binding post to BNC impedance analyzer test fixture?

[coppercone2]

So if you have a impedance analyzer and you can connect a BNC to banana binding post adapter to the measurement point, what frequency is this useful for?

I calibrated the impedance analyzer using a BNC short, BNC open (dust cover) and BNC 50 ohm load. When I attached the dual binding post to it, it read a capacitance of 4pF roughly when open circuit. When I added a 2pF capacitor to it, it read 6pF.

So I can easily stick a piece of braid as a short circuit (or maybe a cutout of copper sheet that conforms nicely around the binding posts like a shorting jumper on old HP equipment), leave it open for open circuit, then put say a 1/8 watt leaded resistor between the posts to get a '50 ohm' calibration. I have not tried this yet.

But anyway, any conjecture on when this stops being accurate (or gets highly inaccurate)? I measured a 330pF 1% capacitor on it and it measured correctly. I am using a old HP BNC to binding post/banana adapter. I wanted to know the theory of what starts going wrong as you increase frequency with this setup.. (its for a 4191A impedance analyzer with bandwidth of 1MHz to 1GHz). I also cross tested the APC-7 cal kit with the crappy BNC 50 ohm load and it seems to be within a few percent, and I am using a direct APC-7 to BNC adapter. It seems like a good enough idea for leaded parts...

I wanted to know if there is some kind of debilitating problem before I try ordering some precision through hole components that maybe have some frequency data to test on it (might save me a few dollars to post this).

[Wallace Gasiewicz]

I don't know what sort of fixture or connector you are using, Are you using a APC to BNC or do  you have the HP fixture?
They had several fixtures for the 4191 A, but also the manual states that the APC7 could be used for a user built test fixture.
There was even a "probe" 
Their "spring clip" adapter advertised good to 500 MHz not 1 GHz  Binding post adapter good to 250 MHz  More practical HP binding post adapter good to 125 MHz See page  3-34 and 3-35 in Manual.
I suppose you have it but here it is anyway (maybe for the ease of others reading this)

https://www.keysight.com/us/en/assets/9018-05401/user-manuals/9018-05401.pdf?success=true

I suppose you could make a decent test fixture that accepts Banana and calibrate out most of the error. Probably much less than 100 MHz.
Maybe you could make a binding post design that would work at 100 MHz reasonably. Theirs uses a "guard"
Just guessing about the 100 MHz.
BNC  open short load stuff is all over the map and some are not very good. I have a bunch of brand name 50 ohm loads that are not good to 100 MHz.

If you find a reasonable good source for high quality "standards" for caps and inductors and resistors, let me know,

Wally

[coppercone2]

Well I notice that those fixtures attempt to reduce inductance as much as possible (loop area?)

is this because there is complex coupling between the DUT and the fixture that you can't calibrate out easily since a fixture with a higher inductance has a higher loop area/more radiation/coupling (since its starting to be treated as distributed analysis)? (i.e. you do a 50 ohm calibration but when you stick a 500 ohm resistor in there the turns count on the coil is so high it starts working with the fixture to do complex/nonlinear things so the results are useless because the part interacts with a test fixture in a way that is not calibrated?)

its mad annoying that those parts are mad $ for something that looks like some plates lol, I wanted to know why I need to invest so much. You would at first glance think you can calibrate everything out. I am scared of the news groups because those people are like primal engineers with venom sacks still attached (TNG: Genesis, S07E19), it looks like a rough place

but this machine does look like it fell off the enterprise

[joeqsmith]

Some time ago I used the NanoVNA ($50) to attempt to analyze a power distribution network.  During part of the demonstration,  I used a breadboard and plugged in some components to form a simple RLC network.  I left the long leads on the parts and ran a sweep to 50MHz.   I then repeated the test with the component leads cut short.   

It's not exactly what you are asking but it may provide you some idea of what to expect. 

[Wallace Gasiewicz]

I only know enough to be dangerous.
I think the test fixtures are designed with the same sort of idea in mind as designing COAX. in that the impedance is constant over a big freq range.

