Component Tester board, for SA-NA impedance caps inductors filters

[oz2cpu]

Here is my design shown
it is a Component Tester board, for Spectrum Analyzers or Network Analyzers
you can measure the impedance of capacitors inductors filters, what ever,
at any crasy frequency range your equipment can handle,
this pcb is tested from 0 to 3GHz :-)

explain video:
https://youtu.be/y79nOd8GLWI

[oz2cpu]

AVX 22201C106KAT2A

10uF 100V 2220 size
XC is 170mOhm at 100kHz
ESR is 0.9mOhm
SRF is 1.4MHz
attached is my DC voltage tests also.
and also my component test pcb for SRF using SA, this way i can test from kHz to GHz

[oz2cpu]

> Can you share more about the design? I noticed you included an attenuator at the in & outputs; I guess to get a better impedance match.

sure can, if you send me your email, i send you the gerber zip, so you can order this design,
here is a youtube link to a detailed explainer about this :

https://youtu.be/y79nOd8GLWI

i keep one with 0R on the centre, and then use that to calibrate the analyzer tracking generator, for a flat line,
now swap to the tester one, where an inductor or capacitor is mounted,
you can also use this to trim filter designs for best performance, see how different caps in parallel behave (parallel resonance)
the two 5dB attenuators each side of the DUT, is to keep the NA / SA-TG tester in good match,
and to feed the test area with a resistive load from both sides,
else you create resonance problems with the cable length, different cables would then create different results,
this methode is much better :-)
this PCB is flat to over 3GHz. i also design RF stuff, so i need to see how my parts perform.

[oz2cpu]

VERY old proto type,
this was just to measure SRF, but cable length do affect the results,
so i made the next one, with attenuators = Much better

[oz2cpu]

some of my experiments

[oz2cpu]

some experiments with low pass filter for a switchmode supply
just to show how power full a tool like this test pcb really is

[mawyatt]

Very nice Thomas 

These will allow folks to "see" the effects of frequency, temperature, voltage & current on various passive components. Should work well with the popular low cost nanoVNAs also!!

Thanks for posting.

Best,

[Hamelec]

Very interesting Thomas,
what is the thickness of these PCB boards?
to have a proper 50R Line...
thanks

[oz2cpu]

>what is the thickness of these PCB boards?
>to have a proper 50R Line...

1.6mm

[Kean]

I ordered 10 of the test PCBs via PCB Partner earlier this week for free and just US$1 shipping and they are already on their way by DHL
https://www.eevblog.com/forum/eda/pcbpartner-new-sponsor/

[oz2cpu]

ha ha how cool is that
the SMA connectors cost bit
but maybe cheap ebay types will also work fine,
please post some curves of experimental parts your perform investigations on
and tell us what you learned.

[nctnico]

I got the boards last week. I created some calibrators for VNA use:



For the 50 Ohm 'load' I used two 100 Ohm 0306 resistors which have the contacts on the side to reduce self inductance.



A while ago I got a bit of an epifany. I was doing some EMC pre-compliance testing and I got really fed up with screwing & unscrewing SMA connectors for connecting different probes and remembered that I had bought a bunch of SMA-SMB adapters so I started using those to have a plug & play solution. For this reason I have fitted the boards with edge mount SMB connectors. These are a bit harder to find but they are available.

[oz2cpu]

i come from designing telecom equipment, where SMA is the smallest we would use,
we would never use SMB, but i dont see why, it is not like they fall of in a hurricane,
ya ya on the best NA you might be able to see a tiny movement, when you touch the cable a bit,
I like the way SMB can turn so you dont get twisted cables, for many EMC test probes SMB is much smarter over SMA

try and test the decoubling capacitors you like to use the most :-)
and lets see if they perform well in the freq range you normally use them to,

[Jay_Diddy_B]

Hi,

I would like to thank oz2cpu for the inspiration to build a version of this project.
This is my version.

Schematic



I have 5dB attenuator at each of the board.

PCB


I got the boards made in a panel by JLCPCB. The result was a 3x3  array of boards, a total of 45 boards for $
I used low cost edge mount SMA connectors from eBay.

Calibration Fixture



I made a calibration fixture with a 1 resistor. If I calibrate with this the vertical scale on the VNA is db
I also made a board with a 1uF 1206 ceramic capacitor.


Results



This picture shows the impedance of the 1uF capacitor.

The marker is at -14.58 dB and approximately 1MHz.

-14.58 dB = 0.186

C = 1 / 2 x Pi x F x Xc  = 855nF


Here is the impedance of a Nichicon 100uF 100V VR(M) series electrolytic. The ESR is around 100m



Many thanks to oz2cpu.

Best regards,

Jay_Diddy_B

[oz2cpu]

ha ha cool Jay
but why ? the gerber files are free to dl and order from :-)
really no need to re-do this specially when you are adding errors, remember i have been designing GHz rf stuff my entire life for a living..
also : you placed R1 and R2 wrong, they must be on a straight line
I only use two parallel resistors on +3GHz stuff
the single resistor to gnd attenuators as i have shown on my pictures perform good enough for most people
but more important : your isolation distance to ground from your rf tracks are way too short, and will affect the impedance
this means you need to trim the rf track width
use this link
https://chemandy.com/calculators/coplanar-waveguide-with-ground-calculator.htm

more distance i have in be design, is smarter, this way there is much less manufacture tollerance from pcb to pcb
and also when playing with different parts to measure, less chance to short

[mawyatt]

Like the idea of using a 1 ohm reference calibration

Recall a typical 1206 has ~0.4pF residual capacitance, which shouldn't have much effect below a few 100MHz with a 1 ohm shunt.

