I have a number of RF72 hard disks. These are 1GB DSSI disks (DEC rather grandiosely called them Integrated Storage Elements, or ISEs). Most of them work well, but I have one that will not spin up. The console firmware in a VAX 4000-500 does not see the disk if I do SHOW DEVICES. I decided to see if I could work out why it won’t work.
I have removed the drive module to see if I can find a fault that I can fix, the picture of the board is below. The drive module uses a Motorola 68000. The blue connector on the right connects to a ribbon cable coming from the head-disk assembly (HDA) for the actuator and read/write, and the 8-way socket half way down the board and towards the left supplies power to the spindle motor. The power is driven by the large MPM3003 power transistors that can be seen at the top of the board. There are a lot of surface mount components, if one of these turns out to be faulty then I could be stuck as I don’t have the means to replace them reliably.
Clearly the disk has not been treated well because one of the locking clips on the DSSI connector on the left of the drive module is missing. There are some dents and scratches on the HDA itself too. The shock mount assembly is also distorted, so it must have taken a heavy blow.
In order to be able to test the disk on the bench I needed to get power to it. These disks do not use the normal 4-pin Molex connector used in PC disks. Instead they use a 5-pin connector, a friend identified it as part of the Molex Mini-Fit range. The 5th pin is “Power OK” signal and appears to be a 5V signal. So I made the little cable below to connect the disk to a standard PC power supply. Note that the 5V pin on the PC side goes to both the 5V pin on the DSSI side and also to the 5th pin to provide the Power OK signal. I tested this cable successfully on a known good RF72 disk drive.
Having got the drive module on the bench I did some tests and found the following:
One or two of the electrolytic capacitors have a fairly high ESR, but I found this to be the case on a known good drive module too.
When I apply power there is a click as if it is trying to spin up the disk but the fault light comes on immediately and nothing else happens.
The bad drive module will not work if I put it in a working RF72. It fails in the same way.
The drive module is not sending power to the motor. Looking at the power transistors (MPM3003), there is no signal on the gate pins.
There is clock activity on the MC 68000 on the drive module.
A possible issue with the obvious physical damage is that the ROM could have become unseated. So I re-seated it, and while I was at it I took the opportunity to dump the ROM, the image is here. There is a timestamp in the image of 17-MAY-1991 10:12:41 at address 0x6830. The re-seating did not make any difference.
I have not tried putting a known good drive module on the HDA of the bad disk just in case something in the HDA is damaging the drive module.
Clearly more investigation is required, I think the CPU could be working because the Ready LED does come on briefly and you can hear an attempt to start the disk. Possibly the DSSI interface is not working because the drive module is not being recognised by the VAX console firmware.
A few years ago I acquired a KDA50, which consists of 2 modules, the M7164 and the M7165, an RA72 disk, two RA70 disks, an RA60 disk drive, an operator control panel (OCP) and some big cables, with one I/O bulkhead for installing disks and controller in separate cabinets.
The controller and the disks use SDI, the Standard Disk Interconnect, which was part of the Digital Storage Architecture, and allows controllers to connect to large system disks (large for the time anyway). SDI allowed disks to be connected via long cables, so the disks could be housed in separate cabinets. SDI encoded signals serially, so it seems to be an early pre-cursor to SATA. SDI appears to date from the early 1980s.
I decided fairly quickly that I was unlikely ever to have the time and resources to get the RA60 going so I passed that on to a friend. The other stuff sat in my pile for a long time and I kept promising myself I would look at them one day. Well one day finally arrived a couple of days ago and I decided to get everything out and see what works.
To begin with I brought the disks in from the cooler storage area and allowed them to warm up to ambient temperature overnight before doing anything with them at all. I first decided I would see if the disks spin up when powered up outside the machine, by connecting them to a regular PC power supply. The disks did not spin but I saw some LED activity. I was not too disheartened because I thought that perhaps they might need to be connected to the KDA50 controller before they do anything, and that did indeed turn out to be the case. It turns out that they really need to be attached to the KDA50 controller before they will do anything.
So I needed to get the KDA50 installed in a machine. I wasn’t sure which machine to use. I decided to try the VAX 4000-500, mainly because it is handy. This turned out to be a poor choice. I think when I first got the KDA50 I may have installed it in this machine and it appeared to work, it displayed a cycling pattern on the LEDs of both modules, which indicates that it is not talking to the host yet. This turns out to be normal until you actually boot an operating system.
