|
Re: HP 478A Thermistor Sensor. Input return loss and matching
¿ªÔÆÌåÓýThe HP 478A Thermistor Sensor ?Option H72 and others state? only to be used with HP 432 ? So it sounds like the new HP 478A Thermistor Sensor. Without the option might also not be useable / compatible ?with the 431 ? So has anyone a handle to the serial number of the HP 478A Thermistor Sensor. that go with the HP 431?? and Those for the HP 432 Paul ? From: [email protected] [mailto:[email protected]] On Behalf Of ebrucehunter via groups.io ? I was under the impression (although possibly incorrect) that resistance and capacitance elements found in early 478A sensors were needed corrections for 431 power meters that utilized a 10 kHz bridge circuit. ? Bruce, KG6OJI |
|
Re: HP8757A Display Problem (Scalar Network Analyzer)
I think the 8757A uses an HP 1345A display module, which is used in multiple
different instruments. You can isolate the display subsystem by unplugging the
26 signal flat flex cable, at either end. This will cause the 1345A to go into test
pattern mode, that should look like this:
?
?
The service manual for the 1345A is available online from the usual sources.
?
I recently repaired one of these displays that apparently had a functional Y axis, but
the X axis was dead, thus just a vertical line with varying intensity. As part of my debug process, I took scope measurements of all the nodes in the Y axis starting at the pre-amp chip at U28 and continuing to the final transistors that drive the deflection plates. All the scope images, test pattern, and faulty display can be accesses at: ?
?
Many of these signals have a significant DC offset (10V to 100V) with an AC
component (the actual display data) of 20mV to 40mV P2P. My scope traces
either show that in two separate pictures, or on the same picture with two
channels, one showing average DC level, and the other trace shows the AC
coupled version of the signal with 20 mV/Div .
?
From your description, it sounds like your issues are earlier in the signal path,
but this test pattern is an easy starting point to localize the fault to be either in the display section, or in the 8757A supplying display commands over the flat flex cable. ?
None of the analog signals representing frequency or amplitude are directly
connected to the display. They are processed in the rest of the analyzer
and a list of display commands (a little like HPGL) are sent over the flat flex.
Depending on the options for the 1345A display, it ether has a display memory
that contains the display command list, and repeatably plays it over and over, or it doesn't have this memory, and it is the responsibility of the host instrument
(8757A) to continuously send the data to refresh the screen. ?
With display memory, the host instrument only has to send the display commands
when the display has to change. This is best for generally static displays. Without
the memory, the host instrument processor has to continuously send display
commands, so a heavier load, but more appropriate for a continuously changing
display.
?
Philip
? |
|
Re: HP8757A Display Problem (Scalar Network Analyzer)
This afternoon I read more of the manual. I then checked the "X-Y-Z Analog" board that drives the video - gently "banging" (tapping with a plastic rod!) the rear corner of that board suddenly but intermittently changed the vertical position of the display. I dismounted this board and I saw that someone in the past had re-soldered 4 of the preset adjustable pots - that's an indication that they also maybe had this problem in the past, so I checked around them and properly cleaned those joints and re-soldered them together with the pins for the wites that connect to the backside of the board.
?
This solved the problem of the display being off-centre - it now appears in the correct place (also with the test pattern selected), although the problem of the sequencing/cycling of the 4 different 'video' channels being far too slow is still there.
?
I checked the CPU board and the battery - the battery is still good, and I can store the setup and recall it perfectly okay.
?
So I think the problem that remains is the speed of the 'frame sequencing' if that's the correct terminology, and maybe that is on the CPU board since that sends the outputs to the display board. I have no further ideas. |
|
Re: HP 478A Thermistor Sensor. Input return loss and matching
When I fitted a ferrite bead to my early 478A I found that the VSWR curve became smoother but it degraded from about 1.068:1 at 50 MHz to about 1.075:1 with the ferrite added. This is unavoidable as the resonance was slightly improving the VSWR at 50 MHz.
?
A VSWR of 1.075:1 is still OK for measuring the output of a typical 1mW reference at 50 MHz but I had a go at improving this by fitting an external matching network at 50 MHz. Obviously, the matching network should provide minimal insertion loss.
?
The complex input impedance of the early sensor at 50 MHz is about 1000pF in series with 50 ohms. So the simplest matching network would be a series inductor of about 11nH.
?
See the plot below that shows the input VSWR of the early 478A sensor (complete with an added ferrite) in the unmatched state
against what happens when matched with an external series 11nH inductor with an unloaded Q of about 55 at 50 MHz.
?
You can see that the response is about the same if I set port 2 of the simulator to have the same port impedance as my early 478A sensor with the ferrite added. Without the matching network, the insertion loss is about 0.005dB.
?
The resistive loss of the matching inductor means that matching it makes no improvement in the insertion loss but there should be an improvement in mismatch uncertainty.
?
By changing port 1 to 52R (VSWR of 1.04:1) and adding a long lossless 50R transmission line I can graphically proved an indication of the mismatch uncertainty caused by a 1mW source that might have a VSWR of 1.04:1. See the second plot.
?
