开云体育

I have a number of tracking generator projects in the works, over years, for the 8566 SAs. When I acquired a couple almost-working 8568s a few years back, I figured I'd make one for them too, if I got them fully working, so started design and parts collection, and rough layout and construction. This was a low priority project, until I acquired a couple more 8568s and 85662s this summer, which led to the fix-a-thon I mentioned earlier - to get some fully working. Having three usable SA systems, I restarted effort on the TG project, and was really close to done on the design and build, until this 8568 sweep-stopping etc problem showed up.

The TG isn't quite done, but is usable and works well. It started life as an HP8444A, which was part of a large HP141T SA system I got many years ago, but have since decommissioned and parted out, except for a few pieces. Now, this isn't your typical 8444A - all that remains of the original is the chassis/case, power supply, output amplifier, leveling system, and a bandpass filter. The original 8444A type was used for a number of HP SAs that had pretty much the same frequency plan as the 141T system, such as that in the 8555A, which was the main plug-in. The 8568 is quite similar, and the 8444A has been used with it too, but with limited capability - it can't normally match the high resolution and narrow spans the 8568 can provide.

I wanted this TG able to allow the 8568 to be all it can be, and do all it can do, TG-wise, so I built a whole new frequency control system into it. It uses all three of the local oscillators present in the 8568. Normally, only the the first LO is accessible externally, so simple mods are needed to get the second and third LOs out and delivered to the TG for processing. I had planned all along to use the second LO - it allows for seamlessly using spans below 1 MHz, and is available inside on an SMB coaxial test point that just has to be brought out.

When I started the frequency control design years ago, I didn't have detailed info for the 8568, and I naively and wrongly assumed that all the LOs were fully synthesized and "exact," but recently discovered that wasn't the case. All my work on this part of the plan (I'll explain later) was down the tubes, and I had to incorporate the third LO information. This required another, fairly simple mod to pick off a little bit of power from the third LO in one of RF modules - adding an SMB connector, a couple resistors, and cabling out to the rear panel. The third LO info is what allows the system to reach the narrowest IFBW resolutions all the way down to 10 Hz.

So, to reconstruct the center frequency, the third LO (nominally 280 MHz) is mixed with the second LO (nominally about 1750 MHz +/- a few MHz) to make 2050 MHz. This 2050 MHz, mixed with the first LO (2050 to about 3550 MHz) provides the center frequency ~0 to 1500 MHz as the difference, which is low-pass filtered out, amplified, and leveled. The tricky part is that the second LO includes the fine tuning info for narrow spans, and info from the third LO, used in its synthesis.

The third LO is not directly controlled or synthesized - it's a crystal oscillator that may vary quite a bit, say +/- 50 kHz - but since it is used in the overall second LO synthesis, the variation gets cancelled out at the last conversion from 301.4 MHz IF to 21.4 MHz IF. This was the crucial piece of info that I was missing way back when I started. I thought the 280 MHz was right on, and simply making 301.4 MHz to mix with the second LO would do it. I had a slick arrangement all rigged up, with an oddball OCXO I happened to have, that was close enough to modify to run at 100.466666... MHz, then run through a cool amplifier-SRD frequency tripler from a 1960s PLO unit. I spent countless hours on these two pieces, perfecting it to make a precise, clean, 301.4 MHz, only to find that it wouldn't work, because the 50 kHz or so variation in the 280 MHz was hopelessly beyond even the coarse tuning range of the OCXO, let alone the fine tuning for tracking.

I realized that even if I could have made a new oscillator setup, or put in a DDS system, it still would not accommodate and correct for the possible drift in the 280 MHz XO - there are no specs on this, so all I would know is that the actual frequency can be pretty far off nominal, and drift an unknown amount. The only simple way to go then, is to get that third LO signal and use it to make the proper approximately 301.4 MHz that includes the error.

So, mixing the third LO with 21.4 MHz from a regular VTXO is what finally made it all work - beautifully.

That's all for now. I hope you can appreciate why I used all three LOs, and that it's not all that hard to get them. Next time I'll explain the details of how I did it. The frequency plan is just how to make the desired output frequency, but there's a lot more to making it good, and making the TG operation invisible to the SA.

Ed

It's been chilly out again, and the 8568/85662 has been running stable since last night without any sweep-stopping, so I took the opportunity to work on the TG some more. I took some screen shots to record the dynamic range and look for adequate isolation. Only one came out legible after freehand shooting of the upside down display. It's at

/g/HP-Agilent-Keysight-equipment/photo/78668/0?p=Name,,8568,20,1,0,0

in a new photo album called "8568 TG project." I think I managed to get the image right side up, at least on my PC. My apologies for the poor photography or if it shows upside down.