However, before installing the KDA50 again I decided to check the KDA50 User Guide. I found that both boards in the set have two jumpers, W2 and W3 that are supposed to be removed in a Q22/CD slot. I checked the enclosure manual for the VAX 4000-500, it is a BA440, and all the slots are Q22/CD. Having not checked for this before, I asked on the VCF forums for advice. The jumpers are present by default, presumably because the KDA50 is for an older generation of VAXen with a serpentine Q-Bus. I think I may have been lucky the first time, possibly because I had a TK70 with a dummy card in the CD slots before the KDA50. Unfortunately the jumpers are soldered in and not the type with a header, so I had to desolder them. Rather than remove them completely though I desoldered only one of the leads, lifted the link while bending it out of the way and finally putting a little bit of insulating tape over the holes left behind.
With this done I installed the boards in my VAX 4000-500, I got the cycling LEDs and the following from the firmware console:
I think the “?” for the UQSSP Disk Controller is because there was no actual disk attached.
The next step was to connect the disks. It was at this point I realised that the BA440 enclosure isn’t really designed to be used with SDI hardware. There are no molex power cables to get power to the disk drives, there is nowhere for the OCP to go and no connection for the OCP to the CPU. The VAX 4000-500 requires a separate enclosure for RA-series disk drives, this is confirmed in the manual. I realised that I would have to switch to my MicroVAX 3400, which is housed in a BA213 enclosure. I hadn’t switched this machine on for quite a long time, so I was a little concerned as to whether it would work. I needn’t have worried, it powered up fine.
I installed the KDA50 in the MicroVAX 3400. I then had to remove the DSSI disks, the DSSI media faceplate and the DSSI OCP before I could install the RA72. The problem was how to get power to the RA72. The RA72 uses 4-pin molex power connector, and the 3400 uses DSSI power connectors “natively”. The TK70 tape drive uses a little adapter cable (part number 17-01937-01) to convert from DSSI to the ordinary 4-pin power connector, so I took the TK70 out. Sadly it meant I could only test one disk at a time because I couldn’t find a splitter cable. Finally I installed the RA OCP and connected it all up, with the disk connected to the KDA50 by Port A. Unfortunately I don’t have the RA media faceplate and I didn’t bother to remove the internal cables from the I/O bulkhead at first, so the setup looked a bit ugly but it was functional:
I am not entirely clear about all the buttons on the OCP and without a media faceplate I don’t have the labels to know for sure. The OCP is shown in Table 1-4 of the BA213 manual. The bottom button halts the CPU and puts you back to the firmware console.
When I powered it on I got the following from the firmware console:
This was very encouraging and I was able to boot VMS from a boot node and verify that the disk is OK.
I tried the two RA70 disks next. I had to swap the mounting rails round though so that the end of the disk with the SDI and power connectors faced the front of the machine. One of the disks worked and one did not. The curious thing is that the two RA70 disks showed up in the console like this:
UQSSP Disk Controller 0 (772150)
and like this:
UQSSP Disk Controller 0 (772150)
I don’t know where the names DUA33 and DUA64 came from. I know the firmware has some SET HOST/DUP commands, I have used them with my DSSI disks to change things like that, but those commands don’t work with this controller and disk combination, not even if I try SET HOST/DUP/UQSSP. I feel it must be possible but I don’t know how.
The MicroVAX 3400 is back to using the DSSI disks, but the KDA50 is now permanently installed in it too. I would like to find the RA media faceplate for the BA213 enclosure that goes over the OCP. This is part number 70-24534-01 and is also described as a ‘BA213 RA DISK OUTER PLATE ASSY’. I would also like to find a couple of 17-01937-01 adapters to convert the DSSI power connectors to standard 4-pin power connectors.
Recently I fixed the power supply of my VAXmate, but found that there was a problem on the monitor board, which may be what caused the power supply to fail in the first place. My suspicion immediately fell on the flyback transformer, which is a common failure in vintage screen displays.
I removed it and tried doing a ring test on it to see if there are any shorted coils. Here is a picture of it, with the pins numbered, as close as I can tell, according to the description in the Figure 14-4 of the Technical Description.
VAXmate flyback transformer from the monitor board with pins numbered
My probing with a multimeter suggests that pins 1-2-3-7 form one winding, and pins 5-6-8 form a second winding. This does, however, correspond to the pin numbers shown in the Technical Description.
