LTspice, Capacitor "Series Inductance" and HF Spikes

I am simulating the LT8362 SEPIC 5V-output circuit shown on page 24 of the datasheet:

http://www.analog.com/media/en/technical-documentation/data-sheets/8362fa.pdf 

You can download my LTspice *.asc file and associated symbols and libraries here:

https://cl.ly/qlBw/LTSPICE_TEST.zip 

I use the MacOS X version of LTspice, but I tested the above files in the Windows version and the simulation is the same.  If you are using Windows, but be sure to put my files into your "Documents > LTspiceXVII > lib" path.  Rename your "ISO16750-2.lib" file first though, since I made alterations to mine in order to make the simulation run faster.

My *.asc file is pretty much what you see on page 24 of the datasheet, except I added Vin and Loads to the circuit.

PROBLEM

After you copy my files to your computer, start the simulation and probe the OUT voltage and you will see HF spikes reach 30V!  (See attached "5Vout_650pSeriesInductance.png.") That's the problem.  Halt the simulation, then right-click C7 (22uF cap) and delete the "Series Inductance" value of 650p. Click OK and run the simulation again.  Now the OUT voltage will increase to 5V in about 900us, and all will be well with the rest of the simulation.  (See attached "5Vout_NoSeriesInductance.png.")  In other words, it seems the simulation is not running properly when Series Inductance is added to the capacitors.  (There is no series inductance on caps in LTspice by default.)

Of course, I added the "Series Inductance" because all components have it, and 650pH is a reasonable estimate of what that 25V 22uF SMD capacitor has (according to TDK spice models I've reviewed).  And yet, if I add even that tiny 650pH of series inductance, the result in the simulation is HF spikes reaching 30V.  Without series inductance on the cap, the simulation runs as expected.  I am curious why this is happening.  It's pretty clear that something is off with the simulation seeing that Linear Tech would not put forth a demo circuit in the datasheet that they have not tested themselves.

No, I have not yet built the circuit.  I am in the design stage running simulations.  I would appreciate hearing your thoughts and advice.

Thank you.

attachments.zip
  • dulcevida, I sincerely appreciate your kind assistance in bringing this to the attention of Analog's LTspice expert.  I look forward to hearing more.  In the meantime, perhaps I could ask you another, related question...

    BACKGROUND INFO

    My application is a switch-mode power supply for 12V automotive use, supplying 5V @200mA (peak) to an MCU and CAN Transciever circuitry.  Sustain current draw by my circuit when the vehicle's CAN is awake will be around 80mA.  (I know this because I have an existing product on the market with linear power supply which I used in my testing of current draw.  We are redesigning the power supply to ensure operation during cold start, using the LT8362.) When the vehicle's CAN is sleeping, the total device current (including PSU) draw will drop to less than 4mA.  Vcc-min for the CAN transceiver and MCU is 4.50V.  The aim is to maintain a steady 4.7V-5V to the circuit even during cold-cranking (where the vehicle's battery voltage could drop to as low as 2.8V).  So long as I don't add Series Inductance in my capacitors in LTspice, the LT8362 circuit simulates as expected.  A reverse polarity protection diode (DFLS260) at the input will cause a further voltage drop to as low as 2.44V, which is the lowest the LT8362 will see at Vin.  (DFLS60 is 60Vmax but is protected by a TPSMC36CA TVS diode with Vc=49.9V, which is placed before the Schottky in the circuit.)

    In my LTspice (version IV for MacOS X) simulations (which does not consider the PCM layout, I know), I am finding the Peak-to-Peak noise (ripple and HF spikes) is 7.6mV when in BUCK mode (>6V at Vin), and 28mVp-p while in BOOST mode (when the car battery voltage drops to 2.8V).  I honestly do not know what the acceptable level of noise is for a CAN transceiver's Vcc, but even 28mVp-p is only 0.6% of 5.0V, which I guess can be considered "low."  I did simulate other variants of my circuit using a BL02RN1-R62 Ferrite BEAD and 47uF EEU-FR1H470 capacitor, which yields 880uVp-p in BUCK mode and 3.2mVp-p in BOOST.  Another circuit variant I tried uses a 0.33-ohm resistor (actually, three 1-ohm resistors in parallel, which would be cheaper than a single 0.33-ohm resistor) on the output, followed by the same 47uF electrolytic lowESR capacitor, yielding 517uVp-p noise in BUCK mode and 2.4mVp-p in BOOST. Output Voltage never dropped below 4.996V in any of the simulations regardless of input voltage dropping to 2.8V and with up to 200mA load current.  But again, these simulation results were performed without any Series Inductance on my capacitors, so actual peak-to-peak noise will likely be higher than this.

