Topics

Is my Tek 468 beyond repair?


Ray
 

hmmm...

Looked at the TI datasheet: fmax for the 74HCT112 is typically 60 Mhz. So it is worth trying.

I somehow got the price wrong by a few cents, at mouser it is actually 68 cents in single quantities and they still have 788 ea in stock.

Ray

On 09/21/2020 3:19 PM tgerbic <tgerbic@gmail.com> wrote:


On Mon, Sep 21, 2020 at 10:54 AM, ron roetzer wrote:


The 74HCT112 is only guaranteed for 24MHz clock. The 74S112 is rated for 80MHz minimum. The 74F112 is rated for 100MHz min.
Makes sense to look at the rated speeds. Tek picked the S type for a reason, and it is clock speed. If the oscillator is replaced, the original part numbers would work as designed, as they did for the last 30+ years. Pick F if you cannot find S. Digikey and Mouser have both types in stock. I did not see them at Allied, RS or Farnell.


There are only 4 actual pins, but their positions are mapped onto a 14 pin DIL package?
Yes oscillators often come in 8 or 14 pin footprints, so the pins are referred to by DIP numbering.

Regards
Tony




Brian Cockburn
 

Hi,

When I look at the Nexperia 74HCT112 data sheet I would guess that it will be fast enough. You may have to buy a few and hand select. But, and it's a big but, they look to only be available in surface mount. No DIP part. :-( Going for a 74S112 from a NOS vendor would be the best solution I reckon. And. just in case a reminder is needed. eBay and various Chinese vendors sometimes end up delivering re-marked parts (whether this is deliberate or inadvertent is a discussion for another day), so testing/validating the part on the bench is essential IMHO. You don't want to be installing a re-marked LS part that will fade out as the temperature goes up, if it even managed it at the start.

Cheers, Brian.


Harvey White
 

You may want to check the other variants of the "74" class, such as ACT, and ACHT.  Note that the T = TTL compatibility on the inputs, but TTL to CMOS (straight) does not.  2.4 volts minimum output is not 75% of VCC on the CMOS circuit, so logic 1 is iffy on straight TTL -> CMOS).  So for talking to CMOS (only) circuits, the HC, C, and AHC (etc) parts should be ok.  IF you're putting outputting to a TTL gate, or getting something from a TTL gate (most important) then you want HCT, ACT, etc., as long as the clock frequency is sufficient.

As for the surface mount stuff, IIRC the pinouts are identical, but the package size is not.  There are companies making surface mount to DIP adaptors that aren't too expensive if you can't design/make your own.  A single sided board is sufficient.  IF you design your own, and send it to a Chinese fab place, then duplicate the pattern and you're likely to have over a hundred for 5 dollars plus shipping.

Harvey

On 9/21/2020 8:02 PM, Brian Cockburn wrote:
Hi,

When I look at the Nexperia 74HCT112 data sheet I would guess that it will be fast enough. You may have to buy a few and hand select. But, and it's a big but, they look to only be available in surface mount. No DIP part. :-( Going for a 74S112 from a NOS vendor would be the best solution I reckon. And. just in case a reminder is needed. eBay and various Chinese vendors sometimes end up delivering re-marked parts (whether this is deliberate or inadvertent is a discussion for another day), so testing/validating the part on the bench is essential IMHO. You don't want to be installing a re-marked LS part that will fade out as the temperature goes up, if it even managed it at the start.

Cheers, Brian.





Roger Evans
 

Philips 74HCT112 in 16 pin DIL package are available from a UK seller on eBay at £2.99 plus 40p postage, not anyone I have bought from but I don't think I have had any dead or fake parts from UK sellers offering almost immediate delivery (2-3 days). Also in 16 pin DIL from France, six for 10 Euros, Texas Instruments, delivery in a couple of days so shouldn't be direct from China.

I also measured the output of a 50MHz can oscillator in 8 pin outline that I happened to have to hand. About 4.4V pk-pk with a 390R resistor either to ground or to Vcc. I thought the latter was a better match to sinking the TTL input currents.

Roger


John
 

Philips 74HCT112 in 16 pin DIL package are available from a UK seller on eBay at £2.99 plus 40p postage, not anyone I have bought from but I don't think I have had any dead or fake parts from UK sellers offering almost immediate delivery (2-3 days). Also in 16 pin DIL from France, six for 10 Euros, Texas Instruments, delivery in a couple of days so shouldn't be direct from China.
Roger, saw both of those. Decided to go for 74S112's as per original design. These were NOS and 10 pieces for a fiver and local to me.

Interesting that your crystal holds its voltage with a 390Ω load resistor across it. I have added a crystal to the shopping list. The total came to about £30 and included sockets, 74S112, 8085 processor, crystal and some caps. I know there are other faults but I hope it doesn't come to too much more. If this gets the computer running then all well and good. Of course, its not all about money but also the satisfaction of bringing a piece of vintage kit back to life, but I am not sure how much more will be worthwhile spending on it....