My ring tests showed good ringing between the following pin pairs:
Where the ringing looks like this:
Good ringing from the flyback transformer
I also saw some poor ringing between the following pin pairs:
Where the ringing looks like this:
Poor ringing from the flyback transformer
I can’t be completely sure that the poor ringing means the transformer has failed because I don’t know how the transformer has been wound, and it could be a low number of turns in the particular part of the winding.
I have a MicroVAX 3100 Model 95 that I acquired some years ago. At 32 VUPs it is one of the faster MicroVAXen.
From the front with RX33 and TZ30
Showing 32MB of installed memory
Some time ago I stopped using it because memory modules would suddenly stop working. At the time, I checked the power supply and some of the voltages seemed to be a little out of tolerance, in particular the output marked as 5.1V seemed to be producing 5.3V. I put the machine to one side because at the time I had little knowledge of repairing power supplies.
Recently I decided to take another look at the power supply with a view to getting the machine up and running again, given it is such a comparatively fast machine. I checked the ripple on the outputs but what I saw were spikes. The spikes seemed to reduce in amplitude after a few seconds. It turns out that I was measuring the ripple with poor technique, but I don’t know what the original ripple was now, because I replaced the majority of the electrolytic capacitors before realising my error.
I decided to replace any vaguely suspect electrolytic capacitors, which included the two big smoothing capacitors as one had a high ESR and the other appeared to be bulging slightly. I discovered that one of these would not fully discharge, which I found when I tried to measure its ESR after having had it powered on. Thankfully the meter had some protective diodes, which now need replacing.
After replacing the suspect capacitors I excitedly put it all back together, only to find that it would not power on. It turned out, that following some other sparks I managed to get from the not fully discharged smoothing capacitor I had ended up frying the UC3842N pulse width modulator. Fortunately I had a spare on hand (UC3842AN), replacing it fixed the power supply, and also seemed to resolve the problem of the capacitor not discharging.
Sadly, I also managed to break the connection to the power LED out to the front of the PSU, so now I won’t have that working.
After resolving my measurement mistakes the ripple on the 5V and 12V outputs was about 20mV and the voltages seemed to be right. So I put the machine back together again.
Unfortunately it seems that one of the memory modules is not quite right because the firmware reports an error:
KA51-A V2.6, VMB 2.1
Performing normal system tests.
? Test_Subtest_40_06 Loop_Subtest=00 Err_Type=FF DE_Memory_count_pages.lis
16 MB RAM, SIMM Set (0A,0B,0C,0D) present
Memory Set 0: 00000000 to 00FFFFFF, 16MB, 32768 good pages, 0 bad pages
Error: SIMM Set 1 (1E,1F,1G,1H)
SIMM_1E = 16MB ?? SIMM_1F = 16MB SIMM_1G = 16MB SIMM_1H = 16MB
Memory Set 1: 01000000 to 01FFFFFF, 16MB, 0 good pages, 32768 bad pages
Total of 32MB, 32768 good pages, 32768 bad pages, 112 reserved pages
However, VMS seems to think everything is OK.
$ sh mem
System Memory Resources on 10-MAY-2020 13:24:53.80
Physical Memory Usage (pages): Total Free In Use Modified
Main Memory (32.00Mb) 65536 21316 42491 1729
So I am hoping the machine will now work reliably.
In previous posts I have been describing my attempts to find a fault in the H7270 power supply of my VAXmate. After a long time studying and testing it I finally found the fault. It was a shorted diode (marked D24 in my reverse engineered schematic) on the secondary side that was part of the +28V supply to the monitor board.
The reverse engineered schematics are here. Note that they probably still contain errors, especially on the secondary side.
H7270 Primary Side
H720 Secondary Side
The pictures with the part labels are here.
VAXmate H7270 Power Supply – Primary Side With Parts Labelled
VAXmate H7270 Power Supply – Secondary Side With Parts Labelled
The problem was that the power supply did not appear to start at all. There was no blip of the fans, no clicking from repeated attempts to start etc. However, with an oscilloscope it was possible to tell that it was attempting to start, but shutting down after only about 20ms, with the SCR (D19) on the primary side detecting an overcurrent. The trace below was taken by lifting R32 and powering the UC3842N (E3) from a bench power supply with the mains still coming in through the AC inlet. It shows the SCR triggering.
