TSP #225 - Agilent E5052A 7GHz Signal Source Analyzer Teardown, Repair & Experiments

foreign e5052a which is a signal Source analyzer now if you haven't seen my video on phase noise that would be a good one to watch before this one then I describe what phase noise is from a theoretical point of view its implications on various applications how it's measured and even some of the more advanced measurement techniques that an instrument like this uses which is cross correlation techniques in short the cross correlation technique allows this instrument to measure phase noise below its own internal synthesizer phase noise which is interesting and if you watch that video you'll see exactly how that is done now this unit is defective even though it's actually quite old I don't want the video that I did on the phase noise I took apart and showed you what's inside of the e5052b which is the follow-up on this one but I think they're very very similar that one doesn't have a floppy for example and might be some improvements of the internal circuits so let's go ahead and turn this on and see what kind of fault it has even though it's an old instrument these are extremely valuable and definitely was repairing so let's take a look well first thing first is complaining the CMOS battery is dead not surprising these motherboards are quite old we can fix that later and by the way the screen looks really reflective but it's actually not it does have a matte finish the angle is just really straight on I think it makes it look worse than it is so let's do some basic measurements right now we're in Phase noise measurement mode as you can see and we're measuring gibberish because there is nothing connected the input has no power going into it and it says insufficient RF level which is correct now I have the input of the instrument connected to the road Insurance SMB v100b which itself is a very clean synthesizer I'm going to enable the output at one gigahertz here we go instead of this this is a one gigahertz tone coming in 0 dbm and you can see the carrier has been detected automatically very close to one gigahertz that's to be expected because both of these instruments have oven control crystal oscillators so they should be in pretty good agreement which they are now I haven't set up the phase noise measurement we just want to do some preliminary testing if you look at the power here it says that the power is -4 dbm does give us a reasonably good shape of the phase noise is probably correct but if you look at the minus 4 dbm that's almost certainly wrong because I have a very good cable connecting the rodent Shores and the engineer here and that cable may have a DB of loss at worst at one gigahertz but it's not going to have 4 DB of loss so let's change the frequency and see what happens I go up by 100 megahertz and look at that I just went up by 100 megahertz it got detected perfectly but the power already changed by quite a bit and this cable does not have these kind of issues so something is certainly wrong if I change the frequency let's go all the way to let's say two and a half gigahertz yep if the frequency is being detected no no problem if I go all the way to six gigahertz which is the highest the road Insurance can do and you can see it's almost perfectly six gigahertz and the power is still minus 3 dbm let's go the other way let's go down to let's say 900 megahertz yeah look at that at 900 megahertz it says it's -5 dbm that's wrong 800 is not minus 9 dbm minus six nine again now minus 13 at 500. minus 16 at 400 yeah this is bad okay so there is something wrong here now you may be tempted to say that well done conversion has an issue but this instrument actually measures frequency power and phase noise in three completely different ways it's not like a regular Spectrum analyzer the architecture is quite different you will go over it very briefly so these are separated in some ways makes it a lot easier to debug this particular problem so there's some subtleties that we need to take care of there are also front-end attenuators but I don't think the front and itinerary is an issue in this case because it's changing widely between different frequencies but we can see if the attenuator works so let me set it to one gigahertz and actually testing the itinerary is really not that meaningful anyway because the antenna is not in the path of the power detection but anyway you will see that in a second so here's 5db if I change it to zero you can see that the MX absolutely no difference in the power measurement this is not even in the path yeah it has no effect the phase noise obviously changes because the amount of power arriving at a phase noise measurement part of the system does change but not for the power detection it doesn't in a very different way so let's go to the block diagram so we can explain this so we'll give us some guidance on how to debug it so it turns out that there is a service manual for the e5052a but unfortunately there are no block diagrams in it and no real information about the internal architecture of the assemblies I guess they really wanted to keep these things a secret but there is an instrument block diagram for the e5052b which is going to be good enough for us to work on as I think as I said they're very similar so here's the RF input coming in from 10 megahertz to 7 gigahertz it's going to get split into two using this resistive splitter one goes down and one goes up let's worry about the downside a little bit later the signal going to the top is split once again into two pads One path going in here and the other path going in here and if you look in the front of the instrument there's actually two loopbacks those are these two loopbacks so you have two completely independent inputs now into