The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Measurements >> RF Measurements >> S-parameter measurement
https://designers-guide.org/forum/YaBB.pl?num=1280129455

Message started by aaron_do on Jul 26^th, 2010, 12:30am

Title: S-parameter measurement
Post by aaron_do on Jul 26^th, 2010, 12:30am

Hi all,

I have designed a LNA and down-conversion mixer, and I added a GSG probing point at the output node of the LNA for on-wafer measurement. The LNA output impedance is approximately 2 kohm and it roughly matches the mixer input impedance of 2 kohm. I'm only quoting the real part of the impedances. So anyway, in simulation, I turned off the mixer (which made the input impedance infinite), and ran 2-port s-parameter simulations on the LNA, with port 1 having a reference impedance of 50 ohm and port 2 having a reference impedance of 2 kohm. My thinking was that I would be able to run actual measurements on the LNA using 50-ohm input and output ports, and then convert them to s-parameters with 50-ohm and 2kohm port impedances.

When I ran the measurements using a vector network analyzer, the magnitude of the s-parameters were apprpoximately as expected (S11 and S21 in dB), but when I convert them to s-parameters with 50-ohm and 2-kohm port impedances, the magnitude of S21 is way off (S11 is correct but S21 is totally wrong). My only explanation is that there is considerable inaccuracy in the delay measurement of the VNA, or perhaps the calibration is wrong. One problem is that I was only able to calibrate up to the end of the cables connecting to the probes, rather than the probe tips themselves, but i don't know if it would affect the measurements that much.

So my question is, should I expect such inaccuracy with the delay measurement of the VNA, or perhaps does anybody have another explanation for the measurement discrepancy? I have yet to measure the rest of the front-end. Also, how much uncertainty is added by only calibrating up to the end of the cables rather than the probe tips? Or is the problem due to the large mismatch between the actual 50-ohm output port, and the desired 2 kohm output port impedance?

thanks,
Aaron

Title: Re: S-parameter measurement
Post by rfcooltools.com on Jul 27^th, 2010, 9:42pm

Aaron,

All s-parameter files are referenced to a homogeneous port impedance . the s-parameters referenced to 50 Ohms can be exercised with any port impedance and give the correct result when they interact to any port.
http://rfcooltools.com

Title: Re: S-parameter measurement
Post by aaron_do on Jul 28^th, 2010, 11:43pm

thanks for the reply,

but I didn't quite understand it. Do you mean to say that I should be able to convert the S-parameters to any port impedance and get the correct result? If this is true, then the only reason why the conversion gave strange results would be because I was not able to calibrate my system up to the probe tips (for on-wafer measurement).

However, I suspect that the more mismatch you have between the actual port impedance and the desired port impedance, the greater the inaccuracy of the conversion. Is this not true?

I have taken measurements at the output of the mixer, and although there is some discrepancy with simulation, it is not as bad as what was suggested by my LNA's 50-ohm to 2-kohm conversion. I guess I will try and see if there is a way to calibrate up to the probe tips.

thanks,
Aaron

Title: Re: S-parameter measurement
Post by rfcooltools.com on Jul 29^th, 2010, 10:08am

Aaron_do,

I am guessing that the measured s-parameter file is mostly correct, but the conversion to 2k is where I believe the problem lies. Why and how are you doing this?

What I meant to convey is leave the s-parameter file at 50 Ohms. Then in your simulator put what ever port you wish to connect to the s-parameter file and then look at the results. Are they correct?

http://rfcooltools.com

Title: Re: S-parameter measurement
Post by aaron_do on Jul 31^st, 2010, 8:58pm

Hi,

yeah that's what i did. The conversion is correct since I got the same results by doing the conversion from within the VNA. Only thing i'm wondering is if this method is valid for such a huge discrepancy between the actual port2 impedance (50 ohm) and the desired impedance (2kohm). I guess most likely the reason is because I could only calibrate the system up to the cable ends rather than the wafer probe tips. There is probably around 3-5cm of unaccounted-for transmission line which at 2.5 GHz may be ruining the accuracy of the conversion.

thanks,
Aaron

Title: Re: S-parameter measurement
Post by rfcooltools.com on Jul 31^st, 2010, 10:21pm

Aaron_do,

Depending on the simulator, you can add a t-line with a negative length in series with the port in question this seems undo the un-embedded length of trace. this seems to work in ADS and Cadence's spectre as long as the Q on each end isn't large.

http://rfcooltools.com

Title: Re: S-parameter measurement
Post by aaron_do on Aug 3^rd, 2010, 7:46am

I was wondering if something like that would be possible. I'll give it a go.

thanks,
Aaron

Title: Re: S-parameter measurement
Post by rfcooltools.com on Aug 3^rd, 2010, 2:33pm

Aaron_do,

In the past I have used negative inductance and negative capacitance to try to de-embed other non-tline structures.
where for reference:
L~=length*(ln(2*d/a))*(uo/2*pi)

a: Wire radius (for a trace Width=Thickness then a = Width/sqrt(pi))
d: Distance, d ≥ a
l: Length
uo: 4*pi*1e-7

usually L~= 1nH/(mm length)

http://rfcooltools.com

Title: Re: S-parameter measurement
Post by vp1953 on Aug 5^th, 2010, 11:29am

Hi Aaron_do,

Why not calibrate to the chip of the wafer probe using a ceramic calibration standard? These are expensive but I think less expensive than the probes.

Sometimes the connector between the cable and the probes have loss, and these cannot be de-embedded by a cable length change.

Title: Re: S-parameter measurement
Post by vp1953 on Aug 5^th, 2010, 11:31am

HI Aaron_do

>The LNA output impedance is approximately 2 kohm and it roughly
>matches the mixer input impedance of 2 kohm

Were they conjugate matched?

Title: Re: S-parameter measurement
Post by pancho_hideboo on Aug 6^th, 2010, 8:19am

aaron_do wrote on Jul 26^th, 2010, 12:30am:

One problem is that I was only able to calibrate up to the end of the cables connecting to the probes, rather than the probe tips themselves,

Use "Port Extension" of Network Analyzer at least, although it is not accurate.