Improving the LM3x7 regulator circuit

[Christer]

Originally posted by sobazz
What is the purpose of R7? SIMetrix seems to like the circuit better when R7 is removed.

It's probably there to provide a dummy load that
draws the required minimum current from the regulator. Even
if your simulator doesn't like (why would that one be a problem?)
your regulator will probably like it.

[sobazz]

Yes, I thought so. But with a dummy resistance of 150ohm in order to draw 100mA of current the simulation shows around +-0,5v ripple...

[sobazz]

I'm looking forward to Fred's real world results.

[Fred Dieckmann]

What I ended up building is a one transistor addition to the three terminal regulator. The three transistor circuit doesn't add much to the performance and is not worth pursuing IMHO. Any good small signal transistor with an HFE of several hundred will work. Use a 1K resistor to program 1.25 mA through the transistor the error current through the voltage divider that sets the voltage will be 1.25mA divided by the Hfe of the transistor. You will want at least 100 times this current through the resistor divider to minimize the error. The transistor should have a base resistor of about 100 ohms to limit fault currents from discharge the bypass capacitor connected to its base and to provide better RF stability. All the bypass caps in the schematic are metalized film except the output electrolytics. About 0.1 to 0.47uF on the input to the regulator and 1 to 5 uF for the cap to ground connected to the base are fine. These are not that large size caps if using a metalized film cap. Stacked layer Polyesters would be about the smallest and polypropylene the largest (but with better performance). This circuit is not that critical for cap values. Put a 0.5 ohm resistor in series with the electrolytic output cap to ground. The main idea of all this was to use film caps of fairly low values with the transistor to make the impedance at the AJD terminal lower. There should be a large amount of leeway in the parts values for this circuit while still achieving a great performance improvement over the data sheet bypass circuit. I did not show the preloading resistors on the outputs but I would shoot for a minimum of 25 and preferably 50 to 100 mA. The output impedance is proportional to this current load and higher currents give lower output impedance. They also give more ripple on the raw supply so a balancing act is required to optimize both of these deign goals.

[Jocko Homo]

You could replace the 1K resistor with one of those CCS diodes.

Oops.....they cost >$1.

Maybe you ingore that..........[joke]

Jocko

[Alcaid]

C1,C4 = 0.1-0.47uF MKP/MKT
R1,R6 = 1k
R2,R7 = 100R
R3,R8 = 220R
R5,R10 = 0R5
C2,C5 = 1-5uF MKP/MKT
C3,C6 = something like 4700uF?
R4,R9 = 2k5 (for 15,45V output)?
Q2 = ?

Is this correct values? What transistor will you recommend?

[jcarr]

Fred: In the quest to make a silk purse from a sow's ear, interestingly enough, I also made a souped-up 3-terminal regulator circuit strikingly similar to yours. However I placed C2/C5 in your circuit directly at the base of Q2, to lower the drive impedance into Q2.

The cap used here turned out to be surprisingly audible, so I would suggest doing some swapping and listening. My own preference seemed to be tied to a good high-frequency impedance curve on the cap. OS caps and tantulums sounded better to me than any normal electrolytic tried, including audiophile as well as industrial grades.

hth, jonathan carr

[Fred Dieckmann]

"Is this correct values?"

Absolutely not........... The idea is to make R3 R4 large to get a good low frequency corner for the filter. R3 should be 10K or above (probably not greater than 50K for the transistor base current to be small in comparison to the current through R3) I am using an A970 and C2240 for the transistors but anything with a beta of 400 or more at 1 mA will work. BC550 and BC560 would probably work very well and as would the 2N5087 and 2N5210. The Hfe for the ZTX450 and ZTX550 might be a little low but might work with adjusting the value for R4 and R9 to compensate for the larger base current. The electrolytic on the output should be in the 220uF to 1000uF range and don't leave out the series resistor. Quality and not quantity is what one should look for in the caps used.

Sorry I was vague, it was 4 in the morning.........

Fred

[CheffDeGaar]

Originally posted by jcarr
Fred: However I placed C2/C5 in your circuit directly at the base of Q2, to lower the drive impedance into Q2.

The cap used here turned out to be surprisingly audible, so I would suggest doing some swapping and listening. My own preference seemed to be tied to a good high-frequency impedance curve on the cap. OS caps and tantulums sounded better to me than any normal electrolytic tried, including audiophile as well as industrial grades.

Is it a way to form a LF filter to avoid HF noise entering the base of Q2 ?

Have you tested with an additional cap across R4/R8 (> 100uF) ? According to ALW (see here), this cap seems to do a nice job.
Well, it's not exactly the same purpose, but I was wondering...

Edit : Sorry Fred, was typing while you replied... My additional cap question is in the case C4 is directly connected to Q2's base...

[dimitri]

Sorry guys, but it is unrealizable dream to improve the regulator which initially has 5-25kHz bandwidth …