Evaluating old CRO probes

[jdutky]

I have two old Tek oscilloscopes that belonged to my father, one 475 that dates to at least 1975, and a 2213 from around 1982, both with the original passive voltage probes that he used for work (two P6075A probes for the 475, and two P6120 probes for the 2213, all x10 probes, non-switchable). One of the P6075A was marked "BAD" by my father some time prior to 1995, but, though he kept detailed notes, I have no other indication of what malfunction he observed. I have tried to evaluate the probes myself, but while I am not a rank amateur (I have a basic, college level understanding of electronics, understand Ohm's and Kerchhoff's laws, know how to handle a soldering iron, multi-meter, and even a bit about using an oscilloscope) I am barely a journeyman. I am likely to use the wrong terminology for some things, but I hope that my meaning is clear.

My question is what should I be measuring about the probes to determine whether they are "good" or not? I have used a fairly nice multi-meter to measure the resistance and capacitance from several points: the probe tip ("tip"), the grounding clip ("clip") (with the GND REF button pressed and not), the BNC center pin ("center"), and the BNC ring ("ring"). I have measured continuity from tip to pin, from tip to clip, and clip to ring. I have measured resistance from tip to center, and tip to ring (with GND REF button pressed). And I have measured the capacitance from tip to clip. I have also compared my measurements against a newly purchased third party probe (cheap P2200 off Amazon), and a refurbished Tek probe (cheap P6075A off eBay).

I have found that the contact between the ground clip and the probe shell/BNC ring has been unreliable on these old probes, but the probe that my father marked "BAD" seemed to be revived after I removed the ground wire and cleaned the spring clip that contacts the outer housing of the probe. Are there other measurements that I should be doing on a scope probe to determine if it is working?

Thanks in advance

[Vovk_Z]

I can see at least two possible reasons:
1) There is some rupture in the wire, possibly intermittent.
2) Some other problem which leads to a bad amplitude-frequency characteristics.
If first - then it possibly is easier to repair. If second - you need a good enough functional generator and oscilloscope to check. I'm not sure if it is easy to repair (osc probes are a bit of art).
P6120 is a 60 MHz probe (?) - it is not too wide range, so you may just buy another.
P6075 is 200 MHz probe - so it hurts to loose it, and there is a sense to try to repair.

[David Hess]

What matters on an oscilloscope probe is transient response which is tested using a fast pulse generator with the probe connected to the 50 ohm termination so with a 25 ohm source.  For this test to work, the signal must be connected to the probe tip coaxially without the ground lead; at 200 MHz, a BNC to probe tip adapter can be used.

[jdutky]

So, when you say "fast pulse generator" what are we talking about? I have an inexpensive 0-3 MHz function generator that can make pretty ugly pulses that are about 200 ns wide at 3 MHz, but I'm guessing I'll need something that goes up to at least 100 MHz with a nice sharp rise and fall with a width of maybe 10-20 ns.

[David Hess]

So, when you say "fast pulse generator" what are we talking about? I have an inexpensive 0-3 MHz function generator that can make pretty ugly pulses that are about 200 ns wide at 3 MHz, but I'm guessing I'll needs something that goes up to at least 100 MHz with a nice sharp rise and fall with a width of maybe 10-20 ns.

You need a nice sharp rise *or* fall with a transition time significantly faster than the bandwidth to be tested (rise or fall = 0.35/BW) and a minimum of pulse aberration.  Tektronix recommended faster than 600 picoseconds for testing up to 100 MHz so figure 300 picoseconds for 200  MHz.  These types of instruments are sometimes called reference level pulse generators.  I use a Tektronix PG506 but it is marginal for 200 MHz.

AC or LVC logic can achieve this directly with some care in construction.  If you just want to buy something, then Leo's pulse generator is an excellent option.

A BNC to probe tip adapter should also be used.  Some probes come with them.

[jdutky]

wow, that's a couple orders of magnitude faster than I was thinking of. There's no way I'll be able to coax that kind of performance out of my cheap, old BK 4001.

I have an old TM503 that I got for a song at the local university surplus outlet, and I've been looking at 500-series modules, including some of the PG50x modules, but everything I've seen on eBay (and listed as working) has been asking for real money. If I'm going to drop $500 on something it is going to be a modern digital scope.

