After building a linear current limited power supply (1.2V-26.9V, max 1A) which is powered by an SMPS brick, I am in the process of testing the load regulation.

When connecting an adjustable constant-current load at the output (see attached schematic where I have circled this positive output point in red) I've found the load regulation to be over 3% in some cases. I found that when connecting the load and taking measurements directly off the LM138K TO-3 package the load regulation was significantly lower (0.1442% to 1.6184% @1A depending on the output voltage).

I am thinking that because I used smaller traces when designing the PCB (see attached), and wires going from the LM138K TO-3 package to the PCB, and wires at the output, there are several points where voltage is being dropped (if I connect the load at the circled output point).

When connecting the positive of the load to the circled output point, and taking my voltage measurement directly off the positive output on the TO-3 package, the results were not the same as having both the load and meter directly on the TO-3 output. Why is this?

I would think if voltage is being dropped between the output of the TO-3 package and the circled output point, if I take my voltage measurement at the TO-3 output, it would be the same as having both the load and DVM directly connected to the TO-3 output.

 

There are many trace width calculators on the web, mostly based on the same equation for calculating the resistance of a copper trace based on its dimensions and temperature, and how hot you can allow it to get. Copper is a resistor, and it creates voltage drops and dissipates heat. Usually these are trivial amounts, but in power supply design it can cause the problems you are seeing. The traces from the 338 output to the load and from the load to R3 pin 1 are critical. For a 5 amp output with 1 oz. copper, I'd start with 0.5" wide traces. 1/10th inch per amp is a good rule of thumb for a starting point. If you don't have room for the fat traces or want to rework you existing board, you can solder solid wire on top of your narrow traces.

ak

 

There are many trace width calculators on the web, mostly based on the same equation for calculating the resistance of a copper trace based on its dimensions and temperature, and how hot you can allow it to get. Copper is a resistor, and it creates voltage drops and dissipates heat. Usually these are trivial amounts, but in power supply design it can cause the problems you are seeing. The traces from the 338 output to the load and from the load to R3 pin 1 are critical. For a 5 amp output with 1 oz. copper, I'd start with 0.5" wide traces. 1/10th inch per amp is a good rule of thumb for a starting point. If you don't have room for the fat traces or want to rework you existing board, you can solder solid wire on top of your narrow traces.

ak

Thank you for your help. I will refer to the trace width calculators and your recommendations for trace width sizes for future layouts.

For the current board, I am thinking another option of directly connecting a thick wire to the screw terminal where the LM138K TO-3 package is mounted to the heat sink would also work? (since the load regulation was decent when I clamped the meter and load to this point during my tests)

I can see how voltage being dropped across the thin traces would effect load regulation measurements as described in scenario one below. What I don't understand is scenario three below:

Senario One:
The load and meter are connected to the circled output point. There are losses accros the thin traces and that is why the load regulation measurements at this point are so bad.

Senario Two:
The load remains at the circled output point, and the meter is connected directly to the LM138K TO-3 output. For a given load, the voltage on the meter is higher than senario one as it is measuring before the voltage drops between the TO-3 output and the circled output point where the load is connected.

Senario Three:
When the meter and load are both connected directly to the LM138K TO-3 output why is the regulation much better than senario two? I would have thought if voltage is being dropped between the LM138K TO-3 package and the circled output point, that the measurements in senario two and three would be very close. Why is this not the case?

 

From the On Semi data sheet. http://www.mpja.com/download/31800ps.pdf
Load Regulation
The LM317 is capable of providing extremely good load
regulation, but a few precautions are needed to obtain
maximum performance. For best performance, the
programming resistor (R1) should be connected as close to
the regulator as possible to minimize line drops which
effectively appear in series with the reference, thereby
degrading regulation. The ground end of R2 can be returned
near the load ground to provide remote ground sensing and
improve load regulation.

 

Senario Three:
When the meter and load are both connected directly to the LM138K TO-3 output why is the regulation much better than senario two? I would have thought if voltage is being dropped between the LM138K TO-3 package and the circled output point, that the measurements in senario two and three would be very close. Why is this not the case?

Because the regulator doesn't care where the meter is connected. Only the high-current path to the load affects regulation. When you move the load closer to the regulator,or decrease the resistance between the two, regulation improves. In scenario #2 the meter is not connected at the load, so whatever it is measuring, it isn't the true regulation of the circuit.

ak

 

Because the regulator doesn't care where the meter is connected. Only the high-current path to the load affects regulation. When you move the load closer to the regulator,or decrease the resistance between the two, regulation improves. In scenario #2 the meter is not connected at the load, so whatever it is measuring, it isn't the true regulation of the circuit.

ak

I think I understand now. Previously I thought that the *only* improvement in scenario 3 would be the sum of the voltage drops between the TO-3 package output and the circled output would no longer be present (as being present in scenario 2).

While moving the load closer to the regulator (as in scenario 3), the voltage drops are eliminated, however regulation by the LM138K is further improved (beyond what was lost in those voltage drops), for an even greater improvement over scenario 2, correct?

If this is the case, and if the improvement by moving the load closer to the LM138k is not *only* the sum of the voltage drops across the thin traces no longer being present, what else is happening (other than the voltage drops being avoided) by moving the load closer that improves regulation?

 

I think I understand now. Previously I thought that the *only* improvement in scenario 3 would be the sum of the voltage drops between the TO-3 package output and the circled output would no longer be present (as being present in scenario 2).

While moving the load closer to the regulator (as in scenario 3), the voltage drops are eliminated, however regulation by the LM138K is further improved (beyond what was lost in those voltage drops), for an even greater improvement over scenario 2, correct?

If this is the case, and if the improvement by moving the load closer to the LM138k is not *only* the sum of the voltage drops across the thin traces no longer being present, what else is happening (other than the voltage drops being avoided) by moving the load closer that improves regulation?

Would anyone have an answer to the above question?

Thank you