LM7905 noise issues (STMicroelectronics)

[cj9]

Hi all, happy new year!

I have some LM7905 negative voltage regulators and have tried them on PCBs, breadboards, with and without filter caps - but for some reason get 2v noise on the output if the input (DC) is in the 8v to 15v range.

It's quite an odd problem, so I just wanted to see if anyone has encountered this or have any suggestions?

[wkb]

Sounds as if the LM are oscillating.  If I remember the application notes correctly you should mount a couple of capacitors as close as possibile to the  LM to prevent this

[slateraptor]

I have a suggestion: How about a schematic; your description couldn't possibly be any more ambiguous.

[PStevenson]

slateraptor, I'd imagine he's just riffing on a breadboard following basic datasheet schematics, sometimes they neglect to mention/or just overlook the bits about close proximity filter caps  (a condition I call  "datasheet blindness" which is something you get after looking at 3 billion characteristic graphs, formulae and so on, on a computer screen)  - yeah follow wkb's advice and try different values and you should sort that noise out.

[slateraptor]

PStevenson, last week, I stood by and watched a bunch of so-called professionals needlessly burned three days of operational downtime and approx US$100k as a combined consequence of "datasheet blindness" and troubleshooting without a clear description of the problem at hand. So you'll have to excuse me if I'm liable to not take anything for granted.

[cj9]

Yes, I completely understand with the datasheets! It's just surprising to have so much noise when a lot of the time, they work perfectly with no capacitors.

I had followed the datasheet on my PCB and breadboard, turned out 22uF cap got rid of most of the noise, when the recommended 2.2uF or 10uF did almost nothing - maybe there was something I missed in the datasheet.

Thanks for your help!

[PStevenson]

PStevenson, last week, I stood by and watched a bunch of so-called professionals needlessly burned three days of operational downtime and approx US$100k as a combined consequence of "datasheet blindness" and troubleshooting without a clear description of the problem at hand. So you'll have to excuse me if I'm liable to not take anything for granted.

well that is a disgrace, I would be a hell of alot more diligent if I could be responsible for $100k!
I wasn't having a go at you sorry if it appeared that way it was said in a more of a jokey way.
for the record if I had stood by while that had happened, I would be in a chemically induced coma recovering from it!

@cj9 - as someone who works mainly in the field of none critical devices I tend to put 2 100n caps very close to the reg and 2 100u (either side in both cases) and then just experiment backwards till I get ideal values

[alm]

Maybe the datasheet specified tantalum electrolytics, and you used aluminum electrolytics? You need a much larger value (most datasheets recommend at least ten times the value) aluminum cap because the ESR is much higher. The large difference between 10 uF and 22 uF is surprising, was there a large difference in ESR? Maybe the 10 uF cap was dried up or the 22 uF was a low ESR type?

[Nico]

In addition to the Cin and Cout capacitors (both should be electrolytic), did you apply any load?

The device is specified with a minimum load o  5mA. A no load condition could generate an oscillation.

[ciccio]

I cannot find LM7905 regulators on ST's catalog. Are they L7905?
In my experience I've seen that negative regulators made by ST have many times a non-predicable behavior, and I try to avoid them when possible.
I believe that their internal schematic is different from the one used by the "original"  uA79xx series by Fairchild, and the results are sometimes "weird".
Sometimes they operate correctly only if the two diodes (from output to ground and from input to output) shown on many schematics are installed.
Sometimes, when switched on on an an heavy load, the output is higher than the nominal value.
Other times, when used in bipolar supplies, they stay off at switch-on, resulting in a positive-only supply.
I ended up following the policy of the my first employer: stuck with Fairchild, National, Motorola (ON).
Best regards

[slateraptor]

I cannot find LM7905 regulators on ST's catalog. Are they L7905?

Yes.

http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00000450.pdf

[Greg J]

The clue is in the ST. They suck as a company, and their chips suck too.
(and they are horrible at supporting small businesses and hobbyists - so I slag them as much as I can without feeling guilty)

[Bored@Work]

The clue is in the ST. They suck as a company, and their chips suck too.
(and they are horrible at supporting small businesses and hobbyists - so I slag them as much as I can without feeling guilty)

I find them more approachable then e.g. Atmel when it comes to microcontrollers. Their STLink programming hardware is more robust, has more features and is cheaper than the Atmel equivalent. And with their DISCOVERY series of eval boards they are making a serious attempt to attract those on a budget. Although they don't understand the open source tool thing.

