Repair Progress on the H7826 Power Supply

I have been working on the H7826 power supply that came in the TURBOchannel Extender of my DECstation 5000/240. Normally, after a visual inspection, I will just apply power to see if a PSU works. This one, though, was not in a very good state, with a lot of dirt and corrosion, and bits of heat sink all over the board. You can see more details in my earlier post.

So, the first step was to clean it up and replace many of the electrolytic capacitors that are apparently known to fail. Most of the capacitors I removed were in the output area, although I did remove a couple of small electrolytics near the two big input smoothing capacitors. After removing the capacitors in the output area and cleaning it up it looked like this:

I did not find evidence of leakage, but the dark area under one of the capacitors is a bit suspicious. I tested the continuity around the dark area and it seems fine. I then replaced the capacitors and cleaned the corroded heat sinks as best I could. The results are in the pictures below:

So then it was time to see if the PSU works. I reassembled it, making sure the fans were attached, as well as the status LED. I applied power and….. nothing. I was disappointed, but not surprised.

I did some probing around to see at which point the PSU is failing. I found the voltage across the big smoothing capacitors was a healthy 330VDC or so, so they are fine (they were not fine on the H7878 in the DECstation itself). I then checked the primary side of the transformer and I could see a signal. However, when I checked the secondary side of the transformer there was no signal at all.

This could be a problem with the transformer itself, or something else after the transformer. It was impossible to check without removing parts, so I started by removing the transformer. Here is the board without the transformer (primary side on the right):

H7826 With Transformer Removed

In the two pictures that follow I arbitrarily numbered the pins 1 to 6 from left to right. This is the primary side, where the windings are 1-6, 2-3 and 3-4:

H7826 Transformer Primary Side

This is the secondary side, where the windings are 1-2, 3-4 and 5-6:

H7826 Transformer Secondary Side

While I was at it I got out a ringing tester that I made a while ago (thanks to a member of the classiccmp mailing list) to make sure that the transformer is working and does not have any shorted windings.

This is the result of the ringing test on pins 1-6 of the primary side:

H7826 Primary 1-6 Ringing Test

On the secondary side pins 1-2 and 5-6 gave this result:

H7826 Secondary 1-2 and 5-6 Ringing Tests

The other windings all looked like this:

H7826 Other Ringing Tests

I am told the latter result may be because there are insufficient turns on the windings being tested, and that these results show that the transformer is working and there are no shorted windings. That is a huge relief because I suspect that the transformer would be very hard to replace.

The next thing to check was the output rectifier. In circuit, one of the diode networks (an MBR 3045) appeared to show some shorts. That meant I had to remove them, to test them out of circuit. The problem here is that the rectifiers are screwed into a large heat sink and the parts around the heat sink make it impossible to reach the screws. This meant that I had to remove the whole heat sink. To make matters worse, the heat sink is fixed to the board by two pegs. I could not see how the heat sink was connected to the pegs

To do this I first had to de-solder the three diode network parts and the temperature sensor. Once I had done that I tried gently levering the heat sink off the pegs, in the hope that the heat sink was just held on the pegs by friction. Thankfully this turned out to be the case. I used a little bit of WD40 to help the process, letting it soak in first, and I soon had the heat sink and the 4 components attached to it out of the board. The pictures below show the pegs and the heat sink:

I used a Peak DCA55 tester to check the diode network parts and they seem to be working fine, so that is not the cause of the transformer not showing any outputs. When probing the holes in the board, the ones for the rectifier network nearest the output still appeared all to be shorted together, but I traced that to two low-value resistors, so that is not the problem.

So I am still stuck not knowing why there seems to be no signal on the output windings of the transformer.

Posted in Retro-Computing | Tagged | Leave a comment

Imaging RD51 From a DEC Rainbow 100 Model B

I recently acquired a DEC Rainbow 100 Model B, with a working RD51 disk. The disk is running MS-DOS 2.11. I want to preserve the contents of the disk before it fails. A while back I acquired the MFM Emulator board designed by David Gesswein and built it. Mine is an older revision of the board and it works really well. I have also bought the later board but I have yet to build it.

