Accuracy of a 15v rail

Projects General Guidance
1

Hello

I am trying to repair an old PCB and it has a 15v rail on it that currently is sitting at 14.7v

The LM2676-ADJ regulator appears to need a voltage divider on the feedback pin to set the output voltage.

Trouble is, the original regulator isn't made anymore... so I used this replacement which I think might possibly be the problem (I thought it matched).

LM2574HVM-ADJ/NOPB

The output is set by a voltage divider (I think)

In this case, it is 1k and 11.3k.

But, the voltage out is actually 14.7v, so I wondered if these resistors were out of tolerace or something (apparently it used to be a perfect 15v).

I swapped these resistor out for 0.1% tolerance resistors, but I get the same issue.
I tried a 12k (just out of interest) instead of the 11.3k and I got 15.48v out.

It appears this divider is very sensitive to the resistor values.

I then swapped the 1k and 11.3k for resistors off a different working board (same PSU) and got 14.9v. All resistors measure correct value on my bench multimeter.

Any ideas what else might be causing the drift or incorrect output?
I have no idea how to actually calculate the voltage divider. I will look into the datasheet now.

15v PSU
15v PSU1441Ă—585 16.1 KB

2

I think this is the regulator.
The differences are too great.

Just found this which is probably a far better match

3

Use a 10k 10 rev fine adjustable trimpot instead of those two resistor.
Using 100x larger resistor in parallell will decreas the value of its resistor by 1 percent.

4

Very, very few solid state devices are currently being made. All were made in the millions on a one-time assembly and testing production line. All such devices can still be found and purchased. Google found many well known suppliers of such devices with your device still for sale.
You either did not want to buy or did not look.

1 Like
5

What happens if you add a small load maybe 100mA or more.

6

It is supposed to be sensitive to the ratio of those resistors...
Just slightly adjust one of them to get the correct output voltage.
Or accept the deviation.
For things like audio amplification it won't matter.

7

The voltage divider ratio is designed with the 1.2volt to 1.25volt bandgap feedback voltage of the LM2676 in mind (1/(11.3+1) = 1.2195).
Because that internal bandgap has a bit of tolerance, your 15volt supply can be off a bit too.
I don't see how this should be a problem though.
Old analogue gear with 15volt rails can usually tolerate between 12volt and 18volt.
14.7volt is actually very close to 15volt and shouldn't give any problems.
Leo..

8

Hi, @baffled2023
Is there a reason you need EXACTLY 15.0Vdc?
14.9Vdc is well within most equipment specifications.

Have you measured the ripple on the output, no load and loaded?

What was the original chip part number, have you looked at its specs?

Are you over thinking the problem?

0.1V in 15V is 0.1 /15 * 100 = 0.66%

Tom.... :smiley: :+1: :coffee: :australia:

9

The problem is, the circuit further down the line does division calculations using this 15v rail.
This board belongs to a friend of mine, and is 12+ years old (the PCB, not him).

He wants to know why his readings are slightly off in his logging software.

I removed the 15v regulator and pumped in a clean 15v and his processor starting reading accurately again.

So, first off... poor design I would say.

I just replaced the regulator with one that is identical, but has a higher current ability. Just out of interest.
Same voltage.

So I just changed the 11.3k to an 11.5k and got 15.04v. So this looks like a winner.

I could cut the trace on the PCB so that I can test the regulator unloaded (there is no other way to achieve that test). But I don't want to mess up his PCB.

The math says 1k and 11.3k should equal an output of 15v, but hey ho.

I also replaced the 10nf capacitor that connects to the boost pin, as apparently that can influence the output, but that seems fine.

10

Texas still showing it as Active production with alternatives

11

Hi, @baffled2023

Then there must be a calibration feature, as not all 15V power rails will be exactly 15.00Vdc

You should do some Ohms Law and work out what value a 2 resistors and a low value trim potentiometer will be to make the output adjustable over a small range to cal 15Vdc.

Tom.... :smiley: :+1: :coffee: :australia:

12

Trouble is, how do you add a trim pot to a 0603 footprint. I am not making new boards, so anything I do is a bodge extra.
The way would be to make it something like an 8k with a 5k trim I would think

13

Bad downstream design then.
Power supply chips usually have a ± 0.5volt specification, which includes temp and load variations. And the noise of a switching supply.
If you need a more stable voltage, the it must be followed by a second analogue reference voltage stage.
Leo..

1 Like
14

I have fixed it, but the PCB is badly designed.

I am not sure why they used a variable reg if they wanted 15v?
What does having a variable output gain you? (assuming you want dead on 15v).

They make this regulator in a 15v fixed output. Surely that would have been better than the -adj version

15

Hi, @baffled2023

Because if you look at the spec for a fixed 15V regulator, it would probably fall outside the requirements for your circuit.

By making it adjustable, with a calibrated volt meter, you can set it to 15V to YOUR tolerances.

Does this 15V supply power any other components on the board?

Tom.... :smiley: :+1: :coffee: :australia:

1 Like
17

And what is the voltage after an hour of operation (where temperatur increased from ambient to 20 degC above ambient?). And at hogh and low load?
You really cannot expect a voltage regulator to be within 0.1V...
You reach that accuracy, but that requires compensation curcuits and proper calibration.

18

This topic was automatically closed 180 days after the last reply. New replies are no longer allowed.