Veraxis
Active Member
Hello All,
Lately, I have been working on building a replacement for a somewhat rare cable assembly for communicating between a HP 141T/8555A/8552 mainframe spectrum analyzer and a HP 8445B "Automatic Preselector" add-on box, and I thought I would document my work here for others to follow.
A bit of background first: the 8555A RF plugin for the 141T mainframe is a 10MHz to 18GHz microwave spectrum analyzer. They are very capable units, but one of their quirks is that the harmonic mixing used to achieve this high frequency range also generates a large number of spurious mixing products on the display. The best solution to this quirk, after the slightly clumsy "signal identifier" mode built into the plugin, is the 8445 preselector, an add-on box which uses a 1.8GHz-18GHz YIG tunable bandpass filter in line with the input to track the sweep of the spectrum analyzer and block out any out-of-band signals which might produce spurious mixing products, albeit at the cost of about 6dB of loss.
Here's the problem: in order to get the 8445 preselector to "talk to" the spectrum analyzer, one has to have a very special mixed-contact D-sub cable assembly to link the two boxes together, known under the the HP part number 08445-60007. Given how old these units are, probably the majority of these cables have either broken, gone missing, or been thrown out by mistake over the years. Nowadays, finding the cable that goes with the box is pretty rare, and so the cable alone can often go for as much as the preselector itself.
So I determined to make my own. From a bit of research into mixed-contact D-sub connectors, I determined that the type I needed was a "17W5P" type D-sub connector, which has 17 total positions comprised of 12 normal pins and 5 holes for either large power contacts or coaxial signal lines. In this case they are for coax lines, and it appears from the box that only 1 of the potential 5 lines is being used ("AUX B" on the spectrum analyzer side).
First, I went and ordered some blank connectors with solder cups for the pins.
Then I soldered wires into all of the solder cups and covered them in shrink tubing, being very careful to keep track of which pins were connected where. Fortunately, the bodies of the connectors I got actually have the pin numbers molded right into the plastic, which was really helpful. I wanted to use stranded wire for added flexibility and durability under repeated bending, but I only had white stranded wire, so I was extra careful to triple-check all of my connections with a multimeter.
The coaxial connectors were assembled separately on some RG-174 cable, but then the bodies just pop right into the holes in the d-sub. Be careful, though, once the c-ring on the coax body pops into the plastic of the connector, it's not coming back out!
I then put a few inches of some larger heat shrink tubing over the full bundle of wires around each end. This helped bunch everything together for when I had to put the backshells over the connectors. I would be careful not to use too much heat or for too long when putting shrink tubing over RG-174, as it might melt the polyethylene dielectric in the cable. My recently purchased 850D heat gun came in handy on the shrink tubing and controlling the heat.
I then put some braided cable sleeving over the full length of the cable.
Then I put the assembly into a backshell. It is actually a "37 position" size backshell, which is the same size shell as a 17w5P mixed contact d-sub.
The fully assembled cable with the backshells on it.
The cable on the back of my spectrum analyzer for a quick test. After some additional research, I later realized that the original cable was 18 inches long, so I may have been a bit too generous to make it 24 inches long. I probably could have made it 12 inches long with no problem.
In the first image you can see the trace without the preselector for a 5GHz signal on the 2.07-6.15GHz band at 200MHz/division. The (very faint) peak on the right is the real 5GHz signal and the peak on the left is a spurious harmonic mixing product at around where 3.85GHz should be. The reasons for this are a bit complex, and require an understanding of how the harmonics of the 8555A's LO and IF frequencies add up. On the next plot, with the preselector connected now, we lose about 6dB on the 5GHz signal as expected, and the spurious signal is eliminated.
I took my time making this, so it probably took a few hours to make. It took even longer to document my progress and set everything up to test it. In small quantities, even the parts alone to make this cable are pretty expensive, so I probably didn't save that much money in the long run, but I feel good about having increased the total number of usable cables left for these units.
Edit 7/12/2018:
Here is my parts list for these cables (IIRC, I believe I used the crimp-type coax connectors, but the solder-type should also work if you don't have access to hex crimp tools for the contact):
17W5P housing, solder cups (male) --mouser/digikey (3017W5PCM99A10X)
17W5S housing, solder sups (female) --mouser/digikey (3017W5SCM99A10X)
37 position size die cast backshell --mouser/digikey (165X02639XE, 8655MH3701BLF)
female coax contact (crimp type, female center contact) --mouser/digikey (131J21019X)
male coax contact (crimp type, male center contact) --mouser/digikey (132J21019X)
female coax contact (solder type, female center contact) --mouser/digikey (131J20029X)
male coax contact (solder type, male center contact) --mouser/digikey (132J20029X)
22AWG stranded wire (19x34) --mouser/digikey (400R0111-22-9, 55A0111-22-9)
2mm, 5mm, and 10mm shrink tubing -- ebay/amazon (from kit)
PET expandable braided cable sleeving -- ebay/amazon
Lately, I have been working on building a replacement for a somewhat rare cable assembly for communicating between a HP 141T/8555A/8552 mainframe spectrum analyzer and a HP 8445B "Automatic Preselector" add-on box, and I thought I would document my work here for others to follow.
