HP 4145B Semiconductor Analyzer: Repair with photos (dead ROM) [FIXED]

[Berni]

Notable posts in the tread:
- Small character ROM image:  https://www.eevblog.com/forum/testgear/hp-4145b-semiconductor-analyzer-repair-with-photos-(display-garbage-dead-rom)/msg1702649/#msg1702649
- Big character ROM image: https://www.eevblog.com/forum/testgear/hp-4145b-semiconductor-analyzer-repair-with-photos-(display-garbage-dead-rom)/msg1743677/#msg1743677
- Conclusion to the repair: https://www.eevblog.com/forum/testgear/hp-4145b-semiconductor-analyzer-repair-with-photos-(display-garbage-dead-rom)/msg1743662/#msg1743662




So i managed to score a HP 4145B Semiconductor Analyzer on the cheap and it is quite the boat anchor.



Because these things keep there firmware on floppy disk its pretty good that i got a system disk in it. However it was soon clear that we have a problem. While it boots fine it seams to be only able to display the character "a" in the wrong place, yet seams to still be drawing lines just fine. Clearly there is some issue with the graphics circuitry.



So off we go with the lid and we are kinda mooned. It looks like someone has opened the cover before, so this might be a failed repair attempt. Uh oh...



After a lot of screws later we are in like flin.



Around the back there is nothing interesting apart from the very important floppy drive. Also found a loose screw down here... thats not fun



So looking for the graphics stuff i have a look around the VERY hard to pull out cards. No sight of the graphics hardware, but i did find the Motorola 68000 CPU in here



Running out of ideas i go on to try and take out the CRT if it has any clues. This ended up being quite involved as you have to take apart the front pannel.



Has taken me quite a while to figure it out as i didn't have the full service manual of the thing.

[Berni]

Eventually i got it out, turns out you have to undo some screws around the frame and 3 screws at the back of the CRT module.



And finally its out, and AHA here is the "graphics card" on the left side of the module!



Its quite a neat assembly actually, turns out this whole thing is a standard part by the name of "HP 1345A Digital Display". Service manual here: http://vectorvgapro.com/PDFs/HP/HP%201345A%20OSM%201985.pdf
It is basically a CRT module that runs from 5V and +/- 15V while having a 16bit parallel interface. Capable of drawing vector graphics and vector text (No raster images tho)



The neat thing about it is that if you just power it up without connecting the data cable it goes into a self test mode and displays a test pattern. Once again its clear that it draws lines just fine, yet text is a mess of "a" characters again



So pouring over the nice service manual of it shows well how it works. It uses a special vector drawing chip that outputs 8bit data to a pair of X and Y DACs (Off to the right) for steering the beam. That self test turns out is done by a 8bit MCU sitting on the bus, and it seams to be the only job for it! A whole MCU just for a self test feature!

So after poking my scope around i noticed that the outputs of the character ROM chip(Shown in red) are always 0 while the address lines are all moving about. Reseating it in the socket does nothing, it gets power and all, so i conclude that this chip is dead.



And here is the top side of this vector graphics board. Red arrow is the dead character ROM.





To further confirm it i tried to unplug the chip and run it, indeed i got garbage where the "a" characters used to be, and if i tie one of the lines on it to 5V i get different characters in place of the "a". So yes this thing is toast.

I have tried to find ROM dumps online but so far i could not get anything. So i looked for alternatives. One option is getting a new display module, as of right now these can be found on ebay for about 300 EUR. Since that's almost 3 times as much as i payed for this whole thing that's not so great. Another option is to buy a NewScope5 LCD retrofit kit: http://www.simmconnlabs.com/1401/2501.html But that costs 400 USD. At least that gives me a modern LCD with color but still thats a bit steep.

Does anyone know where to find a ROM image of this chip?

As a last resort i did get the idea to try and guess the contents of the ROM trough brute force. Its only a 512 byte ROM and only half of it is actually used up. The input to it should be ASCII values (The char set is in the service manual), i just don't know what the output should be. Essentially this ROM looks up the address of a character drawing in the larger 2KB ROM, while also generating some control signals for the vector chip. Maybe a CPLD could be used as a ROM emulator to quickly change contents while running. Once i have the correct ROM image from that i can then burn it into a actual memory chip and replace it. Sounds like a lot of work still...

