Hot!1nf capacitor on VUSB3V3

Author
uzslm
Super Member
  • Total Posts : 347
  • Reward points : 0
  • Joined: 2016/06/28 18:46:27
  • Location: 0
  • Status: offline
2018/12/01 18:37:20 (permalink)
0

1nf capacitor on VUSB3V3

my pic is pic16f1459,runing as an usb device in FS speed which was powered by PC's USB 5V;
for the pin VUSB3V3,an 1nf and a 0.47uF capacitors have been added;i believe this has been a reference to the design of "USB Keypad Reference Design"
https://www.microchip.com....aspx?appnote=en565508
today i check the most new datasheet,in section 29.7,only a single capacitor(0.22~2.2uF,typical 0.47uF) is required for the VUSB3V3;
then can i just leave the 0.47uF,and remove the 1nf cap from my board?(this will save sapce and reduce elements amount.)
however if there is any good effects,i will consider about keeping both on my board.
 
Thank you in advance for any advice!
Regards,
uzslm
#1

17 Replies Related Threads

    qhb
    Superb Member
    • Total Posts : 7898
    • Reward points : 0
    • Joined: 2016/06/05 14:55:32
    • Location: One step ahead...
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/01 18:40:42 (permalink)
    +1 (1)
    You can use a single capacitor, so long as it has very low ESR.
    Adding a smaller capacitor in parallel is an attempt to make up for a larger cap with not such good ESR.
     
    #2
    uzslm
    Super Member
    • Total Posts : 347
    • Reward points : 0
    • Joined: 2016/06/28 18:46:27
    • Location: 0
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/02 01:05:07 (permalink)
    0
    both are ceramic caps in 0603,X7R,Samsung;
    i am not sure of their ESR;but we can be sure that ESR is lower with 1nf present....
    #3
    Jerry Messina
    Super Member
    • Total Posts : 339
    • Reward points : 0
    • Joined: 2003/11/07 12:35:12
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/02 06:05:46 (permalink)
    +1 (1)
    both are ceramic caps in 0603,X7R

    In that case I doubt the 1nF is worth keeping. It'll give you a slight dip in impedance up at 100MHz (very slight).
    The parallel combination with the 0.47uF will also give you an anti-resonant point around 70-80MHz, so the impedance actually gets worse.
     
    If you want to use two caps you'd almost be better off just using two 0.47, but honestly the single 0.47 is likely fine.
     
     
     
    #4
    NorthGuy
    Super Member
    • Total Posts : 5101
    • Reward points : 0
    • Joined: 2014/02/23 14:23:23
    • Location: Northern Canada
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/02 07:27:03 (permalink)
    +1 (1)
    Usually small bypass capacitors are used when working with high frequencies and fast edges. For FS USB, single 0.47uF cap is fine.
    #5
    coffee critic
    Super Member
    • Total Posts : 279
    • Reward points : 0
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/03 14:47:42 (permalink)
    0
    I wouldn't get rid of the 1000pF cap.  Extra HF bypass will clean up a lot of transient conditions that are not even seen by the larger cap.  The extra cost component cost is $0.005.  Pick and place cost is about the same. 

    Attached Image(s)


    n_*$
    #6
    Jerry Messina
    Super Member
    • Total Posts : 339
    • Reward points : 0
    • Joined: 2003/11/07 12:35:12
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/04 04:46:31 (permalink)
    +2 (2)
    Look at the vertical scale on those two graphs. The impedance of the smaller cap is about 1/2 an ohm around 100MHz. If you superimpose that on the other graph you'll see it's not really doing squat (well, very little).
     
    Above the resonant point the impedance is set by the package size inductance. It goes against intuition, but IMHO you're almost always better off choosing the largest cap value for a given package size. If you need a lower impedance just use multiples of that value. That helps avoid the "anti-resonance" peaks that occur from the L-C interaction mixing values.
    #7
    uzslm
    Super Member
    • Total Posts : 347
    • Reward points : 0
    • Joined: 2016/06/28 18:46:27
    • Location: 0
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/04 04:59:26 (permalink)
    0
    i am not sure if 1nf is completely necessary for FS speed,seems that the signal frequency is at 48Mhz;
    i tend to use single .47uF,and remove the smaller one;although both caps have been running on many boards;
    and,after all,it is suggested in datasheet.
    post edited by uzslm - 2018/12/04 05:12:33
    #8
    Jerry Messina
    Super Member
    • Total Posts : 339
    • Reward points : 0
    • Joined: 2003/11/07 12:35:12
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/04 05:45:58 (permalink)
    +3 (3)
    and,after all,it is suggested in datasheet

    For no valid reason (in my opinion).
     