That Leo USB pulse generator looks like a nice piece of kit, though, and it's at least in petty cash range (if tending toward the high end). I was thinking about building my own, but that's pushing the limits of my current skill set. Of course, if you don't push your limits then how can you expect to learn anything?

(also, I was wondering what those BNC dodads were. I got a couple of them with the cheap probes I bought, just in case the old P6075As were permanently dead. Thanks for clearing that up without requiring me to make any further displays of the depth of my ignorance)

[David Hess]

wow, that's a couple orders of magnitude faster than I was thinking of. There's no way I'll be able to coax that kind of performance out of my cheap, old BK 4001.

Before I got the PG506, I modified the sync output on one of my function generators by replacing the original 74LS140 dual 4-input NAND line driver with the best performing replacement that I could find which was the 74S140, but it was still too slow for 100 MHz testing.  The internal layout also prevented a clean edge so I abandoned further modifications.

That Leo USB pulse generator looks like a nice piece of kit, though, and it's at least in petty cash range (if tending toward the high end). I was thinking about building my own, but that's pushing the limits of my current skill set. Of course, if you don't push your limits then how can you expect to learn anything?

I just remembered another source for fast and high quality edges.  The horizontal and vertical sync outputs on a VGA port are both fast and clean, if you can get the probe connected with a minimum of ground lead length.

[srb1954]

I have an old TM503 that I got for a song at the local university surplus outlet, and I've been looking at 500-series modules, including some of the PG50x modules, but everything I've seen on eBay (and listed as working) has been asking for real money. If I'm going to drop $500 on something it is going to be a modern digital scope.

When looking for TM500 series modules also look out for the SG503 leveled output signal generator. This can be used to verify the bandwidth of the scope and probe or check for gain vs frequency variations that might indicate problems with probe compensation. The SG503 covers up to 250MHz so it is well suited for verification of the 2213 performance although it would be barely sufficient for checking out the ultimate performance of the 475.

There is a higher frequency version SG504 that covers up to 1GHz but this requires a separate external leveling head which is often hard to come by.

[jdutky]

While I do have an immediate application for these two scopes, their main value is sentimental. I had considered getting he equipment to properly calibrate them, but the price, while lower than the sentimental value, far exceeds the price of a brand new digital oscilloscope (the Siglent 1204X-E is only $741 on Amazon!).

I had also considered using micro-controller to generate the high speed pulses, but the idea of using a video signal is even better (though I'm not sure which is easier at this point: I have a couple Raspberry Pis lying around which I know work, but getting a known functional computer with a VGA port is going to be iffy). Sadly, it doesn't look like HDMI has H and V sync signals (or, if it does, they are encoded in the differential data signals).

[David Hess]

Here is an example of the type of circuit that you could build using AC or LVC logic.  As the video discusses, these types of pulse generators have other uses like time domain reflectometry.  I often use mine to do quick tests of attenuators, coaxial cables, and oscilloscope probes; a VNA is even better for this sort of thing but I have fast oscilloscopes and no VNA.

https://youtu.be/9cP6w2odGUc

The cleanliness of the pulse will depend on how well the circuit is constructed.  Dead bug or Manhattan style construction over a ground plain is required and not particularly difficult.

The photograph below shows how the fast rise and fall circuits of the PG506 are constructed.  Note that the BNCs have build in 50 ohm termination resistors in the shape of a disc.  It achieves 600 picoseconds or better with an exceptionally clean pulse.

[jdutky]

So if I had a fast rise time circuit (similar to the one in the PG506) then I could drive it with my cheap signal generator and get a nice, crisp pulse to test my scope and probes with? That seems like it might be in my reach.