Parts will take a few days to arrive so in the meantime I will replace the caps that I can and remove the two 74S112s and clean the holes so the board is ready for the sockets and replacement ICs when they arrive.


satbeginner
 

"I hooked it back up to the board using jump leads, powered up the 5V rail but there was no output. I hooked up the sig gen again and got output from both 74S112s. "

Do you have the onboard 5V checked at the oscillator as well??
And the ground to the oscillator?

5V onboard does not mean it is 5V everywhere.

Just my 2cts

Leo


John
 

satbeginner, that is a good point, thank you, but yes, just to confirm 5V and GND were present at the oscillator and the 74S112 chips when I ran the tests.

Just another update on the crystal saga (the last one for now!), I ran the crystal test for about 10 mins yesterday but decided to do a further soak test this morning. Must have ran it for about an hour. Results initially were the same as yesterday: The top peak hit about 4.2V without load, about 3.5V when loaded down with a 470Ω resistor. However after an hour, the top peak was at just +2.6V. Not sure exactly how long it took to drop to that level, but wouldn't that be rather borderline for TTL?

In any case, replacement crystal ordered so will be able to compare when it arrives.


satbeginner
 

If you don't want to wait that long, you can temporarily add a NPN switching transistor as a simple pulse inverter.

Input via 1k to the base, 10k between base and emitter, emitter to ground.
Pull up resistor of 1k from +5V to collector, collector as output to the 112's

This way you ignore the lower output level of the oscillator.

Quick and dirty, I agree...

Leo


John
 

Its probably a bit academic since I have ordered replacement parts, but funnily enough I was just wondering whether a transistor (or JFET) could be used to buffer the output and reduce stress on the crystal. Just needs one that can comfortably handle the switching speed I guess. Perhaps worth a quick test....


Roger Evans
 

John,

A bipolar or JFET buffer is quite feasible but the collector (drain) load resistor has to be quite a small value. With the semiconductor turned off the resistor has to charge up 4 gate input capacitance plus stray capacitance in around 5nsec ( 1/4 period at 50MHz), 20pF total capacitance requires around 220R load resistance.

Roger


SCMenasian
 

A fairly robust way of buffering would use a fast PNP transistor and a fast NPN transistor with the bases connected together and the emitters connected together; PNP collector goes to the Vcc supply; NPN collector goes to ground. Input to the (connected) bases and output taken from the (connected) emitters.
Shoot-through is impossible and current drive (both positive and negative) can be quite large. Voltage gain is unity, reduced by the EB drops. Efficiency is quite high. As long as you don't need voltage gain, this should work quite well. BTW, this also works very well for driving power mosfets.

SCM


John
 

Replacement 5V crystal arrived today. Extremely fast dispatch but its output is only 1.2Vpp and nowhere near the expected TTL level!

I modeled satbegginers idea in LTSpice but couldn't get it to work properly. The output was barely above zero. Scm's idea works well enough though provided NPN collector goes to Vcc and PNP collector goes to ground! I used 2N2222 for the NPN (transition frequency=250MHz) and 2N2907 for the PNP (transition frequency=200MHz).


SCMenasian
 

You're right - my fingers were typing faster than my brain.

SCM


tgerbic
 

John,
What is the part number and manufacturer of the oscillator?

Under what condition did you measure the voltage? If open circuit, the oscillator is likely defective or expecting a pullup of some sort, or it could be a remarked marginal oscillator. If in circuit, it still could be a bad FF input. One other thing to check. If the replacement oscillator is not the same make/part number as the old one, that extra forth pin might actually be an output enable, instead of a No Connect pin. That is probably a long shot.

I just looked at some photos I took of my A19 board and see it has a Dale XO-33A oscillator on it. It is not what the manual/schematic shows and this part has different specs than the K1122A oscillator. Looking at the spec, it indicates it can sink 16ma of current to make a low and source .4ma for a 2.4V high. It might be worthwhile to repeat the oscillator output test with the load resistor connected to +5 rather than ground. The oscillator on my board appears to pull a high level down rather than drive a low level up to a TTL level. I assume that would mean the oscillator is actually pulling down the high level on the FF pins, not trying to drive the FF pins up to a high. I had suggested much earlier that you try pulling pin 8 of the oscillator up through a resistor to +5 to see if it improves the clock level. Did you try that? Might try a 1K pull up with the scope monitoring the pin and see if you get a clock with a better TTL signal.

Another long shot to try when you have the FFs and the oscillator out, ohm the trace to ground and +5 (power off) and see if there is any resistance. Should be an open circuit. Proceed to add parts and measure as you install them. Could also measure the resistance of the trace before you remove the old FFs to compare with later. Could be an internal short in one of the FFs, some problem between nearby traces or a cold solder joint on one of the FFs. Are the voltage levels all ok on all the pins of the FFs (with the oscillator removed and/or installed)?

However.... Something is not right here as I would say you (and us) are spending too much time analyzing and speculating on this one problem. If you have a good new oscillator and have good new FFs, the circuit should just be fixed and work as it did 30 years ago. I am sure there are hundreds of these scopes, with this design, still working today. Mine is. I would not spend time designing new circuitry unless you wanted to design a new circuit to try to improve Tek's design. You should not have to design anything unless the parts are defective and cannot be replaced. Just my opinion.