this instrument and this is the attenuator that we were changing and you can see once you enter the instrument from the front panel everything after this path is now replicated there's two completely independent down converters two fully independent synthesizers lo2 and lo1 generating 10 megahertz to 7 gigahertz to down convert the signal into basement now even the reference for these synthesizers the oven controlled Crystal references are from two different vendors and even mounted perpendicular inside of the chassis so that vibration effects and even gravity effects become uncorrelated it's really important that these two down conversions are uncorrelated because that information is used to subtract their effect from the original signal the signal here and here which is from the input are completely correlated but the signal afterwards is now has two effects on top of each other the effect of the loss and the effect of the original signal and by looking at this DSP after they are digitized you can subtract the uncorrelated from the correlated effect and therefore go well below the phase noise of lo2 and lo1 which is pretty clever this is a pretty old technique and there have been improvements on this since this was originally invented now if you look at the other path down here we take this signal and we split it to the bottom side and we go through these two diodes which are of course positive and negative Peak detectors and these two positive and negative Peak detectors tell you the maximum minimum value of the signal and you can digitize that to find out the input power and therefore the power is measured using this part of the circuit and it has nothing to do with the top which is very helpful because this is where the complication actually is in terms of the design and then afterwards we have a data converter here it's low speed ADC which is going to digitize the value of that there's also a switch in front of this a to d allowing it to digitize also the temperature monitor this temperature monitor is probably coupled to these detectors so that the temperature can be canceled out and as the temperature inside the unit changes you can correct for that and calibrate for that now the signal coming from here is also tapped off it looks like and it goes through a divider comes down and goes through a counter so indeed they're tapping off a little bit of power over here in order to run this counter so they can measure the frequency as well in this counter is of course comparing it against the internal reference itself so we have three separate measurements being done the phase noise up here the frequency over here and the power down here and this is what I was saying is that it's a little bit of a red herring try to chase a 10 units up here because they have nothing to do with the power measurement we should be really focusing in this area of the instrument to figure this out there's also some very low voltage and very very low noise and very accurate sources here so used to measure PCOS we don't have to worry about that here but that's also available so now that we understand this we can go back and take it apart and find every one of these components so we see where we need to pay attention to all right so here's the top of the instrument this is basically all the PC stuff just the basic power supply we have the hard drive which definitely needs to be imaged and replaced hopefully replaced with an SSD because it's most likely end of life and we have the motherboard underneath I think by removing this we should be able to change the battery of the motherboard so that error with the CMOS dying goes away that should be definitely easy to do all the actual RF measurement is on the other side and the PC and the RF are separated by a plating between them so there's no noise coupling to them anyway so let's change the battery before we go to the RF and here's our CMOS battery all the way down there should be pretty easy to replace and from the bottom instrument we can see the signal path no pun intended going from the RF input so here's the RF input coming in goes into this block which I think is a DC block then it goes into a power supporter the first one in the chain that we saw in the block diagram too on the right side we go into another Power splitter and then we get two signals coming out which go to the front of the instrument and then back in go into the two separate attenuators and after that they got both Bend in and go to the other side where we have all the other down conversion but on the other side of the resistive splitter we go into another Power splitter once again and this time we have the positive and the negative detectors as two separate sma-based component which is nice easy to find them and then we have the temperature sensor right over here I'm not so sure how well this is thermally coupled but I think it's just in the proximity probably good enough and then we have the two analog signals coming from the two detectors going into the board where I think that the 80d and the amplifiers underneath here we also have another signal which is the frequency divided divided signal I believe which goes into further divisions probably underneath this cage as well so really all the signals we need to take a look at are here which is super convenient and we should be able to quickly measure that so if I unplug these two and digitize them myself I should be able to measure exactly the detector output coming from here and seeing if they make any sense as a function of the input power and as function of the input frequency and see if we can detect the problem or not and if the problem is for example one of these detectors being bad we should be able to find that pretty quickly if these produce perfectly