LDO strange problelm

T Chap

I have experienced a weird problem today and am still stumped. I have a pcb design that uses a 3v3 LDO regulator, the input is 5v from the main controller. The previous batch of this board worked fine. I just ordered a new set of panels, with a few new parts included on the panel, and a copy and paste from the old board that worked fine(Eagle). I got in the boards, built 4 today and all 4 have the same exact problem. The output of the regulator sits at 3.46V. There is an IC on the board, Qprox QT1040 capacitive sensor, 4 channel. It outputs on 4 pins a pulse at a set frequency to the copper sensor areas. One of the problems on the regulator output is that it is showing a spike at the same frequency as the pulses that are sent to the copper sensors. Writing this, I realize that the 4 outputs are staggered I believe, so it may be possible with a scope to see if any one particular channel of 4 is in sync. The output of the sensor channels also contain the pulse( sawtooth ) when high, but the active low signal is clean. If I remove the Qt1040, the voltage goes to 3v3. On an older board with the exact design, there is no problem. Like I mentioned, this board is just a copy and paste to a panel containing a number of boards.

At first I assumed that there was a short between one of the sensor pad traces and the 3v3 rail, causing the sawtooth on the rail. So I removed all the caps and resistors in the path from the QT to the pads, but no luck. The output is open col, with 10k pullups. My test circuit works great with other boards of the same design. Can anyone suggest how an LDO can output a higher voltage(3.46) on 4 different boards, on which the QT on all boards(old and new boards) is from the same reel, the regulators are all from the same tube?

LoopyByteloose

Which LDO regulator are you using? Can't help you without reading the product PDF.

These may all be in one tube, but different batch numbers that would indicate different silicon.

Phil Pilgrim (PhiPi)

The regulator could be oscillating due to using the wrong type or value of input or output caps. You have to pay attention, not only to the capacitance, but also to the ESR. LDOs are particularly finicky in this regard.

-Phil

RDL2004

I agree with Phil. You don't say exactly what you're using, but I've used the LM2940 before and it will definitely act strange without the correct (low ESR) output capacitor.

T Chap

http://www.ti.com/lit/ds/symlink/lp38691.pdf


Thanks guys. I have used the exact same parts and schematic on hundreds of these boards and never had a problem. Apparently, I did in fact re-route this latest board, there is no difference in part placement, only the traces are slightly different from the autorouter, and a slightly different ground plane end result.

The sawtooth on the 3v3 output is in sync with the sense pulses that are outputted on the Qprox sensor pins to the copper pads. There are no shorts anywhere, else the QT1040 would not operate correctly, if any one of the sense pins were pulled up or down, that channel would not respond. However all of the channels respond correctly. I can only conclude that this is cross talk due to proximity of one of the large pads and the 3v3 traces, so that the output is compensating by adjusting in sync with the QT pulses, thus the average voltage is higher on a meter( 3.5VDC, but on the scope the output is a sawtooth. Since I cannot find the culprit, I can easily go back a revision and re-order a batch that is tested good, patching a lot of these boards is not an option.

Phil Pilgrim (PhiPi)

So what kind of caps are you using with this regulator? Please provide part numbers.

-Phil

T Chap

http://www.digikey.com/product-detail/en/TCJA106M010R0300/478-3127-1-ND/827213

This 10uf 3216 is in the in and out. There is also a one of these 0603 .01uf caps on the in and out:

http://www.digikey.com/product-detail/en/06035C103KAT2A/478-1227-1-ND/564259

When the QT1040 is removed, the voltage is solid.

The condition is identical on 4 boards. Sawtooth in sync with the QT pulses, average voltage of 3.5VDC.

T Chap

I have discovered that placing my hand over the QT or any some of the copper sensor apds causes the sawtooth to stop and the voltage to go to 3v3.

Leon

That indicates that it's oscillating. A friend of mine had a similar problem many years ago and I found that putting a scope probe on the output made it give the correct voltage.

Phil Pilgrim (PhiPi)

Was this criterion met in your layout?

regulator datasheet wrote:

OUTPUT CAPACITOR: An output capacitor is required for loop stability. It must be located less than 1 centimeter from the device
and connected directly to the output and ground pins using traces which have no other currents flowing through them.

-Phil

T Chap

EDIT. These are met. A trace goes to the .01 and 10uf, and a separate trace goes to the QT.

There is an obvious unrouted trace, VSS. But I have ran a jumper from one side of the section to the other. I tried several different points of connection for the jumper with no difference.

T Chap

Oh well, this may be too much trouble to salvage. I placed a .1 right at the QT, no effect. Some efforts seem to stop the problem but it is not consistent.

T Chap

Success! I placed a .01 directly at the QT1040 supply pins, and a 10R at the LDO output and GND, works perfectly now.

Ron Czapala

RDL2004 wrote: »

I agree with Phil. You don't say exactly what you're using, but I've used the LM2940 before and it will definitely act strange without the correct (low ESR) output capacitor.

Rick,

Interesting! My spec sheet for the LM2940 says to use a .47uf cap on the input and at LEAST a 22uf on the output. (I usually use a 100uf)

**COUT must be at least 22 μF to maintain stability. May be increased without bound to maintain regulation during transients. Locate as close as possible
to the regulator. This capacitor must be rated over the same operating temperature range as the regulator and the ESR is critical; see curve.

Leon

A tantalum capacitor on the output would be best, they are available with specified (low) ESR values.

Phil Pilgrim (PhiPi)

He's using an LP38691/3 (see post #5), not an LM2940. 'Different requirements altogether.

-Phil

RDL2004

The diagram I posted comes from an older NSC data sheet. National was not always consistent with their information. I think the point is that with many LDO type regulators the ESR value of the output capacitor is very important. As far as capacitor size is concerned, I've always pretty much followed the old "rule-of-thumb", 1000uf per amp for the input filter and 100uf per amp on the output (along with a couple of 0.1uf or so right at the regulator input and output).

Phil Pilgrim (PhiPi)

Rick,

You're right: there are at least two considerations for the caps: regulator stability and amount of load. Some regulators I've seen even specify a stability band for ESR; IOW the ESR can be too low, as well as too high. And even though a regulator might be stable with a 1uF ceramic cap, it does not mean that 1uF would provide adequate filtering for transient load swings.

-Phil

jmg

T Chap wrote: »

Success! I placed a .01 directly at the QT1040 supply pins, and a 10R at the LDO output and GND, works perfectly now.

Do those changes one at a time, and change the 10R upwards. You need to identify exactly what trips this, and how much margin you have.

Often hidden in the specs, is a minimum load requirement on some LDOs, and below that, their Vo will creep above spec.

Phil Pilgrim (PhiPi)

jmg wrote:

Often hidden in the specs, is a minimum load requirement on some LDOs, and below that, their Vo will creep above spec.

That could be the case here. Minimum load with his regulator for regulation is 100 uA. The CD4050 draws maybe 90 uA, and the QT1040, 100 uA at its slowest rate. Although that totals to more than 100 uA, the "Output Voltage Load Regulation" percentage is only characterized from 1 mA to full load.

Maybe that regulator is powering other stuff, though, which would render this speculation moot. But if not, it does seem like overkill to have a 0.5A regulator providing mere hundreds of microamps.

-Phil