Hi all,
So I'm assembling a bench power supply for general use. Output variable up to 12V, probably 1 or 2 amps.
I'm using a generic Chinese 12V, 10A supply (this link is to a similar one. The one I bought was surplus and no longer available.) The voltage regulator is an ebay LM2596 with constant current.
The problem is that the output is very noisy, both from the 12V supply and the LM2596. I've attached a picture of the scope output.
Every 13.6 us (73.5 kHz) there is a regular output spike from the 12V supply (Yellow trace - Volts peak-peak is 1.4). That translates to a similar voltage spike on the output of the LM2596 (Blue trace, set to 3.3V, with 2.6 peak-peak spikes.) Also, sometimes there is a spurious spike on the output that is not on the input as shown on the scope.
My question is how to filter out these spikes? I think a filter between the 12V supply and the voltage regulator will clean up most of the trouble. I tried a capacitor but it made negligible difference. I don't want to connect sensitive 3.3v or 1.8v electronics to this with such large spikes.
Chinese 12V, 10A supply (this link is to a similar one. The one I bought was surplus and no longer available.)...
...
Every 13.6 us (73.5 kHz) there is a regular output spike from the 12V supply (Yellow trace - Volts peak-peak is 1.4)
You could probably clean that up with an L-C filter, but I'd say get a better power supply. Although.... A proper voltage regulator will also act as a filter.
If you plan on using a variable voltage regulator to vary the output-voltage, your main power supply should be around 14 or 15V to allow the regulator to "work" up to 12V out. (The regulator spec is called "drop out voltage", and if the voltage-drop across the regulator is less than the minimum, it will fail to regulate.)
The voltage regulator is an ebay LM2596 with constant current.
That's not a voltage regulator, it's a current regulator (made with a "voltage regulator" chip).
...If you are building a power supply, you should be familiar with [u]Ohm's Law[/u], and if you understand Ohm's Law you should understand that a "normal" power supply* should put-out constant (or regulated/controlled) voltage and MUST NOT put-out constant current. If the power outlets in your house were constant current, you'd have a disaster!
It would also be helpful if you know how to calculate power (Watts).... Power = Voltage x Current, or with Ohm's Law and some algebra you can derive P = V2/R or P = I2 x R.
LED power supplies are a special case where we need to control/regulate the current. LEDs (like all diodes) are non-linear... Resistance drops drastically when you reach the operating voltage. That means a slight increase in voltage causes a drastic increase in current an if it's not limited/controlled, stuff burns-up! With LEDs, we regulate/limit the current and the voltage "falls into place."
This suggestion is only a "shot in the dark" but is the powersupply unloaded during your testing? Some supplies are slightly unstable at no load. I really don't think a load will do much but it might be worth a quick test.
If you are determined to use this supply you should look more closely at the capacitor route. At high frequencies a capacitor starts to look like an inductor. Grumpy_Mike describes this on his site.
You will need a ceramic or special film capacitor. Maybe a ultra low ESR electrolytic but I doubt it.
Thanks all for the replies. I've tried a handful of things, but not much improvement.
DVDdoug:
That's not a voltage regulator, it's a current regulator (made with a "voltage regulator" chip).
It is a voltage regulator with constant current "cap", as in you can set constant voltage, or constant current. Most lab bench power supplies have a constant current feature. Also, the 12V PS can actually go up to ~14V, so I can easily get out 12V without worrying about the dropout.
TomGeorge:
Have you got an earth connected at the mains input end?
In the first test, it was not, but subsequently (see the attachments) I connected it. It made no appreciable difference.
JohnRob:
This suggestion is only a "shot in the dark" but is the powersupply unloaded during your testing?
It was unloaded, but see the attachments, a load made no difference.
JohnRob:
If you are determined to use this supply you should look more closely at the capacitor route. ...
What capacitor(s) did you try?
I've tried various capacitors, see the attachments.
"V-" is the negative DC output, which for this test I connected to earth ground. All the previous tests had the DC output floating.
This is the best, so far, but is adding more capacitors the right thing to do?
Maybe I need to learn about LC filters. I've looked at online LC filter calculators, but I don't know what the terms mean, ie, cut-off frequency, impendence, quality factor, number of poles, ect...
Hi,
Can you post a picture of your project please, so we can see your component layout.