When Dave was discussing his power supply kit and expressing the need for 12 bit ADCs and DACs I was about to suggest some STM8L micro, because the do have them built-in (as opposite to the STM8S series). However they aren't available in DIP.

Their past is another thing. One part (Thomson) was owned by the French government in the past. And boy, did Thomson produce junk components. If I remember correctly Thomson were also the clowns we have to thank for SCART becoming a European standard. And sometimes their bureaucratic past shines through. E.g. check the "license" that comes with their eval boards. Laughable.

[IanB]

If I remember correctly Thomson were also the clowns we have to thank for SCART becoming a European standard.

That's a bit unfair. SCART may be an ugly fat connector with a thick chunky cable, but it was universal plug-and-play AV connector for many years during which the USA was still messing around with a hodge podge of composite, S-video and component video alternatives.

And give me a choice between SCART and HDMI and I will happily squash HDMI. The encryption and commercial lock-down on HDMI is evil.

[Bored@Work]

That's a bit unfair. SCART may be an ugly fat connector with a thick chunky cable, but it was universal plug-and-play AV connector for many years during which the USA was still messing around with a hodge podge of composite, S-video and component video alternatives.

That incredible PoS? I have given up counting how often I had to cut a wire in a SCART plug just to get rid of some crosstalk. Or the other way around, finding out that some cable happened to miss the RGB signals. Or how often I had to use hot glue to glue one of the fucking plugs into place because the plug kept falling of or was about to damage the socket. If it was up to me I would hang the SCART inventors from their necks, using SCART cables.

[ciccio]

The problem of SCART is that it works, but only most of the times. When it does not work, it's an hard problem...
The plugs love to detach from the sockets, due to the missing locking mechanism and their huge mass and size, and to obey to Murphy's law,  so they will fall down in the inaccessible space between the wall and the rear of the cabinet on which the TV is sitting.

It must be noted that ST's other partner was SGS, that was an Italian company, once owned by a company owned by the Italian government, that has it's roots in ATES, one of the first European Company to manufacture semiconductors, but this is a long story.
When ST was incorporated,  I had big doubts on the possibility of such an Italian - French cooperation to work and survive more than a couple of years, because in the late 70's I  was involved in an Italo-French cooperation  in the nuclear energy sector, that failed for incredibly stupid reasons...

[amspire]

I have been staying out of this thread, because I think if cj9 wants help, there just has to be better information from cj9.

The issue is simple. Working ST regulators don't have 2V noise on the outputs, otherwise they would be useless.  So either this is a faulty batch, or there is an issue with cj9's tests. My intuition is that used correctly, the 7905's will probably work.

Slateraptor early on asked for a circuit, and cj9 didn't provide one.  A photo of the test setup would be better, and details of the different capacitors that have been tried. Ideally the two caps and the 0V pin of the 7905 should be star connected. If there is something odd about the way the capacitors were connected, then that may encourage oscillations. What is the power source? Regulated DC, or rectified output from a transformer. Does the DC input have any ripple? Is there any load on the output of any kind? Is there no load? Does the oscillations go away if there is a load of more than 5mA? Have different multimeters been tried?

If everything is right, then the next step is for cj9 to source some new 7905's - a different manufacturer would be good - and see if the same problem is there.

Otherwise I can see this topic moving through discussions of everything bad anyone called Thompson ever did.

Richard.

[SeanB]

I used to get stuff made by Thompson CSF. Lovely stuff, but the price............. Wounded buffalo is nothing!

Did dissemble onr to see what they had done about the one component they used that was made from Unobtanium. When I looked , they had used a diode that I could order in large lots, so I did as they had ( hey, it was a Manufacturer approved mod, so did not need the 500 pages of documentation and such to approve it) and it worked. One I had to substitute, as they were a special order, and had a $100 each cost, minimum order 10k units for them to set up the wafers in a fab. I was looking for 2, and 4 as spares. Good thing TO3 cans all look alike without part numbers on them.

[cj9]

Sorry about not replying until now.

The reason I posted this is because 2v noise seemed quite odd, with the wrong capacitors or no capacitors I've never seen such an unstable output. The input is DC from a lab power supply, so it's not oscillating with the input signal.