This one shows the full ~0 to 1500 MHz span, at IF and video bandwidths low enough to get near the noise floor of the 8568 - each sweep took 500 seconds. The top A trace is the TG output near 0 dBm reference level, while the bottom one is the stored B trace maximum hold values, with the TG in its "RF OFF" mode. It exhibits better than 90 dB dynamic range between the states. The small spurious signals and ripples in the noise floor are due to leakage through the coaxial relays that rout the center frequency mixer output and amplifier output signals within the TG. When the cable from the TG out to the SA is disconnected, these disappear, leaving just the SA's noise floor. Other experiments confirm that the TG is virtually invisible to the SA, due to good isolation of all the LO signals, and proper containment of the TG's internal operating signals.

Ed

Hi Ed,
I’ve done pretty much the same thing a few times with asst SA’s. ?You’ll find that the narrow bandwidth on the SA really isn’t that useful with an SA/TG combination as the DUT will determine the resolution. ?
You could do a filter tracing with a signal generator and a diode detector, which obviously has unlimited bandwidth.
The schema I used was mixing the 1st LO with a synthesized LO at the high IF center frequency. ?I’ve used a TCVCXO as the reference and slide the frequency to get it close to the SA center frequency. You’ll need either a real good mixer and/or an isolator to keep the high IF LO from lifting your baseline. ?They’re fun projects; more an exercise in shielding than design. ? Main thing: have fun with the project and don’t do too much hair pulling.
Jeff

Ed,
I just recalled a weird intermittent failure involving the 85662A A3A4 and A3A8 brds. The sweep ramp goes to both brds. if I recall correctly. The A3A8 digitizes the sweep ramp and when it hits the proper value it creates an EOS ( End Of Sweep) signal that then goes to A3A4 and the 8568 A22 brd (8566 A16brd). If the EOS is sent early then the Sweep is too short. If the EOS is late or not sent then you only have one sweep and no reset of the sweep. I have replace many 1826-0448 DAC’s for low order bit failures, if I recall correctly the replacement DAC may be 1826-0698?
The DAC’s were id’d as MSB and LSB ramp, sweep, sweep time, and EOS ramp DAC’s.
Also the failure symptoms were similar for too short of a sweep on the 8566 and the 8568, but drastically different between the 8566 and 8566 for the too long of a sweep.
Best regards,
Don Bitters

Ed,
How versed are you in the freq. tuning scheme of the 8568? Are you aware that the scheme involved toggling frequency shifts between the 1st and 2nd L.O.’s every 20MHz of freq. sweep? This was done to keep mixing products out of the 21.4MHz IF path.
I have used the 8568B and 8444A opt 059 in 10Hz and 30Hz RBW’s in very narrow spans or zero span with 5 - 20 min. stability, but only after 24 hr. warmup/runtime. The HP-Agilent TG solution was a different box with both freq. and sweep lock circuitry., which the 8444A opt 059 did not have.
I wish you great success with your 8568 TG.
Don Bitters

Now I'll talk about the methods used for processing the LO signals. There are many ways to do it, but I used particular arrangements depending on what I had on hand in the junk parts department. My rule of thumb for TG system invisibility is to acquire the signals and get them to appropriate levels, while providing about 100 dB or more of reverse isolation between all activity in the TG, and the SA's LO sources.

The first LO (1LO) is fairly strong, at around 6-8 dBm, so I went without amplification, and used a cascade of one-octave 2-4 GHz isolators. Four of these, followed by the original 2-3.8 GHz BPF from the 8444A, followed by one more isolator, provides 1LO for the output mixer.

2LO is fairly small, around -26 dBm, so needs amplification. Three narrow band isolators, tested for good performance at 1.75 GHz, are used in this chain. First, an isolator, then a 30 dBg amplifier, then another isolator, then a 1.75 GHz BPF, then another isolator, provide 2LO for the IF mixer.

3LO is also fairly small, around -26 dBm - I picked off only a small amount from the 3LO module in the 8568, to minimize loading and interference with the signal that goes to the PLL synthesizer in there. This is the part I discovered late in the game - the need for 3LO rather than a fixed 301.4 MHz source.

It turned out that I had some big old RF modules from an AIL brand synthesizer I junked out twenty years ago, and had only kept these parts intact all this time because they had nice BPFs built in. The BPFs were the wrong frequencies, but close enough to be readily re-tuned for what I needed. The first module is an amplifier/filter/amplifier, and was ready to use pretty much as-is, by re-tuning the BPF and peaking circuits in the amplifier stages for 280 MHz operation. One drawback is that these modules were built for -12V power, but the 8444A PS only has -10V, and I didn't want to make extra power supplies. They work OK on -10V, but have less gain available. They can easily be modified if necessary, for more gain.

