Announcement

Collapse
No announcement yet.

MPP preamp is vastly improvable

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • MPP preamp is vastly improvable

    and it just involves adding 3 components (plus two pairs of bypass C's and R's = 7 components).




    The value of the damping resistor Rdamp is R1 in parallel with R12 (R1 || R12). R1 is hardwired to GND, but R12 is wired to the virtual ground at the (-) input of the NE5532 op amp.

    The problem is that the NE5532 cannot enforce the virtual ground at the critical moment when the coil decay goes below the ON voltage of D5.. Even though the slew rate of the op amp is rated to 13V/us, the 33 gain brings it down to 0.4 V/us The actual signal changes at 2.5V/us or above.

    While the op-amp tries to establish a virtual ground R12 is "floating" and not contributing to damping the coil's energy, causing ringing (oscillations). The virtual ground has to be set as soon as possible. In the MPP, the practical value of R1 || R12 that avoids oscillations is much lower than the ideal Rdamp (sqrt(L/4C)) because R12 is floating for too long. A lower Rdamp results in a much longer sampling delay than would be possible.

    The solution is to add a low-gain stage at the preamp's input that can cope with the signal's slew rate. The low gain (typically 1 - 3) preserves more of the op-amp's slew rate. The NE5532 is designed for a gain > 3, otherwise it's not stable (it oscillates). It also should have a lower input noise than the NE5532 because it will add to the preamp chain. However, the extra noise is compensated for by the ability to sample much earlier, where the signal is stronger and the S/N ratio much higher. I would suggest an LM4562 (2.7nV noise, 55MHz, 20Vus slew rate) for this supplementary stage.

    To ilustrate the difference between the MPP and a properly design preamp I've run two simulations showing the state of the virtual ground in both cases.

    Simulated TX circuit:

    - Coil inductance: 500uH
    - Coil capacitance: 200pF
    - Critical damping resistance: 790 Ohm.

    R1 = 3K6 Ohm, in parrallel with R12 (1K Ohm) gives 782 Ohm, which is a bit below critical damping.

    The behaviour of the MPP virtual ground is shown in red, the preamp with the additional stage in blue:



    We see that the MPP oscillates at Rdamp, but the modified MPP does not. Also, the response of the preamp is much faster since the virtual ground is quickly established.

    In order to avoid oscillation, the MPP requires R1 || R13 to be much lower than Rdamp, in this case about 470 Ohm (R1 = 900, R12 = 1000). The sampling delay increases by a whooping 8us, and, for the same TX current, the transient peak goes down from 500V to 330V.




    The proposed preamplifier for the MPP will be in the next post.
    Attached Files

  • #2
    Proposed preamplifier:



    The first stage quickly establishes the virtual ground, solving the problem.

    R2 can be increased up to 3K, in which case R11 = 18K and R1 = 17K in order to achive a 1000 x gain. This way the bandwidth increases by 70% (from 200KHz to 350KHz).

    Though TLC081 has been used in the sim, any unity-gain stable op amp with noise < 4nv,slew rate > 13V/us and BW >= 10MHz can be used instead, as long as it can operate with +-5V.
    Attached Files

    Comment


    • #3
      A modified preamp + integrator.

      The EF is in fact an offset to the input, so instead of a differential integrator we can cancel the offset at the preamp (bye bye to matched components!). JFET op amps required for doing this (TLC081/082 recommended as U3a,b and U11).
      U2a,b should be low noise, high slew rate and ample phase margin (unity gain stable). The LM4562 foots the bill. Do not use NE5532 because it's not stable enough at low gain.
      The clamp diode D12 is necessary to achieve virtual ground as fast as possible.
      C4 may even not be necessary, since offsets are cancelled by Q3 and the TLC08x parts have very low offset.




      The relevant section in the MPP is the following:


      Attached Files

      Comment


      • #4
        I made a mistake, U2b should be non-inverting.

        Attached Files

        Comment


        • #5
          Great contribution Teleno, thank you.

          What sort of image is hidden here (not visible for me):

          Comment


          • #6
            Is it not time to also start removing the 1K input resistor (moodz auto blocking mosfet) or it will be too much improvements.

            Comment


            • #7
              Originally posted by WM6 View Post
              Great contribution Teleno, thank you.

              What sort of image is hidden here (not visible for me):

              It was the relevant section in the MPP which you can see in the post that followed. Somehow it got deleted by chance.

              Comment


              • #8
                Originally posted by eclipse View Post
                Is it not time to also start removing the 1K input resistor (moodz auto blocking mosfet) or it will be too much improvements.
                The 1K input resistor is absolutely necessary and is part of the damping resistor (in parallel with R1, to be determined by the user according to the coil).

                You can try to make it work with a blocking MOSFET but I can tell you that biasing it requires control (variable with temp.) and the signal will be underdamped at the end of the transient. Plus it adds 800pF to 10nF "improvement" capacitance to the coil. I'm just trying to make the best of the MPP while keeping it simple.

                Comment


                • #9

                  Comment


                  • #10
                    Originally posted by eclipse View Post
                    This is the modded MMP with the same coil on 1us and 0.5us targets:



                    Tried the autoswitch already, it doesn't get even close to seeing the targets.
                    Attached Files

                    Comment


                    • #11
                      Ouch didn't expect that. Thanks for the feedback on this I was going to try it it on a real board.

                      Comment


                      • #12
                        We could name this version as "MPP Gold"

                        The timing logic will have to be modded to allow 5us dampling delays or even less.

                        Comment


                        • #13
                          Originally posted by Teleno View Post
                          This is the modded MMP with the same coil on 1us and 0.5us targets:



                          Tried the autoswitch already, it doesn't get even close to seeing the targets.
                          Are you charting TP3 reply#4? What did you use for coil inductance and capacitance?

                          Comment


                          • #14
                            Originally posted by green View Post
                            Are you charting TP3 reply#4? What did you use for coil inductance and capacitance?
                            Yes. Simulation file in the attached zip.
                            Attached Files

                            Comment


                            • #15
                              Originally posted by Teleno View Post
                              Yes. Simulation file in the attached zip.
                              thanks

                              Comment

                              Working...
                              X