I connected it up as you can see in the picture below. In this case I connected the floppy power connector to the emulator board, and of course the RD51 to the MFM connectors.

Imaging RD51 With Gesswein Emulator

You can see that the BeagleBone Black that runs the software is connected to a network cable. I use PuTTY to log into the BeagleBone Black over SSH and run the commands needed. To image the disk I used the following commands:

./mfm_read -d 3 --ana --emu /sd/<file>.dsk --ext /sd/<file>.ext --note "<description>" | tee /sd/.log

After that I shut the machine down, and discharge the reservoir capacitors with a resistor, just to speed things up. The later revision has a jumper for this I believe.

Now is the time to run the Rainbow from the emulated disk. To do that you just connect the MFM connectors to the edge connectors on the emulator board, connect the power and power up.

This however does not automatically start a disk emulation however. I have not set that up because I want to use it to emulate different disks at different times. So I still need the network cable to be able to login and run the emulation this time. The commands to emulate the disk for the Rainbow are:

./mfm_emu -d 1 -f /sd/<file>.dsk

Note that I have to use drive id 1 for this to work on the Rainbow. Here it is working:

Running Rainbow With Gesswein Emulator

The Rainbow thinks it has a hard disk, but really it is just an emulator!

Posted in Retro-Computing | Tagged | Leave a comment

Corroded H7826 Power Supply

Now that I have got the DECstation 5000/240 working again it is time to move on to the TURBOchannel Extender that came with it. This is in a terrible state, with lots of corrosion all over the place. The worst of it though is the H7826 power supply. A lot of the heat sinks have corroded, leaving metallic flakes all over the board, presumably capable of creating shorts.

In some areas there seems to be some traces of a liquid which seems to be holding a lot of dirt and some of the flakes that have come off the heat sinks. I suspect this could be leaked electrolyte from some of the capacitors. However, none have obviously leaked, although they are mostly in contact with the surface of the board, so they could have leaked from underneath.

I will have to source some new heat sinks and find a way to clean the board thoroughly, and probably replace all the electrolytic capacitors. Only then will I dare to try to power it on. In its current state it is just too risky.

I don’t think the pictures below really do the problem justice, but hopefully they give a flavour of what state it is in.

Posted in Retro-Computing | Tagged | 1 Comment

Booting VAXen From The Network

Sometimes it is very useful to be able to boot a VAX from the network. The two most common scenarios are

  1. Your machine is diskless. As time goes on this is going to be more and more common as SCSI disks die out.
  2. You have a disk in the machine and you have the VMS media, but you don’t have devices to load the media from.

In these scenarios booting from a running instance of VMS is the easiest way to get your machine running, and the best bit is that you don’t even need it to be running on physical hardware. I use the wonderful SIMH emulator to run VMS, and I use it as the boot node for getting other VAXen running. You need your boot node to be in a cluster and then add the real VAX as a satellite to the cluster.

I am assuming that you have VMS running on SIMH. If you need to understand how to install VMS then Phil Wherry’s page is a great start. Some notes to consider when following these instructions follow:

  • Make sure you select to install DECnet Phase IV
  • You may want to include DECwindows support too if you want to run it on a real diskless VAX.
  • When prompted for the SCSSYSTEMID you can use the suggested value of 1025, which equates to a DECnet node address of 1.1. However, if you are likely to want to get on HECnet, or run this machine as part of an existing DECnet network then you may want to use another value. To set your node address as N.M, the SCSSYSTEMID is calculated as N * 1024 + M.

Once VMS is up and running, login as SYSTEM, register a DVNETEND and a VAXCLUSTER license. The first step is to set up DECnet, but this is only so that you can communicate between the boot node and the satellite using DECnet. To configure DECnet run @NETCONFIG. For the node address use the address that corresponds to the SCSSYSTEMID you used above, so if you used 1025, then enter a node address of “1.1”. You do not need the node to operate as a router (and you need to register a DVNETRTG license if you do). I then choose all the defaults when asked about accounts. Finally edit the SYSTARTUP_VMS.COM file so that DECnet is started automatically (uncomment the line START/NETWORK DECNET).