The second module was a mixer/amplifier/filter/amplifier/mixer - the exact function I needed, except for the second mixer, which I deleted. I tweaked it up for 301.4 MHz, and added a PIN circuit and a gain stage at the output.

So, the first module amplifies and filters the 280 MHz 3LO, which is then mixed in the second module with 21.4 MHz from a VTXO. The resulting 301.4 MHz is amplified and filtered, then amplified some more. The PIN stage provides gain control for the ALC.

The 301.4 MHz goes to the IF mixer, along with 2LO, to make 2050 MHz, which goes through an isolator, then a 2.05 GHz BPF, and another isolator, then to the output mixer, where it gets subtracted from 1LO to make the center frequency.

You can probably tell by now that I like using isolators, and have a pretty good stock of them. If I didn't, I would not choose to go this route, because to acquire them new would be very expensive, as with most packaged microwave parts. These functions and high isolation can all be readily attained by using lots of attenuation and lots of amplification, properly distributed, in a number of arrangements. It's also likely that so much isolation is unnecessary, depending on filters, amplifier and mixer characteristics, and signal levels - I've probably overdone it by 30-40 dB in some spots, but it's expedient to just have lots. With experimentation, the various sensitivities of the LO ports can be determined for all operating conditions, and just enough isolation can be applied. Even with little isolation, it doesn't hurt anything - it just limits the available dynamic range.

Next time I'll talk about the output section, low frequency limitations of the original output amplifier and mixer, and the low-band amplifier I had to build to reach nearly DC.

Ed

Thanks Don,

That is very interesting regarding the sweep digitizing process - I'll have to look into that.

Regarding the LO shifting process, I was vaguely aware of it, seeing it mentioned years ago, and apparently you can tell the 8568 not to do it for some situations like using the 8444A. Including the second LO info in the TG makes it so I don't have to care about that - it always gets the right answer.

The tracking stability of my TG/SA setup is looking very good - all that remains for instability is drift in the TG's 21.4 MHz VCXO, and the 85662A's down/up converter 18.4 MHz XO, and the drift in the 3 MHz crystal filters.

Ed

I have added a new picture, showing the low frequency end performance:

/g/HP-Agilent-Keysight-equipment/photo/78668/1?p=Name,,,20,1,0,0

The upper trace A is the TG output near 0 dBm, up to 1 kHz. The lower B trace is the stored maximum hold, with the TG output "OFF." The line spurs are artifacts due to ground loops in the experimental test and repair setups - I'm still working on the sweep-stopping issue, etc. In earlier tests before this came up, the TG exhibited no line/PS stuff at all - only the residual line noise of the 8568 was evident, about 10 dB lower than it looks here. It should be OK once everything is buttoned up properly.

The TG/SA arrangement exhibits nearly 100 dB dynamic range, and is at the limit of on-screen capability. The near-DC (f=0) response of the 8568 at the narrowest IFBW hits the noise floor by about 100 Hz, so it's fully usable from there up. The TG output is of course, coming from the low-band (~0-500 MHz) amplifier. The original, main amplifier rolls off below around 300 kHz, so is incapable of reaching the bottom end of the 8568's frequency range.

Ed

I have posted up some more pictures of the 8568 TG project here:

/g/HP-Agilent-Keysight-equipment/album?id=78668

The pictures show kind of what it looks like so far. Also included is a comparison at 1 dB/div between flatness of the TG/8568 combo, versus typical 8568 independent flatness. The TG shot has some tilt correction applied to compensate for internal and external cable loss - it would drop about a dB lower toward the top end otherwise. The TG/8568 combo result looks a lot like the 8568 example alone, and fits within the 2 dB p-p spec.

This is for the original amplifier and leveling system, more or less - I tried to keep the new PIN leveling section compatible with the rest for now. Ultimately I expect to redo the whole thing to include the low band amplifier's directional detector, but that needs more complexity. For now, in the low band mode, the leveling is open-loop, using manually adjustable power level. Even so, it turns out to be quite good, around the same 2 dB p-p flatness, from nearly zero up to around 400 MHz. Doing true leveling all the way down to nearly DC becomes problematic, and needs an auto-tuning control loop, so that will come later. I mainly want to get it usable and more or less complete for now. Ultimately, I'd like to build a single amplifier/detector module to do the whole works, instead of the two-band system.

