Now I notice that the Krohn-Hite is nicer than I thought. It does actually have a 5:1 vernier control for fine adjustments. I'd never noticed or used that. It's a cool geared control that feels great.
The Tek does not have vernier control, and could use it actually, it's hard to set the lowest frequencies closer than 1 Hz on the bottom of the 20-200Hz scale. You breathe on the control and hope and keep trying to get what you want or give up with something close enough. The Tek also (and like many similar units) has only a linear dial, which isn't good for sweeping and setting the low end.
The Genrad has vernier control, but it's just a friction control on the edge of the frequency dial that only works if the frequency dial is completely flat, and mine is just enough unflat to make it nearly impossible to use. This is not a true geared vernier like the Krohn-Hite. Anyway, the Genrad main knob itself turns so fluidly it seems to allow very small adjustments and I was thinking of bringing it back out of storage for that reason.
And now on closer examination I see the Krohn-Hite control is not linear like the Tek as I was thinking, but fully log, giving better control at the bottom of every range. And now with the 5:1 internally geared vernier, I've decided it should probably be my standard, thought the smaller and lighter Tek is often more convenient.
And the Krohn-Hite is one nicely made unit otherwise too. You can remove the side and bottom panels for full access, with the electronics on circuit boards inside an aluminum rod frame. I see from the main circuit board that it dates from 1976. Curiously that's when I first came across Krohn-Hite gear, which I used in my Undergraduate Senior Thesis experiment in auditory perception. The Krohn-Hite gear we had in the Psych lab was tube and it looked like it dated from 20 years earlier. But it worked great, you could just rely on it.
Anyway, I was hoping the power supply voltage adjustments would fix the problem. That's what typically goes first on old gear--the power supply. But no, the +/- 22 volt test point voltages were well within spec +/- 0.2V specification. Pretty good for a 46 year old unit. The - is at 22.13V and the + is 22.03V. Perhaps the slight offset causes trouble downstream, but that is best adjusted later.
I then checked the DC voltage. The required test points seemed hard to find, I concluded that what looked on the board like "O1" was actually "Q1" which is where I had to dig up and attach a pair of resistors (approximately 330 ohms) to "short out" the oscillation. Then I couldn't find F1 and instead adjusted for zero offset at F0. Later I found F1 and discovered I had adjusted in the wrong direction. Anyway none of these adjustments had any affect on the cutoff frequency for oscillation, though slightly different on each of the two bottom bands.
Using a pair of Fluke 8060's, I monitored the oscillator output frequency (which were exact), output levels (which declined throughout the bottom of the bottom two bands), and the measured voltages.
Studying and thinking about the schematic was what really set me on the right course. I did a lot of that during the two days of investigation. My first thinking was that the big 32 Meg Ohm resistors in the bottom two bands, which were clearly unlike all the rest (possibly glass rather than metal film, but fully encased) had gone defective. But that couldn't easily explain the amplitude drop with frequencies in the bottom two bands. I began to favor a theory in which one of the 100 uF caps in the front end had gone bad. One was a drain regeneration bypass, and the other a coupling capacitor. I really thought the bypass had the greater chance of being defective. They are all lab equipment grade Sprague electrolytics in grey metal cases--the unit is chocked with them.
It occurred to me I could clip in bypass caps "on top of" these caps (actually only the wired clips would be on the cap leads, the replacement caps would be on my bench).
But, first, I thought I'd measure things with my Rigol scope. Best to measure before taking actions which could hurt things I figured. And it was another "opportunity" to re-learn the Rigol scope, which I found baffling at first.
Finally it was clear that all points I could clearly measure, like F0 and F1 and the sides of the coupling cap, basically lost signal at the same time. There was no point you could say "the signal died here." Well that makes sense because they are all coupled by feedback.
But the very front end at Q1, I couldn't measure signal at all. I figured the signal is too low or the impedance too high. But whatever garbage it put out, or measurement on my Fluke, seemed to to vary with oscillator frequency. So I figured the "signal" was OK there, but I couldn't be very sure.
I found a bunch of 10uF film capacitors in my junk box, and wired them together to make 10Uf and 40 Uf. I hooked the 10uF to the front end capacitor and it seemed to make the osciallator work at low frequencies, but the level was very variable. Hooking 40uF to it killed the oscillation. Hooking 40uF to the coupling capacitor fixed it, there was only a small decline probably thanks to 40uF not being the specified 100uF. But I figured the built in cap had simply lost capacitance, NOT failed completely, or else how could the oscillator work at all???
I clipped out the bad capacitor, leaving enough lead on the board to possibly mount new capacitor there. Or unsolder from the bottom, which would be more risky but give a better "repair." I'll decide what to do when I get the new capacitor.
I measured the capacitor with my capacitor meter, and it measured only about 3uF. Well, no wonder it wasn't working at low frequencies!!!
That big resistor is 680 ohms, which biases the differential bipolar oscillator amplifier. It looks like at least a 2W resistor and my back-of-envelope calculation said it could have as much as 10W--but that can't be true. Anyway, it's significant source of heat right next to this capacitor, which I've seen has a 105C rating (should be OK) but with a mere 2000 Hour lifespan (!!!).
The lifespan may almost double with each 10C decrease, so it's probably larger than that in practice. But I think a better design would have somehow put that resistor farther away from that capacitor.
But it may also be that at 40 years of so, ALL of these electrolytics are due for replacement anway.
Anyway, I'll just fix the coupling capacitor for now, unless it is clear then that something else needs replacement. It turned out I could have ordered an "identical" specification Sprague replacement that looks the same (but might be slightly smaller). But Vishay (who now owns Sprague) offers a unit with 4000 hour lifespan and 125C rating, specially designed for long life. I also got the 40V rating rather than just 25V. (The rail spread is actually 44V. When the circuit is operating correctly it works at less than 25V.
Anyway, the new higher temp, higher reliability, and higher voltage cap (same capacitance) from Vishay is now on order from Mouser. I notice the similar Sprague unit with same specs as mine is basically the same size. My older Sprague capacitor is somewhat larger, it appears. So the new one should fit regardless of the upgrades, which is usually how it goes, the parts keep getting better, at least if you order the newer ones.
This experience illustrates something I've found before in my career as a computer programmer (in science and engineering mostly). I usually find things mostly by thinking "what could cause this." In the case, a frequency dependent error (loss of low frequencies) would be most likely to result from a defective capacitor, than, say a transistor or resistor or connection.
Often one can step computer programs through debuggers (which in my career mostly didn't work at all, just pure frustration...though I once built one I thought was great) until one is blue in the face without learning anything useful. In this case I used a scope for several hours and could have gone one for many more without learning anything (because of the feedback loop). It's thinking about things that is the most effective diagnostic instrument.
No comments:
Post a Comment