good signals across all powers and frequency then the problem could be underneath this could be the digitizer or the amplifier will have to dig in further so let me set that up okay so let's do some proper tests on these detectors so I have the input coming in again from the rodent Shores smbv100b we're going to apply a low frequency am modulation of that signal so basically the amplitude of the signal is going to shrink and grow shrink and grow at a low frequency that's perfect for the detectors to follow the Peaks the negative and the positive peaks of that signal that's going to be interesting so I have taken the two outputs of the detectors you can see I've disconnected them from the main body of the PCB and these two are now going to the channel one and channel 3 of the road Insurance mx04 series now this is a 12-bit ADC based instrument with a very low noise and high dynamic range perfect to look at the detector signals which are unamplified we're looking at the raw signal so if I turn the instrument on we should be able to see the am modulation envelope from the detectors all right here we go I'm gonna turn the RF power on and look at that that looks pretty good so here's the positive and the negative detectors and they look pretty symmetric and they look pretty good so I'm a little bit puzzled now so this is a frequency at one gigahertz let's change the frequency like we did with the actual instrument I'm going higher oh the amplitude is not stable at all so right now this is two gigahertz let's go back down below one gig yeah there's something going on so here's 800 megahertz you can see the amplitude is shrunk so much wow this is 500 megahertz and it is completely gone this is bizarre but the detectors are giving very nice symmetric signals I'm beginning to think the problem is not with the detectors which would be unusual but we have to now go a little bit earlier in the chain and see what the issue is so here's 300 megahertz 200 megahertz becomes large again very unusual okay let's change the setup and measure some s-parameters of the entire front end see if you can find the issue so we loaded the signal here and here the output of the detectors these two cables and they were symmetric but they still show that frequency dependent issue so we should step back one and measure perhaps here and see if the s-parameter from here to here looks good now this side is should be terminated normally so that shouldn't be an issue is this side that's actually the problem and let's find out if this has a nice flat frequency response as we expect it is possible that someone put too much power in here and partially damage these resistors in this divider here and that could explain the problem we're seeing if there's this sort of partially damaged you could see that frequency response potentially but it's a little bit puzzling nonetheless but let's measure here and for that measurement we can simply just use the znl20 we'll connect Port 1 to the input of the instrument and Port 2 over here to the output side of this divider and that will give us the S parameter of the front end let's see how that looks well I was just about to remove this cable so we can measure the S parameter here and look at what I found look at that that connector is completely broken from that power divider circuit and that of course explains the frequency response because you don't know that kind of connection that is making you could have notches in its frequency response which is what we were seeing so yeah no need to measure us parameter right now we just have to remove this and see if this is repairable at all otherwise we'll have to find a replacement for it and here's our divider module this is of course fully passive and yeah there's something definitely going on with that these are nice and solid I think we're going to disassemble it a little bit more I didn't disconnect this from here for obvious reasons because there's a lot of torque on this if I twist that I'm worried that I will do more damage so let's completely disassemble it before we take any more action all right so let's see what we have here so this is where the connector has separated and you can see that it looks like we might have had some cold solder so that connector wasn't really well soldered down and it must have failed over time and I can clean this up a little bit and I think this should be easily repairable there's nothing unusual about this kind of power divider it's actually just a resistive divider I'll show you the other side in a second and this is where that was so yeah there's a bit of a cold solder joint in here but I think we should be able to fix that and on the other side of this of course we have our resistor divider no problem here this only needs to work up to seven gigahertz so it's very low frequency anyway there is nonetheless still a little bit of matching you can see they have added a tiny bit of series inductance over here to absorb the capacitance of the surface man components and match it nicely over the Broadband I like the transition this coaxial vertical transition a lot of viewers around there nicely done yeah so I think I'm just going to try and Reflow this very carefully obviously not to damage the connector but uh yeah then we will measure with s parameters if this parameter of this looks good I think this might fix it no matter minor point if you look at this solder joint versus this one and this one you can clearly see that there is a very big difference so this problem must have originated from the factory because yeah this is clearly not not sorted correctly and there's a dip in it and that's what probably has slowly caused it to fail maybe some uneven heating of this entire structure and that would explain this problem and