Spartan® Family FPGAs

[joshua5526]

Hello. I with to power a XC6SLX16-3FTG256C with a small linear regulator. I had picked out a LM1117MPX-3.3/NOPB from National Semiconductor http://www.national.com/ds/LM/LM1117.pdf second sourced by ADP3339AKCZ-3.3-R7/BKN by Analog Devices http://www.analog.com/static/imported-files/data_sheets/ADP3339.pdf for the 3.3V rail.

I now see that a regulator like this requires a high ESR cap > 300mOhms while the Spartan6 requres a low ESR cap. With further research I see that Digikey has some special low ESR linear regulator such as anyCAP and PowerWISE serise. However if I go with a more specility regulator I'm not so sure I will be able to match up a second source.

I am wondering if anyone else is useing a more "normal" linear regulator and is somehow getting around the high ESR requirement, if if someone could point me in the direction of finding a small second sourceable regulator. I am trying to stay away from switching in order to reduce design complexity and have something more standardized. I am budgeting a minimum of an amp of power for both 3.3V and 1.2V.

[austin]

j,

Bypassing the regulator right at the regulator with the capacitor that makes it happy is the first rule.

Then, by the time the power (and ground) planes (wires, paths) get to the FPGA, you place the recommended bypass capacitor there (to make the FPGA happy).

Now, you are all done.  As long as the distance from one cap, to the others is more than a few inches, even of a solid power and ground plane, they are isolated by the pH/square, and the ohms/square of the planes/paths/wires.

[eteam00 · Accepted solution]

Austin is a master of such things.  So take my advice as secondary to Austin's.

Personally, I am quite risk averse, and a linear regulator which requires a 300mOhm ESR output cap strikes me as a device which is susceptible to oscillation.  There are a number of linear regulators which have earned a reputation for being "squirrelly", or unstable, and so I try to avoid them.

There are sooooo many linear regulators on the market, why choose one with such a red flag?  Note that the Analog Devices regulator does not have the same stability warning as the NatSemi (now TI) regulator.

When you purchase a capacitor, do you shop for a minimum ESR cap?  No!  You shop for a capacitor with a maximum ESR, and this is how capacitors are specified.  Here's a workaround for regulators which may have an instability problem:  add a stub of interconnect (trace) between the regulator output plane and the capacitor.  This bit of circuit board etch will provide enough series R and L to help settle the regulator, and you no longer need to worry about a capacitor which is "too good".

If this sounds hokey to you, then you should consider an alternative to the (dirt cheap) LM1117.  Note that the ADP3339 (a 1.5A component) is considerably more expensive than the LM1117 (a 800mA device).

Just my two pence...

-- Bob Elkind

[austin]

Bob,

I did not even think about that, but you are spot on:  any regulator that requires a special capacitor right at its terminals might be of the "squirrelly" kind.  I would just google that part number, and stability, or oscillation, and see what pops up.

Xilinx did have a power central website, with approved regulators.

But, that seems to have gone away.

Now we have pcb user's guides, but we do not have specific vendors listed.

We did have Intersil, TI, and other vendors who we had approved.  I guess it is just too hard to keep up with all them anymore.

[eteam00]

We did have Intersil, TI, and other vendors who we had approved.

I believe that all of the major manufacturers are equally capable of producing both good and not-so-good regulator designs.

Some suggestions:

LP38690/LP38692

MCP1826

ADP3338/3339

MIC37100

TL1963A

TLV1117LV  <---- note this version of the 1117 regulator is stable with low-esr ceramic caps!  And cheap, too!

ZLDO1117  <---- note this version of the 1117 regulator is stable with low-esr ceramic caps!  And cheap, too!

-- Bob Elkind

[joshua5526]

Thank you Austin and Bob for the quick response. You have both given me something to work with.
~Joshua

[joshua5526]

I did the Google search Austin recommended and found a few pages warning about the LM1117. I also went though the data-sheets of the suggested for the parts Bob recommended. Almost everyone would work. 1 Amp, SOT233 package, Ceramic bypass happy. For future readers, here are the datasheet links.
MCP1826 $0.98 http://ww1.microchip.com/downloads/en/DeviceDoc/22057a.pdf
ADP3338/3339 $2.80 http://www.analog.com/static/imported-files/data_sheets/ADP3338.pdf
MIC37100 $1.25 http://www.micrel.com/_PDF/mic37100.pdf
TL1963A $1.83 http://www.ti.com/lit/ds/symlink/tl1963a-33.pdf
TLV1117LV $.27 http://www.ti.com/lit/ds/symlink/tlv1117lv25.pdf
ZLDO1117 $.28 http://www.diodes.com/datasheets/ZLDO1117.pdf

[sujithkuroor]

I planned to use TLV1117LV in my spartan 6 LX4 board. But how we can meet the start up ramp time requirement of FPGA ie .2 ms to 50 ms?

[austin]

s,

I = C dV/dt

One places a suitably large capacitor on the regulator.

For example, for a 1A limited regulator, I = 1 amp.

Then if C = .001F (1000uF, or two 470 uF), for a 1.2v supply:

1 = .001 * 1.2 / dt

solving for dt

dt = .0012, or 1.2 millisecond.

The capacitors we recommnend almost surely cause you to meet the ramp up time. (Because regulatorws cannot source infinite currents!).