|
I was able to source some CS5208-1 from Kynix. Had good results ordering obsolete/obscure parts from them before:
?
Also finished designing some adapter boards to bring the A3 out of the chassis and allow for "hot" testing.?
They connect together using Samtec SFSD .050" wiring harnesses, which come in all sorts of lengths. Hopefully 24" of wire doesn't mess up signal integrity too much.?
?
?
?
?
I just ordered parts and boards. If its successful I'll share gerbers and BOM so others can use.?
?
In the mean time, does your firmware disassembly have any useful hints as to what that error means? Looks like its pointing out a particular .cpp file in the FW related to the A3.?
Also, if anyone has info on LEDs that would also be great. Wish I had a working unit to look at and compare which LEDs are lit/not lit. Of course service manual doesn't say anything.?
|
Appreciate the suggestion. I reinstalled the A3 and tried cycling all the PSU connectors, including the motherboard one. I also brushed a tiny dab of Deoxit D100 on each male pin and cycled the connectors a few times.
?
Unfortunately, no changes.?
?
I also checked all of the power supply test points that are made accessible. All of the voltages seemed present and at a reasonable level:
?
?
One thing I observed on the back of the A3 board are the presence of red LED. These seem to be connected to different power rails on the back of the board.
?
?
I noticed that 1 LED was very dim and the other was not lit. These are visible with the outer cover off, through the chassis air holes where the A3 is slotted.
?
The 2 LEDs in question are the bottom most ones, which are connected to the 2 TO-220 voltage regulators U18 and U19.
Not sure what the "normal" condition should be; are both normally lit when no issues are present?
?
?
U19 is an ON semi CS5208-1. Datasheet is as follows:
Compare to U18 which is an LT1085:
?
The CS5208-1 is not available anymore, but the LT1085/1084 are. As far as I can tell, these are nearly identical regulators?
The CS5208 is rated to higher currents (6V, 9A) while the LT1085/4 is tolerant of higher voltages (30V, 3/5A). However, the pinout and output divider adjustment equations are identical?
?
I can get the LT1084 on Digikey. Maybe its worth swapping both U18 and U19 out with new LT1084 and seeing if anything changes?
|
The first thing I recommend that you do is to look at all the PSU cables in the PSU area including the big multiwire one that loops back to the motherboard.
?
As a first attempt, try partially removing each connector from its header and walk it up and down a few times on its header pins to try and remove any oxidation. Do this for all the connectors in this area including the big multi-way connector that loops back the CPU area. I think this one has lots of brown wires. Wiggle and walk them all in their pins and retry the unit. It could be that you have excess voltage drop in the connections to the A3DSP board.
?
?
|
Hi all,
?
Great thread, lots of useful info including suspected power supply issues on the A3.
?
I recently picked up an E5052A at auction. I've worked on similar era Agilent boxes but new to this model.?
Unit was sold has having a known issue on startup. There is a strange startup message which I believe is related to the A3 DSP board:
?
?
This happens on power up. The analyzer will then show the following runtime error and exit; it never makes it to the measurement screen:
?
?
The first few things I did was to backup the original HDD (20 GB Travelstar) and set that aside.?
?
I used the ghost factory backup image in the recovery partition to make a new hard drive, which booted just fine.?
?
Same error message on power up with original FW version 2.00. Updating to 2.51 does not change anything.
?
I'm convinced that the error "m_pA3drv -> Initialize()" is likely a hardware error, telling us that the A3 ADC/DSP board fails to initialize.?
This is only a guess; I see that jmr has disassembled the firmware. Your software skills are far superior to mine; maybe this error message makes more sense to you?
?
What I'm working on now is to make a wiring harness so I can extend the 3 x 32 (96 position) header for the A3 ADC/DSP.?
This way I can power the board outside of the chassis and check things like power supply rails, clock distribution for the ADCs and FPGAs.?
?
By the way, is the A3 board a huge hassle to remove for anyone else? The board is wedged extremely tight in my unit. I had to use a screwdriver to pry against the metal shield bolted to the board. I did it very carefully so no damage... it was impossible to remove otherwise.?