Ed

The output leveling is looking pretty good now. According to an HP8484A/HP436A, the power level flatness of the TG alone is within about 0.3 dB p-p, from 10 MHz to 1500 MHz, with tilt compensation applied. It looks like the 8568's flatness (spec 2 dB p-p) is the limiting factor, so there's not much point to improving the TG any more. The adjustable tilt compensation works great, and can accommodate the 1 dB or so internal loss, and up to a couple meters of external DUT cabling. I've set up the variable power range for about >+2 to <-12 dBm, which in conjunction with the step attenuator, allows setting to any level from +2 to -82 dBm.

I have modified the original leveling system quite a bit, mostly to speed it up, get a wider power range, and incorporate the tilt compensation. The remaining stuff now is mostly about speed - as the variable power level is set toward the lower extreme, the detector diode signal gets very weak, and the response slows down, causing jagged amplitude on screen. When the 8568 sweep time is extended from the fastest 20 mSec, to around 50 or 100, everything smooths out nicely. I don't think the 8444A was intended to keep up with these kind of rates, so needs some fixing up.

Ed

I've been working on residual line noise, which required some mods in the 8444A, mostly remounting the power transformer properly with insulators, adding a common-mode choke on the output signal line, and isolating the output connector from the front panel. The transformer changes took care of most of line stuff internal to the 8444A, while the others were for the common-mode noise from the 8568's back panel, which is connected to the 8444A back by four coaxial cables. There's lots of crap circulating in the interconnects between the units - the RF, IF/display, TG, and line cords.

I've managed to get the line noise down almost to what it looks like on the 8568 all by itself with no input, but with the TG all hooked up and running and connected to the 8568's input. The line spurs are setting around -95 dBm, primarily 120, 240, and 300 Hz. This is a few dB above the noise floor. It was quite an exercise, with lots of fooling around and experimenting and frustration.

Ed

On Sat, Nov 17, 2018 at 02:00 AM, Ed Breya wrote:

The first LO (1LO) is fairly strong, at around 6-8 dBm, so I went without
amplification, and used a cascade of one-octave 2-4 GHz isolators. Four of
these, followed by the original 2-3.8 GHz BPF from the 8444A, followed by one
more isolator, provides 1LO for the output mixer.

Hello again Ed. Can I resurrect a point from this old but very interesting thread? I'm a bit puzzled as to why it is necessary to add four isolators between the 8568B's LO1 output and the original 8444A circuit. The HP designers didn't see a need for them so what did they overlook?

Thanks, Alan

Alan wrote:
"I'm a bit puzzled as to why it is necessary to add four isolators between the 8568B's LO1 output and the original 8444A circuit. The HP designers didn't see a need for them so what did they overlook?"

As I recall, the original 8444A had one isolator, and the bandpass filter. This was good enough for the HP141 system with limited dynamic range and IF RBW, It was also good enough for the 8568A, since its usable RBW in this case is also limited. For my TG system, I wanted to preserve the 8568's full capability, so a high degree of isolation is needed on all the LO signals. I did not do actual tests to see just how much is needed, but applied my rule of thumb (from experience with other TG designs and experimenting), that 100 dB or better pretty much assures "invisibility" of the TG to the SA in all conditions. In reality, less isolation may be OK, but I provided for lots, with options to delete some eventually if it was OK. You can typically count on about 15-20 dB for each isolator, and 15-20 dB for the mixer. Amplifiers can provide good isolation too, so in a more conventional approach with attenuation and gain, you get a dB per dB of attenuator, and maybe 20-50 dB for an amplifier per 20 dBg, depending on the design. Sometimes the off-state reverse isolation of an amplifier matters too - it's typically very poor. In most TG setups, when the amplifier is shut down, so is everything else, so there are no offending signals present anyway. In more complicated, multi-band systems, this may not be the case.

Regarding isolation, remember that each LO port on an SA leads into the IF chain, with not much isolation, and is just as valid a signal input as the main input to the front-end mixer. Insufficient isolation doesn't hurt anything - it just limits the usable dynamic range due to possible spurious signals and baseline lift, which is the raising of the apparent noise floor. In fact, an easy way to assess the isolation situation is to just start with little, and see the results. Then add more until it's acceptable. You can also drive the LO ports with an external generator signal (not much power is needed, say -30 dBm or less will show a lot), and figure out the sensitivities.

Ed

Oops - forgot about the filter. The BPF can provide a great degree of isolation, depending on its CF and BW, relative to the offending signal frequencies. I'd have to look back at the details on it before commenting. Later.

Ed

On Fri, Jul 3, 2020 at 07:32 PM, Ed Breya wrote:

In fact, an easy way to assess the isolation situation is to just start with
little, and see the results. Then add more until it's acceptable.

That sounds a nicely pragmatic way to proceed: just plug the 8444A into the 8568B and see if the performance is compromised. What a great answer.

Regards, Alan