Home » R/E/P » Dave Hecht - Master Tech » Filter caps--perhaps a crazy notion
 Filter caps--perhaps a crazy notion [message #320791] |
Sun, 02 March 2008 16:00  |
Sam Lord
Messages: 119
Registered: October 2004
Location: Kentucky |
Active Member |
|
|
|
Pardon if I'm having a senior tech moment here--just turned 50, they happen a lot. I have a question. Please shoot it down fast if it has no merit! What happens to say a typical V+ regulator circuit if you *lift* the ground (cathode) of the cap that lies before and parallel to the output--not a coupling cap, understand. This ground lift would of course be *before* bringing up the supply voltage. Let's say the circuit is just an LM317 with a typical load, and the cap is a 10uF electrolytic with anode paralleled with the load and the cathode grounded, just like countless such circuits. How would the circuit behave? Would the cap still perform its energy storage function? The reason I ask is an idea I've had for a while, and need to know if there is any use in it. Thanks for any help, Sam
|
|
| |
| Re: Filter caps--perhaps a crazy notion [message #321240 is a reply to message #320796 ] |
Mon, 03 March 2008 23:15  |
Sam Lord
Messages: 119
Registered: October 2004
Location: Kentucky |
Active Member |
|
|
Thank you Larry! We (my old company, now defunct) used lots of 317s, and the 10uF cap stuck in my head from ancient application notes. However, the ulterior motive of my question was really about alternatives to capacitors for certain functions. Specifically, one designer had urged us to try using a copper bus bar, specifically a solid copper cylinder, in series with the output of a voltage source. These wouldn't replace caps one applies as needed for stability, but would drop the dynamic impedance of the voltage source output. It could be a regulator output or an unregulated rail, for instance. Since regulators already have one or more references to ground, additional large caps on the output(let's say 470uF alongside the 10uf) only store energy. So is the ground reference necessary, and if not, why not just use a series charge reservoir? Of course the speed of this dipole could cause many a source circuit to oscillate, but still...
I know this sounds rather bizarre, but if the the reservoir is not too large (causing the regulator or rails to see a virtual shorted output), then one could perhaps give the output of the voltage source a faster peak current (almost zero ESR). I don't know how to model a series buss bar--SPICE won't do it, I think. I really appreciate your input. I'll understand if you say, "what on earth for?" I'd also like to possess your lab, stellar gear... Sam
|
|
|
| Re: Filter caps--perhaps a crazy notion [message #321429 is a reply to message #320791 ] |
Tue, 04 March 2008 14:05 |
 |
 Larrchild
Messages: 2677
Registered: June 2005
Location: Fort Lauderdale, Florida |
Platinum Member |
|
|
Ah, ok..(and thanks for the kind words!)..
What folks usually do to protect from a short on the output leg is this:
A big honkin diode pointing backwards across the regulator.
Very common.
For good reason!
Regarding the buss bar idea, I'm not sure.
I'm trying to understand better.=)
Attachment: 7805a.gif
(Size: 3.42KB, Downloaded 635 time(s))
Larry Janus
http://tubeequipment.com
|
|
|
| Re: Filter caps--perhaps a crazy notion [message #321464 is a reply to message #321240 ] |
Tue, 04 March 2008 15:40 |
 |
Andy Peters
Messages: 782
Registered: April 2004
Location: Tucson, AZ |
Gold Member |
|
|
| Sam Lord wrote on Mon, 03 March 2008 22:15 |
Thank you Larry! We (my old company, now defunct) used lots of 317s, and the 10uF cap stuck in my head from ancient application notes. However, the ulterior motive of my question was really about alternatives to capacitors for certain functions. Specifically, one designer had urged us to try using a copper bus bar, specifically a solid copper cylinder, in series with the output of a voltage source. These wouldn't replace caps one applies as needed for stability, but would drop the dynamic impedance of the voltage source output.
|
You could go right into a power plane.
| Quote: |
It could be a regulator output or an unregulated rail, for instance. Since regulators already have one or more references to ground, additional large caps on the output(let's say 470uF alongside the 10uf) only store energy. So is the ground reference necessary, and if not, why not just use a series charge reservoir?
|
"ground" provides a (hopefully) solid reference. You don't want the cap charging or discharging through the ground, only through the rail. If this happens, the rail moves, which goes against the whole point of the decoupling.
| Quote: |
I don't know how to model a series buss bar--SPICE won't do it, I think.
|
A simple resistance would do it. A more-complex model would include inductance, but that should be vanishingly low (otherwise, what's the point?).