?
Giving the A3 a once over with shield removed, are a few CPLDs and a tssop FLASH chip, probably for the Sharc DSP. Hopefully its not a data corruption issue on one of these factory programmed devices...?
|
I also designed a 10 MHz LC resonator version with the aim to run the resonator at fairly high power. I've a fair bit of experience designing discrete low noise gain blocks so I replaced the noisy MMIC with a discrete BJT design and the first attempt is as below.
?
I think I can improve on this by several dB by reconfiguring the output network of the gain block (to allow higher power) and also I think I can improve on the resonator if I order some better parts for the resonator. It currently manages -165dBc/Hz at 10 kHz offset but I think the next version will be better than this.
There's a lot of spurious signals below about 500 Hz and this is due to pickup in my workroom from nearby equipment. I'll try and do something about this. I don't think it is getting in via the DC power feed.
The noise floor is very lumpy once it gets below -180dBc/Hz on the E5052A and again, I'm probably exploring the limits of the E5052A again here.
?
This is an LC oscillator not a crystal oscillator so the close in phase noise is relatively poor. I'm not that interested in achieving low phase noise below about a 1kHz offset but I can probably experiment with the gain block to minimise flicker noise.
?
?
?
?
|
Thanks.
I had some free time today and experimented with a discretely designed gain block in place of the MMIC. I think I'm hitting the limits of the E5052A now but I managed -170dBc/Hz at 10kHz offset and about -178dBc/Hz for the far out noise floor. I couldn't improve on about -160dBc/Hz at 1 kHz offset. This is a slight improvement but I think the E5052A is becoming the weak link. I've seen the PN2060C phase noise analyser by BG6KHC on qsl and this looks to be an interesting piece of equipment for $849.?
?
I've linked to it below.
?
Welcome to BG6KHC’s website
?
Has anyone got any experience of this device?
|
Hi Jeremy,
Nice noise floor. You are still at 20dB/decade at 10Hz slope so playing with sustaining amplifier and reducing Fk might improve things a lot.
I will email you, perhaps it's getting a bit off-topic.
Leo
|
Hi Leo.
I dug out my old homebrew 10 MHz xtal oscillator and measured it with the E5052A. One of the design goals for this was to be able to use it as a low noise LO for the E5052A so I could measure phase noise for carrier signals below 10 MHz.. But it was also an exercise to see if it agreed with theory and how low I could get the noise floor.
?
The circuit does run the crystal with a very high drive level and I didn't bother trimming it to 10 MHz.?
?
I used a basic MMIC gain block as the active device so I think I could improve on the close in noise by several dB if I used a lower noise device. I might revisit this at some point although the performance below is already good enough for my needs with the E5052A.
Regards
Jeremy
?
?
?
|
Isn’t one solution to side modes using multiple (fiber) loops of different lengths, such that fewer frequencies fit?
(Borrowed thinking in terms of laser resonators and etalons where sometimes specific choices are made that permit only a single mode within the gain bandwidth of the amplifying media)
|
Thanks, Jim!
?
We have gone through all the usual suspects when choosing the phase noise kit for this round. There was also NoiseXT (is it former Aeroflex?)
Cross-correlation is a great advantage but you have to start with decent phase noise LOs if you want to keep the R&D cycle sensible.
I have no idea how people design long term high stability oscillators in one lifetime!
Leo
?
On Sun, Mar 2, 2025 at 09:51 PM, Jim Ford wrote:
Watching this subject with interest.? Leo, you may want to check out OEWaves, Synergy Microwave, and Holzworth Instrumentation (now part of Maury Microwave) for ultralow PN sources and Rohde & Schwarz, OEWaves and Holzworth for measurement instruments.? Also, there are several source mfrs making sapphire resonant cavity oscillators.? You probably already know about these companies, but throwing the names out there just in case. Microchip (was Microsemi, was Symmetricomm, was Ball Efratom; correct me if I got that progression wrong) as well.
?
I bought one of your GPSDO models a few years ago; it's great!? Looking forward to getting one (or more) of your pulsers at some point.