You can now see there is a tiny advantage in having the matching network as there is much less uncertainty ripple (at 50 MHz) in the red trace compared to the green trace. However, in reality, I think a typical 1mW source from an HP power meter will have a fairly accurate 50R resistive part of its source impedance. So I'm not convinced that there will be much advantage in having this matching network inline. This is because of the way the 1mW 50 MHz reference circuit is designed in a typical HP power meter. It's likely to look like (maybe just over?) 50R resistive in series with a tiny reactance.
?
However, I have gone ahead and built a matching network and I can describe and show it it a later post. I hope this stuff is interesting and not too nerdy :)
?
?
?
?
?
?
?
?
--
Regards
Jeremy |
|
Re: HP 478A Thermistor Sensor. Input return loss and matching
Here's a comparison showing the VSWR of an early sensor (VNA measurement of my HP 478A) against a model based on the schematic of the later sensor. You can see that the VSWR at 50 MHz should improve to about 1.05:1 with the later sensor type and there is no resonant blip at 53 MHz.
?
I do have the later type of sensor here but it is still factory sealed (see the image below to see it has never been used) and it seems a shame to disturb it just to measure the VSWR. You can see in the image that the red internal washer for the N connector is supplied external to the sensor. I'm not sure why this is. I doubt it has fallen out on its own but it is lying flat just below the black outer of the N connector. I think this sensor is actually over 30 years old but never been unpacked and used.?
?
I also have a brand new unused (also made over 30 years ago) 432A meter with sealed cable and paperwork from HP and I keep this carefully stored for the day my other 432A or 478A finally fail BER.
?
You can also see that the efficiency chart on the side of the sensor shows the efficiency of the later sensor type is very close to 100% at 0.01GHz and the efficiency drops to about 98% by 2 GHz.
?
My early sensor also has a minor VSWR wiggle and (in)efficiency bump at about 850 MHz but I've not investigated this as it is almost certainly caused by something inside the metal body containing the two thermistors and I don't want to disturb this.
?
?
--
Regards
Jeremy |
|
Re: HP 478A Thermistor Sensor. Input return loss and matching
¿ªÔÆÌåÓýIt would be reasonable to try 61 mix NiZn ferrite with a 300C curie temperature, and mu of 125.
From: [email protected] <[email protected]> on behalf of jmr via groups.io <jmrhzu@...>
Sent: Saturday, May 17, 2025 8:34:17 AM To: [email protected] <[email protected]> Subject: Re: [HP-Agilent-Keysight-equipment] HP 478A Thermistor Sensor. Input return loss and matching ?
The reason the efficiency also suffers with a blip at about 40-60 MHz is that some of the incident RF from a 1mW reference source will end up in the 2.7R damper resistor via the resonant path to R1. This means the DC substitution (used by the HP 432A meter)
can't correctly indicate the incident power in this frequency range unless the efficiency change is known across this tight frequency range. It's the equivalent of the sensor losing efficiency and in my case this was a loss of over 1% efficiency at the peak
of the bump. This was always really annoying for me because it meant I had to have an efficiency chart showing how the efficiency sharply changed by over 1% across 40-60 MHz and this required lots of cal points.
?
Adding the ferrite achieves two things. It removes the bump in the input VSWR and it also flattens the efficiency. At first, it might appear that adding the ferrite would suck even more energy from the thermistors as a ferrite bead looks resistive up at
VHF.? This would make the efficiency even worse. However, as long as the ferrite introduces 'lots' of series resistance (ESR), there will be a big mismatch loss at 50 MHz so only a tiny amount of incident RF will get burned up in the high resistance of the
ferrite. The efficiency won't be perfectly flat but there should be a big improvement in efficiency across 40-60 MHz and I was able to achieve this.
?
Obviously, the choice of ferrite is fairly critical here as it has to provide a high ESR at 50 MHz and it also has to be fairly immune to temperature changes. This means that the initial permeability needs to be reasonably stable wrt temperature changes
across about 10-50degC.
?
The latest manual for the 478A sensor lists this ferrite as E1 but it doesn't give a part number for it. It would be really nice to know what the official ferrite material is.
?
?
--
Regards
Jeremy
|
|
Re: HP 478A Thermistor Sensor. Input return loss and matching
The reason the efficiency also suffers with a blip at about 40-60 MHz is that some of the incident RF from a 1mW reference source will end up in the 2.7R damper resistor via the resonant path to R1. This means the DC substitution (used by the HP 432A meter) can't correctly indicate the incident power in this frequency range unless the efficiency change is known across this tight frequency range. It's the equivalent of the sensor losing efficiency and in my case this was a loss of over 1% efficiency at the peak of the bump. This was always really annoying for me because it meant I had to have an efficiency chart showing how the efficiency sharply changed by over 1% across 40-60 MHz and this required lots of cal points.
?
Adding the ferrite achieves two things. It removes the bump in the input VSWR and it also flattens the efficiency. At first, it might appear that adding the ferrite would suck even more energy from the thermistors as a ferrite bead looks resistive up at VHF.? This would make the efficiency even worse. However, as long as the ferrite introduces 'lots' of series resistance (ESR), there will be a big mismatch loss at 50 MHz so only a tiny amount of incident RF will get burned up in the high resistance of the ferrite. The efficiency won't be perfectly flat but there should be a big improvement in efficiency across 40-60 MHz and I was able to achieve this.