However, I am not clear why the SCR was taking so long to trigger, in the trace the 15V peaks on the Q1 source have been going for about 15ms, so quite why it triggers the SCR only after that time is not really clear to me. It seemed to me that the peaks on Q1 source were not causing the SCR to trigger, but it turns out that they must have been the cause.
There didn’t seem to be any problem on the primary side, so on the advice of several members of the classiccmp mailing list I checked the secondary side more carefully. It was clear that the secondary side crowbar was not getting triggered, because this is the trace:
I desoldered and tested the rectifiers on the secondary side, and they tested fine. However, the +28V supply for the monitor board uses a rectifier constructed of discrete diodes (D23 and D24), and it turned out that D24 was shorted. I have replaced both of them to be safe.
It was a bit of a problem to get to the failed diode because it is under the large heatsink. I had to remove the heatsink, but I ended up doing it twice and the second time ended up damaging the vias a bit, so I didn’t completely remove it.
I also replaced capacitors C54, C55 and C56 in the secondary side as they had a high ESR, two of these are part of the +28V supply. Along with a few others that had a marginal or high ESR.
The complete list of replaced parts on the PSU is in the list below:
C10, C33, C50, C51, C54, C55, C56, all because of a high ESR.
D24, the shorted component.
D23, just in case as it is paired with the failed D24.
E3, a UC3842N, replaced with UC3842AN, but only due to errors damaging the original.
I was a bit concerned as to why D24 had failed. As it drives the monitor board I examined it. I could not see any visible damage, and it does not appear to present a short circuit. I checked the electrolytic capacitors on the video module and I found 5 where the ESR is marginal. They were all 15uF 16V parts, and they are marked in the picture below. Number 4 has a better ESR but I decided to replace it anyway since all the others have become marginal.
VAXmate Monitor Board With Replaced Capacitors Marked
I reassembled the machine and powered it on. It made some reassuring beeps and a few lights came on, but there was no image. A few moments later I noticed a burning smell, so I quickly powered it off. It seemed to me that the smell was coming from the monitor board, so I disconnected it and powered the machine on again. It seemed to work (apart from the video display of course), the diagnostic LEDs did not indicate an error, the floppy disk drive was accessed and it seemed to react to keypresses. There was no more burning smell.
It is clear that something has failed on the video module that caused the power supply to fail. It seemed as if the burning smell may have been coming from the flyback transformer. If that is the case then I suspect that this machine will never work again as I am unlikely to find a new flyback transformer. I took the monitor board out again and another physical examination does not show anything visibly wrong. I wasn’t completely sure that the EHT lead was making good contact with the anode of the CRT though.
I would like to acknowledge the help of all of the following people from the classiccmp mailing list for helping me to find the problem in the power supply. In alphabetical order they are:
Plus others on the mailing list who preferred not to be named, but they know who they are.
Recently I had occasion to dig out one of my VAXstation 4000 VLC machines. Before using it I decided it was prudent to check the power supply, a H7109-C, to make sure there were no leaking or bulging capacitors.
It was a good job I did, because one of the capacitors had leaked. So I took the opportunity to check all the electrolytic capacitors and found several that had a high ESR. I have now replaced them all and cleaned up the leak. I also noticed that a ceramic disk capacitor was split (see photo). This capacitor did not measure a stable capacitance, so I replaced that too.
In the photos below I have numbered the replaced capacitors and give their values here:
470uF 25V, this is the one that leaked
3.3nf Ceramic Disk, this is the one that is split.
After replacing the capacitors I powered on the machine and it worked fine. The ripple was about 50mV on both the 5V and 12V outputs.
Recently I bought a set of 8 memory modules that were advertised as “DEC VS4000”. They had a part number of 50-19464-02. This is one of the modules:
50-19464-02 Memory Module
I had assumed this was for the VAXstation 4000 VLC, but it turned out not to be the case, the VLC takes modules that are physically narrower. However, they did fit physically in a VAXstation 4000 Model 60. Unfortunately they did not work in that machine and seemed to make the power supply think there was a short circuit.
On suggestion from the classiccmp mailing list, I dug out my DECstation 2100. I discovered mine had 10 of these exact modules installed, out of a possible 12. So now I know that this memory is for a DECstation 2100, and each module is 2MB. I tested the memory in the 2100 and it all worked fine. So now I can have 24MB in my 2100.