?
I'm working on a synthesizer myself based on a long spool of optical fiber to get the ultralow phase noise.? The tradeoff is sidemodes, which are spurs in the microwave output.? I have an idea for a way to knock those down.? My goal is a low SWaP-C synth mainframe in a 3 rack unit box that can be expanded from 1 to 8 outputs DC to 18 GHz or so.? I just don't have the room for multiple 4 RU boxes (HPAK 8340, 8662, etc.)? Stay tuned!
?
Jim Ford
Laguna Hills, California, USA
|
Thank you Jeremy,
?
I might pick one up to play with.
?
I have spoken to Jeremy Everard last year about shared low phase oscillators development and ended up sending him the actives I was using that they were going to test at the department.
He was using fully kitted out FSWP26 at the time but it was not perfect for low phase noise work.
To be honest far out noise between -170 and -180 is not a problem since cross-correlation gets it cleaned ?up within a minute or two.
It's the close in at 1Hz which takes forever to stabilise — I usually have to wait 30-40 minutes for it to reach -120dBc/Hz at 1Hz for 10MHz carrier.
Also 10-100Hz is a very problematic area where a lot of hash is holding cross-correlation stubbornly high for longer.
Leo
?
On Sun, Mar 2, 2025 at 10:56 AM, jmr wrote:
The best thing about the E5052A is that it is very fast. What isn't so good is the qty of internal spurious terms, especially the one from the backlight at 60 kHz. I don't think it can compete with modern alternatives in terms of the noise floor it can achieve. The E5052B is going to be better I think. Hopefully the internal spurious are improved on the E5052B as well.
|
Unlikely to have been from a switching reg -- 1kHz is pretty low. Close-in spurs are often suppressed by some form of cancellation, and a little drift can cause the reappearance of spurs at around that level (or more). That's where I'd look first, anyway. But if it's gone, be happy (unless or until it comes back). :)
Tom
--
Prof. Thomas H. Lee
Allen Ctr., Rm. 205
420 Via Palou Mall
Stanford University
Stanford, CA 94305-4070
toggle quoted message
Show quoted text
On 3/3/2025 1:42 PM, Jim Ford via groups.io wrote: In my case it was a Keysight E8257D PSG that showed spurs >-60 dBc at around 1 kHz offset. ?Our project had about a 6 month lull, and when we got back to the lab and set up the benches again, no such spurs could be found! ?I figured the original spur was from a switching regulator, but who knows?
|
70 MHz Surface Acoustic Wave (SAW) BPFs used to be common for satcom. ?10 kHz is very narrow, though.
I’m snickering a bit at the differences in spur performance in sig gens because I’ve seen exactly that issue before not too long ago. ?In my case it was a Keysight E8257D PSG that showed spurs >-60 dBc at around 1 kHz offset. ?Our project had about a 6 month lull, and when we got back to the lab and set up the benches again, no such spurs could be found! ?I figured the original spur was from a switching regulator, but who knows? ?Either that particular generator had a problem and we have been using a different one since then, or KS fixed that when it went for calibration. ?Cue the Twilight Zone theme!
Jim
toggle quoted message
Show quoted text
On Mar 3, 2025, at 11:46?AM, jmr via groups.io <jmrhzu@...> wrote:
? Hi Jim, thanks. I obtained the cal certificate for the other E5052A today. It was calibrated less than a year ago and this one shows an even lower noise floor.? It's a slightly newer unit and it has some hardware changes although I'm not sure if these changes are the reason it is slightly cleaner than my E5052A. The service manual shows the hardware changes wrt serial number. It also has the same internal spurious frequencies as mine and they appear to be a bit higher in level. It has the backlight spurious at 60 kHz as well.
?
Looking in the service manual, it appears to describe this test at 70 MHz where a (low phase noise) PSG sig gen is used as the signal source at 70 MHz and it is passed through a 10 kHz wide BPF. Presumably some form of crystal filter? This will then shave off all noise down to the thermal level in the stopbands of the filter.
?