    Here's a graph showing the interaction of a 0.47uF + 1nF (both 0603 mlcc). As you can see, adding the 1nF does give you a lower impedance right at the 100MHz mark, but it does very little good otherwise. It actually makes things much worse in the 50MHz-80MHz range, which is likely where your primary osc freq is!
     

     

    Attached Image(s)

    #9
    uzslm
    Super Member
    • Total Posts : 347
    • Reward points : 0
    • Joined: 2016/06/28 18:46:27
    • Location: 0
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/04 06:34:44 (permalink)
    0
    Thank you so much for these explanation and drawings!
    #10
    mbrowning
    Just a Member
    • Total Posts : 1154
    • Reward points : 0
    • Joined: 2005/03/16 14:32:56
    • Location: Melbourne, FL
    • Status: online
    Re: 1nf capacitor on VUSB3V3 2018/12/04 07:42:53 (permalink)
    +3 (3)
    Jerry Messina
    and,after all,it is suggested in datasheet

    For no valid reason (in my opinion).

    Agree. This is an antique recommendation from the through-hole days, and was perhaps a little valid >10years ago when chip ceramics weren't as good and you had to go to larger packages (with higher ESR/ESL) for larger values. I haven't used a through-hole cap in nearly 30 years and 15 years ago it was clear that using a range of values in the same package for digital decoupling was a waste of parts and money.
     
    In our broadband RF designs we do use a range of 0201 caps for power supply filtering, but we design these filters in Genesis with Modelithics models.
     

    Oh well - there's always next year
    #11
    NorthGuy
    Super Member
    • Total Posts : 5101
    • Reward points : 0
    • Joined: 2014/02/23 14:23:23
    • Location: Northern Canada
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/04 07:52:29 (permalink)
    0
    uzslm
    ... for FS speed,seems that the signal frequency is at 48Mhz



    12 MHz.
    #12
    coffee critic
    Super Member
    • Total Posts : 279
    • Reward points : 0
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/04 09:39:24 (permalink)
    0
    I will stick with my original recommendation.  0.1uF caps were fine when CPUs and logic were sub 10MHz.  A smarter person than me told me once to never bypass on the inductive side of the curve.

    n_*$
    #13
    uzslm
    Super Member
    • Total Posts : 347
    • Reward points : 0
    • Joined: 2016/06/28 18:46:27
    • Location: 0
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/04 20:58:45 (permalink)
    0
    Sorry,the clock for the usb module is 48MHz;bit rate of usb FS speed is 12Mbit per second.
    while bit rate of usb high-speed is 480Mbit per second,then does that mean the core usb module's clock is at 480*4=1.92G??
    I was actually planning to upgrade to HS,but found cost is higher including the MCU and more complex design;
     
    for this module,Microchip provides  a varity of footprints:DIP,SOP(wide),SSOP(wide),QFN;i've tried all of them,finally i used SSOP;DIP and SOP are pretty big,QFN is smallest;
    someone told me that QFN is rarely used in industrial or automotive designs;the reason i chose ssop is that it is samll enough,and more popular(easy to get),and i can solder by hand to have a test;qfn is too samll,for me very difficult to solder by hand;
     
    As a general rule,the cap should be place as close as possible to the VUSB3V3 pin,but how to define "close"? 1mm,5mm or 10mm?for smaller package like ssop or qfn, it is pretty hard to place the cap in 1mm or 2mm due to  high density of routing.
    #14
    Jerry Messina
    Super Member
    • Total Posts : 339
    • Reward points : 0
    • Joined: 2003/11/07 12:35:12
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/05 05:13:59 (permalink)
    +2 (2)
    but how to define "close"?

    "Close as possible" means just that. Increasing the distance will increase the inductance, making your cap less effective. But, welcome to the real world. Your only other options are to mount it on the opposite side of the board, use a smaller package, etc.


