LM 338 regulator for GC with power on delay

[ajju]

3 days on EAGLE..this is what i cooked up.
I couldnt locate all required footprints and yet to explore how to create custom footprints.

OK, the pcb is for a single half of LM338 based regulator for LM3886, includes multiple input transistors in the ratio 1x .1x .01x, Power on delay, and provision for SMD tantalum output transitors for improving power supply impedence figures.

ajju

[sek]

Hi Ajju,

I'm very interested in working out a schematic and a board for the LM338 as a regulator in a GC supply. Actually I'm planning to integrate both the PSU and the GC on a single board with SMT components (to keep it tiny).

Creating custom parts in eagles library editor is actually easy and should be done in the first place - my experience. I haven't sorted all the parts out yet, because your board view makes this rather difficult.

Would you like to share your schematic with us (either as a picture and/or as the .sch and .brd file)?

thx,
Sebastian.

[ajju]

Ooops Sorry ...what am i saying..! please excuse that momentary lapse of reason...

Please replace transistors with Capacitors in my starting post..!!

ajju

[ajju]

Hi Sek,

This is the schematic I started with...!!

ajju

[Arius]

Don't forget the usual high freq 0.1uF decoupling cap at the regulator output.

Try to place the small value caps nearer to the regulator terminals.

If you must use solid tantalum caps, be sure to derate them to 50% of rated voltage. Solid tants have a nasty habit of blowing up. I've seen that happen too many times. Polymer tants are much better.

Good luck.

[ajju]

Many thanks for the inputs. I'm aware of the nasty nature of Tantalums.! I'm using them just because they are the best decoupling caps around and also offer good size to cap ratio..!

Will try to accomodate a O.1uf on the output as close as possible to the output pin.

ajju

[diy_audio_fo]

Very few notes:

1) Put a (at least (and beside high freq decoupling)) 1000uF capacitor at the output of the LM338 - but very close to LM3886 (see LM3886 National application note)

2) The 47uF capacitor at the adj pin is unusefull beause of 100uF at the base of Q1 (the equivalent capacitor at the adj pin is 100uF * gain of Q1). In any case if you keep the 47uF capacitor put a diode in the same manner as D1 at the adj pin.

3) To get a 24 V DC output voltage you need (at least) a 24 V AC transformer (if you need detailed calculation, ask).
If you let a +10% fluctation of mains voltage you get
24*sqrt(2)*1.1=37 V peak at the input of the regulator, near the 40V rating of the regulator. Be aware of the no load - full load regulation of the transformer!

4) Put a reverse (anode to ground) diode at the output

5) What about the negative regulator?

Regards

[ajju]

Originally posted by diy_audio_fo
Very few notes:

1) Put a (at least (and beside high freq decoupling)) 1000uF capacitor at the output of the LM338 - but very close to LM3886 (see LM3886 National application note)

The idea was to use this as a dedicated power supply board. The LM3886 will sit on another PCB and so will the 1000uf Caps.

2) The 47uF capacitor at the adj pin is unusefull beause of 100uF at the base of Q1 (the equivalent capacitor at the adj pin is 100uF * gain of Q1). In any case if you keep the 47uF capacitor put a diode in the same manner as D1 at the adj pin.

Yes, you are correct. A diode is indeed essential to allow a fast discharge path for C1 . I missed that in the schematics.

Q1 does not operate in the active region at all times. Only on power on Q1 has a role. During power on Q1 will conduct for a time approx. determined by rise time of R3 C10. Q1 eventually gets cut off. (Q1 also might have a role to play following recovery from an output short or a very heavy load.)
The 47uf has been added and is essential to reduce ripple and noise effects.

3) To get a 24 V DC output voltage you need (at least) a 24 V AC transformer (if you need detailed calculation, ask).
If you let a +10% fluctation of mains voltage you get
24*sqrt(2)*1.1=37 V peak at the input of the regulator, near the 40V rating of the regulator. Be aware of the no load - full load regulation of the transformer!

asking...!
I thought i can settle for a 22V or 20V transformer, with some tradeoff in ripple at heavy loads.

4) Put a reverse (anode to ground) diode at the output

Why is this reqiured...? is it to protect from inductive spikes..?

5) What about the negative regulator?

Negative regulator will be the same..
The transformer will have a split secondary and each secondary will have its own regulator. After the regulator, at the output the GND and +24V out of the two seperate regulators will be tied to gether to form the ground, and to get the dual supply.

However for this its a must tha we use a transformer with a split secondary.

-ajju

[diy_audio_fo]

> Q1 does not operate in the active region at all times. Only on power on Q1 has a role.