over vibration and time yeah eventually it fails so I think that the Reflow went fairly well and the connection looks really good so now we should be able to close this back up I cleaned it up as much as I could you can see the cavity in the back piece over there to ensure that you're not coupling to these traces and make them essentially as independent as possible so let's reassemble this and now we can measure its s-parameter by itself and if this looks good then we should be able to put it back and here's our s-parameter setup very simple we're going to use the road and short Network analyzer now it doesn't matter which two ports I use for measurement and this was the damage Port which was the common Port the way it was set up in the instrument so we have Port one here and Port two here and the unused Port is terminated now this is a resistive divider so of course all the ports need to be 50 ohm swimming and it all resist all dividers require that and at the same time because it's resistive we're going to lose half of our power inside of those resistors the benefit is that this is broadband and it works down to DC although it's not really used down to DC in this instrument as a result we expect from here to here they have an S21 of nominally about -6 and it would be the same from here to here let's take a look and here's the response and it looks good this is near DC this is eight and a half gigahertz and of course this is wider than the instrument itself is supposed to do but nonetheless we can measure up to eight and a half and we can see that the loss is about minus 6.3 and if I go down this knob definitely needs acceleration it doesn't have it so I'm gonna have to turn it a lot to move that marker forward and the marker over here this is a 1 DB per scale so we do have a little bit additional loss right down the middle but that doesn't really matter and then you can see it picks up again and I bet that network with a little bit of inductance is to compensate and that's how they get this addition of peaking towards the end but at the lowest point we're about minus 6.7 or so and that of course can be calibrated that's part of the calibration of the instrument across power so I think it looks good we should put it back in the unit and try it out and all back together nice and solid here we go the moment of truth I have the synthesizer set to 200 megahertz and zero dbm let's enable and nice we're reading minus 0.3 dbm which is pretty good let's go here's 300 400 500 700 800 sorry 800 it was used to read minus 17 or 15. so it's working now here's a gigahertz nice let's go to 2 gigahertz good now this is not just a cable loss here's three gigahertz and 4 gigahertz five gigahertz and six gigahertz yeah so it definitely needs to be recalibrated I need that power gradient is definitely going to have to be readjusted but it looks good the main problem where it was just giving us completely garbage result that definitely has been addressed which is very good so the only other thing to do is to measure phase noise with a very very low phase noise Source just to make sure that part of it also works and then we can call the success and here I'm using the signal Hound pncs1 which is a one gigahertz reference precisely for this kind of verifications ultra low phase noise source and here I'm running 30 correlations starting from 10 Hertz all the way to 40 megahertz now starting from 10 Hertz and running 30 correlations you can see how slowly this thing actually goes and does one round of correlations but you need to do that if you want to see the true phase noise of this Source especially for frequencies below 10 kilohertz and offsets and it takes a long time now you look at the performance of this at one kilohertz offset we're sitting almost at -140 DBC per Hertz which is really quite good and at one megahertz we're already at -150 and there is a saw filter inside of the pncs1 and above about 20 megahertz or so it really cuts out everything else and there is literally nothing left and we're sitting at minus 167 DBC per Hertz it's really really quite clean integrated Jitter from 20 kilohertz to 20 megahertz you can see it's about 25 for them to Second which is also very good and that's exactly what you need in order to be able to test the performance of something like the e5052a now keep in mind that if I connect the pncs to a regular Spectrum analyzer you will not see these results at lower frequencies especially because you would be simply measuring the Spectrum analyzers internal phase noise and that's exactly what that instrument is for but this one it can do cross correlation go well well below that although it is slow for a modern let's say phase noise analyzer you can go down to minus 180 185 DBC pairs which is really quite impressive but nonetheless I think the e5052a is still an excellent instrument and it seems to be fully functional and the power measurement here although it's a little bit obscured by the cursor there it says minus 9.8 plus 9.8 or so which is also correct that source is supposed to put that about 10 dbm so I think yeah I think it all works very nice and there you have it I hope you enjoyed this repair I'm going to put the cover back on and it will be ready to get back to all tests to learn this to me for this repair video I'm always grateful for that and let me know what you think in the comment section and even though the problem itself wasn't fundamentally very difficult to fix if you still need to know where to look and the analysis of the block diagram the symptoms that's part of the repair process and I hope you enjoyed that I'll see you next time foreign

Channel: The Signal Path

Views: 17,438

Keywords:

Id: rmoD_7Uc8Ss

Length: 20min 36sec (1236 seconds)

Published: Sun Apr 23 2023