Post by thread:[OT] A final question about decoupling caps.

Today as i was installing decoupling caps on a circuit board i thought to myself: "Why dont the IC manufactures just incorperate a decoupling cap right in the ic itself?" Any one have any ideas? ********************* Andy writes: The largest size we can get on a die is only a few pf. And it takes up as much space as many thousands of transistors. Furthermore, it is not a very good cap, since there is significant resistance in the internal track which, along with the dielectic, limits the "Q". Additionally , any caps that are used for coupling (RF ckts) internally also have a stray coupling factor to substrate of about a fifth of their value (it varies some). Lastly, the substrate for a silicon IC is not conductive so you have to use interconnect as the ground leads as well as the hot lead inside the device and it can lead to undesireable intercoupling. Whew!! Actually, frequency compensated op amps do use internal capacitors, but they are very small..... and can only be used in special areas. In summary -- the technology for making acceptable capacitors is not the same technology for making acceptable transistors... Andy in Dallas (too wordy for a Sunday morning...) ______________________________________________________ Get Your Private, Free Email at http://www.hotmail.com

Capacitors are VERY area-expensive to integrate. ---------- > From: Lynx {Glenn Jones} <spam_OUTjones_glTakeThisOuTEFN.ORG> > To: .....PICLISTKILLspam@spam@MITVMA.MIT.EDU > Subject: [OT] A final question about decoupling caps. > Date: Sunday, January 17, 1999 09:12 > > Today as i was installing decoupling caps on a circuit board i thought to > myself: "Why dont the IC manufactures just incorperate a decoupling cap > right in the ic itself?" Any one have any ideas? > > ---------------------------------------------------------------------------- -- > A member of the PI-100 Club: > 3.1415926535897932384626433832795028841971693993751 > 058209749445923078164062862089986280348253421170679

you can buy sockets for your chips that have decoupling caps built in. >Capacitors are VERY area-expensive to integrate. > >---------- >> From: Lynx {Glenn Jones} <jones_glKILLspamEFN.ORG> >> To: .....PICLISTKILLspam.....MITVMA.MIT.EDU >> Subject: [OT] A final question about decoupling caps. >> Date: Sunday, January 17, 1999 09:12 >> >> Today as i was installing decoupling caps on a circuit board i thought to >> myself: "Why dont the IC manufactures just incorperate a decoupling cap >> right in the ic itself?" Any one have any ideas? >> >> >--------------------------------------------------------------------------- - >-- >> A member of the PI-100 Club: >> 3.1415926535897932384626433832795028841971693993751 >> 058209749445923078164062862089986280348253421170679

At 01:07 PM 1/17/99 -0800, William Chops Westfield wrote: > >you can buy sockets for your chips that have decoupling caps built in. > > Those are not nearly as nice as they look at first glance. Look at > the long leads on the cap! > >I think you'll have to explain how to connect pin 8 to pin 16 of a typical >DIP using "leads" that are any shorter than that! > >(this *is* why the newer generations of fast chips have moved the power pins > to the center of the package.) > >BillW No problem. I route a solid ground to the ground pin. I mount the bypass cap at this point. Then I run a track from here to the package's VCC pin. The same idea is seen in the Murata and Panasonic three-lead EMI supressors. The EMI is forced to travel to the cap, before it can get to anything else. Distribuiting the L and C differently in the circuit forms a Tee filter, rather than a high pass filter (well, actually a notch, but at the wrong frequency) The total lead length isn't any less, but look at the circuit it makes. The noise from the IC has only one place to go, which is to the + side of the cap. With the capped sockets, the cap is in the middle, with lead inductance on both sides, and nothing at all to prevent the noise from going out to the system. It really does make a large difference, but it's kind of like a tuneup in a car, $100 racing plugs don't help you if your timing is off. You have to get a solid ground to everyone that makes any noise, and you have to provide low Z paths between devices that exchance high speed signals.

At 12:43 01/17/99 -0500, dave vanhorn wrote: >At 12:12 AM 1/17/99 -0800, Lynx {Glenn Jones} wrote: >>Today as i was installing decoupling caps on a circuit board i thought to >>myself: "Why dont the IC manufactures just incorperate a decoupling cap >>right in the ic itself?" Any one have any ideas? > >because it would be hugely expensive in silicon. who's been talking about "silicon"? most ics are quite a bit more than the small silicon chip. eg. in most but the smallest dip packages, one could imagine that there's enough space in the package (not on the die) for some nFarads. it's probably just too expensive. ge