I know there are fixtures for the four connector LCR meter fixtures available from Tonghui and I have one that works fine at 1 MHz. It does not look that complicated to make one and they did it for far less than the used HP ones are going for.  I do not know any sources for other test fixtures for APC connectors.

At the prices they have on e pay for the HP APC fixtures, I think you could get a machine shop to make a few...if you had the APC of course. Or design a board and get some boards made and have nice connectors made from solid silver, it would seem to be cheaper.

I wonder if anyone has just replaced the APC with a different system, I do not know. Apparently all the connections are made to the one APC connector internally, making the stray impedances less but making connection more critical, I guess...

[coppercone2]

BTW I highly recommend making racks because then all those 'pains' that people complain about with older equipment go away. All you need is a cable storage bin to make any connection that you want. It might save you money in the long run. Good wheels too to make it easy to move around.

I was able to do a similar experiment with my banana adapter. You take the capacitor with long leads and let it stick out of the binding post adapter, take a measurement, and then you stick it  in so the capacitor is closer to the binding post with the leads hanging out the other end. The resonance point does move around.

I think I will order a few cheap but good parts and perform measurements that have the part soldered in a connector like a terminator and compare it to the part in the banana adapter. Actually I might have a eBay leaded inductor kit that I got a decade ago that had some parts that IIRC had quite a high operating frequency range in the VLF band to test this, I think I was waiting for more then a decade to find a use for that lol

https://en.maritex.com.pl/product/attachment/90429/749fb5a97de04985f84d9d6a0b8a1e80

*SRF of inductor is either 180Mhz or 200Mhz minimum, I don't have the part label but I think its the 0.9uH from the datasheet, the damn label on the paper got lost

[joeqsmith]

I had forgotten that I also show the same circuit built up on a PCB with surface mounted parts which further improved the performance (shocking!!  ).

You could maybe make up a custom fixture easy enough but no harm just starting out with the BNC to banana adapters.  For what you are trying to do, these may prove good enough. 

[coppercone2]

Well I put a strip of copper sheet in there for a short, left it open as a open and put a 51 ohm metal film resistor as a load, in order to do the calibration in the adapter rather then with the BNC standards that made the fixture read 4 pF.

Now the fixture reads 0.06pF after calibration, and the difference is

1) calibrated with BNC short, BNC dust cover and cheap 50Ohm BNC load that seems to agree with a APC-7 cal kit BNC load OKish.
inductor reads ~880nH at 1MHz, and 1205nH at 55 MHz

2) calibrated with copper strip, resistor and nothing
inductor reads 885nH at 1MHz and 905nH at 55MHz.

So the difference between the two calibrations (one that 'compensates' for the fixture is a difference of 425nH and 20nH with a difference of  54MHz. The inductor claims to have a min SRF @ ~200MHz. Defiantly a big difference, but there is still too many unknowns since the calibration resistor is just a random cheap 1/4 inch metal film with the thin leads lol, but it certainly looks more linear given the high SRF of the inductor.

so 2% change vs 31% change with same parts. If that inductor is really as stable as they say with a high SRF, I am inclined to trust the crappy calibration then the coaxial calibration?


What would be a nice through hole resistor to attempt to play with this fixture more that has good RF parameters.. it needs to be 50Ohm. I would prefer a leaded part and not to solder leads to a SMD part (i hate this). I also need to measure out a thicker copper sheet to act as a shorting standard because I used like copper foil and it was pretty dodgily cut (it reads like 20-30miliohms after cal when you put it back in, the shorting BNC (amphenol) reads approx 15-20 miliohms) at 1MHz.

[coppercone2]

Unless there happens to be a long RF SMD resistor that fits between the binding posts naturally (so like a 1/2 inch long 50 ohm resistor that is SMD would be acceptable, since thats not that different from the metal plate, if a wired one is not good enough.

[coppercone2]

It looks like the SRF is 304.5MHz (from ind to cap), the inductor must have been a 5% 1uH inductor with a datasheet minimum SRF of 200MHz. The De5000 shows 1.0uH @ 1KHz

[TimFox]

Unless there happens to be a long RF SMD resistor that fits between the binding posts naturally (so like a 1/2 inch long 50 ohm resistor that is SMD would be acceptable, since thats not that different from the metal plate, if a wired one is not good enough.

