Unstable Linear Voltage Regulator

[laingalion]

This is my first attempt at a H-Bridge amplifier... and electronics in general.

My application is very specific and I am drawing a lot of power, 25+ amps RMS at 15V from a lithium battery at roughly 20 kHz.
Earlier tests went well but I tried to draw more current and the battery voltage became unstable. This translated to a huge ripple on the 5V control side (see attachments)
Disregard the value of the regulator in the schematic, I used it because it had the same footprint. What I am actually using is a 7805:
http://www.newark.com/stmicroelectronics/l7805c-v/ic-linear-voltage-regulator-5v/dp/89K1378
I also have a 4.7mF capacitor at the battery terminals, that helped a lot.

I can limit the amount of current draw so that the battery voltage does not go unstable, but I was curious as to how to prevent such a huge distortion on the 5V side of a regulator.
I read in the datasheet that a resistor on the input would help but I don't think it would help THAT much.
Is there some unstable feedback happening with the regulator when it tries to regulate itself?

A very beginner question, bare with me and thanks in advance.

[Jay_Diddy_B]

Hi,

You can try this:

1) Increase the value of C1 to 100uF or 220uF Low ESR 105C Electrolytic capacitor.

2) Add a ceramic capacitor  between pins 1 & 2 on the regulator. The capacitor should be physically close to the regulator IC.




I would try these first.

Regards,

Jay_Diddy_B

[Andreas]

Is there some unstable feedback happening with the regulator when it tries to regulate itself?

A very beginner question, bare with me and thanks in advance.

What comes into my mind is that your regulator GND wiring might be wrong; picking up the ripple on the power GND line.
The GND pin should be attached to the 5V (logic) load and not anywhere on the high current path.

With best regards

Andreas

[nickm]

The waveforms look so close its probabily power supply ripple is getting right through.  R-Cs will help as long as they're placed well below 20kHz.  10ohms and 10uF is around 2kHz.  Increasing the output capacitance of the LDO too much can make the PSRR worse because the open loop gain crosses over much earlier, but too small output capacitance can result in stability problems.  If you have an assortment of caps try messing around with a range of values and see what you get.  If you're worried about stability try adding a 1ohm in series with your 10uF cap on the output. 

[AndyC_772]

There's a clue in the 7805 data sheet. Its power supply rejection figure is quoted at 120 Hz, but the noise on your 12V supply is nearly three orders of magnitude higher in frequency than that.

It looks to me as though the 7805 just can't respond fast enough, and you need a regulator with much wider bandwidth and/or a passive filter network on the 12V input to the regulator.

[SeanB]

LC filter on the input will help a lot, very few regulators will handle high frequency noise on the input well, often passing it straight through or worse oscillating.

[peter.mitchell]

don't put the cap on the battery terminals, put it as close as you can to the input of the regulator!

[Jon86]

You're gonna need a better regulator. The 7805 is very basic, and isn't going to like all that high frequency stuff.

[laingalion]

What comes into my mind is that your regulator GND wiring might be wrong; picking up the ripple on the power GND line.
The GND pin should be attached to the 5V (logic) load and not anywhere on the high current path.

I think you are right. I considered this and split the paths of the powerside and the control side, but they still share a few inches of wire which is part of the battery.

1) Increase the value of C1 to 100uF or 220uF Low ESR 105C Electrolytic capacitor.

2) Add a ceramic capacitor  between pins 1 & 2 on the regulator. The capacitor should be physically close to the regulator IC.

Will do, thanks for the advice

The waveforms look so close its probabily power supply ripple is getting right through.  R-Cs will help as long as they're placed well below 20kHz.  10ohms and 10uF is around 2kHz.  Increasing the output capacitance of the LDO too much can make the PSRR worse because the open loop gain crosses over much earlier, but too small output capacitance can result in stability problems.  If you have an assortment of caps try messing around with a range of values and see what you get.  If you're worried about stability try adding a 1ohm in series with your 10uF cap on the output.

This regulator is not an LDO (I think?) but I understand that your suggestions also directly apply to the regulator I am using. I will try your suggestion, hopefully i have a resistor small enough sitting around. Thanks

There's a clue in the 7805 data sheet. Its power supply rejection figure is quoted at 120 Hz, but the noise on your 12V supply is nearly three orders of magnitude higher in frequency than that.

It looks to me as though the 7805 just can't respond fast enough, and you need a regulator with much wider bandwidth and/or a passive filter network on the 12V input to the regulator.

LC filter on the input will help a lot, very few regulators will handle high frequency noise on the input well, often passing it straight through or worse oscillating.

Yes, it appears like the control side waves are responding directly to the supply wave. What is confusing me is that the waveform are mirroring. When the high side voltage goes up, the low side voltage goes down and vise-versa.

don't put the cap on the battery terminals, put it as close as you can to the input of the regulator!

I have the 10uF capacitor right next to the regulator. Are you suggesting I put the gigantic 4.7mF capacitor next to the regulator? I will try that. Thanks

You're gonna need a better regulator. The 7805 is very basic, and isn't going to like all that high frequency stuff.

Im afraid this may be the case. Do you have any suggestions for a relatively cheap regulator? Something like an isolated switching regulator?



Thank you all for your replies. Once I get home from work I will try your suggestions and report back.

[tszaboo]

2) Add a ceramic capacitor  between pins 1 & 2 on the regulator. The capacitor should be physically close to the regulator IC.

NO! and again NO!
7805 and other old regulators became unstable because low ESR ceramic capacitors. Remove them, use newer type if you need ceramic, or other capacitor.

[Jon86]

The LM338 has a 30dB ripple rejection at 20KHz, you might want to try something like that.
You don't need a switching converter though, a linear regulator will do just fine.

[Jon86]

Also, LM2941.

[laingalion]

The LM338 has a 30dB ripple rejection at 20KHz, you might want to try something like that.
You don't need a switching converter though, a linear regulator will do just fine.

But looking at the 7805 datasheet:
http://www.farnell.com/datasheets/1696820.pdf
On page 37, at 20 kHz there is a 70 dB rejection

NO! and again NO!
7805 and other old regulators became unstable because low ESR ceramic capacitors. Remove them, use newer type if you need ceramic, or other capacitor.

On page 28 of the datasheet it says "A 0.33 µF or larger tantalum, mylar or other capacitor having low internal
impedance at high frequencies should be chosen" (referring to the input capacitor)
Is internal impedance the same as ESR or am I confusing the two separate terms?



I am out of electrolytic and large ceramic capacitors so I will re-test once I grab some tomorrow. I will try stick them right on the legs of the regulator...
Thanks for your help thus far everyone. Very informative. I will report back tomorrow.

[nickm]

http://www.ti.com/lit/an/slyt202/slyt202.pdf
http://www.ti.com/lit/an/slyt194/slyt194.pdf

Check out these two documents.  The first shows the ripple rejection falls off as loop gain decreases.  The 70dB of loop gain they are showing is without any external capacitance so they are really pushing up where the dip in PSRR occurs.  With the reasonable amount of output capacitance the PSRR will fall off much sooner so you might be somewhere in the minima where decreasing or increasing output capacitance with improve psrr.

The second one illustrates what affects stability if you're concerned/interested in that.