1. Home
  2. Electronics Design
  3. Regulator minimum capacitor ESR

Regulator minimum capacitor ESR

Greetings,

I'll be using a TI TPS767D301 LDO to power my DSP. I find it easier to deploy ceramic caps rather than Ta, since it's less trouble fussing with CAD libraries, making sure the parts I can order fit the pads, etc., or rather than building my own Ta cap lib.

The point is, the LDO wants a min ESR of about 50 milliohms, but a parallel combo of many ceramic caps is likely to be lower than this.

Is it Ok to put a resistor at the LDO output terminal to add ESR there? The weird thing about this, of course, is that the load would be on the other side of the resistor, which isn't great. OTOH, the load is still connected to the low ESR capacitance, which is good.

Anyway, it's easier to add the resistor there, rather than in series with the caps, or spend a few more hours switching to Ta caps, which may still have too little ESR with many in ||.

--
_____________________
Mr.CRC
crobcBOGUS@REMOVETHISsbcglobal.net
SuSE 10.3 Linux 2.6.22.17
Reply to
Mr.CRC
Loading thread data ...

For regulator stability I'm pretty sure that putting the resistor where you propose is OK. Weird, but unless that DSP is sucking way more current than I think, OK.

--
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?

Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com
Reply to
Tim Wescott

Besides, I heard from a customer that the Great Tantalum Capacitor Shortage is back on. I haven't checked news sources (or distributor stock) to verify, though.

--
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?

Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com
Reply to
Tim Wescott

In these cases I connect the load directly to the LDO and put a resistor in series with a 10uf capacitor close to the LDO. Having ceramics further downstream doesn't seem to hurt.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------
Reply to
Nico Coesel

e

Your placement degrades DC regulation more or less depending on load current. Something like this would be more in line with reality: Please view in a fixed-width font such as Courier.

. . . . . --------- . | . | . OUT|-----------+------------->

. | | . | [0.06R] . REG | | . | --------------- . | | | | | . | =3D=3D=3D =3D=3D=3D =3D=3D=3D =3D=3D=3D . | | C | C | C | C . --------- --------------- . | | . | | . -----+----------------+------------->

. | . --- . /// . . . . . .

Reply to
Fred Bloggs

50 mOhm is NOT a lot of resistance. Unless ALL the ceramic caps are RIGHT next to the regulator, I doubt you will have any problem. If most of the ceramics are at the load chips, the distributed resistance to those caps can be enough. Possibly an inch or so of modest circuit trace from the regulator would be 50 mOhm.

Jon

Reply to
Jon Elson

Yeah, that makes sense. I suppose I'll put the 2-4 22uF bulk caps in a gang like that.

But I still worry about all the other little 0.01-0.1uF guys sprinkled around paralleling to less than 50 milliohms.

--
_____________________
Mr.CRC
crobcBOGUS@REMOVETHISsbcglobal.net
SuSE 10.3 Linux 2.6.22.17
Reply to
Mr.CRC

=3D=3D=3D

Actually you bring up a good point about the other decoupling caps sprinkled around. This actually strengthens the argument for keeping that series 0.06R + bulk caps in shunt with the load as recommended. The reasoning is as follows: the regulator does not even see the loading of the other decoupling caps except at higher frequency, and at the higher frequencies the ESR + bulk caps combination looks like a simple ESR alone, then the phase shift of the decoupling caps only comes into full effect when their combined reactance equals the ESR, which is to say when the high frequency output current splits evenly between the ESR and small decoupling caps. Typically the properly compensated LDO will have an open loop crossover 0dB frequency of 10's of KHz, so to be conservative, you can limit the total decoupling capacitance to be such that the combined reactance does not equal the ESR until say 1 MHz or so. Then the induced phase shift will not occur until the loop gain is well below 0dB ( like -40dB) and stability is not corrupted. Numerically this is expressed as ESR=3D1/ (2*pi*freq*Ccplg) or maximum Ccplg=3D 1/(2*pi*freq*ESR)=3D 2,6uF ( for ESR=3D0.06 as recommended and freq=3D1MHz). This is a very generous allotment of 0.01u decoupling caps- about 260 of them! Another word of caution is to back off that excessively large filter cap and use the 10u recommended. Too large a bypass cap is as bad as too high an ESR for stability purposes. Use the recommended 10u.

Reply to
Fred Bloggs

Just use an electrolytic as the data sheet suggests. Note Jon Elson's comments and don't worry.

Reply to
John S

The AVX ceramic X5R dielectric 10u at 10V has an upper limit ESR of .

5 ohm, so this would be right in the middle of your stability range, no external R required. External bypass caps at the load you can handle are now at 0.3uF total.
Reply to
Fred Bloggs

to

h

the

ll

That's not how that works.

Reply to
Fred Bloggs

1 oz copper is about 500 uohms/square. So a 10 mil trace 1" long is about 50 mohms. 
--

John Larkin, President       Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators
Reply to
John Larkin

er to

with

or

re?

on the

still

hnology.com=A0 jlarkin at highlandtechnology dot com

Yep- but that in series with say a 0.01u decoupling cap has a corner frequency above 300MHz. Sounds kind of absurd doesn't it? The decoupling caps do not degrade the stability for the reason stated in my post, they do not introduce phase shift of any significance within the 0dB bandwidth of the feedback loop, there's no need for series resistance, or even 1" traces all over the place, just put it down wherever.

Reply to
Fred Bloggs

Ok, I buy that.

That too.

Uh-huh.

Fine.

That's the min. recommended. There's 3x22uF Ta on my reference board. I doubt the DSP will have huge current surges, as I don't plan on turning on/off major blocks of functionality in my application. Though, I do want to design the board to be flexible enough to work with other people's low-power apps.

So I think I'll put in locations for several caps, and just populate as necessary. With a resistor option as well.

--
_____________________
Mr.CRC
crobcBOGUS@REMOVETHISsbcglobal.net
SuSE 10.3 Linux 2.6.22.17
Reply to
Mr.CRC

And for any sensible modern PCB design, the power is going to be distributed to a high pin-count chip via a solid Cu plane, so the resistance can be very low.

All of the caps are thus effectively right at the regulator.

--
_____________________
Mr.CRC
crobcBOGUS@REMOVETHISsbcglobal.net
SuSE 10.3 Linux 2.6.22.17
Reply to
Mr.CRC

If you wanted to add 50 mohms, a trace would do it.

But why don't you use a regulator that's happy with ceramic caps? Or add a tantalum, which usually makes things happy?

--

John Larkin, President       Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators
Reply to
John Larkin

The LDO is easy to use since it has the two supplies, one adjustable (needed to get 1.9V for full speed 150MHz clock), with reset timeout. Combined with another supervisor chip, it can perfectly sequence the power for the TMS320F2812 with every detail as required.

It is easier to make the TPS767D301 happy than research and deploy another dual LDO.

Even switching to some Ta or good Al caps wouldn't be hard. I'll just make some cap footprints that can fit anything.

--
_____________________
Mr.CRC
crobcBOGUS@REMOVETHISsbcglobal.net
SuSE 10.3 Linux 2.6.22.17
Reply to
Mr.CRC

OK, why don't you use an LDO that's happy with ceramic parts?

Or use an LDO with a powergood and an enable.

So use a tantalum cap.

So you don't want a solution.

Reply to
krw

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.