Seems like a good experiment would be to test various resistors up to ~1K and see how well this 1 ohm calibration holds up

Also, another thanks to oz2cpu for the fixture design and files

Best,

[Jay_Diddy_B]

Hi,

I used the chemandy CPW calculator when I determined the size of the traces and the spacing.



I used a 2mm wide trace and a 0.4mm gap. This is very close to 50 given that the dielectric constant of the FR4 has a wide tolerance.

TDR measurements

I have a Tektronix 11801A with an SD-24 sampling head. This allows me to measure the impedance.

Test Board



The two ends of the test board where connected together with an 0805 zero Ohm resistor. There are discontinuities caused by the zero Ohm resistor.

The board was connected to the SD-24 with a Goreflex 50 cable.

The TDR was measured with the other end of the board terminated in a 50 termination and open.

TDR measurements



This with the board terminated. The impedance is 50.7  on the left. This is the Goretex cable. The impedance has some ripple going through the board and then becomes 50 when the signal arrives at the termination.
The board measures 46 at the location of the zero Ohm resistor.



With board unterminated the reflection from the open end can be seen.

Given I am using cheap connectors from eBay this is okay.

The signal should travel across the board 50mm in about 400ps, then the signal bounces back.

Jay_Diddy_B

[oz2cpu]

i too find it amazing how accurate the results are from cheap jlc :-)
yes a 2mm trace will solve this, so you did compensate it (of course !!)
you are not affright the short distance will short when you play a few times on the same pcb ?
I actually reuse mine many times, and will continue until the copper go loose

[Jay_Diddy_B]

you are not affright the short distance will short when you play a few times on the same pcb ?

In my day job, I regularly change QFN parts with pads on a 0.5mm , about 0.2mm spacing. I do most of soldering under a stereo microscope. I am not worried about shorts. They would show up during the testing and can easily be fixed.

By using the 'panel by JLCPCB' feature and the 3x3 array, I got 45 boards for $9.60 USD. I combined the order with some other boards to optimize the shipping costs.

I will probably recycle the SMA connectors and toss the PCBs if they become worn out.

Regards,

Jay_Diddy_B

[Jay_Diddy_B]

Like the idea of using a 1 ohm reference calibration

Recall a typical 1206 has ~0.4pF residual capacitance, which shouldn't have much effect below a few 100MHz with a 1 ohm shunt.

Snip ...

Best,

Hi,
I will share analysis of the 1 Ohm calibration method. I will illustrate the analysis using LTspice.

If we look at the baseline measurement circuit:



There is a 50  source on the left and a 50 receiver on the right. In between there are two attenuators with a nominal attenuation of 5dB each. The result is 10dB of attenuation.

Now if I introduce the 1 calibration resistor:



The output is reduced to -38.186dB.

This can be normalized to 0dB by increasing the source by 38.186dB:



Now if I replace the 1 calibration resistor with 0.1 :



The output becomes -19.694 dB

Which is 0.1036 a 3.6% error

If the DUT is 0.01



The output is -39.662dB
Which is 0.010397 a 4% error.

So this method is good for measuring low values of impedance.


Higher values of DUT

The circuit can be reduced to Thevenin equivalent like this:




This can be analyzed using DC:



This shows that the best performance is obtained if the DUT has impedance of 4 or less.

The series connection is better for testing higher impedances.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi,
The Norton equivalent circuit reveals that the error is asymptotic at 4% for value less than 1 .



When the DUT is 1 the error is 0$ as the resistance of the DUT reduces to 0 the error increases to 4%.

Regards,
Jay_Diddy_B

[mawyatt]

Thanks for the analysis, makes sense that the results would linger around the calibration value for best low shunt DUT results. Also make sense to use a series type DUT for higher values.

Using a calibration to 1 ohm seems like a good value for typical ESR type measurements, and having a scale in dB ohms is a plus

We just received our fixtures and awaiting the SMD resistors and SMA connectors, and also combined the order to save shipping. The other custom PCB is one that will allow our TH2830 LCR meter to use a high bias voltage, and now have the concept working with the Hioki IM3536

BTW when you ordered your test fixtures from JLCPCB as a panel did you have to cut the fixtures out?

Anyway, thanks for the additional information, and thanks to Thomas for starting this concept

Best,

[Jay_Diddy_B]

Snip ..

Using a calibration to 1 ohm seems like a good value for typical ESR type measurements, and having a scale in dB ohms is a plus

Snip ..


BTW when you ordered your test fixtures from JLCPCB as a panel did you have to cut the fixtures out?

Anyway, thanks for the additional information, and thanks to Thomas for starting this concept


Best,

When I ordered the PCBs in the 'panel by JLCPCB' format they come scored. They are easy to separate. I have used this service several times if I need a large quantity of a small board.

Jay_Diddy_B