    So why did I choose the LT8362 for my SEPIC application?  Because Vin-max = 60V.  That is important because my TVS choice is Vc=50V.  Any switcher chips with Vin lower than 50V are not acceptable.  And when considering jump starts and Load Dumps, the TPSMC36CA with its Vc=50V is really the best choice for circuit protection in a 12V vehicle application.

    I am still new to switching power supply design, and I've been running a number of simulations in LTspice to determine the best component picks to make for the purpose of building a test circuit.  I am well aware that the PCB layout plays a major role in peak-to-peak output noise and the overall performance of the circuit.  But for now, I am pondering 5V output noise in LTspice in conjunction with my coupled inductor choice.

    Even though I don't expect more than 200mA peak current draw, I am adding a safety margin and designing this power supply for a maximum current output of 5V@300mA, which again would be very short-term peak current draw.  Sustained draw will not exceed 80-90mA (while the Ignition is ON and while CAN is awake).  Doing the LT8362 datasheet SEPIC calculations for a 0.3A 5V output, I get the following:

    • Dmax = 0.667 (based on Vin-min = 2.8V and my Schottky diode choice of DFLS260-7)
    • Rt = 20k-ohm (2MHz operation), Bias pin tied to GND
    • Isw-max-avg = 900mA
    • Isw-peak = 1.2A
    • Isw = x * Isw-max-avg = 0.6A ("x" should be between 0.5 and 0.8, so I chose 0.65)
    • Delta-IL1 = Delta-IL2 = 0.5 x Isw = 0.3A
    • IL1peak = IL1max + 0.5 * Delta-IL1 = 750mA
    • IL2peak = IL2max + 0.5 * Delta-IL2 = 450mA

    Based on the above calculations, the minimum inductance for a coupled inductor is low enough to qualify use of either 2.2uH or 4.7uH (per inductor).  (The LT8362 datasheet's 5V example circuit uses 2.2uH.)  After running LTspice simulations and viewing peak-to-peak noise on the 5V output, I decided on a 4.7uH inductor.  The Wurth inductor shown in the 5V-output datasheet example is physically too large for my application at 12.5 x 12.5mm, and no surprise since that example circuit is designed for a 1A output.  So in my test circuit in LTspice, I am now using a Wurth 744877004 coupled inductor (4.8uH, Irms=2.55A, DCR=64m-ohm), which is 7.3x7.3mm in size. (The Wurth 74489440047 is Irms=1.85A, which should be adequate for my 0.3A application, but there is no LTspice model for that inductor.)

    MY QUESTION

    When comparing inductor brands, I am finding a somewhat large pricing disparity.  Wurth seems to be the Mercedes Benz of inductors, as shown in this coupled inductor comparison (all roughly 7.5x7.5mm in size and all SHIELDED):

    • WURTH 744877004, 4.7uH, Irms=2.55A, DCR=64m-ohm, Price=US$1.53EA per 1000pcs
    • EATON DRQ74-4R7-R, 4.7uH, Irms=1.67A, DCR=51m-ohm, Price=US$0.84EA per 1000pcs
    • COILCRAFT MSD7342-472ML, 4.7uH, Irms=1.74A, DCR=51m-ohm, Price=US$0.80EA per 1000pcs
    • BOURNS SRF0703-2R2M, 4.7uH, Irms=2.00A, DCR=66m-ohm, Price=US$0.36EA per 1000pcs

    Do you know why the BOURNS is so much cheaper than the WURTH?  They all are coupled inductors.  

    One thing I greatly appreciate about the WURTH coupled inductors is the fact I can conveniently simulate them in LTspice, but price really does matter in my application.  I am inclined to go with the BOURNS based on price, but I want to know why the other coupled inductors are so much more expensive at the same quantity. Any guidance you can offer on coupled inductor selection for my application would be greatly appreciated.

    Thank you.

  • 0
    •  Analog Employees 
    on Apr 11, 2018 1:40 AM

    Yes, I see what you mean. I was able to duplicate your findings. As soon as I hear back from our LT Spice expert, I will report our findings.

