When I got this machine it was not fully functional. But even if it was, the cooling fans made an infernal noise. I suppose I can't do delicate analysis work with a vacuum cleaner next to my ears. Of course I powered it up to check the initial condition of the machine and out of curiosity, to check what was on the hard disk drive.
There was nothing really interesting. I found some files for analyzing the AFX Bus. From my knowledge this is something specific to the Sun Microsystems SPARCstation 5 computers. I believe only one card was ever designed for the AFX Bus. It's a video card of some sort that is able to output accelerated high resolution 24-bit color graphics. I might preserve these files just for the sake of uniqueness. There were also some files related to some sort of SRAM analysis. But that's about it. The rest was just empty space.
Anyway, enough with the historical facts and let's move on to more interesting things.
The main issues that I encountered are listed below.
Let's take them one by one.
One of the logic analyzer cards was detected but not identified by the operating system. It appeared under the Master Frame section but it was unusable. I decided to check the known issue with these old HP cards before doing any other hardware level debugging.
The issue is easily identifiable and is well documented on-line: the glue used to hold the plastic rails on the PCB becomes corrosive in time. I don't know the cause for this corrosion. Either the glue contains substances that break up in time and attack the solder compound or it traps water which in turn reacts with something in the glue to provoke the corrosion.
Under the rails there are mostly circuit traces covered entirely by soldermask. But there are also a few vias very close to the rails. Some are even close enough to be in direct contact with the glue. In the following picture you can see two vias that appear to have signs of corrosion.
The solution is to first remove the plastic rails and clean the area thoroughly. Once everything is clean, the damaged vias need to be identified and cleaned. My method is to use plenty of flux paste and eutectic solder to melt down the existing corroded solder compound. Then I extract the damaged compound with the vacuum pump. Following is a good cleaning of the area with IPA. At the end I fill the via with fresh eutectic solder.
The operation is cumbersome since removal of the glue residues under the plastic rails is a very tedious job. In addition, even with a high temperature soldering iron it is still very difficult to melt the corroded solder from the vias. Corrosion does not have good heat transfer properties.
Normally it would be good to reattach the plastic rails to the PCB. But I didn't do this since I don't know of any good thin double adhesive tape that would not have any impact on the solder in the future. I might do some research on this topic in the future.
So now I have to be very careful when inserting or removing these logic analyzer cards since I can easily mow down the small SMD components in case of card jamming in its cage.
While here I decided to also replace the aging power supply ripple filter axial electrolytic capacitors. I have used Vishay 125 ALS series parts of the same capacitance rating as the originals but with an increase of operating voltage to 16 V. However the new capacitors have a maximum temperature rating of +105 °C. Here are the old capacitors made by Nichicon.
And here are the new ones.
I'm happy that the repair job went straightforward and the card is now correctly detected and identified by the operating system.
In the end I removed the plastic rails on the remaining three logic analyzer cards. Fortunately there were no corrosion signs on any of the problematic vias. So I got away with just a good cleaning with IPA.
I also replaced all 3300 uF / 10 V capacitors with new Vishay parts as detailed above.
These logic analyzer cards are now ready for many hours of hard work.
Quantum hard disk drives of the early 1990s are not that reliable. I mean, sure, they are fast and relatively good performers, even though a bit noisy. But I know they fail. The ProDrive series in special have a rubber actuator stopper that, in time, turns to sticky goo. The actuator tends to stick to this rubber part and the drive will refuse to start since the head block is stuck in parking position. However the 16500C unit comes with a different model of drive, the Fireball 540AT. This is a standard 3.5" IDE drive.
The hard disk drive format structure is a bit weird. It looks like an MS-DOS FAT16 structure with no partition table and byte swapped data blocks.
My initial idea was to use a Compact Flash card in a 3.5" to 2.5" CF to IDE adapter so that I can install it in an adapter bracket in place of the old mechanical drive. Unfortunately it didn't work for me. I tried a bunch of CISCO 512 Mb cards and I had no luck. Then I tried some modern Transcend cards from 256 Mb to 4 Gb. No luck with these either. Whenever I install one of these cards, the 16500C just hangs at the power-up self tests screen. It also displays a Software Error message and some weird error codes on the screen with no further explanation. I couldn't find any reference to these codes anywhere on the Internet or in the manuals.
