EEVblog Electronics Community Forum

Electronics => Repair => Topic started by: SaabFAN on May 05, 2016, 02:32:29 am

Title: Philips PM5190 Function Generator - Repair and Upgrade
Post by: SaabFAN on May 05, 2016, 02:32:29 am
I just got a PM5190 Function Generator made by Philips back in the late 70s, early 80s.

From the description of the seller and what I saw when I turned the device on I guess that the digital part is faulty.
So before even measuring anything, I took out the EPROM and stuck it into my TL866CS.
Each time I read the chip, I get different results, so it either lost its content, or was damaged by what happened right before it stopped working.
Voltages on the board are OK:
5V Analog = 5.01V
5V Digital = 4.91V (a bit low, but cannot be adjusted, and considering the power-requirements for TTL-Chips, I'm guessing this to be normal - The Service-Manual says that Voltages down to 4.75V are ok)
+20V = 20.01V
-20V = -19.99V
I haven't measued the noise on the rails yet.
The 5V Digital-Rail is not regulated by the way. It comes directly from the transformer, gets rectified by a bridge-rectifier and is then distributed among the chips on the digital board.

The seller described what happened to the device like this:
It was connected to some device, I'm guessing a tube-based amplifier because he owns the website Mikes Röhrenbude, and someone pulled an anode-cap. This caused quite a spark and after that the generator didn't work anymore.
The primary symptom of the problem is the digits going crazy once any key is depressed.

Other problems I found with the device: The Mains-Filter has vomited its resin and needs to be replaced (it also starts to sizzle after being connected to mains for a while), and the input-keys have to be cleaned.

Does anyone here own a similar instrument and can upload the EPROM-Image? So far my google-searches have come up empty.
If someone still has the Test-ROM available, that would probably also help quite a bit.
So far, I haven't seen any unobtanium-component inside the device. All the ICs are available on ebay, although some at ridiculous prices, or are waiting in my parts-box. Some are socketed, but the majority is soldered in, so I'm hoping that I don't have to replace too many of these ICs.

The entire device is built around a Intel 8035 Processor and uses direct digital synthesis to produce a sine-wave from which the other 2 functions are derived: Square and Triangle.
Title: Re: Sick Philips PM5190 Function Generator
Post by: picburner on May 05, 2016, 04:47:34 am
Here is a copy of both EPROM and PROM.
Even in my PM5190 the filter has gone in the same way.
I had to redo all the welds of the final stage because cooked by heat.
Title: Re: Sick Philips PM5190 Function Generator
Post by: SaabFAN on May 05, 2016, 06:28:54 pm
Thx for the images.
The EPROM ist pretty much down for the count. I need 6,25V VCC to get a positive verify.
That the TL866 only has 21V Program-Voltage doesn't help either (the chip requires 25V!). So there's a new chip incoming - 27C16. I hope that it'll be able to drive the data-bus. I can't quite remember which one has the larger fanout: CMOS or TTL.^^

In the meantime I'm going to verify that the rest of the digital board is working. For that purpose, I'm going to put a supercap on a 16k SRAM and build an adaptor for the programmer, because I have to tie the WE-Pin to GND.
If there's a problem in the rest of the digital parts, I think I'll get a 8749, as I can get it for quite a bit less than a new EPROM, Dual 4bit Latch (HEF4508), and the CPU. Enough to justify the additional work required to build a programmer for the chip. :)

How would you recommend I proceed in checking the digital circuits? Just apply 5V with a lab Power-Supply (can't use the internal one, because of the broken mains-filter), or first check the ESD-Diodes (if they already have them in there - With one or two exceptioins, there's only LowPower-Schottky or standard TTL 74-Chips in there.
And I would like to avoid desoldering them all to check them :)
Title: Re: Sick Philips PM5190 Function Generator
Post by: Kleinstein on May 05, 2016, 08:16:41 pm
You could give it a try with 5 V, or the raw voltage (e.g. about 9 V) before the regulator from an external supply.

Parts to check would be caps in the supply for high ESR  and short for tantalum ones.

From the description on how the damage occurred, one could also suspect a damage in the output amplifier.

I once repaired such an old FGEN: it had a burnt transistor in the DAC circuit. AFAIR its a 9 Bit DAC using discrete transistors that get quite hot. 
Title: Re: Sick Philips PM5190 Function Generator
Post by: SaabFAN on May 06, 2016, 03:50:52 pm
For the time being, I have removed the Line-Filter to continue testing.

