels from a constant location source as the beam was rotated, so the pattern is a mirror image of the beam pattern, for these circumstances.  Also, my best DXing direction is 120 indicated, 137 true, which points roughly towards San Diego.  I think I see the edge of the reflecting hill dropping off with increasing azimuth from there.




2006 September 27  Took some data with room temperature and heated resistors and tried to use it to solve for DSP-10 noise temperature.  Got results that were consistent with 600 - 1000 K (except one that was nonsense) but decided that I needed a better designed experiment.  Put some noise temperature reduction code into the EME post processing software, then made a note to take it out.  The list recommended another approach.  Back on the todo list.

2006 October 6  Drove to Foster City, Ca (San Francisco area) for the AMSAT Symposium and Annual Meeting.  Gave my paper Software Radios, an Enabling Technology for Satellite, Space, and Ultra-Weak Signal Applications to goo reviews on October 7.  <Paper to be linked from here, from Phase Five, and from EME of which it is a rewrite>.  Driving home October 8, we tried to work Echo from the car using equipment we had with us.  John, KG6HCO, operated the 2 meter mobile rig at 65 watts and an FM HT on 70 cm while I instructed verbally from the driver's seat roaring down I-5.  We heard a few squelch breaks (maybe) but were unsucessful.  Monitoring at home in future weeks, heard activity for several seconds to a minute when Echo was overhead a few times.  Need a better antenna.

Took several more EME2 captures during September and October.  They have not been post-processed.
 
2006 December 19 0200 Z  Worked W6GL on CW, 144.200.600, 7 watts.  Established contact on airplane scatter then it dropped into the mud.  Steve, WB2WIK on a hill between us (Chatsworth / Woodland Hills) came in and intermediated.  While he talked to each of us, we listened for the other but heard little except for a few more fortuitous airplane reflections.  K6EMF/m La Crescenta also came in.

2007 January 1 UTC.  Hooked up a straight key and called for SKN contacts on 144.100 and 144.200.  No responses.  (Made a conventional SKN contact on the 20 meter QRP frequency 14.060 on the conventional TS-680 radio.)  Did contact W6QE at 2007 January 2 0137Z on 144.200 CW but quickly moved to SSB.

2007 February 12 - March 11   In Hillsboro, Texas for my mother's knee replacement surgery (bi-lateral means "both").  QRT except for e-mail and a 40 meter Rockmite.  Only had the Rockmite on for a few days of one of the weeks.  Heard some traffic including W6GL calling me from California 2007 March 3, 0410 Z (arranged by rapid e-mail exchanges), but antenna was not good enough to be heard.  Did some DSP-10 planning.

2007 May 12  Viannah, KG6GXW, graduates from Franklin & Marshall College in Lancaster, Pa.

2007 June 22  Tried installing DOS-only VPN from a Windows 98 disk.  The disk is now unusable because I can't find the book with the security code.
In the Windows 98 VPN I do have, edited autoexec.bat and added the line 'command.com'  This causes it to boot up into DOS just like a computer does but it still wouldn't load the DSP-10 via the Keyspan.  I think there is still a timing issue.

Using a dumb terminal program I can talk to the EZKIT through the Keyspan (USB serial port) so I know the Keyspan is working in both directions.  This means it should be possible to write my own Mac based (Xcode or Cocoa) loader.

Did not try EZSLOW because I already use EZFAST L which is already 9600 baud.  It's handshaking, not bit sync that appears to be the problem.  Also, even after the DSP-10 is properly loaded, my "resume" script doesn't work from this DOS-running window, so the communication issue must exist there too.  In both cases it appears to get past the first step then wait forever for a second ack, as if it hadn't started listening quite soon enough.

2007 June 23 Field Day  Bought a Radio Shack 350 watt inverter and ran all needed computers, lights, and DSP-10 from a 75 AH storage battery, either direct or via that convrter.  Made 14 2-meter phone (SSB and FM) QSOs in about six hours proving that the battery would buy me plenty of time to lug the generator out and get it going in case of disaster.

Detailed FD07 report:

I ran the whole office on my KWH battery (advertised at 75 AH / 12 V) for 6-1/4 hours.  Voltage was still 11.4 on rx and 11.1 on tx at the "end."

No good indication of how far down I took the battery.  I would bet one of 1/4, 1/2 or 3/4 but really have no idea.  What I do know is that I can run everything over here (including the printer) for at least six hours before I have to start messing with the generator.  That's the number I wanted.  I'd bet 8 or even 12 hours in a pinch.