> At 08:35 AM 1/17/99 -0500, ryan pogge wrote: > >you can buy sockets for your chips that have decoupling caps built in. > > Those are not nearly as nice as they look at first glance. Look at the > long > leads on the cap! > The last place I worked used some very strange decoupling caps. They were fitted underneath the IC, without using a socket. They were very thin with the same outline as the IC and thin pins that could be pushed through the same holes as the IC. This got around the problem of long leads with the socket type decouplers. Regards Mike Rigby-Jones EraseMEmrjonesspam_OUTTakeThisOuTnortelnetworks.com

>The last place I worked used some very strange decoupling caps. They were >fitted underneath the IC, without using a socket. They were very thin with >the same outline as the IC and thin pins that could be pushed through the >same holes as the IC. This got around the problem of long leads with the >socket type decouplers. Those are nice. Excellent caps, but putting a conventional cap at the GND pin and routing the power to it first, then to the chip, actually uses the trace inductance to help in the filtering :) At 300+ Mhz, it's significant, and FCC scans up to 1GHz. Still, If I HAD to use socket caps, that's what I'd use.

Mike, Q-Cap made those devices and they worked very well. They also made bus strips and custom devices. It's been awhile. I have'nt followed this thread very closely but now days, as speed increases, there is a trend to use multiple low-ESR caps (ie: 10 0.01uf caps to make 1 0.1uf). There was a good discussion of this in the comp.arch.fpga news group. - Tom At 09:09 AM 1/18/99 -0000, Michael Rigby-Jones wrote: {Quote hidden}>> At 08:35 AM 1/17/99 -0500, ryan pogge wrote: >> >you can buy sockets for your chips that have decoupling caps built in. >> >> Those are not nearly as nice as they look at first glance. Look at the >> long >> leads on the cap! >> >The last place I worked used some very strange decoupling caps. They were >fitted underneath the IC, without using a socket. They were very thin with >the same outline as the IC and thin pins that could be pushed through the >same holes as the IC. This got around the problem of long leads with the >socket type decouplers. > >Regards > >Mike Rigby-Jones >mrjonesspam_OUTnortelnetworks.com

On Mon, 18 Jan 1999 09:09:25 -0000 Michael Rigby-Jones <@spam@mrjonesKILLspamNORTELNETWORKS.COM> writes: >> At 08:35 AM 1/17/99 -0500, ryan pogge wrote: >> >you can buy sockets for your chips that have decoupling caps built >in. >> >> Those are not nearly as nice as they look at first glance. Look at >the >> long >> leads on the cap! >> >The last place I worked used some very strange decoupling caps. They >were >fitted underneath the IC, without using a socket. They were very thin >with >the same outline as the IC and thin pins that could be pushed through >the >same holes as the IC. This got around the problem of long leads with >the >socket type decouplers. > These are "micro-Q" capacitors, originally made by Rogers Corp., now called CCI at +1 602 967 0624. The capacitors mount in the same holes as the DIP chip or socket, making connections to the diagonal power pins (such as 7 and 14). My last order with them was for some 303A28 and 303A40 parts. These are each 30nF, one for a 28 pin DIP, and the other for a 40 pin DIP. Cost was $1.16 for the first, $1.27 for the second. As Michael points out above, these parts get around lead inductance and save board space. Harold Harold Hallikainen KILLspamharoldKILLspamhallikainen.com Hallikainen & Friends, Inc. See the FCC Rules at http://hallikainen.com/FccRules and comments filed in LPFM proceeding at http://hallikainen.com/lpfm ___________________________________________________________________ You don't need to buy Internet access to use free Internet e-mail. Get completely free e-mail from Juno at http://www.juno.com/getjuno.html or call Juno at (800) 654-JUNO [654-5866]

In a message dated 1/17/99 12:14:40 AM Pacific Standard Time, RemoveMEjones_glTakeThisOuTEFN.ORG writes: << Today as i was installing decoupling caps on a circuit board i thought to myself: "Why dont the IC manufactures just incorperate a decoupling cap right in the ic itself?" Any one have any ideas? >> Hi All, Since I am in the business of making machines that make Ceramic capacitors I'll field this one. One problem with capacitors is that the value of the capacitor is directly related to its size. Integrating a capacitor of sufficient value to act as a bypass cap would require that the IC's package grow by quite a bit.. Open your Digikey catalog and check the values VS Size. There are IC sockets that have caps built into them as long as your chip uses a standard Power/Gnd pin configuration (Typically the lower left pin for ground and the upper right pin for VCC) One of the things my company is pioneering is the use of very thin Dielectrics (less than 5 micron) to make very small chip caps. If you half the thickness of the dielectric you increase capacitance by a factor of 4. You also reduce the voltage tolerance. Life is made of trade-offs. We are starting to see the use of 01-02 chip capacitors. Thats 10 thousandths by 20 thousandths. They look like pepper when sprinkled on a sheet of paper. I'm not sure how they handle them but thats not my department! Dave Duley Dreitek Inc. http://www.dreitek.com