This revealed up 1,24V Ripple at about 3,3Mhz on some data-lines and horrible looking waveforms. The Chips are all Lowpower-Schottky with a few HEF-Chips dropped into the mix.
But even 74LS chips should produce cleaner Square-Waves at 5Mhz.
Another thing I discovered: The keyboard/display-controller (Intel P8279) introduces MASSIVE Ripple (1,36Vpp) on the !RD and !WR-Lines, as well as the Interrupt-Line. The ripple-voltage gets a little bit lower when I remove the CPU, but not much.

Next obvious step: Replace all the capacitors and see if that helps.

Btw. The Supercap-Buffered SRAM, built according to the circuit-diagrams on this website: https://hackaday.io/project/8109/gallery#22dc46aba9a83782f50ab59f1306e4d0 (https://hackaday.io/project/8109/gallery#22dc46aba9a83782f50ab59f1306e4d0) works pretty well.
Title: Re: Sick Philips PM5190 Function Generator
Post by: SaabFAN on May 06, 2016, 06:33:46 pm
Here are some Screenshots from the Scope

I've probed pins 1 to 20 with the device running "normally" directly after startup.
Pin 8 is the RD-Pin, which tells the connected P8279 that a data-read is requested. Pin 10 is WR, telling the 8279 that there's data incoming.
Upon startup I can see a short pulse on WR, after which the front display starts to flash the first digit - All digits display 0 at this time (see this Video: https://youtu.be/Q0UMYOMQdPU (https://youtu.be/Q0UMYOMQdPU) ). Once a key is pressed, the Square-Wave LED lights up fully, while the Sine-Wave LED lights up at about half power and the triangle-wave LED glows very dim. At the same time, the Digits go crazy. From that point forward, nothing else happens.
This happens with every keystroke and is only stopped by power-cycling the entire device.
Interestingly, once this failure-state is reached, the Interrupt-Line is constantly low and there's a bit more ripple-voltage on the supply-rails.

First Screenie is Pin 8 - RD
Second one is Pin 10 - WR
Third one is Pin 11 - Address Latch Enable - As you can see, the amplitude is not constant, which I find very odd
Fourth one is Pin 14 - One of the Data-Lines of the Bi-Directional bus. It more or less looks the same on the other data-lines, except for pin 15, where I could see no signal (measurement-error or there's really no signal in that mode, idk.)

Edit:
Little note to the previous post: The SRAM-Circuit I built uses a Zener-Diode instead of the 2 white LEDs to turn on the Transistor when the Programmer increases the VPP-Voltage from 5V to the programming-Voltage.
Title: Re: Sick Philips PM5190 Function Generator
Post by: SaabFAN on May 09, 2016, 12:25:49 am
Does anyone have the TEST-ROM for this device?

So far I have confirmed that the CPU seems to be working. It sets up the Display and then waits for any input. Once the Input arrives, signaled by !INT going LOW, the CPU sends several Write-Commands to the P8279 Display-/Keyboard-Controller. These are followed by Read-Commands, but the Interrupt-Line never goes back to HIGH again. Instead, the CPU  seems to be sending a specific set of instructions over and over again.
Without a Logic-Analyzer that has more than 8Channels, I'm not going to see what's the problem, however. And this month's budget for T&M-Toys is already spent.
So the Test-ROM would be pretty helpful. It would show if there's any problem with the CPU itself (RAM-Issues), or the other components without me needing a logic-analyzer or replacing chips blindly :)

I also measured the current-consumption of the mainboard + display-board: The take about 650mA. Quite a lot and would explain the noisy voltages despite the 4 caps I desoldered measuring OK in terms of capacity and ESR.
Also the 5 47µF Tantalum-Caps I added (I put them in basically everything, because I have about 1000 of them in my parts-box^^) almost didn't make a difference and the Manual also states that the Display-Unit causes quite a bit of interference.
Title: Re: Sick Philips PM5190 Function Generator
Post by: SaabFAN on May 13, 2016, 09:39:36 am
Well, it seems that the CPU was damaged.
Once I replaced it with a "fresh" MAB8035HL6PB (Philips variant of the Intel 8035 µController) it started working again. I was able to set the frequency, waveform, amplitude, dc-offset, etc.

