THS3491: Frequency response performance

Stefan Lintner

Part Number: THS3491

Hi team,

a customer of mine is interested in the frequency response performance of the THS3491RGT. When testing the EVM the measured frequency response does not align with the specs from the datasheet. Here is the setup:

Agilent E5061A Networkanalyzer
2x 1m long RG58 cables with an N and SMA connector each.
Meaurement is calibrated regarding S11 and S2 -> cables should not affect the measurement
+-15V power supply to EVM
Input level is -20dBm to measure small signal response

Attached you find the values of S11 and S21 from 300kHz to 1.5GHz. You can see that at 300 kHz the value of S21 aligns with the specs from the datasheet but decreases strongly at higher frequencies.

Also the datasheet only mentions typical values for the frequency response. Are there any upper/lower limits that are guaranteed to be met? The customer would like to use the part up to 450MHz.

Thank you for your help!

BR,

Stefan

Evaluation Module THS3491 RGT at Pin = -20dBm.xlsx

over 5 years ago

0 kai klaas69 over 5 years ago

Guru 140200 points

Hi Stefan,

have you read section 2.2 of the EVM manual?

www.ti.com/.../ths3491rgtevm

Maybe the resistors of the output resistor network have some manufacturing tolerances? For a small signal response measurement you could replace the output resistor network by a precise 50R resistor, as mentioned in section 2.2 of EVM.

Kai

0 Sean Cashin over 5 years ago in reply to kai klaas69

TI__Genius 13996 points

Hi Stefan,

Without even considering the device, there is clearly a severe impedance mismatch causing 70% reflections back to the input. There are only 3 possibilities: incorrect source impedance; incorrect cable characteristic impedance; and incorrect termination resistance on the board. I see that the your network analyzer has a 75 mode, I would verify it is truly set for 50 ohms. Then, I would switch the cables with shorter ones until I am confident they are not a problem, and verify they too are 50 ohms. Finally, I would verify that the 50 ohm resistor on the board is connected and not damaged.

Best regards,

Sean

0 Stefan Lintner over 5 years ago in reply to Sean Cashin

TI__Expert 4235 points

Hi Sean, kai,

thanks for your quick responses! There is only one resistor in the output network that has a tolerance of 5%, all the others have 1%. My customer will replace this and test it.

Regarding your recommendation sean, here is some additional info from my customer:

Agilent E5061A is a 50 Ohm Network Analyzer
RG58 C/U cables are from Huber und Suhner (See datasheet) and when connecting a 50 Ohm termination resistor there is 30dB alignment (20*log|S11| <-30dB) until 1.5GHz
S11 was measured with One-Port calibration of E5061A and the calibration kit 85032E from Agilent at the end of the RG58 cable together with an adapter SMA f to N f .
Attached is also an excel file of the input admittance (in Siemens) at the SMA connector of the EVM. At 450MHz it is (0.019696 + j0.010975)S which results in an absolute value of the reflection factor |r|=0.26658 or 20log|S11| = -11.483dB. This means that only r2 = 0.07106 = 7.106% are reflected which is an attenuation of 0.32dB.
This does not explain the decrease in gain at 450MHz from -13.121dB to -14.63dB which is 1.509dB

Thanks again for your help!

Evaluation Module THS3491 RGT S11 -Yin.xlsx 22510015_dataSheet.pdf

BR,
Stefan

0 kai klaas69 over 5 years ago in reply to Stefan Lintner

Guru 140200 points

Hi Stefan,

the datasheet of your RG58C/U cable says that the maximum operating frequency is 1GHz. So, this cable is not good enough to make measurements up to 1.5GHz.

You should take a really good SMA cable with a much higher frequency range:

www.thorlabs.com/newgrouppage9.cfm

Kai

0 Stefan Lintner over 5 years ago in reply to kai klaas69

TI__Expert 4235 points

Hi kai, Sean,

here an update from my customer:

Now I have removed the attenuator at the output of the evaluation board. First I measured the resistor values with the Fluke 187:

R9 = 239.1 Ohm               -0.38%
R10 = 239.2 Ohm             -0.33%
R11 = 286.2 Ohm             -0.28%
R12 = 286.5 Ohm             -0.17%
R13 = 55.0 Ohm                +0.18%

The resistors are very precise and therefore I can’t believe, that the tolerances of the resistors are the reason for the different frequency response compared to the datasheet.

Then I have replaced R11 by 27Ohm 1206 and R12 by 22Ohm 1206.
The other resistors are removed (dnp = do not place).
In sum the output resistance is 49 Ohm, which is near 50 Ohm.
It was very interesting, that the frequency response got much better.
I have added the frequency response of this changed evaluation board in an EXCEL sheet.

Evaluation Module THS3491 RGT without attenuator R11=27 R12=22.xlsx

I suppose, that the problem is the capacitance of the PCB in the range between the two 287Ohm resistors, which build a lowpass filter.
So if you would like to improve the evaluation board, put a 50Ohm output resistance and add a standard 50Ohm attenuator afterwards, which has a good frequency response at RF frequencies.

Concerning the RG58 cable, which is only specified up to 1GHz, I'm only interested in the results up to 1GHz. Further I don't believe, that the cable makes any problems in the range of 1GHz to 1.5GHz, since I correct every error of the cables by calibration of the network analyser including these cables.

I believe, this solves the whole problem and you can close this topic.

***************************************************************

However I would still be interested if there are any upper and lower limits for the frequency response?

BR,

Stefan

Amplifiers

Amplifiers forum

THS3491: Frequency response performance