Good things about running with the 486SX/25 (ebay) laptop:
- it works at all
- works fine at ~11 V

Things that aren't as nice as I'm used to with the big surplus PC (586/133):
- Keys up slow, sometimes a 3 second delay.  Not good for contesting!
  - but sometimes instantly as expected
- Refreshes display slow.  Probably an interaction with "keys up slow."
- Slow to take keyboard inputs.  Misses a lot of them altogether.
- painful reaches like for Alt F-10.
- No right Alt key!  (Used for manual CW.)
- General abbreviated key set.  Some keys are FN - something that would be their own keys normally.  (Yes, and no external keyboard input.)

Well, ya get what ya pay for.  Still it works at all and would be fine portable or whatever.

General about the DSP-10.

As the designer says, it's not really a contest rig.  Only one roofing filter.  Overload can cause digital saturation.  This doesn't happen when trying to detect echoes from Venus, but does happen when your next door neighbor keys up his KW.  CQ FD! CQ FD!

It's generally a do it yourself, futz around rig, not a contesting setup with a thousand individualized programmable controls.  Two of them would be kinda cool, however.  Someday.

No trouble running the rotator from the inverter.  While I was at Radio Shack I also got a mast mounted TV preamp.  Now I can watch KCET digital!  (But two meters trashes it.)

2007 June 24 0330 - 0700 Z  Took extensive room-temperature and load-in-the-freezer noise data on DSP-10 in a new attempt to measure noise temperature.  See 2007 September 12 for data reduction.

             
Cold -- resistors in the freezer ~ 254K.               Warm -- room temperature, thermocouple between 100 ohm resistors in parallel ~297K.

2007 July 5  Box Labels -- The hard part was selecting the font.





2007 July 7  Took DSP-10 filter data.

2007 August 17 0353 Z.  WF6A DM14, Bill in Azusa, called CW CQ on 144.200.700.    Had a 20 minute CW QSO.  Marked my "worked/heard on frequency" list at "8/98", that is, heard 98 calls of CQ or ID that were eligible for me to attempt QSOs, answered and completed 8.  Utilization goal is a 1/10 ratio of worked to heard.

2007 August 20 and
2007 September 1.  These are Excel plots of some filter ripple data I took on 2007 July 20.  Posted questions to the list about these.



Power output was taken every 100 Hz using the Bird 10-C slug and a 50 ohm dummy load.  Three things are notable.  It looks like the 5 KHz 1st L.O. synthesizer switching may cause the bumps around 144.196, 144.201, 144.211, etc.  Second, this seems like more ripple than expected.  Here is the same data in a log plot.



Note that the ripple is a little over 2 dB.  I think this is six repeats of a 5 KHz piece of the passband of the 19.665 MHz four crystal filter.  Normal ripple in a passband is usually limited to 1 dB by design, but I may be seeing a passband edge here.

Here is the corresponding information seen as receiver noise looking into a dummy load.  The vertical scale is 1 dB per division.  I would call the difference between the trough and peak levels at 144.202 and 144.205 about 2.5 dB.  These were taken as three snapshots about a minute apart at averaging rate 102.



This is comparable to a set of W7PUA data taken similarly which can be seen at http://www.proaxis.com/~boblark/p527xfil1.gif.  His 19.665 MHz ripple is more like 0.5 dB.  (I also have a few "in box" birdies that Bob doesn't have.)

There may be some sensitivity to this in the settings of L12 and L13 (See schematic QST 1999 September p. 37. Figure 4.).  I remember from the initial alignment that transmitted power output or received noise wasn't very sensitive to those inductors.  Displays like this could be used to tweak those cans in (near) real time if they affect flatness.  I'm not going to do anything about this now but will add a note in Planning that this might help smooth things out if a flatter response is needed for some future project.  This filter and its response are discussed in the QST reference, Figure 5.

Third, you can see some trending in the peaks.  Noting this, I swept the whole band.  In this next plot I took data every 10 KHz (on what was the same 5 KHz sub-frequency so as to miss the 19.665 MHz effect) from 143.8 to 148.2 (into the same dummy load of course).  I have good synthesizer lock over a larger range but didn't think there was much to prove by extending any further out of the ham band.