The circuit I am using is an exact copy of the one on the datasheet, both on PCB and breadboard; I don't think it's worth taking a picture of this as it won't show anything more than that.

I tried putting ST -12v regulators in the same circuit and it works absolutely fine, it's just these 5v ones that have a problem. I will try again with another manufacturer soon.
Another reason why I thought it was odd, was because between -6 and -7v there was noise, between -7 and -12 there was no noise, and below -12 there was 2v noise. This might help to come to a conclusion on what's causing it?

I realised that as it's not being used for a load, I should use a voltage reference. It's actually for an op amp to shift a voltage down by -5.

[amspire]

Look I do understand - you have copied the circuits in the data sheet and you have oscillation. So why post a circuit and photo?

The reason is that we would have seen straight away that you had the regulator connected via wires to a lab power supply, and you had non-tantalum electro's.

It turns out you were not actually following the data sheet, and that was your problem the whole time.

The Thompson data sheet says if you don't use tantalum capacitors, then if the regulator is an appreciable distance from the power supply's main filter electrolytic capacitor, you need to increase the input filter cap across the regulator by a minimum of ten times for stability. So 22uF is the absolute minimum, which means 100uF is a comfortable value - you don't want it becoming unstable as the cap ages and gains ESR. A power supply connected by wires is "an appreciable distance" in pcb design terms, and a lab power supply does not always have a large capacitor across its output.

The exact way the input and output caps were connected to the regulator is also critical, and we are still in the dark about that. A wrong way would have been for the two negative leads of the caps to connect in one breadboard row, and then a wire goes from that row to the 7905 common lead. The best way would be to have the regulator plugged into the middle of a breadboard row, with one cap connected on the front side of the regulator, and the other cap connected the rear side.

These points probably would have been mentioned within the first few posts if we had known exactly the test configuration that you had. I was waiting for a schematic as asked for in the third post before replying and no schematic ever came.

Also if we had known that you wanted -5V for an opamp level shifting, we would have said the 7905 was not the right part, and we would have suggested other references, or alternative opamp circuits that could use the 5v as the level shifting reference instead. There is nothing wrong with a 4.7V or 5.6V zener in fact - 100% safe from instability issues. When you leave us guessing, we struggle to help you much.

Richard.

[ejeffrey]

The reason I posted this is because 2v noise seemed quite odd, with the wrong capacitors or no capacitors I've never seen such an unstable output. The input is DC from a lab power supply, so it's not oscillating with the input signal.

Nobody thinks that.

The circuit I am using is an exact copy of the one on the datasheet, both on PCB and breadboard; I don't think it's worth taking a picture of this as it won't show anything more than that.

You are obviously using it wrong.  One device could be a defective or damage part, but if you have tested multiple ones and they show the same behaviour you are doing something wrong.  These regulators work for everyone else when used properly.  Therefore you are not using exactly the circuit from the data sheet, or you are not operating within the boundaries specified.  The picture is especially helpful: both a closeup up the breadboard and a shot of the full setup.  Since you think you are using exactly what is specified in the data sheet, that will help identify how what you have is different than what you need.  But a schematic including the input power supplies and the op-amp (draw one on a piece of paper and take a picture if you prefer) is also very helpful.  The more detail you provide the better people can help. 

I will try again with another manufacturer soon.

This will only solve the problem if you got a batch of really bad fakes.  Possible, I guess, but very unlikely. Different manufacturer's devices may operate differently when operated out of spec, so it is possible this will even make your symptoms go away.  You still should find out what is wrong so that you don't make the same mistake down the line.

I realised that as it's not being used for a load, I should use a voltage reference. It's actually for an op amp to shift a voltage down by -5.

It is perfectly fine to use a regulator as a low precision voltage reference, but you will need to satisfy the minimum load current.  A 500 or 1k ohm resistor to ground should be enough.  Again, that is only a guess without more information.

[IanB]

The circuit I am using is an exact copy of the one on the datasheet, both on PCB and breadboard; I don't think it's worth taking a picture of this as it won't show anything more than that.

It will most likely help people to see what you are doing wrong. Those regulators when properly configured don't have a 2 V oscillation on the output the way you are describing. If they did the part would be entirely useless for its purpose.