[mnementh]

I vaguely remember reading in one of my newsletters back in the day of someone trying to recover the ROM in a situation like yours; their attack vector was to try differing combinations of under & over-voltage and 1/4-1/2-full speed read attempts until they got data that they could make repeat.

I'll fully admit that I'm grossly ignorant of the nuts & bolts of what they did and how they configured the burner... just that was the gist of their technique.


Cheers,

mnem
*Ifni Bless this mess*

[edpalmer42]

To further confirm it i tried to unplug the chip and run it, indeed i got garbage where the "a" characters used to be, and if i tie one of the lines on it to 5V i get different characters in place of the "a". So yes this thing is toast.

Continue with this idea.  Unplug the dead chip.  Set up an Arduino or RaspPi to drive the output lines.  Step through all 256 values and see what appears on the screen.  Maybe you'll be able to figure out the pattern from that.  The fact that 0x00000000 generates a lower case 'a' is a good starting point.  Any bets on whether 0x00000001 gives a lower case 'b'?

What's the part number of the dead chip?

Ed

[picburner]

Perhaps this file can be useful:

[Berni]

It is a N82S147N Signetics (4Kbit PROM)

Oh wow thanks! That does look like its the ROM i need!

Now i just need to source a replacement chip and find a way to burn it.

[Berni]

I have contacted http://www.hobbyroms.com for programing me a replacement chip so if all goes well i won't even need to get a programmer for it.

Looks like this will save me a good weekend or two worth of work. Thanks picburner! 

[edpalmer42]

I see that the chip has tri-state outputs.  Have you checked that the enable line is active?

I also see that the prom has a a second copy of the data in the top half of the chip.  Maybe forcing the most significant bit high would get the chip working.

Ed

[Berni]

Update:
I ended up buying a new ROM chip and had it programmed by HobbyRoms for 25 USD including shipping. We will see if it works when i get it in the mail.

Yep i poked all the pins right next to the DIP package body with a scope probe, just in case i had a bad IC socket or something. It did indeed have power right on the pin and was permanently enabled as that pin is directly connected to ground.

[mnementh]

Update:
I ended up buying a new ROM chip and had it programmed by HobbyRoms for 25 USD including shipping. We will see if it works when i get it in the mail.

Yep i poked all the pins right next to the DIP package body with a scope probe, just in case i had a bad IC socket or something. It did indeed have power right on the pin and was permanently enabled as that pin is directly connected to ground.

Yes, but according to Ed Palmer's post, lifting that pin and pulling it HI might enable the 2nd copy of the ROM stored in the second register. Certainly would cost very little effort to confirm, and might prevent killing a new part if there's something else wrong in the circuit.

mnem
*unregistered*

[Berni]

I looked at the ROM image and i don't see a 2nd copy of it in there.



According to the schematic the pin tied to ground is address 5 of the ROM. So if you tie this pin high this adds 32 to all addresses that are accessed on it. So 0x000 becomes 0x020 and if you look at the hex of the ROM you see that that would cause the rom to always output 0xDE on its output. So this would make the display always show whatever character is mapped to ASCII code 0.

There is some duplication visible in there but its not at the right address.  There is likely a different styled font mapped to those areas. since bits 5,6,7 run past the ROM and change other control lines deeper down.

EDIT: Also i seen the highest address line move while its running so it was reading that duplicated area of memory.

[edpalmer42]

I didn't do a full byte-by-byte comparison, but it looks like the duplicated section starts at offset 00000100.  However, when I look at the schematic, I see what you mean about address 5.  I made the mistake of assuming that the bytes were just used in sequential order which would have left address 7 unused and probably tied low.  The truth is stranger than that.  I can't imagine why HP would do it that way.

You could still try driving the outputs and see what shows up on the screen, but since you've got a new prom on the way, there's really no point.  If replacing the prom *doesn't* cure the problem, it might be a useful test.

Ed

[Berni]

Yeah i also found it surprising, its possible its a speed thing since these PROMs can be banked inside.