You can clearly see that, during the alignment, it was peaked for 147 MHz and you can see why, because it drops off at higher frequency faster than at lower.  I'm wondering, however, if I should repeak for 146.0 or 146.5 in order to get better performance at 144.0 - 144.3 where most of the interest here is.

This looks very much like the top of the response curve for the four-pole interstage filter discussed at QST 1999 September, p. 36. Figure 3. but I notice that in that plot the low end roll off is around 6 dB per MHz whereas mine from 145 to 144 is more like 3 dB per MHz.  If I'm on the edge of the slope, I'm barely on the edge.  Another note for Planning.

When I initially characterized the rig, I decided that a power setting of 82 gave one watt out of the Brickette.  I now see that this was highly dependent on the frequency chosen for the measurement.  Up in the FM band, I get one watt at setting 80.  At 144.200 it needs 83.


2007 September 12  Processed the noise figure data taken on June 24.  It is all garbage.

Experiment design.  Take ten one-minute noise integrations (4800 Hz Non) writing these down with the time and the temperature in Fahrenheit (which has higher resolution than Centigrade) in each of the following cases:

A:  144.150 MHz 50 ohm
B:  147.540 Mhz  50 ohm
C:  144.150 MHz 50 ohm, through Brikette
D:  repeat A
E:  144.150 MHz 100 ohm.

Per case, ten points cold (load resistor on the end of the antenna coax in the freezer), ten points warm (room temperature in the shack.)

The calculation:  ( Tr + T2 ) / ( Tr + T1 ) = P2 / P1

Solve for Tr

For R = P2 / P1 = 10^( ( dB2 - dB1 ) / 10 )

Tr = ( T2 - R*T1 ) / ( R - 1 )

Wrote a C++ program "noiseFigure" to convert the F to Kelvin and perform this calculation for every pair of numbers in each test (100 comparisons) and to cross compare cases A and D, that is, A cool versus D warm and A warm versus D cool.  Anticipated the need to do statistics on the results to get a good answer.

Results:  All tests and all test points produced nonsense.  (See also http://pages.cs.wisc.edu/~kovar/hall.html .)

Discussion.

In the above equation if R > T2/T1 or if R < 1, a negative answer will result.  This is clearly wrong.  In my data, T2 / T1 is about 295 / 254 (Kelvin) which is about 1.16.  That's about 0.6 dB.  Meanwhile, due to the experiment setup, doing warm and cold in different locations, for example, the differences between cool and warm data sets were 2 or even 3 dB.  This is large when the signal is supposed to be 0.6 dB.

The fundamental problem here is that T2 / T1 ~ 1.2 isn't enough signal to work with unless you have 0.01 dB accuracy and stability in the measurements and the test setup, which we don't.

Use of the "Y Method," a different arrangement of the same calculation, produces similar nonsense.

Y = 1 + ENR / Ftot, so noise figure Ftot = ENR / ( Y - 1 )

Expected ENR in this case is 0.2.  If Y > 1.2, noise figure is less than unity, that is, less than 0 K.


Suddenly, I understood what a "calibrated noise source" is all about.  A typical noise source used for a measurement like this puts out 7000K.  I think I'll acquire a noise source and replan the experiment rather than working on ways to heat a resistor up to 7000K (hotter than the surface of the sun!).


2007 September 12 0509 Z  Thought I was seeing the KH6 beacon in an LTI on 144.170 at beam heading 267 indicated (250 true).  It was too perfect, however.  It was on 500 Hz audio when showing 144.169.500.  I know that my reference oscillator is high, so I expect the signal to be below that, but I don't know its exact frequency.  Doing the same LTI on the dummy load gave the same little peak so its a birdie in the radio.
 
2007 December 15
  Performed the long-awaited re-peaking of the RF (144-148 MHz) and 1st IF (19.665 MHz) to reduce roofing filter ripple and generally get more power output and sensitivity at the low end of the band, 144-146, where the weak signal action is.

First I spent half an hour looking for a proper tweaking tool and finally settled on the only object in the house that would turn the little slugs, a metal eye-glasses screwdriver, an improper tweaking tool.  Marked this with a tape flag so as to accurately count turns.