    For now, I know this issue happens in simulation, only. As you have so eloquently expressed it, Linear Tech, now Analog Devices, would not put forth any circuit if it had not been previously tested.

  • dulcevida,

    Thank you again for sharing your detailed thoughts.  I have 1 additional question and 1 clarification for you...

    QUESTION: Determining INDUCTANCE value for each coil in a coupled inductor using datasheet

    Again, I am new to switch-mode power supplies, so perhaps this question has an embarrassingly obvious answer, but I need to ask it.  Wurth & Coilcraft datasheets make a drop-dead easy to know the inductance of EACH coil in their coupled inductors, but most other coil manufacturers obfuscate that important information for reasons I cannot fathom.  Consider this Bourns datasheet:

    https://www.mouser.jp/datasheet/2/54/RF0703-778039.pdf 

    Rather than just plainly tell us the inductance of each coil, BOURNS gives me PARALLEL and SERIES data instead.  Well, we know that the inductance of 2 inductors is calculated exactly as one calculates the resistance of 2 resistors:

    Series and Parallel Inductors | Inductors | Electronics Textbook 

    So when examining the datasheets of coupled inductors which give only Parallel and Series Ratings, I am inclined to think that one must perform hand-calculations in order to determine the inductance of EACH coil.  Correct or incorrect?

    My design is in need of a coupled inductor that has 4.7uH inductance for EACH of the 2 coils.  For example, in the BOURNS datasheet I linked above, the only instance of "4.7" I see is the Parallel Inductance Rating for Part No. SRF0703-4R7M.  If you look at the "Electrical Schematic" in that same datasheet, you see what defines "Parallel" and "Series" connections.  So it seems clear that the Parallel Inductance Rating of 4.7uH would correspond to the Parallel schematic, which means this:

    Lparallel = 1 / ( 1/L1 + 1/L2 ) = 1 / ( 1/L1 + 1/L1 ), since both inductors are equal inductance

    Solving for L1 yields, L1 = L2 = 2 / (1/4.7E-6) = 9.4uH for each inductor, which is not what I need.

    That would mean I would need Part No. SRF0703-2R2M which is 2.2uH Parallel and yields:

    L1 = L2 = 2 / (1/2.2E-6) = 4.4uH for each inductor, which is close enough to my 4.7uH target inductance.

    Correct or incorrect?

    (The reason I am trying to figure this out rather than just pick Wurth or Coilcraft is because Bourns is much cheaper.)

    CLARIFICATION: TVS protects Schottky Diode at Vin

    I certainly could use a B2100-13-F Schottky which has a 100V rating and then be able to more safely place it before or after the TVS on the input.  But Schottky voltage drop often lower when its rated voltage is lower.  According to the "If vs. Vf" charts (at 25°C) in the respective Schottky datasheets, I see the following voltage drop data:

    • B2100-13-F (Vr = 100V), Vf at 300mA = 460mV (roughly), Vf at 100mA = 360mV (roughly)
    • DFLS260-7 (Vr = 60V), Vf at 300mA = 360mV (roughly)Vf at 100mA = 320mV (roughly)

    Not a huge difference in voltage drop between those parts, but with the cost being similar I might as well chose the absolute lowest voltage drop part available.  This is especially true since that diode will be used for reverse polarity protection on the input, as dictated by the fact the LT8362 doesn't have reverse polarity protection built-in.  And I am using a SEPIC design to ensure a constant Vout=5V even during cold cranking, where the vehicle battery voltage could drop to 2.8V, which means Vin into the LT8362 will be even lower, in accordance with the voltage drop of the Schottky diode used.  In my LTspice simulations I am seeing the voltage drop across the DFLS260-7 go as low as 2.44V when the battery voltage is 2.8V.

    When using a Schottky with a Vr lower than 100V it needs to be protected from Load Dumps.  Most new cars have suppression in the alternator, but the voltage level at which those spikes are capped varies by auto maker.  With suppression, the spikes are usually capped to 50V and lower.  Without suppression, a Load Dump spike could easily hit 100V.  That's why I use a TVS diode in the circuit -- to suppress Load Dump spikes, primarily.  But in this particular design, I place the TVS first in the circuit, before the Schottky, so as to protect the Schottky due to its 60V rating.  If I used any old 1N4004, diode placement would not matter since a 1N4004 is rated for 400V.  But with a 60V Schottky, I need to place it after a TVS.  And I am using a bidirectional TVS in this case to protect the TVS against reverse polarity.  (If polarity was reversed on a unidirectional TVS that came first in the circuit, the TVS would blow.)  Then everything in the circuit after the Schottky will be protected against reverse polarity by the Schottky.  Then there's an external fuse on the wire harness outside the PCB, which would blow in the even the TVS failed (since TVS diodes fail shorted).