Im going to try it when i get the new PROM chip since for now i assembled the unit mostly back together to store it. It was taking up most of my lab while laying out in so many parts. There is some more gear that needs fixing in the mean time.

[james_s]

Wow that thing is very dense, lots of stuff in there. Odd that it uses an electrostatic CRT with vector deflection. I'm curious to see what it does with the new ROM.

[mnementh]

Yeah i also found it surprising, its possible its a speed thing since these PROMs can be banked inside.

Im going to try it when i get the new PROM chip since for now i assembled the unit mostly back together to store it. It was taking up most of my lab while laying out in so many parts. There is some more gear that needs fixing in the mean time.

I stand corrected... the core weakness of armchair quarterbacking is that even if you've done the job before yourself, you don't have hands and eyes on the unit on the bench; so best you can offer is an educated guess.

But that'll never stop a engineer from kibbutzing.

Finite space is always a problem... hence my current battle in the mountains of JENGA. But I shall be victorious!!!

mnem
*hiding out from the Tejas suck*

[Berni]

Wow that thing is very dense, lots of stuff in there. Odd that it uses an electrostatic CRT with vector deflection. I'm curious to see what it does with the new ROM.

Yep it is up there with the most tightly packed gear i seen. I think i have something that even more packed than even this.

Turns out HP got a mention on Wikipedia for its electrostatic vector CRT: https://en.wikipedia.org/wiki/Vector_monitor This XY display they refer to is this HP 1345A Digital Display module. You just send it X Y points over a 16bit bus and it draws them connected by lines. It does make sense that you can draw things much quicker on such a CRT and since HP made scopes back in the day too they had plenty of experience with them.

I have another piece of HP gear that has the same sort of display. The HP 8566B Spectrum analyzer also has a green CRT in that shows vector graphics. I actually looked at the schematics of it to see if it happens to have the same character ROM i need (It didn't, only had a single big ROM for text and the text on it is all single font single style). Turns out the architecture is pretty similar where a ROM table and a DAC are used to electronically drive the beam around. The text, gradicules and multiple waveform traces are all vector drawn by the CPU. However they needed a electrostatic CRT in this because the analyzer also supports a a fast zero span mode. The ADC is too slow to digitize a 3MHz signal out of the detector so what they do instead is draw the gradicules and text under CPU control before switching over the screen into analog oscilloscope mode to draw the waveform and then back to vector mode. So to make a 3MHz CRO you do need static deflection. Oh and as a interesting fact there is no CPU chip in that instrument, the whole computer in there is built out of standard 74 series logic and some RAM and ROM (They do use a few 74 series ALU chips for math tho, but no hardware multiply).

I stand corrected... the core weakness of armchair quarterbacking is that even if you've done the job before yourself, you don't have hands and eyes on the unit on the bench; so best you can offer is an educated guess.

But that'll never stop a engineer from kibbutzing.

Finite space is always a problem... hence my current battle in the mountains of JENGA. But I shall be victorious!!!

mnem
*hiding out from the Tejas suck*

Happens to everyone. I have chased some rather silly red herrings before.

Oh im doing a great job of filling said finite space, if you look closely at the photos you might also find a pair of "HP 6050A System DC electronic load" sitting behind the red tool cart. They are about the same size as this thing, and probably weigh even more when full of modules. One of the modules in there is waiting for a repair, i can also make it a repair thread if anyone is interested.

[bson]

Looking at the schematic, I'd guess U12 is the main ROM containing shapes, and U5 only contains starting addresses into it, as mulitples of 4 since the A0 and A1 inputs to U12 are tied to ground and D0 out on U5 becomes A2.   A10 is D5 xor D6... Upper/lowercase?

How about this?  I'd offer $25 shipped and see if it has an okay ROM...

https://www.waybumarket.top/hp-agilent-8756a-scalar-network-analyzer-vpc-dot-bd-pn-0134566519-p-10060.html

[bson]

I really like the local voltage regulation for the U16 vector processor, using Q1 and Q2...

[Berni]

I already ordered a replacement ROM with the image burned onto it for 25 USD including shipping. It should get here in about 2 weeks.