I had a look at the schematics for my cheap signal generator, because I had already had to replace the main filter caps on the power supply (the 17 year old signal generator, a BK Precision 4001, that I bought for $100 in 2003 was dead from dried out electrolytic capacitors, but the 45 year old oscilloscope, which likely cost a couple thousand dollars when it was new, is still chugging along. I guess that really does show that you get what you pay for), and I was wondering if the sloppy high frequency square wave might be the result of other dried out caps elsewhere in the circuit. I focused on the MHz path, ignoring the common elements, because only the MHz path was showing real problems (probably not a good assumption), and did find that the capacitor used to select the MHz bandwidth regime was well below spec (should have been 220 pF but measured as only 135 pF). I replaced the dead 220 pF cap with a ceramic disc capacitor of the same value (and higher voltage spec), but the slow rise time persists. Looking at the schematic I see that there are a bunch of LM4588 op amps in the signal path (common to all bandwidth regimes) and from the datasheet (looking at the TI CR4558 datasheet, which is all the comes up in a Google search for LM4588, and looks like a similar component) it looks like this op amp tops out around 3 MHz (which is the stated upper bandwidth of the signal generator), so I'm guessing that the sloppy MHz peaks are a result of an inherent bandwidth limit in the signal generation circuit.

Like you, I considered replacing the op amps (and assorted discrete transistors) with similar components that had higher specs (I assume that I will need parts that can handle up to 15-20 MHz in order to get a clean square wave at 3 MHz), but I abandoned that idea when you told me about the Leo Bodnar pulse generator (and as it became clear that I would have to replace a significant portion of the signal generator in order to get a much higher bandwidth out of it). If, however, I can just build the fast rise time circuit, and drive it from the signal generator's sync output, then I'd be perfectly happy to do that, if only as a learning exercise. I'm looking at the PG506 schematics now in order to get some idea of how to build a fast rise circuit.

[jdutky]

After studying the PG503 schematics, and reading a couple of articles about building fast rise time circuits, I'm wondering if I can just build a circuit with two op amps/comparators: one very vast comparator to generate the fast rise time output signal, and another (normal speed, e.g. 3MHz) comparator to condition the input signal into a voltage range suitable for the output fast comparator.

I've see that there is a device from Analog Devices, the ADCMP580, that can deliver a 37 ps rise time on its output, and I can use (I think) just about any common op amp for the input signal conditioner.

With the fast comparator costing $20 each this isn't going to be particularly cheap, but it does give me an excuse to play around with op amps.

I also looked at a range of fast transistors, looking for anything with very fast output rise times, and didn't find anything that was even nearly as fast as the ADCMP580.

(I got the AD part from this article: http://www.starlino.com/build-a-really-fast-pulse-generator-50ps-rise-time-using-an-ultra-fast-sige-comparator.html)

(NM: I just found the EEVBlog discussion of the Jim Williams pulse generator circuit. All kinds of good information there)

[David Hess]

So if I had a fast rise time circuit (similar to the one in the PG506) then I could drive it with my cheap signal generator and get a nice, crisp pulse to test my scope and probes with? That seems like it might be in my reach.

That is exactly right.

After studying the PG503 schematics, and reading a couple of articles about building fast rise time circuits, I'm wondering if I can just build a circuit with two op amps/comparators: one very vast comparator to generate the fast rise time output signal, and another (normal speed, e.g. 3MHz) comparator to condition the input signal into a voltage range suitable for the output fast comparator.

Take a look at how the reference level pulse generators were designed before the PG506, specifically the Tektronix type 106:

http://w140.com/tekwiki/wiki/106
http://w140.com/tekwiki/wiki/File:106_sidiode.png

The transistor which does the switching is driven from its emitter while the base is grounded, which may be called emitter switching.  The other transistor is an adjustable current source which sets the variable output level.

The switched emitter could be driven through a resistor from the output of a fast CMOS logic gate; I have thought about using a 74LVC125 3-state buffer which is characterized for 3.3 volts but can operate at 5.0 volts.

I also looked at a range of fast transistors, looking for anything with very fast output rise times, and didn't find anything that was even nearly as fast as the ADCMP580.

Speed is limited by construction and packaging.  The reason that the ADCMP580 can be so fast is that it implements an ECL or CML output internally with an internal termination so it operates as if it is at one end of a transmission line.  Incidentally by my standards, the ADCMP580 output is not particularly clean, but it is better than average.  At higher frequencies it displays too much overshoot to be a good source.

If we could get dual emitter dual collector transistors like in the old days, then we could get much faster speeds out of a simple discrete implementation.