-a
"On the Internet, nobody can hear you mix a band."
|
|
| |
| Re: Filter caps--perhaps a crazy notion [message #322152 is a reply to message #321464 ] |
Thu, 06 March 2008 11:14 |
Sam Lord
Messages: 119
Registered: October 2004
Location: Kentucky |
Active Member |
|
|
Thank you Larry for the diagram--I do remember using diodes for all kinds of reverse-voltage faults. Does anybody buy less than 1000 1N4001s at a time?? Billions and billions sold!  My worry was mostly that the regulator would see something approaching an AC short on the output as the buss bar is first charged, just like charging a big can--see below.
| Andy Peters wrote on Tue, 04 March 2008 16:40 |
| Sam Lord wrote on Mon, 03 March 2008 22:15 |
...to try using a copper bus bar, specifically a solid copper cylinder, in series with the output of a voltage source. These wouldn't replace caps one applies as needed for stability, but would drop the dynamic impedance of the voltage source output.
|
You could go right into a power plane.
|
Yes, this is what I'm thinking of, a large charged conductor between the source and load. The drawback of the flat power plane is that you can't have lots of usefully large power planes on a board. Also, the plane has practical energy storage limit at a given voltage, dictated by the area and thickness of the plane. I'm thinking 4mils is a pretty thick one. Also if you draw more than one load from the plane, you could expect some unwanted modulation of current going to each load. That's fine and a needed compromise for nearly all ( = tight, tight) digital boards, but not for the cleanest analog signals...
| Andy Peters wrote on Tue, 04 March 2008 16:40 |
| Quote: |
It could be a regulator output or an unregulated rail, for instance. Since regulators already have one or more references to ground, additional large caps on the output(let's say 470uF alongside the 10uf) only store energy. So is the ground reference necessary, and if not, why not just use a series charge reservoir?
|
"ground" provides a (hopefully) solid reference. You don't want the cap charging or discharging through the ground, only through the rail. If this happens, the rail moves, which goes against the whole point of the decoupling.
|
I probably didn't spell out my idea well, but I didn't mean suggest attaching the anode to anything but a rail (for a positive rail)  . I don't mean to replace most filter caps, but rather to add a part which should perform the energy-storage function better, usually leaving a sometimes-smaller filter cap in place.
| Andy Peters wrote on Tue, 04 March 2008 16:40 |
| Quote: |
I don't know how to model a series buss bar--SPICE won't do it, I think.
|
A simple resistance would do it. A more-complex model would include inductance, but that should be vanishingly low (otherwise, what's the point?).
-a
|
I see the model as nearly the opposite, like the power plane (yes without significant inductance). The voltage source would see something like a very big capacitor parallel to the load, except with the cathode floated instead of attached to ground, and without the cap's inductance (although that inductance can sometimes be a help). This means the voltage source (rail or regulator) might need some current-limiting protection, whether intrinsic or external. A power plane is just a single capacitor plate unrolled. OK, roll the plate back up with no dielectric, squeeze it together into a cylinder, attach your voltage to one side and the load to the other, and now you have a buss bar. Filter caps can do two chores, they send noise to ground and they store energy. But in several spots they are used only for the second chore, and this is where I expect the buss bars could work better either as replacements or additions.
I understand that the best route is simply to put this simple part in a circuit and test. I'm just trying to see whether there is any fundamental reason why it would be silly to try. Thanks so much Andy and Larry, it is great to have the varsity here at PSW to help us benchwarmers!
(Note: I sent a very late reply to Dale's post below on April 9th via PM)
|
|
|
| Re: Filter caps--perhaps a crazy notion [message #324680 is a reply to message #322152 ] |
Fri, 14 March 2008 13:28 |
Dale Ulan
Messages: 72
Registered: April 2005
Location: Calgary, Alberta, Canada |
Member |
|
|
Many voltage regulators will oscillate madly without an output capacitor. Some regulators will oscillate, though not as madly, if you have too low of an ESR in the output capacitor. They usually will withstand charging a substantial output reservoir capacitor - read the data sheet! What they will not withstand is a reverse voltage, though. That 1N4001 is there for a reason! I once built a DIY board that blew up in half a second because I forgot the output capacitor. It had a microcontroller and the regulator would go to input voltage (+13.8V) then down to ground. The average was 5.0 volts but the peaks blew up my poor HC11 right away. It got hot. Oops.
Old mainframe and minicomputer boards used to have power distribution bus bars running up and down the boards, especially when a lot of stuff was two-layer. I remember going through some NCR ALU boards that were equipped with those. They still had local decoupling capacitors liberally scattered around the board. On every chip, actually.
Multilayer boards are obviously better for digital circuts because they offer transmission-line performance that you can't get at such a low cost using any other method, allow for sufficient current flow into chips, and offer a lot of EMI resistance. The power/ground plane helps to keep the voltages relatively constant but you still need as much decoupling capacitance around the board. One advantage from a driver's perspective is that you are driving a transmission line so the currents (due to the fast risetimes) tend to balance on a signal trace and its ground/power plane right below it.