But the DC-Offset is way off and depending on the amplitude: Higher Amplitude = higher negative DC-Offset. And I cannot bring the DC-Offset into the positive area. Probably because the DAC (an AD7530J) is also a MOS-Chip and, like the CPU, hasn't survived whatever happened to the device when that anode-cap was pulled.
There's also quite a bit of distortion if I increase the amplitude above 12 to 15V (depending on the DC-Offset) and the scope is also picking up tiny noise-spikes.
So there's probably quite a bit that has to be done to the top-board, which has quite a bit of discoloration around the areas that become really hot (too hot to touch after just one minute).

Attached is a merge of the pages of the Digital board schematic. Can be printed on a A3-Page.
Edit: Added the schematic of the analog board.
Title: Re: Sick Philips PM5190 Function Generator
Post by: SaabFAN on May 20, 2016, 09:18:46 am
I replaced the DAC, the last digital component to be faulty, and some capacitors - some measured only 50% capacity and ESR higher than 1Ohm, but I still cannot bring the amplitude into the positive range.

I have started to touch up the welds on the analog board. Heat every one up with the soldering-iron and put a tiny spot of solder on it - Didn't help though.
Interestingly, only the negative side of the output amps gets hot. The ones for anything moving into the +V-Range gets warm, but can still be touched.
Title: Re: Sick Philips PM5190 Function Generator
Post by: SaabFAN on May 21, 2016, 12:09:34 am
It is working now!
I traced the problem to the Buffer-Amplifier and measuring all the resistances, I discovered that the TrimPot for the DC-Balance (Component 750 on the analog board) had a resistance of about 13,6kOhms where it should have a resistance of 100Ohms. Broken Carbon-Track -.-
To check if that is the source of the problem, I soldered a 100Ohm-Resistor across the 2 terminals and was able to trim out any remaining DC-Offset.
Interestingly, the Trim-Pot R750 isn't mentioned in the adjustment-procedure.

During Adjustment, the device passed all the tests described there (including noise-test - The noise of the Signal-Gen vanishes in the noise-background of both the Rigol and the PM3320A I have here) and I think I will leave it like this.
The 2 big caps next to the Voltage-Regulators are 1000µF caps, which are replacing 22µF axial capacitors. The 2 caps in the front are also 1000µF, which are replacing 470µF decoupling caps, which also made the square-wave a bit sharper - I don't have before/after screenshots of the scope though.
Noise across the board is rather low, but I think a large ground-plane would have improved the boards quite a bit. I guess in the 70s people weren't too concerned about signal-integrity or efficiency: There are some resistors in this thing that heat up to 120°C!
Title: Re: Sick Philips PM5190 Function Generator
Post by: Vgkid on May 21, 2016, 03:51:35 am
Nice job on the repair. I was always leery of that type of trim pot.
Title: Re: Sick Philips PM5190 Function Generator
Post by: Werecow on July 26, 2016, 03:05:48 am
It is working now!
I traced the problem to the Buffer-Amplifier and measuring all the resistances, I discovered that the TrimPot for the DC-Balance (Component 750 on the analog board) had a resistance of about 13,6kOhms where it should have a resistance of 100Ohms. Broken Carbon-Track -.-
To check if that is the source of the problem, I soldered a 100Ohm-Resistor across the 2 terminals and was able to trim out any remaining DC-Offset.
Interestingly, the Trim-Pot R750 isn't mentioned in the adjustment-procedure.

Greetings! I'm working on a PM5190 that has exactly the same problem.  The carbon on R750 actually appears burned at both ends.  It measured ~2k? in-circuit, then went completely open when removed from the board. :-(  Replacing with a 1/2W multiturn seems like a good idea, and should prevent the problem from recurring.
Title: Re: Sick Philips PM5190 Function Generator
Post by: The_Penguin on July 26, 2016, 03:19:10 am
It is working now!
I traced the problem to the Buffer-Amplifier and measuring all the resistances, I discovered that the TrimPot for the DC-Balance (Component 750 on the analog board)

Congrats on a successful repair! Thanks for sharing.
I enjoy reading these, since I haven't done any myself lately.
Title: Re: Sick Philips PM5190 Function Generator + Upgrades
Post by: SaabFAN on November 14, 2016, 08:48:35 pm
I have now replaced the potentiometer I earlier bridged with a 100Ohm Resistor.