[jdutky]

The Leo Bodnar fast pulse generator arrived (sooner than the web site led me to expect) and I hooked it up straight away. The bad news is that the probe that my father had marked as "bad" is indeed bad: there is some kind of intermittent fault about midway down the coax cable; I suspect the center conductor is broken. The good news is that the other probes, and the scope itself work wonderfully, and I was able to see a nice crisp square wave with a 3 ns rise time (which, I think, is what can be expected of the 475. If my math is correct that translates to a bandwidth of around 230 MHz). I still have some parts on order to try building my own Williams pulse generator, but that's just a learning exercise now.

I have to say that I am still in awe of how these old scopes hold up. I'm sure that some of this specific scope's longevity is the result of careful treatment and maintenance by my father, but most of it must be a testament to the excellent work of Tektronix engineers. Just amazing.

[Gyro]

I still have some parts on order to try building my own Williams pulse generator, but that's just a learning exercise now.

You can have no end of fun with that one, especially if you start air wiring it onto a piece of copperclad or directly on the back of a BNC socket. It's very instructive trying different resistor and capacitor constructions, lengths of different flavours of unterminated coax etc as the discharge cap.

Leo's pulse generator is excellent but its rise time is a bit... err predictable   It will be a good exercise to see how closely you can approach its performance (within the limits of what is visible).

[David Hess]

The good news is that the other probes, and the scope itself work wonderfully, and I was able to see a nice crisp square wave with a 3 ns rise time (which, I think, is what can be expected of the 475. If my math is correct that translates to a bandwidth of around 230 MHz). I still have some parts on order to try building my own Williams pulse generator, but that's just a learning exercise now.

Rise and fall time for a 200 MHz 475 should be 1.75 nanoseconds; 3 nanoseconds is way off and indicates a problem.  Double check that the bandwidth limit switch is working properly.

[jdutky]

The bandwidth limit switches (HF Rej and LF Rej positions on the trigger coupling switch) seem to be working correctly. I'm definitely seeing about a 3-4ns rise time on both channels. It's certainly likely that there is something wrong with this scope, as old as it is. In fact I have noticed a specific cross-talk issue between the two channels when either channel is receiving a square wave and is set below 0.5 volts per division; the other channel shows "noise" that is correlated to the rising and falling edges of the square wave on the other channel. The cross-talk is happening after the attenuation circuits because it doesn't change when you change the attenuation on the channel showing the noise. Are you sure that the rise time on the 475 should be 1.75 ns? That would imply a bandwidth upwards of 400MHz! I know that Tek is known for estimating their products bandwidths conservatively, but that's outrageous (and even if the scope can do that, the probes are only rated for 200 MHz, so the scope/probe system shouldn't show that kind of performance).

UPDATE/MEA CULPA: I was doing stupid probe tricks that significantly reduced the quality of the signal. Now I have found my stash of BNC adapters and am going directly into the fast pulse generator with a F-F adapter and the BNC probe tip adapter: the square wave is significantly cleaner, and the rise time is down below 2ns. Wow. So the upper limit on this scope really is something north of twice what it was spec'd at by Tektronix. Wow.

[David Hess]

Are you sure that the rise time on the 475 should be 1.75 ns? That would imply a bandwidth upwards of 400MHz! I know that Tek is known for estimating their products bandwidths conservatively, but that's outrageous (and even if the scope can do that, the probes are only rated for 200 MHz, so the scope/probe system shouldn't show that kind of performance).

Tektronix gives the rise time in the specifications as 1.75 nanoseconds, which is consistent with a 200 MHz bandwidth.  100 MHz is 3.5 nanoseconds.

UPDATE/MEA CULPA: I was doing stupid probe tricks that significantly reduced the quality of the signal. Now I have found my stash of BNC adapters and am going directly into the fast pulse generator with a F-F adapter and the BNC probe tip adapter: the square wave is significantly cleaner, and the rise time is down below 2ns. Wow. So the upper limit on this scope really is something north of twice what it was spec'd at by Tektronix. Wow.

Yea, a BNC to probe tip adapter is required to get consistent results.  Probes are tested in exactly that way and the probe's bandwidth specifications assume a directly connected source with 25 ohm impedance which comes from the termination.  Directly connecting the probe to the pulse generator without a termination makes it 50 ohms but that is close enough to get an idea of what is happening.

