This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TPS546D24A: Noise from TPS546 crossing to other power domains

Part Number: TPS546D24A

Hi,

Related E2E case: (TPS546D24A: Noisy VR, need to verify design and hw/sw configuration, link: https://e2e.ti.com/support/power-management-group/power-management/f/power-management-forum/1305420/tps546d24a-noisy-vr-need-to-verify-design-and-hw-sw-configuration)

We are using the TPS546 on a PCB design involving low voltage and high current.*

Vin=5V

Vout,nom=300mV

Imax~35A*

*design is for high current, although in this test we only pull lower current, Imax ~3A.

Purpose of experiment:

Understand PCB noise which seems to originate from the TPS546 and is observed on other power rails.

Test environment:

  • Tektronix MSO64B oscilloscope with TPP1000 probe (<4pF C), and using barrel-tip measurement.
  • Test PCB “SUT” (system under test), which is composed of 
    • (a) the “OVR” On-board VR, TPS546 installed to supply the “VDD” main rail, 
      • Fsw = 1.1 MHz
      • Compensation hex value: 0x137C420884 
        • #Used TI calculator, PhMa:83, AvMa:-13
    • (b) other voltage sources (discussed as needed below), and 
    • (c) multiple IC components acting generally as a “resistive load”
  • External “EVM” - the TPS546D24AEVM-2PH from TI
    • Fsw = 550 kHz, 2 Phase operated in 1 Phase mode.
    • Compensation hex value: 0x1220C62246
  • 5V PSU - Power supply is a Siglent SPD3303X-E (3 Ch, 3.2A per Ch)
  • Keithley Benchtop DMM, for measuring voltages independently


Variable 1: TPS546 - we have two VRs under test within the same system. 

In the first set of experiments, we use the “OVR” (on-board VR).

In the second set of experiments, we used the “EVM”. In this experiment, we disable the OVR by lifting the output inductor pad, and instead inject power from the EVM, connected through low resistance banana cables.

Variable 2: Observing the voltages with “VDD OFF” vs “VDD ON” (Note, 5V is present on the PCB in all of these captures. This variable just turns the TPS546 off and on)

We captures these signals on the four scope channels (e.g. “C1” means channel 1):

C1=5V=Yellow

C2=0p9V=Blue

C3=3v3=Red

C4=VDD=Green

Measurement method: for each power rail, we used a tip and barrel measurement across a capacitor near the load. For VDD, we measured at a bulk capacitor near the output inductor.

Results:

Experiment 1a, OVR - VDD OFF

Some noise on 3.3V, no exact signature detected. No noise above 50mV. No noise spikes observed.

Experiment 1b, OVR - VDD ON

Above shows 1.1 MHz switching noise on VDD and coupled noise on other power rails. Noise is above 100mVp-p.

Above shows 1.1 MHz switching noise as observed on 3.3V rail (898ns == 1.1MHz) Noise is above 100mVp-p.



Experiment 2a, EVM - VDD OFF

Some noise on 5V, 0.9V, and 3.3V, no exact signature detected. No noise above 50mV. No noise spikes observed.

Experiment 2b, EVM - VDD ON

1.1MHz** switching noise observed on VDD, up to 270mVp-p. **Note the 2-PH EVM is operating in 1 Phase mode at a reported 550kHz, so we see overlapping switching signals detected. Other rails observe some of the switching noise, but amplitude is less. 

Summary

  1. Switching noise evident from both OVR and EVM. 
  2. OVR noise is strongly coupled into other power rails.
  3. EVM is weakly coupled into other power rails.

Sidenote: We have Cout (output capacitance) in the same amount and size as provided by the TI Webench, 6x100uF and 5x560uF.

Questions for TI:

  1. Is noise of this magnitude from the EVM expected, 100mV spikes measured near the load?
  2. How can we reduce noise?
  3. Request: Design review (confidential, need to email/meet separately).

Thanks,

Tom



  •  

    Thank you for sharing all of the details about your experiments and testing.

    The first thing I would ask you to try would be to disconnect all of the ground probes except 1, and see if that lowers the noise level you are observing on that 1 rail.  Repeating that for all 4 rails that you are measuring.

    Why?

    The MSO64B, like almost all oscilloscopes, uses a common busbar for the ground shields of it's BNC inputs, which is also connected to the Earth ground terminal of the oscilloscopes power plug.  This creates a very low impedance connection between the BCN input shields (ground) of two channels, such that most of the resistance between the barrels of two single ended passive probes like the TPP1000 is actually in the ground shields of their wires  As a result, if the external circuits they are connected to are forced apart by a high current flowing through low impedance, most of that resulting voltage difference will appear across the probe ground wire.  With the probes operating in 10:1 input attenuation mode, the oscilloscope will gain that voltage drop up by a factor of 10 and a small ground variation can become a large signal variation.

    This is almost certainly why we see VDD rise and PVIN fall during the SW on-time when there is current being drawn from the input.

    By only having 1 oscilloscope ground connected at a time, we can typically remove that influence on the oscilloscope's displayed results.

     

    The noise spikes on the waveforms appear to be switch transition noise, which is typically between 50 and 200 MHz and caused by the resonance of the PVIN and SW nodes as the Power MOSFETs turn on and off in 2-3 nano-seconds.  That noise is expected as the power MOSFETs are rapidly switching current into and out of the ground, but it can be mitigated with component selection and layout.

    Bypassing PVIN to PGND with an 0402 capacitor whose self-resonant frequency matches the resonant frequency of the SW node for example helps reduce the ringing, as does adding a resistor in series with the BOOT pin to slow-down the turn-on of the MOSFET, or an R-C snubber from SW to PGND to couple any resonant noise back to the exposed pad.

    Finally, good layout technique can minimize cross-inductor capacitance that can conduct high-frequency noise across the inductor, or reduce series inductance on high-frequency output capacitors, minimizing their ability to contain that noise close to the IC.

    For the schematic / layout review, you can click on my profile name and send me a message with your e-mail address so we can assist you with a review.  Have you used the Schematic and Layout review checklist already provided from the TPS546D24A's product folder?

    https://www.ti.com/product/TPS546D24A#design-tools-simulation - 

    SLURB01  TPS546D24A TPS546B24A TPS546A24A Excel Calculator Tool and Checklist 

  • Hi Peter,

    Thank you for your thorough response.

    RE: the scope GND issue, we will try this experiment next time we get in our lab, connecting a single ground from the barrel-tip of the signal of interest, while leaving our other scope channels "un-referenced" without a GND signal sent to the scope.

    Yes, you are right that of course VR switching noise is "expected", however, my opinion is that this should be already mitigated with our Cout capacitor selection. We should review the design and I can add more/different Cout in the next rev as needed.

    I will review the schematic and layout checklists provided by TI. We are in the Rev B design phase now so this is when we want to catch any issues from Rev A and ensure they are not repeated! I have two improvements so far--(1) we had thermal reliefs in Rev A, e.g. on our output inductor, so I am removing those, and although the layout appears generally good in Rev A, I am more closely following the EVM design and datasheet recommendations.

    I'll reach out separately for additional review. Please note we have many TI parts in our otherwise small design as you will see, so I'd appreciate a general review of all TI parts! I hope we can pull in the relevant engineers as needed.

    Thanks,

    Tom

  •  

    I will follow up with you in your private message.

    I can help bring in people from other groups in TI.  If you have a list of all of the TI parts so that I don't miss any going through the schematic, that would be extremely helpful.

  •  

    If we have resolved your issue, can you click "This Resolved My Issue" so we can close out this thread?