HP8753C VNA Sampler Teardown & Reverse Engineering

[coppercone2]

resistive ground plane is right what I was looking for, another fascinating aspect of advanced DIY PCB manufacture

I hit a dead end for applications of resistive wire inductors (i.e. mix R and L for RLC, so its just physically a L and C), but resistance ground plane sounds like something that can involve thick film resistor pastes, nichrome PCB's or other weird things

[JoeyDeng]

Thanks so much for all the information, and for being so patient with a non-professional.  I’ll need to dive into these resources first. Here’s the article I found:
https://www.rfmentor.com/sites/default/files/NA_Error_Models_and_Cal_Methods.pdf

Honestly, without studying this I couldn’t hope to build a VNA myself. Luckily I’m comfortable with math I really enjoy it. I’ll start out using an open-source solution, and eventually move the whole thing onto an FPGA.

[Bud]

Then you need Dr Joel's book
https://www.amazon.com/Handbook-Microwave-Component-Measurements-Techniques/dp/1119477131/

[henrikf]

I have a question regarding the directional coupler you used in the project. In the conceptual diagram in your article, the "coupled forward" and "coupled reverse" ports are left unconnected. However, in the PDF schematic you provided, these are connected to pin 2 and pin 5 of J2. Could you kindly explain where these pins are routed to, and what their function is in the overall design?

In the coupler schematic symbol pin 1 is the coaxial cable signal conductor, pin 2 is the coupled output, and pin 3 is the coaxial cable shield. See the attached image of the coupler footprint.

First of all, his coupler design doesn't work as well as you would expect. Look at his uncorrected S parameters. With high directivity, the short and load S11 traces should not cross.

You are comparing port 1 and port 2 traces which have different frequency responses. Attached is a measurement of the coupler breakout PCB (Input port to thru, coupled, and isolated ports). Unused ports are terminated with about -20 dB return loss loads and the connector return loss is about -15 dB. Coupling drops at higher frequencies due to 0402 packaged resistor parasitics. 0201 resistors and smaller diameter coaxial cable should be used to optimize the performance above 10 GHz.

[EggertEnjoyer123]

I have a question regarding the directional coupler you used in the project. In the conceptual diagram in your article, the "coupled forward" and "coupled reverse" ports are left unconnected. However, in the PDF schematic you provided, these are connected to pin 2 and pin 5 of J2. Could you kindly explain where these pins are routed to, and what their function is in the overall design?

In the coupler schematic symbol pin 1 is the coaxial cable signal conductor, pin 2 is the coupled output, and pin 3 is the coaxial cable shield. See the attached image of the coupler footprint.

First of all, his coupler design doesn't work as well as you would expect. Look at his uncorrected S parameters. With high directivity, the short and load S11 traces should not cross.

You are comparing port 1 and port 2 traces which have different frequency responses. Attached is a measurement of the coupler breakout PCB (Input port to thru, coupled, and isolated ports). Unused ports are terminated with about -20 dB return loss loads and the connector return loss is about -15 dB. Coupling drops at higher frequencies due to 0402 packaged resistor parasitics. 0201 resistors and smaller diameter coaxial cable should be used to optimize the performance above 10 GHz.

Hmm, I just assumed the two ports were identical, and I guess I shouldn't have done that.

I believe that the directivity is still close to zero in the plot you sent. At 15 GHz if you had a short, the source would get attenuated by 10dB before making it to the short, and then after reflection it would be about 30dB lower when it reaches the coupled port. With a load the isolation is about 30dB, so a short and a load would look almost the same.

Also for the mixers you might want to consider looking into building your own. In my other thread I was able to make a 2-18 GHz mixer for a few bucks with decent performance. I think you can extend the performance down to DC if you used ferrites to increase the bandwidth of the microstrip baluns.

I'm also curious as to whether you measured your calibration kit with a professional VNA (since you did do measurements with a professional VNA + cal kit). I don't really think the calibration load you made is good to 15 GHz, but I could be wrong. There are Chinese loads on Aliexpress that go to 18GHz with low VSWR supposedly (they're only $20), I bought two 40GHz ones and my friend at a lab found out that they were good.

[JoeyDeng]

After getting a basic grasp of error models and working through them (btw, this article is really explains things clearly: https://www.rfmentor.com/sites/default/files/NA_Error_Models_and_Cal_Methods.pdf), I realized something I totally missed before:
Even the strongest error model can’t fully separate data where the strace is completely tangled up with no extra parameters to fix errors.
It took me almost a whole afternoon staring at this to finally click!  Turns out, the graph from heneikf shows the short and load setups for different ports...

[JoeyDeng]

btw,I really wouldn’t recommend relying on this stuff made in china for anything mission-critical. thinking back to when I was in middle school in China, I bought a LibreVNA set, hoping to use it myself. But after testing the included load with the VNA in our lab, I found it totally failed to meet specs. The open calibration piece literally had no internal pins — the leakage was terrible. There are so many shady sellers in China  ... You might get one decent unit, but the rest in the batch are probably junk. Just my opinion

[JoeyDeng]

And also the rf source I designed. Hope it works.

