LM7805 climbs above 5 volts why

[dietert1]

What on earth were they thinking feeding the regulators with 31V?

They wanted +/- 15 V for the analog circuitry. So they put 2x 18 Vac PCB transformers to leave some headroom for rectifiers, voltage regulators and mains voltage variation. With light load the transformers output 2x 22 V. Later mains voltage increased from 220 to 230 V and there you are. Since there was no space to have another set of transformers, they just took the same 31 Vdc for the 5V regulators for logic (DACs, shift registers and opto isolators). Instead of LM7805 they used LM340T-5.0 with a datasheet upper limit of 35 V. Now consider another +20 % mains voltage margin.

I solved the heat problem by inserting those very small Polulu switchers as down-converters to 8.5 V. The instrument also got a simple mains pre-regulator that dumps most of the excess power into two mosfets mounted on the enclosure frame, and instead of 31 V there is a constant +/- 22 V for the +/- 15 V regulators.
I also added schottky diodes over all +/- 15 V and +5 V supplies. Without such diodes JFET opamps may get killed after turning off power.

Something similar happened, when i looked into the HP3456A in-guard power supply. Building a good power supply with integrated voltage regulators may appear simple, but it needs some care.

Regards, Dieter

[exe]

simple mains pre-regulator that dumps most of the excess power into two mosfets mounted on the enclosure frame

I've never seen this idea before, do you have a schematic for this?

[Zero999]

I just grabbed LM7805 from National semi and it kept 5V output all the way up to 60V input voltage my PSU could supply. No input/output capacitors and load attached.

If someone designs a circuit based on tests like that or on datasheet limits, there will be a bad surprise later on. I can walk on one foot for a while, but the conclusion to buy one shoe only is certainly bad. The reason for not using a LM7805 with more than 20 V input is heat.
Only recently i fixed a Keithley 213 quad DAC (a rebranded IOTECH DAC488) that has four 5 V regulators running from 31 V input, all without heat sinks, at 55 mA each. That can't work. One of the regulators was burnt, although in theory they have built-in over-temperature protection. All capacitors were dry from heat, so the other three regulators were oscillating. What a mess.

Regards, Dieter

I don't think he's advocating using the LM7805 at 60V, just that based on his test, a high input voltage is unlikely to be the cause of a higher than normal output voltage.

The maxium recommended input voltage for the LM7805 is 25V, so the Keithley is using it out of specification. The power dissipation issue can be fixed with a series resistor.
https://www.sparkfun.com/datasheets/Components/LM7805.pdf

No, it’s not out of spec. Some datasheets recommend 25V, but many do not, leaving only the maximum input voltage of 35V or 36V, which 31V is comfortably below. What the designers at IOTECH clearly didn’t look at were the graphs explaining how much current one can draw at a given voltage differential...

Assuming you're right, I've no desire to spend the time looking at all of the data sheets, a voltage drop of 26V, will only dissipate 1.43W, at a current of only 55mA, which should be fine for the TO-220 package, even without a heatsink.

[magic]

That is not true at all. For keeping output voltage you need the load only for adjustable regulators like LM317 and sort of. 7805 bleeds output current internally through quite low resistance internal feedback resistive divider. Datasheet may specify output voltage at certain current range but it does not mean that below this current output voltage will raise to any significant extent.

Both output voltage and load regulation were only specified for Io≥5mA by National. The 78L05 was specified for Io≥1mA.

This implies precisely nothing about operation with zero load. The parts may work nevertheless, or they may work in 99.9% of cases, or they may work below 50°C, or whatever.

[wraper]

Since there was no space to have another set of transformers, they just took the same 31 Vdc for the 5V regulators for logic (DACs, shift registers and opto isolators). Instead of LM7805 they used LM340T-5.0 with a datasheet upper limit of 35 V. Now consider another +20 % mains voltage margin.

LM7805 and LM340T-5.0 have exactly the same ratings.

[dietert1]

simple mains pre-regulator that dumps most of the excess power into two mosfets mounted on the enclosure frame

I've never seen this idea before, do you have a schematic for this?

See appended sketch. Of course you need to add some protective parts, like zeners for the gate voltage and an over-voltage protection on the input. But it is simple and works well. For the HP 3456A i made something better: A regulator for the transformer magnetic flux that separates in-guard from out-guard consumption.

Regards, Dieter

[SteveyG]

I was about to ask about the load. I had something similar happen with a negative 12-volt regulator: LM7912. It behaved pretty badly before I added a small load. You can see this happen on video to SDG Electronics around the 18 min mark.

See: https://youtu.be/ML-lmuHoh-0?t=1107

 Since watching this video, I have just always added LEDs on the output.

Most likely it was a counterfeit trash as well.

Shouldn't have been counterfeit - It was from Farnell which is one of their specified distributors. I've since had confirmed during a sales visit from some of the UK reps that indeed they may require minimum loading for regulation anyway.

[tkamiya]

There is no longer THE 7805....  You can buy all sorts of LM7805 from Amazon or eBay and every one of them behave differently.  Some has max input of 20V, some 25V, and so on.  I don't trust them to agree with datasheet either.  They may not be a "counterfeit" but definitely not a first rate manufacturing.  Besides.... one should not design circuit so close to the max limit anyway.  Heat generated is so much that energy waist and heatsinking requirement will cost you more than what you save for a cheap component.

By the way....  on similar applications, I've used DC/DC converter module purchased from Amazon.  Applied 12V in, 5V out.  Put a ferrite bead on output.  OK... less spikes.  Put two... not much difference.  Put THREE, it went into full swing oscillations.  37V output at max current.  Fried my components crisp!  Be careful with these....

[exe]

Put a ferrite bead on output.  OK... less spikes.  Put two... not much difference.  Put THREE, it went into full swing oscillations.  37V output at max current.  Fried my components crisp!  Be careful with these....

Not sure how much this article is relevant, but I'll leave it here. It's about designing second-order lc-filters for power supplies. It explains some stability issues: http://ridleyengineering.com/images/phocadownload/1%20second%20stage%20filter%20design.pdf

[dietert1]

For DC/DC down converter i can recommend Pololu modules. Never had any difficulties with them, on the contrary, they helped to solve several problems. For example i was able to pull the switcher from inside an Arroyo precision temperature controller and replace it with a Pololu 9V module that is about 1x1" and rated up to 7 A with proper input. I am using it as a secondary switcher with less RF interference on a nearby 24-bit ADC and supply some 18 to 20 V from a nice linear power supply (external).

Regards, Dieter