I started reading through the various sites describing the IDE to CF conversion for 16500C logic analysis systems. I found out that only specific CF cards are compatible with the HP unit. Since I don't have any of those around I did a quick search on eBay and found some very cheap Dane-Elec CF cards. These should be be working OK in the 16500C machine. I got both sizes: 256 Mb and a 512 Mb. A 256 Mb card should be sufficient for my needs. So let's try the smaller one first. Unexpectedly, it does not work. The machine detects it and is able to format it. But it doesn't boot off it. Thus, moving on to the 512 Mb card. Here it is inserted in a dual CF to 2.5" 44-pin IDE adapter as the master drive.
I secured the CF to IDE adapter to the the 3.5" to 2.5" metal adapter bracket. Then I mounted the metal adapter in place of the old mechanical drive. A new 40-pin to 44-pin IDE connection cable is needed thus I sourced a new one. Since it's a 30 cm long flat cable, I have carefully twisted it and it appears that it fits perfectly between the CF card adapter and the bottom panel.
I like how this CF card replacement job went by. In addition, I got rid of the hard disk drive noises.
The original fans in this unit were Panaflo FBA12G12U. These are spinning at 2750 rpm while moving an air volume of 114.7 CFM. The noise figures for these fans are at around 45.5 dBA. It is quite hard to find low noise fans that move the same amount of air like the originals.
I really wanted a low noise unit so I went with a couple of Noctua S-12A FLX fans. They are spinning at 1200 rpm, moving 63.2 CFM at a 17.8 dBA noise level. It is clear from the start that the cooling performance of the new fans is half that of the originals. On the other side, I did some tests, running the unit on full repetitive acquisition for about half an hour. There is heat generated, but most of it is in the power supply region. The Noctua fans are doing a good job at evacuating this heat on the right side of the machine. But as I said, the performance of the 16500C might be affected, if we are talking longterm usage.
The original fans are wired in parallel to a single connector. Some modifications are required to install the Noctua fans. I cut the original wires and the Noctua extension cords to the desired lengths. Then I removed the Noctua varnish from the extension cord. I cut the yellow rpm sense wire from the extension cords and I have spliced the resulting wires. Next I installed the fans in their respective places and simply connected the fan cords to their matching connectors on each of the two spliced cables.
I installed the machine on my workbench and none of the air admission or evacuation holes are obstructed. Also I expect casual using of the unit in short sessions of one or two hours. I reckon it will hold up well in time.
One interesting fact is that both fans are sucking air from the outside of the case and are forcing it through the spaces between the cards prior to evacuation on the right side of the unit. So if you ever decide to replace the fans in such a machine, be sure to respect the original design constraints. Also, if space allows, it is best to install this machine to your right. Otherwise the heat might be a bit uncomfortable for anybody working next to the unit. I installed it to my right and furthermore I did some tests on remote controlling the machine from the PC. The first tests are very promising.
I actually do not recommend replacing the original fans with any of the Noctua varieties. A better replacement would be a pair of equivalent Papst fans or a pair of new Panaflo fans with undamaged bearings. I don't know what was the actual environment in which HP tested the 16500C machine for optimum operating temperature readings. But I expect it was mostly research laboratories with possibly racks of stacked equipment. If this was the case, then the Panaflo fans would be 100% justified. Then again, I don't know so I did the replacement based on empiricism.
Furthermore, in my room, the maximum temperature rarely exceeds 29 degrees Celsius, and that is during the very hot summer days. And I really do not plan on doing any intensive electronics work during the summer anyway. In general, while working on my electronics projects, I have an ambient temperature of around 23 degrees Celsius. And it is mostly generated by the heat radiated by the tools that I use.
The HP 16500C uses standard 72-pin EDO SIMMs. The official documentation says that the maximum allowed memory capacity is 32 Mb. I have read some reports that people managed to install 64 Mb without issues. I chose to install 32 Mb since I had two 16 Mb SIMMs around that I was keeping for my retrocomputing projects. These modules are made by Siemens.
Installing them is straightforward. After removing all the data acquisition cards, you have access to the main computer PCB. I removed the existing 8 Mb HP branded SIMM and installed the new ones in their respective sockets.
Booting the machine and going to the testing screen reveals the 32 Mb is recognized correctly.
That was an easy upgrade and a very useful one if you plan to use the machine intensively. Especially while using multiple logic analyzer cards connected together. To my knowledge, there is no way to see how the machine manages RAM resources. So there is no way to know exactly how much memory is in use at any moment in time.
I also wrote an essay on this machine. You can see it here: Preparing the Hewlett-Packard 16500C Logic Analysis System
This HP unit is easy to work on, if you know what you're doing and you have the right tools. Otherwise you can easily do damage and brick it forever.
Copyright © 2004- Alexandru Groza
All rights reserved.
VER. 1.0 | REV. A