[EggertEnjoyer123]

btw,I really wouldn’t recommend relying on this stuff made in china for anything mission-critical. thinking back to when I was in middle school in China, I bought a LibreVNA set, hoping to use it myself. But after testing the included load with the VNA in our lab, I found it totally failed to meet specs. The open calibration piece literally had no internal pins — the leakage was terrible. There are so many shady sellers in China  ... You might get one decent unit, but the rest in the batch are probably junk. Just my opinion

I measured some cal kits too, the results are also in the 22GHz thread.

The open is almost ideal (it has some fringing capacitance), the short is basically ideal. The load can just be thrown away, it's garbage. There are honest sellers on Aliexpress though. I bought a 40GHz load for $30 and it measured within spec (had a friend measure it for me with his lab equipment).

Have you also looked into the LMX2820? It is twice the price of the LMX2594 on LCSC but the frequency goes up to 22.6 GHz.

[JoeyDeng]

Im new to microwave. I saw that lmx2594 only sales for 35cny(abt 4.79$) on alibaba, so I asked my friends to get some for me, and the appearance of the chip looks fine and brand new.

[EggertEnjoyer123]

$5 a chip seems way too low to be believable. Then again, I have won chip roulette on Aliexpress before so they might be real (but if they are they're probably factory rejects).

[JoeyDeng]

yeaaa thats true, but if they sales the fake chip, the chip out of spec or the second hand chip on alibaba, u can just get ur money back and keep the goods by providing evidence.

[JoeyDeng]

The load I got from alibaba looks not bad 

[EggertEnjoyer123]

Link?

[szoftveres]

The load I got from alibaba looks not bad 

What type of (how high quality) cal standard you used to calibrate the VNA before measuring it?

[JoeyDeng]

The measurements utilize CETC 41st Institute's 2024-calibrated VNA calibration kits with full NIST traceability. 

[JoeyDeng]

and also the best cal kit I could get🥲

[JoeyDeng]

My friend got it for me in the Chinese market. If you need it, please dm me, I can send you some.

[JoeyDeng]

I truly regret using 01005 components. I tried soldering one capacitor and one resistor, and it felt okay, but still...

[khach]

What is real bandwidth of HP8753 Sampler? Does anybody measure sampler response at third harmonic of VNA source and hack firmware to expand frequency range?

[JoeyDeng]

I saw someone managed to upgrade an 8753C to 6GHz by modify the EEPROM data, but haven't seen anyone push it further. Here's y I think there's a hard limit:

1. The directional coupler in 8753C was only designed to work well at lower frequencies
2. While you could technically use multipliers to get higher frequencies (signal loss isn't even the main issue here...)
3. The real showstopper: The RF diodes in the sampler simply can't handle higher frequencies anymore

If you want to hack a similar architecture VNA for higher frequencies, I'd recommend getting an hp8720

[hendorog]

I saw someone managed to upgrade an 8753C to 6GHz by modify the EEPROM data, but haven't seen anyone push it further. Here's y I think there's a hard limit:

1. The directional coupler in 8753C was only designed to work well at lower frequencies
2. While you could technically use multipliers to get higher frequencies (signal loss isn't even the main issue here...)
3. The real showstopper: The RF diodes in the sampler simply can't handle higher frequencies anymore

If you want to hack a similar architecture VNA for higher frequencies, I'd recommend getting an hp8720

The 8753C receivers work perfectly well up to 6GHz - they have to, as there is a 6 GHz software upgrade for the 8753C. The RF Out only goes to 3GHz, so the 85047A test set contains a doubler to get to 6GHz.

The directional coupler you have there looks like one from a 3GHz test set. Those do not work well above 3GHz.

Getting a 8753C working at 6GHz requires at a minimum.
* A frequency doubler. I made my own. (https://github.com/hendorog/8753_modules/tree/master/hw)
* The 006 option enabled in the 8753C (The simplest way is to execute the POKE command here: https://www.eevblog.com/forum/repair/repair-hp-8753c-network-analyzer/msg989541/#msg989541)
* A directional coupler, bridge or some other three port device. A splitter works at a pinch.
* Another three port device - ideally a splitter - to feed some signal back to the R input

[JoeyDeng]

What is real bandwidth of HP8753 Sampler? Does anybody measure sampler response at third harmonic of VNA source and hack firmware to expand frequency range?

He means the 3rd harmonic (3GHz*3=9GHz).

[hendorog]

What is real bandwidth of HP8753 Sampler? Does anybody measure sampler response at third harmonic of VNA source and hack firmware to expand frequency range?

He means the 3rd harmonic (3GHz*3=9GHz).

Ah gotcha

[khach]

Yes, I am looking for 9- 12 GHz operation mode of HP8753. Of course, S-parameter test set is completely different then original and directional couplers with wider bandwidth. At my HP8753C signal source is degrade by power and not excited 85047A internal freq doubler. Maybe, if  implement some active or PLL like doubler or tripler the possibility of extendimg freq range is good solution. Of course,  bandwidth of sampler harmonic generator SRD need check for higher harmonics present. And sampler schottky diode freq range is critical too.