OPA858DSGEVM: Not getting expected performance figures

Prodigy 80 points

Replies: 16

Views: 154

Part Number: OPA858DSGEVM

I've had in this evaluation board and a blank for testing the OPA858 and OPA855, and my bench tests have been returning consistent results, but not the results I was expecting from what I've gathered off the component datasheets.
Attached are some of my experiment results, along with a photograph of my test setup.

With both ICs, I have performed tests with the circuits relating to Figure 43 & 44 of the OPA858's Datasheet [Page 14], with an input signal of -30dBm. In the tests I've performed last week, the feedback resistor across the amplifier [R2] has been changed from 464ohms, up to 1K5ohms.
As a course of checking the setup, I've tried them with 2 different Network Analysers, and rebuilt the boards several times, including replacing the ICs, yet the results remain the same.

Is our test setup correct [see: attached photo], and are we measuring the right things?

 - To detail the test setup, we are using a Rhode & Schwarz ZNB 8 Vector Network Analyser, calibrated to TOSM with an SMA 50ohm calibration kit, sweeping from 10MHz to 4GHz, and measuring standard S-Parameters.

[spoiler]

[/spoiler]
Bench Test setup

[spoiler]

[/spoiler]
[As Figure 44. OPA858 Datasheet, Inverting configuration]

[spoiler]

[As Figure 43. OPA858 Datasheet, NonInverting configuration with R2 as 464ohm]
[/spoiler]

[spoiler]

[As Figure 43. OPA858 Datasheet, NonInverting configuration with R2 as 464ohm, using OPA855 instead]
[/spoiler]

16 Replies

  • Hello,

    For clarification, you changed all the resistor values on the EVM to match with figures 43 & 44 of the datasheet? What value are you using for RG?

    Best,

    Hasan Babiker

  • In reply to Hasan Babiker31:

    Yes, over the few months I've been working with these amplifiers, I've ultimately ended up replacing all of the resistors on the eval boards I've got.
    At present, the board's been reassembled with the nearest standard component values we have, so RG [R4] is 75ohm.
    The output match [R6 & R7] are set to 160ohm and 68ohm respectively.

  • In reply to user6274350:

    Hello,

    Note that the values chosen on the EVM & datasheet are chosen in a way to provide a 50-ohm matched input/output. This is clear on the non-inverting configuration input, however when using the EVM in an inverting configuration the input impedance of the circuit would be made up of resistors R3 & R4. This can be seen in figure 44 of the datasheet where the parallel combination of 64 & 220 ohms will be around 50-ohms. If you are using an RG of 75-ohms I would recommend adding a 150-ohm resistor in R3 when using the board in an inverting configuration. Also note that using values of 160 & 69 ohms will give you an output impedance of around 47.7 ohms. 

    Best,

    Hasan Babiker

  • In reply to Hasan Babiker31:

    I've changed R3 to 150ohm as you've suggested, and we've also received a fresh OPA858DSGEVM this morning, which I'm measuring for reference.

    With this new EVM, I'm getting an S21 of ~ -0.3dB up to ~400MHz, and by probing the input and output nets with an oscilloscope, seeing a voltage gain of 4.67x [~150mV to 700mV], both measurements done with an input of -10dBm.
    See attached: Results for both boards.

    Do these results look right? Are you able to provide us with your own test data?

    Edited Eval board as Inverting, R3 set to 150Ohm:

    Brand new OPA858DSGEVM:

    Also, as a side note, on Pg.22 of the OPA858 datasheet, where a design procedure is walked through, Equation 1 gives Noise Gain but concludes that, for that particular example, the Noise Gain is 2.79V/V, and equates that to 5.04dB, when 2.79x is equal to 4.45dB.
    Where has the extra 0.59dB come from? Is the Input referred voltage noise involved at all?

  • In reply to user6274350:

    Hello,

    These are not the results that I would expect to see from the EVM. Using a 150-ohm resistor for R3, you should be getting a voltage gain of -6V/V. At this gain you should be seeing greater peaking than what is shown on your plots. I will order an EVM and see if I get similar results.

    In regards to equation 1, it seems like this is a mistake. The gain should be labeled as 8.9dB, however it seems that the conversion was done for a voltage gain of 1.79 rather than 2.79. Thanks for bringing this up and I will look into updating it on the datasheet. 

    Best,

    Hasan Babiker

  • In reply to Hasan Babiker31:

    Thank you for the feedback. I look forward to seeing what you find.

    Something I've found myself, having done a simple maths exercise with the EVM schematic, is that the overall gain of ~-0.3dB seems to actually be correct.
    See: attached

    This basic schematic is effectively the same as that for Figure.1 of the EVM's datasheet [Page.3].
    Without the inclusion of any parasitic impedances, I worked out the supposed gain to be -0.26dB.
    (1V input signal was chosen for ease of numbers).

  • In reply to user6274350:

    Hello,

    My main concern was with the amount of peaking that was being shown in your plots. Especially when it came to the brand new EVM. So I tested the board and was able to achieve the same amount of peaking that was shown on the datasheet. From your picture, it looks like you are using considerably long wires which might be having an affect on your results. Can you try the test with the new EVM again with shorter wires, and also make sure that those wires are being calibrated as well? Also how much current is being drawn from your power supply?

    Best,

    Hasan Babiker

  • In reply to Hasan Babiker31:

    I've remeasured the Eval board on a different analyser, and with the best cables we've got. The results are vastly better now.
    And with +/- 2.5V, the current drawn is 20mA.

    Admittedly, I had been using our generic SMA cables that get used a lot. I'm amazed at the difference!

  • In reply to user6274350:

    Glad to hear that it fixed the problem!

    Best,

    Hasan Babiker

  • In reply to Hasan Babiker31:

    I'm afraid that's only solved part of the problem; I'm still unable to corroborate the results I'm seeing with any maths I have on TIAs. The most glaring mismatch being the measured bandwidth with the GBW product quoted on the 858's datasheet.

    You've implied that the results I'm now seeing are correct, but can you confirm that they're the same or close to what you've measured?
    And can you provide us with any mathematical walkthrough of the Eval board?

    For reference, I've been using the following resources:

    TIA Bandwidth & Noise Calculator, 0508.TIA Calculator - Samir Cherian, E2E Amplifiers Forum
    TIAs: What OpAmp Bandwidth do I need? - John Caldwell, E2E Precision Hub archive
    What you need to know about TIAs - Samir Cherian, Analog technical articles
    TIA Analysis - Erik Margan