Strange ripple on LM350 output

[Rasmus]

So I'm trying to build a lab PSU and it seems I'm getting too much ripple on my output and I don't understand why so I hope someone have an idea to why.

According to datasheet of the LM350 I should get a minimum ripple rejection of 66db with a 10µF cap on the ADJ pin however I'm only getting about 20db.

I'm loading my output with 150R at 20V so I'm drawing around 130mA and getting 15mVpp of ripple on the the output while having 182mVpp on the input. Also the ripple appears to have a much higher duty cycle on the output than on the input and is not symetrical around GND (scope is AC coupled). The input is about 30V so I have plenty of overhead in terms of dropout voltage.

I had the same circuit with a pass transistor on another breadboard but decided to rebuild since I suspected it might be my sloppy breadboarding causing this. When drawing 2A I was getting ripple in the hundreds of mV on the output with 1.5Vpp on the input.

I have attached pictures of my breadborad, measurements and a diagram of my current circuit.

I really hope someone can help me and thanks in advance to anyone who takes the times to answer this.

[Benta]

Use a perboard instead. Those breadboards only bring grief.
What's the minimum input voltage over the 2200 uF caps?

[floobydust]

I don't see a breadboard ground busbar connection (21) for C5 is it even connected?

Best to use single point grounding, the filter caps have high charging currents and what you have done, split bridge rect (-) to the caps and to the LM350 in no good due to the breadboard's resistance. Any hum on GND (for the LM350 cct) can get amplified by it likely what you are seeing.
You would connect the filter caps straight to the bridge output, then have them feed to the IC and circuit and output (-) as a single star for the 4 connections.
I always include the two protection diodes as well.

[Rasmus]

I don't see a breadboard ground busbar connection (21) for C5 is it even connected?

The two power rails are connected in the other end of the breadboard.

You would connect the filter caps straight to the bridge output, then have them feed to the IC and circuit and output (-) as a single star for the 4 connections..

So what you're saying is connect the 3 filter caps and the rectifier all on the same breadboard row?

[Rasmus]

Use a perboard instead. Those breadboards only bring grief.
What's the minimum input voltage over the 2200 uF caps?

´

I will but I like the flexibility of a breadboard in these early stages I'm planing on doing a real PCB in the end

Minimum voltage I guess would be around 28-27V on the negative going swings at full current but I'm not sure since I don't have my pass transistor connected yet

[Terry Bites]

I agree with the other posts. It looks like a rail impedance thing. Waveform 2 looks like the current pulses flowing through the caps, multiply the peak value of that with a few milliohms and you've got ripple. Tighten up your ground paths and clip the scope ground to the same place.

[Vovk_Z]

LM350 as other integral regulators LM317-like are very sensitive to proper wiring. A proper wiring can be described in LM317/LM350 datasheets or other application notes (I can't find right document at the moment).

[Vovk_Z]

Several years ago I was working on my own laboratory power supply (with LM350 ICs) I think I have seen exact description how to wire wires to those ICs to have maximum possible load regulation and ripple rejection.
But somehow I can't find the same information at the moment.

[Rasmus]

Several years ago I was working on my own laboratory power supply (with LM350 ICs) I think I have seen exact description how to wire wires to those ICs to have maximum possible load regulation and ripple rejection.
But somehow I can't find the same information at the moment.

If you stumble upon it at some point a link would be very much appreciated. I will try what has been said on here when I get the time to work on this project again if this turns out to be the problem I belive it will be resolved whenever I get it on a proper board.

[Doctorandus_P]

If you stumble upon it at some point a link would be very much appreciated.

There is nothing special nor magic about routing for these IC's, and the same principles apply to all circuits, although some are much more susceptible to them then others. Some of the rules are:

1). Separate high current paths from feedback signals. High currents cause voltage drops over PCB tracks, and this voltage drop is fed back into a control loop it may cause instabilities. This can of course also be done on purpose, for example a voltage regulator with "remote sensing", but then you have to carefully design the control loop and frequency response, poles and zero's and that kind of stuff.
1a). This is also true for the return currents (usually through GND connections) Currents though GND wires also create voltage drops, and this is why sometimes star ground points are used.

2). Loop inductance. Whenever a wire goes in a circle, it forms a one winding inductor. This inductor can be a cause of phase shifts in your control loop, but it also works as an antenna (both transmit and receive!) The bigger the area circumscribed by the loop is, the more noise it will pick up (such as the ever present mains hum) and also the bigger the inductance will be (Even a straight wire has self inductance!)

3). Parallel tracks (or any tracks near each other) on a PCB form a capacitor, and thus signals can be coupled into nearby tracks.

Everything put together, the routing of the tracks is an important part of a design. It can be the difference between a perfectly working circuit, and a circuit not working at all.

[floobydust]

Best picture I can find of the circuit layout for LM317, LM350, LM1117 etc. linear regulators. RW is the wiring resistance. From On-Semi LM1117 datasheet.
Note the need for a single-point ground, and for years I had it wrong- thinking the voltage sense resistor R1 had to connect right at the output terminals. No, it best connects right to the IC's output pin terminal, which is counter-intuitive.

OP, not shown on the schematic is that the filter capacitors connect first to the bridge rectifier's output, then a two-wire feed from that cap array goes to the circuit input.
You have stop thinking in terms of long runs on breadboard strips, your C5(-) ground is a mile long run and a loop which is the worst possible way to do it.