In the 1990s, I used an -hp- 8753A network analyzer (with reflection bridge) or an -hp- 4191A impedance analyzer (both of which had APC7 connectors at the test port) at frequencies below 100 MHz. 
Specifically, I needed to measure balanced impedances substantially above 50 ohms;  the following system worked well.
We made test fixtures for different ranges, using inexpensive MCL transformers (in flattened-DIP packages) appropriate to the frequency range, terminated in a female connector with 0.200" spacing that matched the MCL package. 
The mating connector was a standard 0.200" male header (with the center pin of three 0.100" spaced pins removed). 
We calibrated the system with simple male headers: either open (nothing connected), short (copper foil between pins), or the appropriate resistor (1206 SMT) at the higher resistance.
Below 10 MHz, we went as high as 13 k, the design value of one MCL transformer.  The 8753A firmware allowed us to set the characteristic impedance, but we had to cheat and tell it we calibrated to 130 ohms, due to firmware limit.
I don't remember if we soldered the 50 ohm end of the transformer to an N or BNC connector, which connected through an APC7 adapter. 
Since the connector and adapter were before the reference plane (0.200" male), the errors were taken out by calibration.

[coppercone2]

I think I Happen to have some chokes/transformer in DIP package (for network cards), IIRC the frequency range was in the area /past the area you are saying. Measuring balanced impedances as actually my next target BTW

[coppercone2]

What do you recommend I measure with a balanced impedance that I could know the system is working properly? Are there any standard objects that might be in the lab that fit the bill? A piece of network cable or something?

[coppercone2]

Also, what BNC did you solder the transformer to? I assume it has to be a PCB  BNC right. A BNCM PCB mount that goes to the mcl that goes to the dip f? I don't think you had a BNCm to pc7, their usually BNCF...? I don't think you used the chassis mount versions right... the body is too hard to solder to without effecting the seals. This sounds like kind of fun

And I gotta say it sounds maybe tempting to try this with a twinax instead of headers too.

[TimFox]

I would try a series of 1206 resistors, of measured resistance, which should be reasonably non-reactive.
When measuring self-resonant frequency of capacitors, I found that one could assume an equivalent series inductance that depended only on the total length of the "capacitor" in circuit.  Specifically, with roughly 0.75" total length, a set of capacitors between 1 nF and 470 nF had approximately 10 to 15 nH ESL, regardless of capacitor type (disc ceramic, multilayer ceramic, or even polystyrene foil).
(These measurements were all made at roughly 10 MHz for a specific project.)

[TimFox]

Also, what BNC did you solder the transformer to? I assume it has to be a PCB  BNC right. A BNCM PCB mount that goes to the mcl that goes to the dip f? I don't think you had a BNCm to pc7, their usually BNCF...? I don't think you used the chassis mount versions right... the body is too hard to solder to without effecting the seals. This sounds like kind of fun

And I gotta say it sounds maybe tempting to try this with a twinax instead of headers too.

It's been a long time, but I believe we soldered the primary leads to a chassis N female connector, then used appropriate adaptors to the APC7.  These are all before the reference plane, and calibrated out.

[coppercone2]

How do you solder the body of the N? That is so big, it makes me want to use a blow torch. Whenever I solder the body of a BNC (let alone a N) the dielectric seems to over heat, and BNC have their rubber gasket expand and make them tough to connect.

PC mount has little studs that you can easily solder to that come out (I guess its a bit of extra inductance). I made like 5 or 6 different things soldering BNC connector bodies, and I would say 5/6 have toasted dielectric. So for low frequencies I switched to 75ohm board mount BNC, the dielectric is teflon so its harder to over heat and you can easily solder to the legs of the connector that are supposed to go to the PC board (this is the easy way to add a pipe to a connector with out a forming tap, if you don't mind a small gap in the shield). Maybe if you spot weld a brass*tube segment (thin) to the N body so it has a elevated ring you can solder that easily...