    0.1uF caps were fine when CPUs and logic were sub 10MHz

    And they're still fine. If you follow that logic, it would suggest that you should use a smaller value cap.
    Replacing the 0.1uF with a 1nF gets you several orders of magnitude worse performance below 100MHz, and the same performance above it.

    Actually, as a "general rule" I'd recommend throwing away all the 0.1, 0.01, 0.001, etc and just use a 1uF cap!
     
    Look at it this way... ideally, there would be no "V" shape to that impedance curve.
    The left side of the curve is Xc, and that's determined by C. The right side of the curve is Xl, and that's set by the package size inductance, not by the value.

    Since the right side of the curve is set by the package size, obviously the only way to change it is to change packages... smaller = better. You can get caps where the X-Y dimensions are swapped and that'll get you lower inductance as well (0306 vs 0603).

    So, the only thing you can change is the left side of the graph, and picking a larger value of C will get you a lower Xc. This will give you better overall performance except for the very small freq range about the resonance point.

    If you truly need to effect the upper freq performance you can use multiple C values, but to avoid the resonant peaks you need to use several C values for each decade,  several decades worth, and multiple quantities for that to be truly effective. You end up with a boatload of capacitors and increased BOM costs.

    A smarter person than me told me once to never bypass on the inductive side of the curve

    I don't know why... I'd argue that we all use the inductive side every day. Perhaps it's because you can't change the performance by changing values?


    #15
    coffee critic
    Super Member
    • Total Posts : 279
    • Reward points : 0
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/05 14:00:20 (permalink)
    0
    Using a 1uf ceramic cap in place of a smaller 0.1uf value would generally not be a good idea. The class 2 ceramic materials (titinates) in these parts this size are practically invisible to HF.
     
    For the beginners who are trying to learn:
    EEVblog #859  First few minutes are bypass 101.  Get more interesting at minute 8.  One big cap answer at minute 11. Multiple cap explanation starts being covered at minute 17.  Minute 27 explains resonance interaction.

    n_*$
    #16
    NorthGuy
    Super Member
    • Total Posts : 5101
    • Reward points : 0
    • Joined: 2014/02/23 14:23:23
    • Location: Northern Canada
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/05 16:35:38 (permalink)
    +1 (1)
    The capacitor being discussed by OP is not a bypass capacitor, but the bulk capacitor for the voltage regulator.
    #17
    Jerry Messina
    Super Member
    • Total Posts : 339
    • Reward points : 0
    • Joined: 2003/11/07 12:35:12
    • Status: offline
    Re: 1nf capacitor on VUSB3V3 2018/12/06 05:27:54 (permalink)
    +1 (1)
    True, but I think some of the general principles are worth discussing since "it's what's in the datasheet" seems to be taken as gospel.
     
    Now, about the video...
     
    Wow. With all due respects to Dave, that has to be one of the worst videos I've ever seen. He should be embarrassed. "Crude test setup" is an understatement. I wouldn't use that bodgy setup to measure audio, much less RF performance. His "mashed potato analyzer" (or whatever it's called) is a joke. He should try doing this using a proper RF test setup and a network analyzer.

    A few comments:

    In the video, around 17:30-18:50 he assumes the impedance curve shape is the same for different values, just shifted in the X axis. That's not how it usually works... it also tends to be shifted in the Y-axis so it doesn't add up to the nice low-impedance "W" shape he shows.

    Also, in his white-board discussion he pretty much completely ignores the real-world effect of the anti-resonance interactions that occur. He briefly makes a quick reference to it later on, but for a beginner's tutorial the importance of it should not be ignored.
     
    Around 26:50 where he's showing real measurement results he states:
    "It's actually a better result just having the 10uF capacitor (vs 10uF+100n+10n)"  (see below)
    He tries to dismiss this as a fluke of "this particular setup with this particular set of caps"
     
    Go figure.
     
    Quoting Larry Smith (Qualcomm/Altera/Sun Microsystems):
    "Putting low ESR capacitors in parallel is like playing with fire.  
    You can cook with fire and heat your house, but if you are not careful with fire, you will get burned."


     
    post edited by Jerry Messina - 2018/12/06 05:30:27

    Attached Image(s)

    #18
    Jump to:
    © 2018 APG vNext Commercial Version 4.5