Now you are ready to setup the boot node as a cluster. To do this run @CLUSTER_CONFIG_LAN. Choose option 1 to ADD <node> to existing cluster, or form a new cluster, then choose to use the LAN for cluster communications. The cluster’s group number just needs to be a unique number with respect to any other clusters you may have, unless you want it to be part of an existing cluster. Select Y when asked if it will be a boot server (obviously, that is the whole point here). For the ALLOCLASS I used the second number in the node’s DECnet address (for address M.N, I used N for the ALLOCLASS).

After the machine has rebooted it is time to add the VAX as a satellite. Login as SYSTEM. I found I had to run the following commands, although I am sure I have not had to that before


Then run @CLUSTER_CONFIG_LAN, select 1 to ADD a VAX node the cluster, then select that the node will be a satellite. Give the node a name and an SCSSYSTEMID that translates to a DECnet address. The hardware address for the new node’s LAN adapter is its MAC address. Once the satellite has been configured, just start up the physical machine and tell it to boot from the network adapter.

There is a full transcript here.

If you have a local disk on the satellite then you can mount the disk and restore the basic VMS saveset to the disk. I then copy the remaining savesets onto the same disk, so that when I boot off the physical disk I say the remaining media is on the same disk. That is how I install VMS onto a physical disk from the network.


Posted in Retro-Computing | Tagged | Leave a comment

DECstation 5000/240 Running Ultrix

Having recently fixed the DECstation 5000/240 PSU, I wanted to see that the machine actually runs properly and can run an operating system. I had previously had another DECstation 5000/240, which I had swapped for something else, but I had kept the disks where I had installed Ultrix and two versions of NetBSD, so I decided to try these first.

I found that the NetBSD 5.0.3 disk will no longer spin up, which is a shame, but may be fixable, so I will look at it again at some point. The NetBSD 1.5.3 disk would boot up to a point and then hang. The Ultrix 4.4 disk booted but the screen went blank once it went to the graphics mode. I suspect that the NetBSD 1.5.3 and Ultrix 4.4 installations don’t like the PMAG-JA graphics card.

I found another spare SCSI disk, a Seagate ST12400N, and installed Ultrix 4.5 on it successfully. Sadly, when I switched it on again later the disk would not initialise. I could hear it seek a few times and then spin down. So I decided to overwrite the Ultrix 4.4 disk, an RZ56, with a new installation of Ultrix 4.5. Here is the setup I used:

DECstation 5000 Model 240 Ultrix Installation Setup

I used an RRD43 that I have. The tray won’t stay closed properly, so you can just about see the head of a screwdriver that I propped against the tray to keep it shut!

The first attempt failed after what looks to have been a file copy verification error, so the disk can’t be too good. A second attempt succeeded, here is the machine working:

Posted in Retro-Computing | Tagged | Leave a comment

DECstation 5000 Model 240 PSU Fixed

I have finally fixed the H7878 PSU from my DECstation 5000 Model 240. When I last posted about it, the PSU was not working very well, it was unable to sustain any load. Following that post I asked for suggestions on the ClassicCmp mailing list.

The first suggestion was to see if there were any significant dips in the voltage across the input smoothing capacitors. But the voltage was rock solid around the 330V mark on the DM DC range of my Digital Multi-Meter.

The second suggestion was that perhaps something was shorted on the outputs, causing the overcurrent sensing to switch it off. Some suggestions talked of checking low value resistors for shorts, there are two 10ohm resistors, and one looks like it has a bit of physical damage, but they both measured fine. I also checked the capacitors across the 5V output, particularly the ones that I had not by then replaced, they seemed OK. Nevertheless I decided to replace the two big ones across the 5V output just in case, particularly the 3300uF one (can’t see a label to reference it) which measured a bit high on ESR, the 8200uF one also had a relatively high ESR, so it was worth replacing too. Replacing them failed to make a difference.

I then had a closer look at the riser board on the output section, as this is probably doing the regulation. I noticed that C407 (120uF 25V) was measuring high on ESR, and then I realised there were signs of leakage. There were also some marks (heat perhaps?) around C408 (120uF 16V). I decided they should be replaced.