A bit of background first: the 8555A RF plugin for the 141T mainframe is a 10MHz to 18GHz microwave spectrum analyzer. They are very capable units, but one of their quirks is that the harmonic mixing used to achieve this high frequency range also generates a large number of spurious mixing products on the display. The best solution to this quirk, after the slightly clumsy "signal identifier" mode built into the plugin, is the 8445 preselector, an add-on box which uses a 1.8GHz-18GHz YIG tunable bandpass filter in line with the input to track the sweep of the spectrum analyzer and block out any out-of-band signals which might produce spurious mixing products, albeit at the cost of about 6dB of loss.
Here's the problem: in order to get the 8445 preselector to "talk to" the spectrum analyzer, one has to have a very special mixed-contact D-sub cable assembly to link the two boxes together, known under the the HP part number 08445-60007. Given how old these units are, probably the majority of these cables have either broken, gone missing, or been thrown out by mistake over the years. Nowadays, finding the cable that goes with the box is pretty rare, and so the cable alone can often go for as much as the preselector itself.
So I determined to make my own. From a bit of research into mixed-contact D-sub connectors, I determined that the type I needed was a "17W5P" type D-sub connector, which has 17 total positions comprised of 12 normal pins and 5 holes for either large power contacts or coaxial signal lines. In this case they are for coax lines, and it appears from the box that only 1 of the potential 5 lines is being used ("AUX B" on the spectrum analyzer side).
First, I went and ordered some blank connectors with solder cups for the pins.
Then I soldered wires into all of the solder cups and covered them in shrink tubing, being very careful to keep track of which pins were connected where. Fortunately, the bodies of the connectors I got actually have the pin numbers molded right into the plastic, which was really helpful. I wanted to use stranded wire for added flexibility and durability under repeated bending, but I only had white stranded wire, so I was extra careful to triple-check all of my connections with a multimeter.
The coaxial connectors were assembled separately on some RG-174 cable, but then the bodies just pop right into the holes in the d-sub. Be careful, though, once the c-ring on the coax body pops into the plastic of the connector, it's not coming back out!
I then put a few inches of some larger heat shrink tubing over the full bundle of wires around each end. This helped bunch everything together for when I had to put the backshells over the connectors. I would be careful not to use too much heat or for too long when putting shrink tubing over RG-174, as it might melt the polyethylene dielectric in the cable. My recently purchased 850D heat gun came in handy on the shrink tubing and controlling the heat.
I then put some braided cable sleeving over the full length of the cable.
Then I put the assembly into a backshell. It is actually a "37 position" size backshell, which is the same size shell as a 17w5P mixed contact d-sub.
The fully assembled cable with the backshells on it.
The cable on the back of my spectrum analyzer for a quick test. After some additional research, I later realized that the original cable was 18 inches long, so I may have been a bit too generous to make it 24 inches long. I probably could have made it 12 inches long with no problem.
In the first image you can see the trace without the preselector for a 5GHz signal on the 2.07-6.15GHz band at 200MHz/division. The (very faint) peak on the right is the real 5GHz signal and the peak on the left is a spurious harmonic mixing product at around where 3.85GHz should be. The reasons for this are a bit complex, and require an understanding of how the harmonics of the 8555A's LO and IF frequencies add up. On the next plot, with the preselector connected now, we lose about 6dB on the 5GHz signal as expected, and the spurious signal is eliminated.
I took my time making this, so it probably took a few hours to make. It took even longer to document my progress and set everything up to test it. In small quantities, even the parts alone to make this cable are pretty expensive, so I probably didn't save that much money in the long run, but I feel good about having increased the total number of usable cables left for these units.
Edit 7/12/2018:
Here is my parts list for these cables (IIRC, I believe I used the crimp-type coax connectors, but the solder-type should also work if you don't have access to hex crimp tools for the contact):
17W5P housing, solder cups (male) --mouser/digikey (3017W5PCM99A10X)
17W5S housing, solder sups (female) --mouser/digikey (3017W5SCM99A10X)
37 position size die cast backshell --mouser/digikey (165X02639XE, 8655MH3701BLF)
female coax contact (crimp type, female center contact) --mouser/digikey (131J21019X)
male coax contact (crimp type, male center contact) --mouser/digikey (132J21019X)
female coax contact (solder type, female center contact) --mouser/digikey (131J20029X)
male coax contact (solder type, male center contact) --mouser/digikey (132J20029X)
22AWG stranded wire (19x34) --mouser/digikey (400R0111-22-9, 55A0111-22-9)
2mm, 5mm, and 10mm shrink tubing -- ebay/amazon (from kit)
PET expandable braided cable sleeving -- ebay/amazon
Last edited:
By purchasing it from the individual boxes, as well as some lucky breaks with local pickups and saving on shipping costs, I was able to assemble a working 8510 for about as much, if not less than what a working 8753 VNA would have cost, especially compared to one of the ones with the highly-desirable 6GHz option installed. The 1.3GHz and 3GHz 8711, 8712, 8713, and 8714 network analyzers would be another viable option at the hobby level, especially if one knows about the 