Tektronix specifies bandwidth of the combined oscilloscope and probe at the probe tip, assuming that a suitable probe is being used.

[Gyro]

Just a thought, you are using the graticule correctly to measure the rise time? That is, using the V/div fine Var adjustment to set the low and high levels to the 0% and 100% graticule lines and then reading off the time between the trace crossing the 10% and 90% lines.

As I say just a though if you're unfamiliar with analogue scopes.

[jdutky]

I've actually only used analog scopes, but it's been almost 30 years since I used one regularly (in college, when I still planned to major in EE). It's not quite like riding a bicycle.

[jdutky]

So here are the results of my tests of two probes: one of the P6075A that I got off eBay, and the other a cheap 1x/10x P2200 from Amazon. The first picture shows both probes connected to the internal calibration signal after I have adjusted them, the P2200 is on 10x but has no sense pin so the scope is indicating 1x. The calibrator or the time base may be a little off, but I've checked it against my BK signal generator and if it's off, it's not off by much. The second image shows channel 1 with the P6074A displaying the signal from the fast pulse generator with the input adjusted to fill five vertical divisions. The signal looks pretty clean, except for that stair step after the rise. Image number 3 shows the same signal at X10 magnification (1 ns per horizontal division) and the 90% rise time looks like 3.6 ns. The fourth and fifth images are the corresponding measurements on channel 2 with the P2200 showing a rise time of about 2.6 ns. I have swapped the probes between channels, and tried one of the other P6075A, all with similar results.

I'm certainly willing to believe that a) this scope needs some adjustment, it has, after all, not been calibrated since 1979, and b) that the probes may be degraded (P6075A) or imprecise (P2200). In performing these tests I have had plenty of evidence that there are significant problems on channel 2 (the level is not centered correctly, many of the setting positions have dirty contacts, and there is some kind of crosstalk problem at high sensitivity ranges). While I have a specific task I'm trying to use this scope for, my real interest in it is nostalgic, as it belonged to my father. If all I were interested in were performing practical measurements I would spring for a Siglent 1204e-x and be done with it (I may yet do that: a good pair of probes from Cal Test Electronics will cost almost $200, and for that kind of money I could have a perfectly nice, modern Rigol 1054Z).

I've got no fear of getting in the scope and doing some cleaning and restoration (okay, I'm a little scared of the of the HV section around the CRT, but I know enough to keep one hand in my back pocket), and I'm certainly learning a lot along the way (and enjoying the process).

[james_s]

That's a nice looking scope. The 475 was one of Tek's legendary analog scopes, they're still quite nice to use when the application is something that an analog scope is suited for.

[jdutky]

That's a nice looking scope. The 475 was one of Tek's legendary analog scopes, they're still quite nice to use when the application is something that an analog scope is suited for.

Thanks. It belonged to my father, who used it for about 8 years as a service engineer for a company that produced laboratory instruments and computers. He hauled that thing back and forth across the US, I think in checked baggage (hard to imagine, but maybe the treated it better than they do normal luggage), but even if he carried it on it's a bit of a marvel (that thing is HEAVY, and even carry on baggage takes a beating over time). He took excellent care of his tools, and this looks almost new.

He bought a brand new 2213 when he went self-employed in 1980 or 81, which is a much lighter beast, but also much less capable than the 475. That scope was also taken very good care of (also hauled back and forth across the US, this time by car for, over a decade), but I'm concentrating on getting the 475 working because it's got the higher bandwidth and the dual delayed timebase.

[james_s]

Those things cost a fortune in their day, I don't know exactly how much the 475 retailed for but the better Tek stuff was typically in the price range of a nice new car. It's the sort of thing that if you owned it personally it was a prized possession and you would take good care of it, especially if your livelihood depended on it.

[jdutky]

Those things cost a fortune in their day, I don't know exactly how much the 475 retailed for but the better Tek stuff was typically in the price range of a nice new car. It's the sort of thing that if you owned it personally it was a prized possession and you would take good care of it, especially if your livelihood depended on it.

The stock 475 was about $2500, which isn't quite a fortune, only about the cost of a new, relatively inexpensive car. It was bought with company money, and bears a company sticker. I'm not entirely sure how my father wound up with it, even with 8 years of depreciation I imagine it was still rather expensive.