    Here is a schematic of my design:

    https://cl.ly/qpTv/LT8362_PSU_12V_rev1.pdf 

    So once I figure out how to properly interpret coupled inductor datasheets like BOURNS, I can finalize my inductor choice and build a working PCB for testing.  I would appreciate hearing your thoughts.

  • 0
    •  Analog Employees 
    on Apr 11, 2018 9:58 PM

    Hi, Kiramek,

    The inductor option you propose seems sensible to me. As long as the base material and saturation current are similar you should be fine.

    Why prices vary so much, I can only speculate, but cannot offer any useful advice.

    You have done a meticulous job using the simulation software for your analysis. I think it is time to build the application and troubleshoot the actual circuit, where if you follow our suggested layout guidelines you might find the circuit works better than expected.

    There is one concern I have, and that is the use of a zener clamp to protect the diode. I don't understand how you plan to connect it, but you might not need it. The diode cathode will be sitting at Vout (5V), and the anode will be swinging to negative Vin. So, the input would have to swing very high for you to need protection for a 60V diode. But, if the concern is real you could use a 100V diode, which I believe is available.

    Best Regards.

  • dulcevida,

    Thank you for confirming that the "parallel inductance" of coupled inductor datasheets is what I need to focus on when picking inductors for my power supply designs.  It is rather curious that some datasheets like the following put the standard inductance values into a separate "Rated Inductance" column along side a Parallel Inductance column, but I do see that the Parallel Inductance column values are roughly the same as the Rated:

    https://www.mouser.jp/datasheet/2/87/Bus_Elx_DS_4311_DRQ_Series-335920.pdf 

    Hopefully, I get the PCB layout to turn out well as that is critical to proper performance of switching power supplies. Unfortunately, the only suggested PCB layout given in the LT8362 datasheet is Figure 5 on page 17, which is for a Boost Converter:

    http://www.analog.com/media/en/technical-documentation/data-sheets/8362fa.pdf 

    I reviewed the PCB layouts of the two LT8362 evaluation kits (DC2517A & DC2628A), but those designs are not SEPIC.  In the DC2517A design files I see it is an INVERTING -12V output, yet the dots on the inductor in the "DC2517A-4-SCH.pdf" schematic (inside the "610-DC2517A_REV04_PDF" folder of the design files) are exactly the opposite of what is shown in Figure 8 on page 20 of the LT8362 datasheet.  I guess one of those is in error?  Also, the DC2517A evaluation board says it's rated for a 2A output, but they use a Wurth 74489430068 coupled inductor with Isat = 2A.  Assuming a Schottky voltage drop of 0.4V, Dmax=0.7337, and my resulting calculations (based on formulas given in the LT8362 datasheet for SEPIC which uses a coupled inductor) show IL1peak=6.73A & IL2peak=3.22A.  But again, Isat for each of the 2 coils is only 2A.  So I am now wondering if IL1/2peak is the value that should define a chosen coupled inductor's saturation current (Isat) spec or not. Hmmm...

    Anyway, while looking over those 4-layer board designs I noticed that there are Ground Planes beneath the main inductor, whereas in the PCB layout for the LTC3637 evaluation kit there is a square cut out in all the ground planes beneath the location where the main inductor sits:

    DC2628A - Design Files

    DC2517A - Design Files

    DC2056A - Design Files

    I have read it is best to not have copper beneath the main coil in switching power supply designs, which has me curious about the DC2517A and DC2628A in light of the SEPIC design I am doing.

    If there are any SEPIC PCB layout files that I can review as a help in my LT8362 design, please let me know.  Studying existing/recommended PCB layout designs would be helpful considering I've not previously built a switching power supply.  Currently I have two designs on the table, one based on the LTC3637 for a 24V automotive application and the other based on the LT8362 for 12V vehicles that we have been discussing in this thread.

    Thank you for your helpful assistance to date.