I tried this with my E5052A using a lower test frequency (10.7 MHz) and a 15 kHz wide crystal filter and I saw about -175 dBc/Hz at offsets above 1 MHz. Not as good as the newer E5052A but the cal certificate did quote an uncertainty of about +/- 4dB for this test type. I really should do the test at 70 MHz as well. The noise floor may be degraded for carrier frequencies as low as 10.7 MHz.
?
|
Hi Jim, thanks. I obtained the cal certificate for the other E5052A today. It was calibrated less than a year ago and this one shows an even lower noise floor.? It's a slightly newer unit and it has some hardware changes although I'm not sure if these changes are the reason it is slightly cleaner than my E5052A. The service manual shows the hardware changes wrt serial number. It also has the same internal spurious frequencies as mine and they appear to be a bit higher in level. It has the backlight spurious at 60 kHz as well.
?
Looking in the service manual, it appears to describe this test at 70 MHz where a (low phase noise) PSG sig gen is used as the signal source at 70 MHz and it is passed through a 10 kHz wide BPF. Presumably some form of crystal filter? This will then shave off all noise down to the thermal level in the stopbands of the filter.
?
I tried this with my E5052A using a lower test frequency (10.7 MHz) and a 15 kHz wide crystal filter and I saw about -175 dBc/Hz at offsets above 1 MHz. Not as good as the newer E5052A but the cal certificate did quote an uncertainty of about +/- 4dB for this test type. I really should do the test at 70 MHz as well. The noise floor may be degraded for carrier frequencies as low as 10.7 MHz.
?
|
Watching this subject with interest.? Leo, you may want to check out OEWaves, Synergy Microwave, and Holzworth Instrumentation (now part of Maury Microwave) for ultralow PN sources and Rohde & Schwarz, OEWaves and Holzworth for measurement instruments.? Also, there are several source mfrs making sapphire resonant cavity oscillators.? You probably already know about these companies, but throwing the names out there just in case. Microchip (was Microsemi, was Symmetricomm, was Ball Efratom; correct me if I got that progression wrong) as well.
I bought one of your GPSDO models a few years ago; it's great!? Looking forward to getting one (or more) of your pulsers at some point.
I'm working on a synthesizer myself based on a long spool of optical fiber to get the ultralow phase noise.? The tradeoff is sidemodes, which are spurs in the microwave output.? I have an idea for a way to knock those down.? My goal is a low SWaP-C synth mainframe in a 3 rack unit box that can be expanded from 1 to 8 outputs DC to 18 GHz or so.? I just don't have the room for multiple 4 RU boxes (HPAK 8340, 8662, etc.)? Stay tuned!
Jim Ford Laguna Hills, California, USA
On Sunday, March 2, 2025 at 06:41:24 AM CST, jmr via groups.io <jmrhzu@...> wrote:
I don't know if this helps, but I dug out the most recent Keysight calibration certificate for this E5052A and there is a noise floor test done at 70 MHz with correlation set to 1.
I'm not sure what is involved with this test or at what power level it is run at but the numbers for the noise floor seem to be very good! I don't have anything here that can replicate this test although I could probably design a low noise LC oscillator that could get close at a 1 MHz offset.
?
I generally design wideband single loop synthesisers up at VHF through UHF and the E5052A is easily good enough for stuff like this.
?
|
I don't know if this helps, but I dug out the most recent Keysight calibration certificate for this E5052A and there is a noise floor test done at 70 MHz with correlation set to 1.
I'm not sure what is involved with this test or at what power level it is run at but the numbers for the noise floor seem to be very good! I don't have anything here that can replicate this test although I could probably design a low noise LC oscillator that could get close at a 1 MHz offset.
?
I generally design wideband single loop synthesisers up at VHF through UHF and the E5052A is easily good enough for stuff like this.
?
|
-120dBc/Hz at 1 Hz is way better than anything I've experienced. I had a look at a plot of my homebrew oscillator and it managed -116dBc/Hz at 10 Hz.? Pretty lame by comparison...
?
I guess you are aware of the Jeremy Everard paper from the University of York. This describes a 10 MHz ultra low noise oscillator that achieves -122dBc/Hz at 1 Hz offset.