>The 47uf has been added and is essential to reduce ripple and noise effects.

Good points. I missed them.

>asking...!
>I thought i can settle for a 22V or 20V transformer, with some tradeoff in ripple at heavy loads.

Vout= Output voltage
Vin-out = minimum differential input-output voltage for regulator
Vr=Vripple
2Vd= Voltage drop on the bridge
Vpeak_sec= Transformer peak secondary voltage
Vsec= Transformer RMS secondary volatge

Let's say:

Vout=24V
Vin-out = 3V
Vr =10% *(Vout + Vin-out)= 2.7V; 10% is a "rule of thumb" number choosen not to stress the transformer with high peak current to charge the filter capacitor
2Vd=2V
-10% = low mains fluctuation

Vpeak_sec= (Vout+Vin-out+Vr+2Vd)/0.9 = 35.22V
Vsec= 35.22/sqrt(2)=24.91V


>>4) Put a reverse (anode to ground) diode at the output

>Why is this reqiured...? is it to protect from inductive spikes..?

No inductive spikes.

National in his design note about fixed or variable voltage regulator says (more or less):

Let's say we have a (heavy) transversal load (plus to minus - not ground referred) . If the positive regulator "starts" before the negative regulator (with no diodes) the output of the negative
regulator is pulled at +Vout so it sees a Vin-out double than normal. In this case you have good chance that the negative regulator never starts (due the foldback current limit) or blows out.

Obviously the symmetric thing happens if the negative starts before the positive.

With the diodes the output voltage is clamped at maximum plus or minus Vd from ground and the Vin-out is limited to a safe value.


>Negative regulator will be the same..
>The transformer will have a split secondary and each secondary >will have its own regulator. After the regulator, at the output >the GND and +24V out of the two seperate regulators will be >tied to gether to form the ground, and to get the dual supply.


In this case which is the Vin-out of the "upper" regulator under short circuit condition i.e. V+ connected to V- (GND of the "lower" regulator)?

Please refer to Fig. 2 and associated text on http://www.linear.com/pdf/dn21.pdf for some hints.

Regards

[ajju]

>-10% = low mains fluctuation

>Vpeak_sec= (Vout+Vin-out+Vr+2Vd)/0.9 = 35.22V
>Vsec= 35.22/sqrt(2)=24.91V

I hadn't considered the mains fluctuation. Its infact operating very near Vin Max.
In this part of the world the power lines are not that clean, if you include +/- 10%
then we are indeed very close..!


>National in his design note about fixed or variable voltage regulator says (more or less):
>Let's say we have a (heavy) transversal load (plus to minus - not ground referred) . If the positive regulator "starts"

True, Quite possible. And the chances of this happening will be accentuated by the fact that there are independent slow start
circuits, whose responses might not be well matched.

>In this case which is the Vin-out of the "upper" regulator under short circuit condition i.e. V+ connected to V- (GND of the "lower" regulator)?
>Please refer to Fig. 2 and associated text on http://www.linear.com/pdf/dn21.pdf for some hints.

Fig2..?? I thought it was best suited for single sided supply with an output determined by ((R4 + R3)*Vref*(2 + R1/R2))/R3.
Ofcourse, yes under shorcircuit, each of the regultors see an inout differential equal to the individual secondary voltages..
The negative half regulator develops a differential of (vin - vref) and the positive half (vin - ( -vref)), which will still
be within the individual in out limits, provided (vin+vref) is contained by design.
Is that what you were pointing out.?
Good point.!

But can this be used for dual supply. I believe, if used for dual supply the regulated outputs wont be symmetric.
the negative rail will have a higher absolute value than the positive rail.

With reference to fig2 on dn21.pdf
Lets reference everything to the junction of the anode and cathode of the output diodes, which in the case of a dual supply will be the ground reference.Assuming this junction to be at 0V(AGND) and referencing everything to AGND...
We have Vout+ = (Vref/R1)*R2 + Vref.
Vout- = -((Vout+ - Vout-) * [R4/(R3 + R4)] + Vref). (as given if we assume R3=R4)
Vout- = -((Vout+ - Vout-) * (1/2 + Vref)
= -((Vout+) + 2*Vref)
= -(Vref((R1/R2) + 1) + 2*Vref)
= -Vref(R1/R2) + 3)

so with reference to GND we have Vout+ at Vref*((R1/R2) + 1)
and Vout- at -(Vref((R1/R2) + 3).

One cost effective and crude way to get around this is to insert diodes to drop some voltage across them on the negative half.

ajju