A cap requires a lot of space on an IC. Gabriel -----Original Message----- From: Lynx {Glenn Jones} <spamBeGonejones_glspamBeGoneEFN.ORG> To: TakeThisOuTPICLISTEraseMEspam_OUTMITVMA.MIT.EDU <RemoveMEPICLISTTakeThisOuTMITVMA.MIT.EDU> Date: Sunday, January 17, 1999 1:09 AM Subject: [OT] A final question about decoupling caps. >Today as i was installing decoupling caps on a circuit board i thought to >myself: "Why dont the IC manufactures just incorperate a decoupling cap >right in the ic itself?" Any one have any ideas? > >--------------------------------------------------------------------------- --- >A member of the PI-100 Club: >3.1415926535897932384626433832795028841971693993751 >058209749445923078164062862089986280348253421170679

|Today as i was installing decoupling caps on a circuit board i thought to |myself: "Why dont the IC manufactures just incorperate a decoupling cap |right in the ic itself?" Any one have any ideas? Placing a cap on the die itself would be 'theoretically' possible, but extremely expensive. One factor which would be something of a 'killer' is that on most of the chip, one wants to minimize capac- itive coupling between layers (since stray capacitance will slow things down and increase power consumption). Unfortunately, if the chip-making process is optimized to minimize stray capacitance, that will make it difficult to get much capacitance where it actually is wanted. A somewhat more plausible idea would be to incorporate a small cap within the package, but have it external to the die. For some app- lications, this approach might make sense (putting multiple parts in a package isn't all that hard as evidenced by the 12CExx parts) but it would almost certainly cost more than an external cap and there wouldn't generally be enough benefit to justify it. While the need for bypass caps may be somewhat annoying, it's usual- ly not that bad. For proto-board projects, bypass caps can sometimes be handy: if the board has power and ground rails running through the center of the chip, placing the cap under the chip's socket allows for very short lead length *AND* the cap leads can conveniently be used to connect the power/ground rails to the PIC's power/ground pins.

Michael Rigby-Jones wrote: > The last place I worked used some very strange decoupling caps. They were > fitted underneath the IC, without using a socket. They were very thin with > the same outline as the IC and thin pins that could be pushed through the > same holes as the IC. This got around the problem of long leads with the > socket type decouplers. > > Regards > > Mike Rigby-Jones > mrjonesEraseME.....nortelnetworks.com The company that does those is CCI: http://www.cci-msc.com/ They also make some nice bus bars for board stiffening and power distribution. Unfortunately the last time I looked at their Micro-Q caps (several years ago), they were prohibitively expensive. If you were to use more than a few of them per board, it was more cost effective to add a power/ground plane and a good bypassing structure to the pcb. However they are very good if you've already got a noisy design that has to be fixed; less expensive (in the short term) than a redesign. In response to other topics in this thread: Its also important to remember that the clock frequency is usually not the frequency that determines the design point for your power distribution system. The rule-of-thumb is: design frequency = .3 / rise time Where the frequency is in Hz, and "rise time" is the fastest switching speed in use in your circuit, in seconds. So at a rise time of 2.5ns (not uncommon) the frequency you should be designing your bypassing for is 120 MHz (even if your PIC is just ticking along at 4 MHz). (Note that the ".3" ranges between .3 and .5, depending upon whose thumb you are using.) The positioning of bypass capacitors is a popular subject, and a good place for information on it is at Howard Johnson's web site: http://signalintegrity.com , where the subject has been addressed several times in his newsletter: http://signalintegrity.com/news/1_6.htm http://signalintegrity.com/news/2_3.htm http://signalintegrity.com/news/2_1.htm http://signalintegrity.com/news/2_26.htm This reprint from PCB Design Mag is good too: http://signalintegrity.com/articles/straight/bypass.htm Note however, that I believe in all instances these discussions regard pcb's with full power and ground planes; the recommendations there may not be applicable to 2-layer, wire-wrap, or proto-board layouts. This is also a popular subject on the signal integrity mail-list, (subscription info at http://www.qsl.net/wb6tpu/si-list/info.html ), and in its archives at http://www.qsl.net/wb6tpu/si-list/ (Threads on this list are often addressed in Johnson's newsletter.) Gary Crowell