Having removed C407 I realised it had definitely leaked. Reaching C408 was going to be hard, so I removed the whole riser. This would make it much easier to repair and clean, and I could see that there appeared to be some cleaning needed. Once I got the riser out I could see some shiny residues on the underside, I am not sure that they can be capacitor leaks, but they did look a bit like it.

I finished replacing C407 and C408 and cleaned everything up. This is the result:

I re-installed the riser and tested it again. I used the MicroVAX 2000 load board on the 5V output. This time all the fans turned immediately, the 5V output looked good, and there was no significant ripple. The 12.1V output measured 12.4V though, I am not sure if that is OK or not.

I partially reassembled the PSU and installed it in the machine. The machine ran fine, I ran the self tests a few times and left the machine running for around an hour or more and it stayed running. So I finished reassembling the PSU, re-soldering back the LED wires and finally reassembled it again.

I replaced quite a few capacitors in the end:

H7878 Parts Replaced

Here it is, working once more:

Posted in Retro-Computing | Tagged | Leave a comment

DECstation 5000 Model 240 PSU Failing Under Moderate Load

The H7878 PSU in my recently acquired DECstation 5000 Model 240 had a shaky start and then stopped working altogether. I found a problem with the input rectifier smoothing capacitors, at least one of them anyway. It was a real job to remove them because they had been epoxied on. I had to cut the first one, somehow I was able to remove the second intact. The one I cut did not have any liquid electrolyte, although the core was damp with it, I am not sure if that is normal or not.

Notice the cardboard disc that was underneath the capacitors. It seemed clean in both cases, so I am not sure if there was leakage or not. The card discs seem to be there to isolate the base of the capacitors from some contacts on the board, so I kept them. The replacement capacitors were slightly smaller so you can see the cardboard discs now.

PSU With New Smoothing Capacitors From Above

After I put it back together I first tried turning it on with no load. The fans twitched once. I assumed that as there was no load the PSU was shutting down immediately. So I put it back in the machine, but nothing happened when I powered it on, the fans did not move at all. I later realised this was probably because the PSU cannot handle the load.

I took the whole thing out again and checked it over. The rectifier voltage across the newly replaced capacitors was looking good. I then measured the outputs of the main transformer. The three outputs seemed to be producing 50VAC, 1VAC and 8VAC. So the inverter and the main transformer seemed to be working fine.

I then realised that the PSU was actually producing voltages on its outputs. I put a 10W 15ohm resistor across the main 5V output and got a voltage of about 5-6V. The ripple looked like it might be high though. So I replaced as many of the output stage capacitors as I could easily reach. These were:

Label Value Voltage
C21 220uF 25V
C36 2.2uF 50V
C26 (see below) 10uF 25V
C24 330uF 16V
C40 120uF 25V
C41 120uF 25V

C26 was a bit different. It was a small purple capacitor made Sanyo, it was marked simply “10 25”. Since it is on the 12V output I guessed this meant that the “25” referred to the voltage. Below are some pictures of this capacitor:

After replacing all these I found that the behaviour was not a lot different, although ripple seemed to be much improved on the +5V output. The 12V fans twitch for a bit and then one of them will start running. The +12V output starts at around 1.5V and then after about 5 seconds or so it rises to +6V, whereupon one of the fans will start running, the +12V output then varies between +5V and +7V, sometimes dropping as low as 4V or rising as high as +8V. This happens when I load the +5V output with a 10W 15ohm resistor, no load for the -12V output, and the fans loading the +12V output.

If I put a bit more load on the other outputs then the +12Voutput won’t rise above 1V or so and the fans don’t even twitch. I tried a 2ohm load on the +5V output and that kills the +12V output, even though the +5V output is rated at 40A. Similarly I tried a 66ohm load on the -12V output and that too kills the +12V output. The -12V output is rated at 0.2A, so I am close to the rated limit in that case.

Basically it seems that the PSU simply cannot cope with much load.

Posted in Retro-Computing | Leave a comment