As for bus bars, yea, but... you are probably better off with local power decoupling everywhere you need it, paying attention what the load is and where it needs to be decoupled too. Having one solid reference (ground) is common. I hear arguments each way about decoupling the power supplies to each other or to ground, especially in +/- circuits.
For inspiration on bus bars, take a look at high-power motor variable frequency drives. Those guys take bus bars and decoupling and capacitance very seriously. When you are switching anywhere from 100 to 5000 amperes of current (or more) you need to pay attention, if you don't you might loose your hearing, or maybe an eye, depending on where the shrapnel flies. I've blown up a few small IGBT's.... 300 amp. They are loud enough.
Anyways, the usual approach is to use bus-bar to keep resistance down, and every half-bridge (consisting of two IGBT's and two diodes) has its own decoupling capacitance, usually very low inductance, and usually between 1uF and 10uF, bolted right to the IGBT terminals. Decoupled where the highest circulating current and fastest edges occur. This is very similar to the approach used in high-speed digital circuitry with local decoupling capacitors, but on a much bigger scale.
Dale Ulan
10000 Cows Recording Studio
Calgary, Alberta, Canada
|
|
|
| Re: Filter caps...Oops, a stupid notion! [message #337942 is a reply to message #324680 ] |
Wed, 30 April 2008 14:01 |
Sam Lord
Messages: 119
Registered: October 2004
Location: Kentucky |
Active Member |
|
|
To Larry Janus, Andy Peters, and Dale Ulan: Here is a copy of a PM I just sent to Dale, with sincere apologies to ALL of you:
"Dale, I'll soon tie up the thread I started on filter caps. I'm embarrassed to say I forgot Gauss' Law, and made a fool of myself. I actually studied it over 25 years ago, but, um, forgot its indication for the futility of storing energy via voltage application to a low-surface-area conductor. Thanks for helping me, and I apologize for suggesting what is patently impossible. Actually I think I'll just post a copy of this note to the thread.
Best regards, Sam"
|
|
| | | |
| Re: Filter caps--perhaps a crazy notion [message #338433 is a reply to message #338423 ] |
Fri, 02 May 2008 03:20 |
J Hinson
Messages: 84
Registered: December 2007 |
Member |
|
|
| dcollins wrote on Thu, 01 May 2008 23:51 |
There's a number of 3 terminal regulators made by Linear Technology that don't require any protection diodes and I highly recommend them.
DC
|
Hi there Dave.
If you mean the low drop out regs such as the LT1085 then, yes, they are a step up from, say, a 317. The way I read the LT data sheet though, a protection diode as illustrated above is recommended for large (1000uF to 5000uF) output capacitance. Hmm? I'll read the data sheet again when I'm more alert.
Anyway, if you're into serious audio then none of the 3 terminal regulators are really worth using for anything but as a pre-regulator to something else at the load. Even the simplest of discrete regs/buffers sound better in my opinion.
Edit: On second thoughts, I take that back slightly. I have used 3 terminals in serious audio as current regulators within a discrete design. And I also frequently use them (LT1085's) in heater supplies for valve gear because of the low drop out.
Cheers.
Possession is 9/10ths of the problem.
|
|
|
| Re: Filter caps--perhaps a crazy notion [message #338625 is a reply to message #338433 ] |
Fri, 02 May 2008 16:04 |
|
Andy Peters
Messages: 782
Registered: April 2004
Location: Tucson, AZ |
Gold Member |
|
|
| J Hinson wrote on Fri, 02 May 2008 01:20 |
| dcollins wrote on Thu, 01 May 2008 23:51 |
There's a number of 3 terminal regulators made by Linear Technology that don't require any protection diodes and I highly recommend them.
DC
|
Hi there Dave.
If you mean the low drop out regs such as the LT1085 then, yes, they are a step up from, say, a 317. The way I read the LT data sheet though, a protection diode as illustrated above is recommended for large (1000uF to 5000uF) output capacitance. Hmm? I'll read the data sheet again when I'm more alert.
|
The big diode is there to prevent a large output capacitor from discharge back into the regulator, which it will do if the input voltage drops below the output voltage (the voltage on the cap). This is standard for any regulator, although as noted many newer regulators have that diode built in.
-a
"On the Internet, nobody can hear you mix a band."
|
|
| | |
Goto Forum:
Current Time: Mon Jul 7 02:26:45 CDT 2008
Total time taken to generate the page: 0.03144 seconds |
Recording Engineering & Production: The nuts and bolts of making records better, not just louder and cheaper. Some of the brightest minds from the creative and scientific world of audio discuss the elements of audio production that interest you most.
Members
Search
FAQ
Register
Login
Home