Terminated the antenna port, set the spectral display to the 7 Hz range and tried for a while to make adjustments based on readings of the noisy noise curves.  Decided fairly quickly that this was hopeless and unscientific.  Pulled out my new KA7EXM power meter and made tables of power output readings for Xmit Pwr 80 at 144.200, .201, .202, .203, .204, and .205.  Because of the LO switching, .200 and .205 were expected to match, being through the same part of the IF at the different LO settings.  For each set of measurements, I looked at the peak-to-peak change.  With the original settings, as seen in "Roofing Filter, dB" above, peak-to-peak was over 2 dB.  Working first with L12, I moved one turn each direction and remeasured.  From the best (least peak-to-peak change) of those three, I moved +/- half a turn, remeasuring, then +/- a quarter turn.  From the best of these, I then repeated this at L13.  From the best combination of L12 and L13, I moved L12 again, then L13, remeasuring each time.  (Yes, like tuning a pi-network, which this is, but lots slower.)  After a couple dozen readings like this, I had settled on L12 at +1.0 turns and L13 at +1.25 turns from the original settings.  The metal screwdriver doesn't matter for this because it is removed while measurements are taken.

Those original settings were made during the first alignment where the whole chain was peaked by ear and S-meter readings near 147 MHz.  I remembered this being straightforward at RF but that the IF coils, L12 and L13 didn't make much difference.  Now, with improved equipment and understanding of what was going on, I could do a better job at IF.  These settings gave peak-to-peak of about 1.0 dB, a big improvement.  I do not know if I have found a local or global minimum.

Next, I went to the RF chain, L8, L9, L10, and L11, to get better performance at the low end of the band than is shown in "Power Out" above.  First I reconfirmed by taking points every 500 KHz from 144 to 148 MHz that the shape of that curve was still pretty much the same as before.  This showed a difference of 4 or even 5 dB between 144 and 147 MHz power output.  Next, at 147.0 I transmitted into the power meter and peaked each coil in turn using the metal screwdriver.  Each coil changed by a significant fraction of a turn.  Finally, when somebody started talking on the Catalina repeater (147.09, 30 miles away) I repeaked L1 and L2 by ear (receive path only), confirming that they had an effect, but not making much net change.

When I remeasured every 500 KHz across the band, it was much flatter now than before, with some possible roll off at the high end rather than the low end.

Went back and did another dozen iterations on L12 and L13, this time to 1/8th turn resolution, about the best I thought I could reliably do, and found new settings (a little surprising, indicating interaction between RF and IF filters) at +0.75 and +1.0 turns for L12 and L13 respectively.  The final IF sweep showed ripple at 0.6 dB, at least from my coarse KHz measurement.

I called this done and put the lid back on.

Then, made the following plots which are designed to be comparable to those above so the improvement (or at least the change) can be easily seen.  These are not exactly the same.  The plots above were made with the Bird wattmeter and inferences.  These were made with the KA7EXM low power wattmeter, freshly calibrated.

First, drive linearity.  These compare to plots from 2005 August 31, near the top of this page.



In the barefoot plot (right) there is a little saturation happening above 96-97.  The Brickette version shows why it isn't very useful to use Xmit Pwr over 95.

The barefoot plot was made with the DSP-10 transmitting directly into the power meter.  For the Brickette, the amplifier output goes to the Bird wattmeter (10C) then to the 40 dB tap which drives the power meter, then to a dummy load.

Amplifier performance parameters from this data are:
Gain:  33.4 dB (even higher than previously thought, uncertainty in the measurement is around 1 dB).
1 dB compression point:  ~ 4 Watts, setting 89.
Power Max:  7.0 measured by Bird 10C, 8.1 measured by the KA7EXM meter.

As we will see next, this could have been 0.4 dB, half a watt higher, if I'd used a slightly different frequency.  The part spec. is 7 watts, so I'm convinced now that it is all working right.

Next we consider ripple in the IF filter.



This compares directly to the 2007 September 1 "Power Out" plot above in terms of rig settings and plot scales.  Note that the frequencies of the peaks and valleys have moved and that the peak-to-peak is about 1 dB, significantly less than before.  There is something going on in the values around 144.206 and 144.211 that has something to do with the LO switching algorithm.  These were taken using the KA7EXM power meter "relative" mode just for the fun of it.

This plot shows six sweeps across the IF passband of interest.

Now we look at this same behavior in the received spectrum.


            .198                                              .200                                               .202                                              .204

This is an enormous improvement over the same plot above and looks exactly like http://www.proaxis.com/~boblark/p527xfil1.gif when you realize that Bob's plot is 2 dB/division and this one is 1 dB/division.  We have the same high end IF roloff (actually, low end in the IF).