?
They used a Symmetricom 5120A as below. Very impressive! I don't think the E5052A can get anywhere near this level of performance.
?
|
Hi Leo
This E5052A is quite old now and it does have a few internal spurious terms. Also the noise floor performance isn't that great even down at 10 MHz.
?
I don't have any oscillators here that can generate really low phase noise close to carrier. A few years ago I designed a 10 MHz crystal oscillator with low phase noise but I think it only managed a noise floor of about -173dBc/Hz from 10 kHz onwards. This was the theoretical noise floor of the oscillator (at least according to my basic calculations) but I don't think I've ever seen -180dBc/Hz even with lots of correlation. I've never used more that 100 for the correlation setting though.
?
I'm fairly certain the E5052A just uses an external reference to discipline the internal reference. Both of the E5052A analysers that I have used have an internal high stability OCXO mounted inside the chassis and this has its own output at the rear panel. The idea is to link it to the REF IN connector with a short BNC-BNC jumper and then it shows up as an external reference even though it is actually inside the chassis. My E5071B VNA uses a similar system for the OCXO option.
?
If this BNC-BNC link is removed then the E5052A uses the main 10 MHz oscillator inside and the OCXO is ignored.
?
The best thing about the E5052A is that it is very fast. What isn't so good is the qty of internal spurious terms, especially the one from the backlight at 60 kHz. I don't think it can compete with modern alternatives in terms of the noise floor it can achieve. The E5052B is going to be better I think. Hopefully the internal spurious are improved on the E5052B as well.
?
The baseband input of the E5052A is normally locked out in software with no mention in the documents how to unlock it. I think Agilent decided to hide this feature, probably because of the internal spurious that spoil the performance and also the noise floor rises a fair bit at low frequencies. There is also an enhanced jitter option which I've played with but I don't have much need for stuff like that. The jitter option is documented and available but the baseband isn't. I had to hack inside the firmware of this E5052A to find out how to release the baseband option. It can't be done via the option menus in the way the jitter option can be unlocked.
?
?
If you look on Eevblog, I'm currently testing one of your LBE-1421 GPSDOs.? I'm getting the same result for phase noise that you did. I'm G0HZU on Eevblog.
?
?
Regards
Jeremy
?
?
|
Hi Jeremy,
?
This is a great win. What is your overall impression of the E5052A?
?
I have R& FSPN8 and 53100A but I would like something I can try to push to lower phase noise levels. I need below -120dBc/Hz at 1Hz and below -180 far out.
FSPN8 gets there but very slowly as its references are not ultra low noise level class. And it does not allow to use external references like 53100A or 5052A.
53100A is OK but has some spurs I can't get rid of and I also like all-in-one instruments.
?
If you were to stick ULN Wenzels or comparable references on 5052A would you reliably get below -120dB @ 1Hz and -180 far out?
?
Thanks
Leo
|
Great!? That's the best kind of repair; when you learn something in addition to getting it working again.
Jim Ford Laguna Hills, California, USA
On Saturday, March 1, 2025 at 05:43:25 AM CST, jmr via groups.io <jmrhzu@...> wrote:
It's all OK now, I took it apart and fixed it :) :) :)
?
On the DSP/ADC module A3 the 5V rail is now 4.97V (was 4.70V typical) and the 3V3 rail is now 3.24V (was 2.97V and very drifty).
?
I disassembled the main firmware and worked out how it does the power on testing of the power supplies and this gave me the pass/fail limits for all the supplies.?
I was able to trace through the power supply distribution and see where it was losing voltage in a poor connection. A thermal camera helped a lot here. It's now all fixed and fine again!
?
I also found out a few things about how it manages the internal 10 MHz reference and how to adjust the high stability 10 MHz OCXO. So that is now calibrated.
?
I am so pleased, I thought it was slowly dying on me. There's no way I could afford to send it to Keysight for repair.
?
I should be able to fix the other E5052A as well as it has a similar issue with a sagging 5V and a sagging 3V3 supply.
?
?
?
?
|
Kevin A.