I have not reverse engineered it in detail, just far enough to understand that this broken ROM holds the offset for the actual character data in the big ROM. Im guessing they did this to be able to use a smaller ROM for character data. It makes more sense for me to just simply use a slightly bigger ROM and not need this one in front.

As for what the duplication is for is im guessing that this is for the lighter stroked font style and for rotated fonts where maybe the vector processor does the rotation by mirroring and flipping signs.

[james_s]

If you haven't already, this may be a good time to dump the data from any othe ROMs in the machine, if one has failed it's not entirely unlikely another one will drop at some point.

[Berni]

I don't have any parallel EPROM programmers (Yeah I'm not very old) to do that. I am looking to get one since i end up dealing a lot with old equipment and i have a Dallas NVRAM that needs replacing in another piece of gear. Any suggestions what programmer to get that works on modern PCs?

Oh and the ROM for the main CPU is annoyingly not sockected so i probably wont go trough the trouble of reading that one. I guess they figured nobody will need to update it because all it does is load the firmware from floppy disk and run it. Its the floppy that contains the actual firmware.

[james_s]

I like the GQ-4X, or if you're on a budget there's the TL-866, but look at the supported device list before buying to make sure it does what you want.

[bson]

Where did you get the image for the ROM?  It has to match the contents of U12 since it outputs an address for it!  One can probably, reasonably, assume there weren't many revisions if any, and so the chance of success is high but still, if any characters look weird or erroneous, or blocks of codes don't render at all this is probably why...

The next time I have it open I'll upload a dump of my 3577's vector graphics ROMs, but it's not exactly an instrument you just pull off the shelf and pop open.

Maybe A10 into U12 is for brightness?  I don't think it has actually has any lowercase glyphs, so the case may be used for brightness and so UC/LC look the same just UC is much brighter.  Just a thought.

[Berni]

I like the GQ-4X, or if you're on a budget there's the TL-866, but look at the supported device list before buying to make sure it does what you want.

Thanks il look into those programmers. That TL-866 looks quite capable indeed. I am willing to pay a bit more for a more universal unit. Programmers that can write PROM chips are pretty rare so il probably not get one for that, but reading of mask ROMs could be very useful for making backups of any memory chips i come across in test gear.

Where did you get the image for the ROM?  It has to match the contents of U12 since it outputs an address for it!  One can probably, reasonably, assume there weren't many revisions if any, and so the chance of success is high but still, if any characters look weird or erroneous, or blocks of codes don't render at all this is probably why...

The next time I have it open I'll upload a dump of my 3577's vector graphics ROMs, but it's not exactly an instrument you just pull off the shelf and pop open.

Maybe A10 into U12 is for brightness?  I don't think it has actually has any lowercase glyphs, so the case may be used for brightness and so UC/LC look the same just UC is much brighter.  Just a thought.

It does have lower case characters, but most of the menus use uppercase.

I can't know for sure if the ROMs match but given that this display module has real HP product number i would assume all of them are the same thing. I can always fallback to brute forcing the correct ROM image manually if the new chip im getting in the mail doesn't work right. I also have an offer for someone for the whole display module in case there is something more seriously wrong with it.(Like that weird vector processor being damaged)

[james_s]

If you find yourself needing to write old bipolar PROMs there are articles from the 70s-early 80s to build manually operated units. The harder part these days though is finding the blank chips, since they were all OTP and haven't been made in years. They're used in a lot of 1970s arcade games.

[Berni]

Ah good point. I did see some simple footprint adapters that let you plug a normal EPROM into the socket of a ROM so that's probably the better way of going about it if i needed one programed in the future.

[JFJ]

... I did see some simple footprint adapters that let you plug a normal EPROM into the socket of a ROM...

Beware that bipolar PROMs (like the N82S147N) were often deployed because their maximum access time was, typically, an order of magnitude faster than that of the EPROMs of the day. So, to replace your ROM, you would need an EPROM (or EEPROM) with a maximum access time of 60ns.

[Berni]

So i got the new ROM chip. Hauled the heavy botacnhor out again, taken it apart and stuck it in...and... Winner winner chicken diner! It works perfectly and passes self tests!

The front panel cleaned up very nicely with some PCB cleaner (IPA+Ethanol) and just look at how nice it looks now. There are still some scuff marks around the front frame but that would need to be touched up with paint.