The performance hasn't changed much, but I don't need to compensate the DC-Balance in the later stages anymore.
Which made me think: Why not improve this thing further?
What I like to improve most is the square wave TTL-Output. The Square Wave of the main Output cannot be improved because of limitations of the amplifier-stages, but the TTL-Output comes directly from a 74LS37 Buffered NAND-Gate.
So I was thinking about replacing the 74LS37 with a 74AC00 NAND-Gate that can produce sharper edges, which makes characterising Filters at higher frequencies possible (Square Wave into the Filter, Output of the Filter into the scope and then use the FFT-Function to see which frequencies have the highest spikes on the screen).

In theory, it should work just fine. The AC-Series has fast rise and fall-times, can source and sink up to 50mA (74LS37 can source 1,2mA and sink up to 24mA - The F-Version has 12mA Source and 64mA sink-capability with rise- and falltimes at about 2,5ns) per Output and has shorter propagation delays and risetimes than the LS-Series (between 1,9 and 6,6ns vs. 12 to 24ns Propagation delay).
Title: Re: Philips PM5190 Function Generator - Repair and Upgrade
Post by: Kleinstein on November 14, 2016, 09:17:59 pm
One could give it a try. Depending on the output level from the comparator one might want to use 74ACT00 instead, or add an pull-up to shift the level a little.
Have an eye on decoupling too.
Title: Re: Philips PM5190 Function Generator - Repair and Upgrade
Post by: SaabFAN on November 21, 2016, 05:06:54 pm
Chips are ordered. Once they are here, I'll fit a socket and then try out the different chips and record the results.
Decoupling shouldn't be a problem. There's plenty of capacitance available and noise on the supply-rails is just barely above the noisefloor of my scopes.

I also have a Centronics Printer Cable on the way, as well as a 24pin Centronics plug, so I can build my own IEEE-488 to IEC-625 Cable - Currently inputs are limited to what I can type in on the front panel, which isn't that much. Programming via GPIB is more or less a must with this device.
I'm also investigating the possibility to use some microcontroller with WLAN-Support to act as a controller for the signalgenerator. There is code out there that turns an Arduino Nano into a GPIB-to-USB-Adaptor. Maybe I can port that code to the ESP8266 if I have the time. :)
Title: Re: Philips PM5190 Function Generator - Repair and Upgrade
Post by: SaabFAN on December 13, 2016, 06:43:50 pm
Now that I finally have a Spectrum Analyzer, I made some measurements of the TTL-Output.
In vanilla configuration, the TTL-Output produces spikes all the way up to 170MHz that posess sufficient strength up to 150MHz to be useful.
I doubt that the AC-Chips can improve this figure given the construction of the board with no ground plane.

Also I'm constantly running into the limits of this device, so I'm selling it on ebay: eBay auction: #252680607480 to get some money to buy an Signal-Generator with arbitrary waveform-ability.
Title: Re: Philips PM5190 Function Generator - Repair and Upgrade
Post by: Mathima on December 16, 2016, 04:36:25 am
Recently I helped fix a PM5190. The sinewave had all kind of spikes on it. Turned out the SINE-ROM had some flipped bits.
I read the contents of the PROM and reverse-engineered it. I figuered out it has just a quarter of a sine wave programmed in to it, as I suspected.
I put everything in Excel and recalculated every wrong value and verified it to the rom i found here on the forum (which is read as signed so it starts at 80 (hex) instead of 0 and drops back to 80 (hex) at adress 128. (check settings before you program that one in a PROM)
After that I corrected the ROM file and tried reprogramming the original PROM to fix the flipped bits. This did not work unfortunatly because the programmer wouldn't allow it.
I tried ordering a new N82S114 PROM but they are next to imposible to get, so I opted for its bigger brother, the N82S115 with 4K memory instead of 2K.
I translated the ROM file to account for the extra adress line A5 in the N82S115 and programmed it. Because the 115 has 9 adress lines instead of 8, one adress line will be floating as its not connected on the pcb. So you need to connect pin 3(A5) to pin 12 (GND) to keep A5 low. In the N82S114 this pin is not connected in the chip so it doesn't matter if it's floating. Both N82S114 and N82S115 are pin compatible with the exception of the extra adress line.

I've included the correct ROM files for both the N82S114 and the N82S115. Be warned, the ROM file that is already available here is probably read as signed and it should have been read as unsigned.

I hope this helps people with the same issue.