The H7874 is the power supply used in the BA4xx and R400x enclosures, such as several machines in the VAX 4000 series. I have a VAX 4000-500 in a BA440. I had some problems with this power supply that seemed to get fixed after replacing some leaked electrolytic capacitors. I still have intermittent problems with the power supply shutting down occasionally, particularly if the machine has not been powered on for a while.
It is an astonishingly complex power supply and very hard to disassemble too. A post on the classiccmp mailing list requested information about this power supply. I have partially reverse engineered the schematic of the 12V output board so I am posting it here. It is unlikely to be correct either.
I have been doing further analysis of the failure of my VAXmate’s H7270 PSU. To start with the schematics are now greatly improved. I am still not sure how correct they are, but here are the latest and greatest:
H7270 Primary Side
H720 Secondary Side
I am fairly sure that the problem is that the primary side is detecting an overcurrent situation. What I am less sure about is whether the overcurrent is real or not. Below is an oscilloscope trace that I think shows this. In this trace the channels are set up as follows:
Ch1: NE555 output, trigger for the one-shot on the negative edge.
Ch2: Vcc (pin 7) of the UC3842 PWM.
Ch3: SCR gate of D19, between R14 and R15.
Ch4: Source of Q1
Clearly the gate of the SCR (Ch3) goes high and the 555 stops oscillating. I am a little unsure if I should really be seeing so much spiking on channels 3 and 4.
I decided to look at how the UC3842 PWM power supply is behaving. Accordingly I set up the channels as follows:
Ch1: NE555 output, trigger for the one-shot on the negative edge.
Ch2: Drain of Q1 (pin 2)
Ch3: Cathode of D6
Ch4: Anode of D7
This is what I got:
I am seeing a lot of large spikes on the drain of Q1. I wonder if they are enough to make R13 detect an overcurrent? I don’t know enough about switched mode power supplies to know if this is expected behaviour.
Now that I have acquired a DSO (a Rigol DS1054Z), I have been doing more work to understand the failure of the H7270 PSU on my VAXmate that I first blogged about here. I still don’t fully understand the problem, but I now have an improved insight into what is going on and a possible explanation. The schematics I refer to in this post are:
H7270 Primary Side
Below is a trace from the oscilloscope during startup of the PSU. What you can’t see here is that Ch1 was steady at about 5V for 15ms before the oscillations start, and this corresponds to a point when voltage on Vcc to the UC3842 has finished ramping up.
CH1: Vcc for NE555 and Vref for UC3842 CH2: Base of switching transistor Q1 CH3: Gate of SCR D19 which shuts down on overcurrent CH4: Current sense resistor (R13)
It looks like after the 15ms that the UC3842 starts to switch the transistor. This also corresponds to when Vcc on the NE555 starts to oscillate. The transistor switches for about 25ms and then stops. I have not been able to work out why it stops. I had been thinking that it was because the SCR (D19) was being triggered, but Ch3 of the trace shows it getting pretty much the same signal throughout the period and there is no change at the time the PSU stops, so I am not sure if the SCR is being triggered or not. After all, the voltage across the current sense resistor is not varying much either.
I was puzzled though as to why Vcc to the NE555 starts to oscillate a lot. My guess was that it is because the output from the transformer has not settled yet. I don’t understand why it is quite so spiky though, I imagined a capacitor would smooth it, but I am not sure. I found one capacitor that looks like it is supposed to smooth Vcc (C11). I took it out of circuit to measure it, nominally it should be 220nF, it measured 335nF initially, but as I left it connected to the meter it dropped in about 5 minutes to 300nF, and kept dropping slowly over time.
I decided I needed to improve my schematic. To help me detect the connections correctly removed the transformer (T1). This is it here:
While I had the transformer out of circuit, I decided to carry out a ringing test on it. Unfortunately, this is where I think I may have found the problem. Four of the windings ring correctly, but one of the windings does not ring, it is the one shown as P1 in the schematic, and it is the two pins on the left in the picture above. However, the Technical Description says that one of the primary windings is operated in flyback mode to provide bias voltage for the pulse width modulator. I don’t know if that has any implication for the ringing test though. Here is the scope trace
H7270 Transformer P1 Ringing Test
Another curiosity is that the pin on the right in the picture does not seem to be connected to any other pin on the transformer. I suppose this might mean that a wire has broken somewhere perhaps. I don’t know if transformers can fail with a nasty smell and no outward signs of damage, but I fear it has failed, and that it will be very hard to replace. I am also concerned that if I could replace it some other fault would damage the replacement too.