[TimFox]

There are insulated BNC panel females with 2 solder cups side by side.  https://www.digikey.com/en/products/detail/amphenol-rf/031-10-RFXG1/2643384?utm_adgroup=Coaxial%20Connectors%20%28RF%29&utm_source=google&utm_medium=cpc&utm_campaign=Shopping_Product_Connectors%2C%20Interconnects&utm_term=&utm_content=Coaxial%20Connectors%20%28RF%29&gclid=EAIaIQobChMI_O6OgpDo-QIVnwqtBh005Q4bEAQYASABEgKzYPD_BwE
A 4-mounting-hole panel N female can take a small solder lug (4-40 screw size) at one of the mounting holes.
A normal soldering iron works with the flattened DIP leads of the transformer into the solder cups or lug.
I'm referring to panel connectors (with female center connection), not cable connectors.

[coppercone2]

ah ok you put a solder post there, that could work nicely. Or you can solder a post to a screw and then thread the hole if you want also.

I have some of the isolated BNC connectors but if you look how they are built, the area where you solder to looks like it has more inductance, the one I have here right now looks like there is a metal shim going down a few MM in to the body, where the chassis mount one its a coaxial right up to the level of the dielectric, so theoretically you can shorter the center pin and if you can solder to the body its going to be lower inductance. I guess it does not really matter too much, and manufacturability and durability is more important.

[coppercone2]

Actually I have a bunch of old waveguide to 3.5mm adapters I found in the trash (like 30 of them) a decade ago. The distance between the center pin and screw hole of the 4x flange is almost perfect for a DIP chip. I wonder if I can hammer a copper wire into the hole like a rivet to make a 'surface mount' solder point for the transformer.

[TimFox]

ah ok you put a solder post there, that could work nicely. Or you can solder a post to a screw and then thread the hole if you want also.

I have some of the isolated BNC connectors but if you look how they are built, the area where you solder to looks like it has more inductance, the one I have here right now looks like there is a metal shim going down a few MM in to the body, where the chassis mount one its a coaxial right up to the level of the dielectric, so theoretically you can shorter the center pin and if you can solder to the body its going to be lower inductance. I guess it does not really matter too much, and manufacturability and durability is more important.

Again, if you apply the open/short/load standards after the transformer secondary, a reasonable amount of stray/unintended impedance before the primary will be removed by the calibration.  I am just telling you what worked for my application 20-30 years ago as a suggestion.

[coppercone2]

well I guess that might start to answer the main question of this thread, which is how much can you cancel before something bad happens. It seems to be at least the amount of a isolated BNC connector and header pins inductance for under 100MHz.

[TimFox]

Using a reflection bridge method (4191 or 8753) will give you the impedance at the plane where you placed the standards.  The BNC is before that plane, extra leads can be after that plane.
To check the method, I suggest trying two 1206 resistors, one 3x lower and one 3x higher than the “load” resistor used in calibration.

[coppercone2]

well I wonder what would happen if you put a meter of romex before the calibration plane. What I am asking is how bad can it get before the calibration does not work (assuming you have enough power for the measurement to over come pure attenuation). Like is HP making those crazy test fixture just to run the output amplifier by a few dBm lower.. or is there a nonlinear error of some kind that manifests itself when it gets bad. Like I could imagine it working after a very long piece of coax, but what if its downright improper. Is interference all you need to worry about, or is there self interference between the DUT and the test fixture based on the properties of the DUT that are different then the load (i.e. like coils).

[TimFox]

HP has described the process in their manuals and app notes.  If there’s too much crap before the reference plane, the error can be too large for the calibration process to yield an accurate correction.
As I remember from the 4191A manual, if you needed to extend the plane far from the APC7 connector through a length of real coax, not Romex, they have a couple of SMA (actually APC3.5) connectors where you should add a corresponding length of 50 ohm transmission line.  This is all quantitative, not handwaving.
Mathematically, this is a “linear” problem, but subtracting large numbers from each other may not give an accurate result.