Thanks for the help guys!






Because the system disk floppy is very important (Contains the firmware that gets loaded in on boot) i made a copy of the original one and stuck it inside the unit for safe keeping, the space at the top of the CRT module is perfectly sized to hold two floppies. Unforcenetly i could not figure out how to read these floppies in a PC because they use a weird format. The way i got these copies is that the HP4145B has a option in the menu that lets you duplicate the system disk by inserting a blank disk.






Now that it works its time to try it out. Luckily i have my triax cable set that was made DIY for a HP 4141B.






Here it is all plugged in. Look at all that triax goodness. One triax is left unconnected because this unit does not have a actively driven ground output.






And here is the money shot. Set it up on the bipolar transistor preset, hooked it up to the first 3 SMU channels and it spits this out at the push of a button.
Mmmm... graphs. Also notice how the display gets brighter close to the graph zero point, the lines are individually vector drawn so the overlapping lines add up and become birther. Those vector fonts also look so lovely clean.

[Berni]

Also i have gotten myself a GQ4x4 USB universal programmer and used it to read out the other ROM. Hopefully i did it correctly. I also wanted to dump the main CPU ROM, but that thing is soldered down and getting that thing out of the multilayer PCB is going to be a nightmare as i don't have a desoldering iron, so i decided not to.

This is for the 8x2K ROM marked U12 on the A2 board.

[james_s]

Nicely done!

[Berni]

Also i found a quicker fix for the switches:




What i simply do is hold the switch down so that the spring bows out and then use a flathead screwdriver to push the bowing bulge in the spring towards the end so that it bends it even more and if you are violent enough keeps enough of the bend to make the button actuate nice and smoothly again.

[texaspyro]

I'd keep a couple more backup floppy sets outside of the machine.  No telling what heat, etc will do to the ones stored in the machine.   Also things like deflection coils and degaussing circuitry if the machine uses those.

[Berni]

Not to worry that part doesn't get all that hot and its a electrostatic deflection CRT so there are no electromagnets at all in it.

I would want to make a real digital copy of it, but i think i would need to buy one of these to do it: https://www.kryoflux.com/?page=home

[james_s]

There are no deflection coils or degaussing on an electrostatic CRT. Still not a bad idea to keep some disks elsewhere though. Floppy disks were never known to be super reliable and pretty much any you find are likely to be 20+ years old now.

edit:
Perhaps you could find someone who has one of those things and send them a copy of your disk to try reading?

[texaspyro]

The disk format may be HP's "LIF" format.  It is basically MSDOS with adjacent bytes swapped.  There are utilities that can copy / create LIF disks on a PC.  Google:  HP LIF format

[JohnPi]

I have an HP 4156A 'Precision Semiconductor Parameter Analyser' which works OK, but a) the CRT is dim, and b) its internal battery (?) doesn't seem to work, so all settings have to be entered (e.g. date, CRT brightness settings etc.) when it boots.

I am interested in replacing the CRT with an LCD, but I can't figure how to take the thing apart -- do you have any more photos or description of how you got yours open ? Perhaps my 4156A is similar in construction to the 4145B ?

[james_s]

Assuming it uses a electrostatic CRT with vector deflection like the 4145, replacing it with an LCD would require considerable engineering effort.

[Berni]

I did take a look around for LIF and yes HP did make a PC utility for copying these disks. Only problem is that it runs on Win 3.1 to Win 98. I'm not a retro computer enthusiast so i don't have one around. I do probably still have enough old PC parts to build one tho.

That HP 4156A is a nice bit of kit. From the looks of it the CRT in there is the same as my Agilent 89410A Vector Signal Analyzer. These are not vector displays anymore. Its just a classical raster scan color CRT like you would see in a TV. Im pretty sure my 89410A has a VGA port on the back and some other similar gear has RGB video on BNCs. That would hint to me that the CRT in there is actually a VGA computer monitor. I don't know the details about the CRT since the only thing that was wrong with mine was a dead RAM backup battery and that was easily sorted by undoing a few screws on the back, pulling the whole CPU board out as a single card, sticking a new battery in and slotting it back in.

