Electronics > Repair
Troubleshooting intermittent fault in an SRS SR530 reference oscillator PLL
DaJMasta:
I've checked and rechecked the logic signals, they aren't changing.
I started trying a more experimental approach - soldering extra capacitors to existing ones and seeing if the fault changed. I tried adding various small and larger capacitances to C336, C306, C308, C310, and C312. While some changes sort of seemed to change the frequency of the fault (and I got the PLL to stop reacting a couple of times), it wasn't by a lot, and the faults were of the same nature.
So I captured the timing of the pulses around the PLL to try and determine if it was the comparator or the PLL that was acting first, and while it seems like it is a reaction to the comparator that causes the PLL to adjust in an unhelpful way (and not the other way around), I captured one event that was clearly bad:
It's noise. Light blue is the comparator input (pin 3 U306), red and yellow are the PLL controller outputs (1 and 13 U306), and dark blue is the signal input (pin 14 U306) and while this is before the short pulse glitch, it shows the PLL controller being incorrectly triggered by noise on the two input lines. Zooming in, it's just a jagged mess, so while mostly common mode between the two inputs, clearly at least some of it triggers a switch.
So where was it coming from? I had suspected noise from outside interference and clapped on the chassis covers before, but it hadn't made a difference. I did the same again, then left only the sig gen connected to eliminate any potential ground loop issues... no dice. Still see it with minimal external connections and the behavior doesn't change when I connect up the scope. I don't have an isolated scope to be able to take that entirely out of the equation, but I'm not really expecting it to be external issues at this point.
I think my next steps are remeasuring the rails specifically looking for high frequency noise that correlates to transitions of the PLL output that have the glitch. I'll also probably just add a small cap across pins 3 and 14 of U306 (maybe 22pF as a start?) for common mode noise reduction.
timeandfrequency:
Hello DaJMasta,
Indeed, this 4th snapshot is interesting : we now see the culprit.
Yes you can check the power rails and also replace all of the caps in the power supply assembly.
But I would also suggest to make a test with VERY clean power supplies.
Lift the power supply pins of U306 and U314 (not the ground pin).
For U306 : use one 9V 6F22 battery + LM7805 regulator + 2 x 100 nF connected to the regulator (see its datasheet)
For U314 : use 2 x 6F22 9V batteries (so supply is ± 9VDC)
Keep the wires from the batteries as short as possible : not more than 10 cm, and twist them together.
Do NOT use lab power supplies for this test.
And then check the loop again to see if it still unlocks.
You have never confirmed if you also changes U313.
DaJMasta:
U313 has been swapped.
I've taken a closer look at the power and grounding and was lead down a rabbit hole by the grounding because the shell of the reference input is directly grounded to the chassis with its own wire and mount, which is not indicated in the manual. Sure enough, disconnecting that ground and letting it only go through the 2k resistor makes it not work, but it means 0-10 ohms between the reference input BNC shell and the circuit grounds and other BNC shells, which took a lot of continuity checking to finally understand.
As for the power rails, I had checked slower speed dips and temperatures for regulators and caps but I checked the high speed noise by time correlating the narrow pulses that are causing issues with each rail - +15V, -15V, +5V, +7.5V, -7,5V and even the 10.24V reference and there is no evidence that noise is present on any of them except for the 5V as identified before. That +5V rail lower speed stepped noise is in phase with the glitches, where its lowest value step is immediately preceding the glitch start. I checked both ends of R330, R336, and R366 to see if the comparator output was dragging down the voltage notably and while all of them look about right, the waveform at the output of U315 (R336) has the lowest high voltage level. Tracing it around, it goes to the signal input of the quadrature (second stage) PLL (U316) and then off board to U708 with a few input pins - maybe the extra fanout is dragging it down a little... but maybe it's more than it should be, since the HC157 input bias currents should be very low. I may swap U708 if I can't find something more suspect. I may also try some lower value bypass caps on these resistors for the +5V rail to see if they will help.
I could also try messing with the capacitor values on the common mode noise reduction and feedback caps on U315, since I was messing with the U314 ones before.
To be thorough, I rechecked every input to the sheet to look for noise or activity correlating to the glitch and there's basically none. !HI, !LO, !f, !2f, !QUAD, !IN PHASE, Phase Voltage, !TRIG2, TRIG2, !TRIG1, and 488Hz measure as to not having any glitch correlated noise, but I see the slightest bit on !RESET. Now this slightest bit is a tiny negative leaning glitch (<50mV) right around ground and not actual toggling, and this part of the circuit really doesn't have to do with the problem I'm seeing, but interestingly, pin 2 of U711 measures normal (low output, no glitch), but perhaps if I'm out of ideas I could try swapping U305.
I've checked and rechecked soldering, I've probed enough relevant points to probably have noticed an intermittent break in a trace, and I am once again scraping the bottom of the barrel to try and find an option.
DaJMasta:
Did some more investigating and technically made progress but it doesn't work yet. I checked on the 5V rail because it was showing that periodic noise, which I would sort of expect from the design, but which didn't seem good. First I checked what it looked like at different frequencies and when lower, it is more pronounced, and approaching 1kHz, it goes away almost entirely (an order of magnitude less or more). Interestingly, the overall lock behavior had changed too, as it was no longer working at 10kHz while the top end of the frequency range was good and the bottom end was even worse than 100Hz (10Hz basically never was locked according to the indicator light). So the noise presence didn't really correlate with the fault, but I slapped on a 2200uF capacitor to see what it did to the noise, and yes, it removed it, but no, it didn't fix the problem.
I went investigating the !HI and !LO lines that go off the board wondering if something on the other end was dragging them down, and though it sort of looks like an output port, it's being driven by U717/U718! What's more, !LO is just an inverted version of !HI and wasn't toggling with it, so I replaced U718, got the toggling back, and nothing changed about the behavior of the lock, best I can tell.
I saw a strange waveform at pin 14 of U327 which really looked wrong, so I replaced U327. Nothing changed, seems like the output of the LM311 is open collector, so the high impedance of the DG211 means that when the switch is switched off, there's basically no waveform on the output. I then rotated some of the pulled chips out with the ones in circuit, the DG211s, the CD4046, and the LM13600s all got rotated to see if something would change and it didn't, so I think the ones I have all probably work normally.
I checked the phase voltage line and while stepping by 90 degrees sort of makes an odd sticking behavior, I think it's compensating for this instead of constantly toggling through software by using the quadrature PLL's outputs which are also being switched in at the same time, and using the fine phase adjust makes the voltage sweep as expected.
I haven't tried messing with the hysteresis capacitor on U315, so maybe that's still something, but since I've been able to "solve" the +5V rail noise problem with bulk capacitance and it hasn't made a difference, I don't know what help that would be. It feels like I need to look off this page to find something, and the third PLL on the vertical board may be the leading candidate since Vvco is part of the feedback loop of the first and connects directly to it.
I say that it's the leading candidate, but I want to reiterate how little confidence I have in that, it just seems like the next most closely related thing to try after literally trying everything else I can think of. I may be 50-60 hours into this repair at all and it is wearing on me, for sure.
Kleinstein:
The LM311 output is open collector - so that is correct.
The 4046 PLLs are known to cause trouble when the input signal at pin 14 is not ideal. This input is not a normal logic input and may react odd when the signal somehow exceeds the supply only a little. On the other side it may handle smaller signals OK. Different manufacturer's 4046 may also behave slightly different in this aspect.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version