?
Obviously, the choice of ferrite is fairly critical here as it has to provide a high ESR at 50 MHz and it also has to be fairly immune to temperature changes. This means that the initial permeability needs to be reasonably stable wrt temperature changes across about 10-50degC.
?
The latest manual for the 478A sensor lists this ferrite as E1 but it doesn't give a part number for it. It would be really nice to know what the official ferrite material is.
?
?
--
Regards
Jeremy |
|
Re: HP 478A Thermistor Sensor. Input return loss and matching
Here's the schematic of the early 478A sensor. You can see there is a 460pF capacitor C3? in series with a 2.7R damping resistor R1.
?
I've added some extra detail to show that there is a short red wire that connects between C1 (inside the metal sensor block) and C3 on the PCB on the back of the block. This wire is probably an inch or so long.
?
Sadly, this causes a resonance near 50 MHz and this is the cause of the blip in the VSWR. I believe that R1 is there to damp out this resonance. To demonstrate this, I've put together a very crude model of the sensor in Genesys and modelled the red wire as a 22nH inductor. I modelled with and without the 2.7R damping resistor R1 and the results are shown below. I also included the real VNA measurement of a real (early) 478A sensor.
?
You can see that the VSWR has a huge resonant wiggle at about 53 MHz when R1 is removed (see the green trace). So I think R1 was added to try and damp out this resonance.
?
Later sensor variants appear to deal with this issue by fitting a ferrite bead over the red wire and the damper network is no longer needed. I've tried upgrading my sensor with a ferrite bead and can confirm that it is well worth doing as it completely damps out the resonance, leaving a nice smooth VSWR curve.
?
I hope the simulation plot is readable. I've enlarged the fonts to try and improve things.
?
?
?
?
?
--
Regards
Jeremy |
|
HP 478A Thermistor Sensor. Input return loss and matching
In the last few weeks, I've had my old HP 432A power meter on the bench along with its 478A thermistor sensor. I normally only use it about once a year to check the 1mW (50 MHz) reference on another power meter.
?
However, I've spent a bit more time looking at the 478A sensor and wondered if it was worth sharing some of the knowledge gained and also some of the options for reducing the measurement uncertainty compared to using a standard sensor at 50 MHz.
?
I've looked at the input match, the efficiency and also at how the 478A has had some changes internally over the years. One obvious option to improve the input match at 50 MHz is to buy a special version with option H75 or option H76. These are probably quite rare and expensive. Another option is to improve the input match of a standard sensor by adding a low loss (external) matching section inline with the standard sensor. I have done this and have achieved good results. I'll show the design process and test results in this thread.
?
The standard sensor has a fairly grim spec for input VSWR at 50 MHz of about 1.3:1. However, it shouldn't be this poor at 50 MHz if you look at what is inside the sensor and model the VSWR. In theory, an older version of the 478A (like mine) should manage a VSWR of about 1.07:1 at 50 MHz. The later versions might be closer to 1.05:1 at 50 MHz. So in this sense, there probably isn't much point trying to improve the match if the aim is to measure a 1mW reference from a typical HP power meter. This is because the source VSWR of the 1mW reference is likely to be much lower than its spec of about 1.06:1. So the mismatch uncertainty should be negligible. Only a calibration house should really be concerned here, but I thought I would have a go at improving my sensor anyway.
?
The following posts show the thought process and the modelling and the test results.
?
I've attached a VSWR plot of the input VSWR of an early HP 478A sensor. This was taken with a lab VNA set to a really slow sweep time of about 20 seconds, although slowing the sweep time from less than 1 second to 20 seconds didn't seem to make much difference.
?
You can see there is a kink in the VSWR at 50 MHz and this has always bugged me because it causes a bump in the efficiency across about 40-60 MHz. I've lived with this 'bump' issue ever since I first got this sensor back in the 1990s. But now I've managed to get rid of it and I'll show what causes it and how to upgrade an older sensor like mine to get rid of it.
?
?
?
?
--
Regards
Jeremy |
|
Re: 1 mw - 0dbm lab calibration
¿ªÔÆÌåÓýHi Jeremy, Nice thread on old stuf I hardly ever use. But now I got encouraged to do some tests as well. You certainly discovered that Appendix II of the O/S Manual of the HP432A gives instructions to install long cable options. So HP is very much aware of the total resistance values. Mine has Option 009 (10 feet cable) but I choose to disregard for the moment, I have the standard cable. The calibration procedure 2 is quite convenient making use of my HP34401 DVM. You need a long stabilization time.? I was wondering: how about thermocouple effects of the different connections in the chain from the thermistors to the internal bridges. For fun I measured the return loss of one of my HP8478B
(1833A13571) sensors. Quite good actually and certainly within
spec. See below. Thanks for the nice work and observations. Best 73 de Harke
On 8-5-2025 22:14, jmr via groups.io
wrote:
|
Paul Bicknell
ebrucehunter
Philip Freidin
Trevor Gale