The CRT you can maybe buy a broken donor unit to rip the CRT out since many of HPs gear share this CRT or you can probably use this LCD retrofit kit to replace it: http://www.simmconnlabs.com/1401/2543.html
Its not all that cheap tho. The NewScope-5 model can replace the CRT module that i had to repair in my unit but its quite expensive so it was more of a plan B for me if i was not able to fix my original CRT. No need for that plan anymore

As far as disassembly goes i would take a guess they reused the same giant SMU cards that they use in most of these semiconductor analyzers and that would mean similar internal construction. If it is the same then all you have to do is take off any corner feet on the back and undo a single screw in the middle of the top panel to get it to lift right off. Same for getting the bottom panel off. This design is pretty common on old HP gear and i really like it (Very quick and easy to get the lid off and on).

[james_s]

You could use Virtualbox to set up an old Win9x install which you can then boot on a modern PC host. I think you can install Win3.1 under DosBOX too, it's another virtual PC type thing aimed at retro gaming but it works to run other software too. I use it to run the antiquated software for the PC control adapter in my (radio) scanner.

[texaspyro]

I think I saw some LIF utils for Linux on that Google search...

[Berni]

I could give virtualization a try but the modern PCs around the house also don't have floppy disk controllers on the motherboard. So that means i have to use a USB floppy drive. This does mean i can use USB passtrough to connect it to the virtual machine, but it also means that i will need win 98 drivers for a USB chipset likely made in the era of NT. And if i do get the drivers working its still highly unlikely that it will actually work because USB storage tends to work more highlevel using sectors rather that direct register access (Since USB is not memory mapped like Firewire or PCIe). Seeing the strict requirement of the LIF tools to run on non NT systems likely comes from doing that direct register access (This broke soooo many win98 drivers from working on newer systems).

So i do still need a PC with an actual floppy disk port on it so that the LIF tool can talk directly to a floppy controller. VM hardware passtrough of non USB stuff gets a bit problematic under windows, so the host PC would likely have to be running Linux and use KVM to run the win 98 virtual machine so that i can get real passtrough to the floppy controller. Then again configuring all this could be a headache so might as well just install win 98 directly on that machine.(Or get one of those Linux LIF tools working, and im not much of a linux geek so my chances of getting that working might not be very high)

I have already tried various forms of low level sector by sector access to the USB floppy drive but i could never get a single byte out of it. And this is what a VM running on a windows NT system would have to also do. Even things like a LPT port is very difficult to emulate under a VM in this way. Floppy drives are also problematic due to how low level they are. A floppy drive is actually so dumb that it doesn't even understand the concept of a bit or a byte. They just spin the disk while spitting out the amplified and square waved signal from the read head, its the job of the floppy controller to turn the modulated signal into actual bits and bytes and hence why oddball floppy formats are such a pain.

[suhaimi861009]

Hi Berni. Can you share the data inside the floppy disc? I think my one corrupted.

[Berni]

Hi Berni. Can you share the data inside the floppy disc? I think my one corrupted.

So far i have no way of reading the system floppy disk in a PC. I need a old PC with a floppy controller or perhaps the cryoflux USB adapter that supports these weird floppy formats.

The only way i was able to make a backup copy for myself is using the actual HP 4145B to write one for me trough the "Make system disk" option in the menu.

[picburner]

If in the future someone needed the 4145B U6 eprom, I just posted it on the KO4BB website.
http://www.ko4bb.com/manuals/93.35.167.81/HP_4145B_Semiconductor_Parameter_Analyzer_EPROM-04145-85203.zip

[vival-ga]

Hi Berni

     old thread, sorry about that. BTW nice job you have done on the cable set and restoration of the unit.

     I just got one 4145B without cables and fixture, may I kindly ask you advice regarding the proper setup..i.e. triax connectors PL75-47 fits this job? or maybe better PL75-29?

Thanks
KR,VB

[Berni]

Thanks

You do need the 3 nub triax connectors ( PL75-29 ) for this one.

I do also have a 2 nub triax (Same shell as a regular BNC) and they do not fit. This is how i discovered about this problem. I do have a old Keithley electrometer that instead needs the 2 nub one.