Noise in DC power supply

[XgN]

Hello,
I'm using home made DC power supply, based on casual LM338T schematics http://www.circuitstoday.com/wp-content/uploads/2009/08/13v5a-adjustable-powersupply-using-lm338.jpg
When connected to a few amps, for example, tube pedal/booster and guitar amp, it's heard the noise, hum for sure.
Measured with oscilloscope, and what I've got:

• 

Very small ripple. 100 Hz. By the way, one square = 0,01V. Tried to change C1 into 10k uF (4,7k was the first). Nothing changed. Any ideas how to remove it?

 

[Harald Kapp]

What you see sure looks like the voltage dropping at the output due to the input voltage falling below the limit required to keep the regulator in operation. The LM338 in particular requires Vout-Vin >= 3V for stable operation. THe frequency of the hum is 100Hz due to rectification.
Increasing the size of the input capacitor is the right way, as long as the transformer is able to supply enough power to fully charge the capacitor.Have a look at the voltage across the capacitor under no-load and full-load. Does the voltage reach peak value under full load?
One option is to use 2 or 3 4700µF capacitors instead of one single 10000µF capacitor. Paralleling these will reduce the equivalent series resistance (ESR) of the capacitor thsu reducing voltage drop due to load or charge current.

 

[duke37]

The ripple could also be due to poor earth layout. The current goes into the reservoir capacitors in big lumps, if this affects the reference voltage then you will get an output voltage ripple.

 

[XgN]

Thanks for answers! Transformer gives 14V AC, ripple could be seen under 5V DC. I will try diferrent transformer. PCB is made according to this website: http://worldtechnical.blogspot.com/2008/12/lm338-power-supply-regulator-138v-5a.html#.VMTZ7v6sV8F

 

[Harald Kapp]

ripple could be seen under 5V DC

The absoulte ripple is not the main factor here. You need to check that the input voltage to the regulator is always at least 3V higher than the output voltage.
If, for example, at 12V output voltage your input voltage is 20Vpeak with 5V ripple, then the input voltage to the regulator is min. 15V which is just fine for 12V output.
If, on the other hand, your input voltage is 17V peak with 3V ripple, then the input voltage to the regulator is min. 14V which is insufficient for a 12V output.

A 14V AC transformer will deliver 14V*sqrt(2)=19.8Vpeak.
A Bridge rectifier will drop 2 diode voltages, which amounts to approx. 2*0.7V=1.4V (at the currents involved here).
The peak voltage on the primary capacitor tehrefore will be 19.8V-1.4V=18.4V.
Assuming a 12V output, the input to the regulator needs to be at least 15V.
The max. allowed ripple is therefore 18,4V-15V=3.4V - much less than the 5V you measure.

 

[davenn]

Take good note of what duke37 said
could easily be earthloop problems

 

[XgN]

I'll try powerful transformer, because when I'm increasing current, the ripples are getting bigger. At some voltage point they become huge ones. Maybe fails my transformer..
Input I think is OK, I wanted to say, that input is 14 AC, and measured ripples at 5V DC output.

 

[duke37]

It seems to me to be unlikely that the ripple low should be a few mV from what is required. You could try a different load current which would change the ripple.

If the problem is poor regulation due to earth loops you could try connecting negative of C4 direct to the output terminal.

 

[Harald Kapp]

Take a scope image of the voltage across the input capacitor for further analysis of the issue.

 

[duke37]

And scope between C4 negative and the output of the regulator.

 

[XgN]

Sory for late reply. Now I'm experimenting with old powerful transformer. After capacitors (which I upgraded to 3x4700uF) the voltage is ~32V. Seems to be too high for LM338. Huge noise and generation. Going to search for another transformer with lower voltage. By the way, added tantalum cap, as written in the datasheet. Tomorrow I'll give scope pictures

 

[XgN]

Took the photos of scope:
across the input caps (3 new 4700uF)

• 

Scope setting - amplifier: "~". I think I have to use this one to filter DC and see waves.
across the C4:

• 

Used more powerful transformer, current under 1A. The more current, the bigger spikes. I thought the caps have to smooth them.

 

[duke37]

Measure the ripple from the negative of C4 to the output of the regulator. If this is negligible then chase around the circuit to find where it is introduced, measuring relative to C4 negative. Changing components randomly will be unlikely to help.

Capacitors will only attenuate the ripple, they will not give perfect smoothing. The regulator is much better for this job.

To look at the ripple and push up the gain, you get rid of the DC so the amplifier gain can be greatly increased and still get the result on the screen.

 

[XgN]

• 

In the output.. The more current or voltage the bigger "hang down". Maybe I should try to construct better regulator?

 

[BobK]

It is not the regulator that is the problem, the transformer is unable to supply the power you need.

Bob

 

[XgN]

Tried with huge one, 100W - the same.. With the load of 1A.. Going to search for new schematics.

 

[Harald Kapp]

Tried with huge one, 100W

The schematic is o.k. It's not only the wattage that counts, it's the output voltage, see my posts #2 and #5. The transformer needs to deliver both enough voltage and current.

Re your post #14: Why don't you measure at the input, as I sugegsted? What you see is not a fault of the regulator. When the input voltage drops below the required minimum, the poor chip can't but folow the input, Where should it get the additional voltaeg from? It's all well explained here.

 

[XgN]

Thanks for answer!. This photo of the input (across input caps)

• 

Made simple fixed voltage regulator with L7809 according to datasheet schematics and it's much more better (very small ripples) on 1A. But my variable regulator doesn't want to be like that.

 

[XgN]

I think, that's my PCB.. Not very good designed.. Searching for a new one. Maybe could someone help me to clear out wich PCB on the internet is designed well. Wihout any ground loops, etc. What about this one? http://320volt.com/lm338k-ile-5-amper-ayarli-guc-kaynagi/ ground seems with no loop. By the way, going to try LM1084. I hope even a little bit better than LM338.

 

[Harald Kapp]

I think, that's my PCB

The PCB may have an impact, but I doubt it's much. when you scope the input voltage as in #19 at different output currents, you will see that the dip becomes smaller with less current. When you scope bot this input voltage and the output voltage simultaneously (assuming your scope has at least 2 channels), you will notice that the dip in output voltage happens when the dip in input voltage is at its lowest point. This is a clear sign that you have an issue with too little input voltage, not layout.
If you had a layout issue, you'd see al kinds of random noise not related to the dips in input voltage.

By the way: when you add a key to your images (horizontal deflection in ms/div and vertical deflection in V/div taking into account any prescaling by the probes) you greatly help us understand what the images show.