I am beginning to think this may be stray Scope Lead pickup rather than actual glitch noise.
Can you leave the scope gnd lead connected to gnd of your project, and connect a 10K resistor from the scope probe tip to the gnd please.
Do not connect the probe to anything else.
Show us the trace.
DO NOT disconnect the mains earth from the power supply earth , they are not classed as double insulated.
TomGeorge:
Can you post a picture of your project please, so we can see your component layout.
I've trimmed it down to the bare bones. It's only the 12V supply with some caps across the output. Here is is with the scope across the output:
TomGeorge:
Can you leave the scope gnd lead connected to gnd of your project, and connect a 10K resistor from the scope probe tip to the gnd please.
Do not connect the probe to anything else.
does your Rigol scope have a square wave terminal on the front (the ones on the web have them in the lower right. If so you should it check the response of your x10 probes (and adjust if required)
If in the last post, 1cm is 100ns then I estimate the high frequency "ring" at ~100 Mhz. Most electrolytic do nothing at that frequency.
So ceramics are the most practical solution. You could use a LC but I think you are asking for trouble with such high frequencies. But you might find the tubular ferrite found in some USB cables might be useful. They are not so much inductors as frequency dependent resistors (kind of)
When adding your ceramics, lead length is not your friend.
When investigating EMI problems I have used a board with multiple ceramic capacitors soldered to copper clad.
I have ensured the probes are calibrated correctly. There is a little "rounding off" at the corners of the square wave, but no ringing.
With the capacitors removed (and just a 10k resistor across the output), the ringing is about 32 Mhz lasting for about 1 ms.
This power supply would be good for lights and motors and such, but I want to use it for electronics (microcontrollers, displays, sensors, ect.). My only fear is that with an output of 3.3v, the output spikes between 1.9 and 4.5v, which would cause problems not to mention it's outside most 3.3v device specs.
I will try the ceramic capacitor on copper clad tomorrow and report back.
TomGeorge:
Hi,
Where is the cover for the power supply?
Tom...
It's tied up with the rest of the bench power supply. The variable voltage regulators, relays, and a whole spaghetti mess of wires was attached to the cover of this power supply. It would have been too much work to remove it all for a picture, and it was too messy to leave on, so I just pulled the cover off so that it would make a nicer picture.
If I ever get this project done, I'll post a finished picture.
I forgot to mention, if you do try the ferrite core from a usb cable, you want to put only one wire through the center of the core (lets assume the + lead). You could also put the (+) wire through the center multiple times if it will fit. This would be between the power supply and the capacitor board.
Note: you might see in the USB cable all the leads go through the core, this style of incorporation is for a different source of EMI (commonly called common mode EMI)
Regarding the capacitor board, it need not be a large as my drawing might have suggested. The goal of this particular style of assembly is to have as short a lead on the capacitors as possible. Even with this setup, the capacitors will start to increase in impedance around 20 Mhz.
So I made a little capacitor bank on some copper clad. I just used some caps that I had on hand:
1x 10uf, 2x 1uf, 5x 0.1uF, 3x 10 nf, and 10x 22pf.
Before:
And after:
But then in a minute, without any changes, it looked like this:
And a few minutes later, again no changes, it looked like this:
It seems that the power supply has some quirks, or maybe there's a loose connection somewhere... I dunno...
The capacitors seem to help, so would the easiest solution be to add some more? 0.1uf? 22pf? something in between?
Also I tried to cut off a ferrite core from a usb cable, but it's just solid rubber
When I touch one of the output wires or wiggle around the terminal block, the noise drastically decreases, but it seems intermittent. There might be a loose connection somewhere, or is this just interference?
You will never get a perfectly clean output from a switchmode supply, unless you go to huge efforts with L-C filtering, metal shielding etc. Switching amps at high frequencies with fast rise and fall times inherently generates a load of mush.
Almost all bench supplies are linear, because there efficiency and size don't matter, but smooth and clean dc does.
Above all else.
When you're working on a design, the last thing you want to worry about is dirty supplies.
Find another use for it, and build/buy a linear psu for bench use.
I've had a Farnell TOPS 3D for > 30 years, and it's still going strong. It'll probably see me out.
( It's the only product I 've ever come across that actually uses LM741's in the 14-oin DIP package - they were old when it was built !)