Regards
Tony


Roger Evans
 

John,

I had a quick look at the DIL packaged 50MHz oscillators available from RS Online and they are all marked TTL/CMOS compatible, from where did you get yours?

Please don't be insulted - you haven't switched the scope input to 50 ohms have you?

Roger


John
 

However.... Something is not right here as I would say you (and us) are spending too much time analyzing and speculating on this one problem. If you have a good new oscillator and have good new FFs, the circuit should just be fixed and work as it did 30 years ago.
Tony, your point is well taken and apologies for over-analyzing. It is mostly about learning and trying to understand on my part and a certain amount of confusion. However, I think I have now discovered why the crystal output was so variable. More below.

I had a quick look at the DIL packaged 50MHz oscillators available from RS Online and they are all marked TTL/CMOS compatible, from where did you get yours?
From an eBay seller.

Please don't be insulted - you haven't switched the scope input to 50 ohms have you?
Not at all. Checked just in case, but my Philips scope does warn and will not allow switching to 50Ω while a probe is connected to the channel. It was not set to 50Ω. I did realise that the FC cable was still connected up and plugged in but both scope and FC were using 1MΩ inputs. Connecting the FC does have some effect on amplitude but not quite that much.

Another long shot to try when you have the FFs and the oscillator out, ohm the trace to ground and +5 (power off) and see if there is any resistance. Should be an open circuit.
I am still hoping I can get the ICs out in one piece so as to be able to make a comparison between old and new and do a "post-mortem" as I agree with the consensus that there is likely a fault with one of them.

It might be worthwhile to repeat the oscillator output test with the load resistor connected to +5 rather than ground. The oscillator on my board appears to pull a high level down rather than drive a low level up to a TTL level. ... I had suggested much earlier that you try pulling pin 8 of the oscillator up through a resistor to +5 to see if it improves the clock level. Did you try that? Might try a 1K pull up with the scope monitoring the pin and see if you get a clock with a better TTL signal.
I did (rather embarrasingly) find a problem with the earth lead on the scope probe which caused a much reduced signal amplitude at times. I also found that when my hand was on the scope probe this also seems to reduce signal amplitude significantly. Both of these issues, while not explaining the abscence of signals on the processor (these were also checked with an LA), did contribute the variations I was seeing while testing the crystals and the resulting confusion. Having sorted that out, I re-ran the tests on both old and new crystals, first with the 1kΩ resistor then with 470Ω, 30 mins each. The new crystal has a slightly higher output (perhaps by 0.2V) without load. Both hold steady at just below 4V with a 1kΩ load resistor and about 3.8V with a 470kΩ load resistor. Neither gets warm to the touch. It seems reasonable to conclude that both crystals are good and the problem is with one of the ICs.

While repeating the tests I did also try pulling up to +5V as suggested rather than pulling down to GND with a 1kΩ resistor but that made no difference to the amplitude of the signal. It only changes the shape of the wave very slightly.

I am still waiting for the remaining parts to arrive at which point I will be able to replace the two ICs, finish re-capping (except for the large can), re-assemble and test the clock. In the meantime I have partially re-capped the board. I still need to tackle removing the two 74S112 ICs.


John
 

Well the good news is that some progress is being made. The board has been re-capped (except for the large can which I have been unable to remove) and the two 74S112 ICs have been replaced and mounted in sockets. Since it tested OK under a 470Ω load, the original crystal was replaced and the board returned back to the chassis and re-connected. The 5MHz clock is now stable and there is minimal ripple on the 5V line.

I have checked the wave-forms on the timer board and found that some are present, others are missing or incorrect.

For example, waveforms at 148, 149, 150, 151 and 157 (all directly associated with the clock) are all correct. A waveform is present at 147 but it does not match the diagram but is more like square wave. There is no waveform at 146/TP120 but the line is high. I couldn't reach some of the points on the board, for example 154 and 155, but nothing was happening at 156. Since be board is heavily interconnected with the CPU board, I am wondering whether this might be the issue.

I am still waiting for a replacement 8085. I checked and the 5MHz clock was present on pin 1 and 2.5MHz on pin 37 but still no ALE. Hopefully the replacement CPU will resolve that.


jrseattle
 

Since there was a request for this in the above discussion, I uploaded an image of the Tektronix 468 service rom to the files section. It's in a directory called "Tektronix 468 Service Rom"


John
 

Thank you for uploading this.


Michael W. Lynch
 

On Tue, Oct 6, 2020 at 01:17 AM, jrseattle2002 wrote:


Since there was a request for this in the above discussion, I uploaded an
image of the Tektronix 468 service rom to the files section. It's in a
directory called "Tektronix 468 Service Rom"
THANK YOU! I have been searching for this SERVICE ROM for some time. Now I have to find the chips, figure out how to load this file and use them. I sincerely appreciate this!

--
Michael Lynch
Dardanelle, AR