The transmit and recieve measurements are consistent.  This is much better than before.

(Note:  I used a different 6 KHz than above in order to stay out of a nasty internal birdie just above 144.205.)

Finally, RF broadband behavior.



I would have preferred to have it a little flatter across the entire band, but since 144-146 looks good, the higher frequencies are FM where 0.1 dB is less important, and other uses of those frequencies will likely be through transverters that will have a little slack of their own, I think I can leave it alone for now.

The fact that I peaked for 147 MHz and got this which is 2-3 MHz lower is due to using the metal tool in real time peaking/measuring.

Amplifier gain from this data is 33.8 +/- 0.2 dB.

The spectral display where I usually listen, 144.0 - 144.3, is much flatter to look at now and the noise level is a few dB higher.  These are pleasing results.  Maybe I can do EME2 in only 5000 points now.

2008 June 6 - QSO with W6GL/m

John, W6GL did some contract work at JPL for a few years.  He would usually leave work late, after I was at home and his route went right by my house near the freeway.  So, he put a 2 meter SSB rig in his truck and we had some QSOs, like this one:



You can see him at 0255 Z (1955 local) hitting 60 over 9 and overloading the front end, then continuing down the freeway getting weaker and weaker.  The scrolling stops when I'm talking.  Eventually (off screen) he would get to the Santa Sussanah pass on the way into Simi Valley and we'd lose contact.

2008 September 6 - Establish DSP-10 Noise Figure.

Reread the 2007 September 12 discussion (see "nonsense" above), then unboxed  the newly calibrated noise source (see Test Equipment for 2007 December 20 through 2010 May 15) to try it again.  Dug out my old files with all the materials about noise measurement and noise sources and found them coated with a fairly thick layer of dust.

Reread the QST 1994 May p. 37 article and did the following algebra:

Y = Non / Noff = ( F + ENR ) / F = 1 + ENR / Ftot

Ftot = ENR / ( Y - 1 ) = ENR / ( Non/Noff - 1 )

Also T = ( F - 1 ) * 290K once F is known.

Took several data points through a nice set of JFW Industries attenuators but a questionable 6 foot BNC hookup cable and made the following calculations:

DSP-10 test @ 144.99.700 USB, 60 second measurements.                           
   Non    Noff    Attn    Pon    Poff      F       F dB     K
42.208    29.92     0    16626    982    15.84    12.00    4302
39.330    29.85     3     8570    965    16.05    12.05    4363
36.745    29.54     6     4726    899    14.90    11.73    4031
33.850    29.64    10     2427    920    15.42    11.88    4182
30.103    29.48    20     1024    886    16.21    12.10    4411
                           
Through Brickette, 875, and Bird 43                           
41.525    29.24     0    14207    839    15.85    12.00    4306
                           
The first measurement agrees well with the 13.4 dB ENR measurement I made when the source first worked (see Test Equipment for 2008 April 8).
The last measurement demonstrates that losses through the amplifier's receive paths, are not significant.

Having expected 600K to 1000K, I was surprised by this 4300K result.  Posted to the DSP-10 list.

Mike KD7TS reported that his DSP-10 had a noise figure of 11.5 dB, which was fine since he used it as a 10 GHz IF.  Based on the measurements with all the attenuators and this one independent expert confirmation, I decided the measurement was accurate enough and I decided that, three years of successful on-the-air operation and QRPpp EME2 notwithstanding, I really needed a good preamp in front of this rig.  Also on the list, Scott N0EDV suggested on of the Down East Microwave kits at http://www.downeastmicrowave.com/Catalog.htm#BM2M.

Hooked up to the TS-680.  With the preamp on (18 MHz and above) "on" goes to S-3 while "off" is S-0, no meter movement.  Cannot make conclusions about noise figure from this information except to note that it is a few thousand K.


2008 October 11 - System Measurements with new Preamp

case             K    dB
DSP-10 alone  3200  10.8   *** better than measured on 2008 September 6.  Possibly because this is a good jumper cable.
system          71   0.95
preamp only     51   0.7   inferred from amplifier chaining equation, but exactly as advertised
station        200   2.2   estimated, based on 1.5 dB cable loss to antenna


Station improvements are detailed in Phase Four.



n5bf/6 DSP-10 page
n5bf-at-amsat-dot-org

updated 2008 October 29, 2010 March 26, 2010 December 18, cbd