Linear reg PSU ripple - too much?

I've measured one audio device with a simple LM317 linear regulator and here's what I'm getting (see the attached pic). The input to the regulator (blue trace) is ~76mV RMS, while output is ~10mV RMS. Am I reading the figures correctly? According to the datasheed LM317 should provide 60-80 dB ripple reduction, so the output ripple should be in the sub mV range.
If these datasheet figures are correct, what can be causing this highish ripple?
The ADJ pin of the regulator is bypassed with the cap.

EDIT: the scope BW is 20 MHz

Could be any one of a number of things. LM317 needs a minimum of 3 volts across it to regulate correctly. it also needs a minimum load current of 10mA to regulate properly. What values do you have for these parameters?

Cheers

ian

Ilya said:

...If these datasheet figures are correct, what can be causing this highish ripple?..

Click to expand...

Poor grounding scheme or too low IN-OUT voltage difference.

These figures are under load (100 mA approximately), and there's more than 5V difference between in and out. So this leaves bad grounding? What should I look for in that respect?

The devil is in the details.  Do you have a schematic?  What voltages are you providing and regulating to?

Ilya said:

These figures are under load (100 mA approximately), and there's more than 5V difference between in and out. So this leaves bad grounding? What should I look for in that respect?

Click to expand...

Ripple. You need 3V even at the lowest part of the ripple. Let's say you read 20V and the output is 15V. If you have 6V ripple, the in-out voltage will oscillate between 2 and 8V. When it's 2V, the regulator cannot do its job.

The data sheet ripple reduction figures are taken with Vout - Vin = 15V. At lower in/out voltages it will not be so good but it should still be better than you measure.

Cheers

Ian

Given it a little more thought, I tend to agree with moamps and would think the grounding arrangement is the cause of the issue. The smoothing capacitorcurrent is probably allowed to interfere with the reference.

The circuit is basically the same as suggested in the LM317 datasheet for reduced ripple PSU (with the bypassed ADJ pin). Input caps are around 2200 uF, and output is around 20 uF, but there're large local caps down the line.
I've checked the other rail (which happened to be a +18V one) and I'm leaning to the grounding issues as well. Since 40V input gives plenty of voltage differential for regulation.
What is on the layout right now is on the left part of the picture (Var A). What would you suggest to reduce the input smoothing cap ripple? Something like ar B?

Wires and PCB traces have resistance, so current flow will cause voltage drops.

That waveform looks like the current charging up the reservoir capacitors.  We can't prevent these voltage drops in traces but we can be careful about where the regulator connects its ground reference so it can ignore the small voltage drops.

Imagine every PCB trace or wire is a resistor, then think about the current flowing in it. On the output side of the regulator, currents flowing should be relatively clean DC so not cause waveforms like you see. On the reservoir cap side of the regulator, all those grounds will have strong AC currents and small voltage drops (thats just how they work).

The regulator 0V reference can be just about wherever you connect it, but be consistent, and keep the cap charging current out of your reference ground trace. 

JR

Ilya said:

The circuit is basically the same as suggested in the LM317 datasheet for reduced ripple PSU (with the bypassed ADJ pin). Input caps are around 2200 uF, and output is around 20 uF, but there're large local caps down the line.
I've checked the other rail (which happened to be a +18V one) and I'm leaning to the grounding issues as well. Since 40V input gives plenty of voltage differential for regulation.
What is on the layout right now is on the left part of the picture (Var A). What would you suggest to reduce the input smoothing cap ripple? Something like ar B?

Click to expand...

Can you draw an exact layout including where the xfmr center tap connects to the ground and what point is connected to the chassis?

abbey road d enfer said:

Can you draw an exact layout including where the xfmr center tap connects to the ground and what point is connected to the chassis?

Click to expand...

Yes, I can. I'm attaching the layout of the other PSU which exhibits the same ripple behaviour. It's 35V in, 24V out. I hope you can figure out the traces. Let me know if they are not clear.
Star point is at the polygon connection. Transformer is off-board in a separate enclosure. There're diode bridge and 2000 uF capacitor inside the transformer enclosure. This feeds into the 28V connector. Device chassis connection is near the star ground. In the external enclosure, chassis is connected to the mains earth only. This mains earth is connected to the device chassis via the PSU cable.

Here's the schematic

Ilya said:

Here's the schematic

Click to expand...

Well, there is a cap multiplier before the regulator. It absorbs at least 1V, maybe more. Is it really 35V instead of 28?
What current are you drawing when you make the measurement?

Yes, the cap multiplier "eats" around 0.7-0.8V. The input (marked 28V) is actualy closer to 40V - it's after 30 VAC transformer winding, so it ends up around 40V after diode bridge drop. The winding is rated at 500 mA. Current draw is approximately 150-180 mA.

Ilya said:

Yes, the cap multiplier "eats" around 0.7-0.8V. The input (marked 28V) is actualy closer to 40V - it's after 30 VAC transformer winding, so it ends up around 40V after diode bridge drop. The winding is rated at 500 mA. Current draw is approximately 150-180 mA.

Click to expand...

Mmmmm... Are you positive there is no connection between the ground of the unregulated block and the earth?

abbey road d enfer said:

Mmmmm... Are you positive there is no connection between the ground of the unregulated block and the earth?

Click to expand...

Well, guess what. I made a really silly mistake. It turned out, the external PSU did have the connection from filter cap to chassis. I removed it and ripple went away.
This whole grounding thing - you just can't turn your back on it. Or it will bite you in the .....
Thanks for your help!

Yes, we must keep the sewage out of the clean water pipe...

JR

It's unclear to me what was the root cause, given it's common to bond the 0V reference to the safety earth to prevent them from being too far apart.

Was this a case of there being multiple connections from 0V to the chassis in this case?

Matador said:

It's unclear to me what was the root cause, given it's common to bond the 0V reference to the safety earth to prevent them from being too far apart.
Was this a case of there being multiple connections from 0V to the chassis in this case?

Click to expand...

Here's the setup. There're 2 chassis: external PSU and devices chassis. Both are metal enclosures and both are bonded together via the PSU cable. The PSU chassis is bonded to the safety earth as required.
Now, bonding filter cap ground at external PSU to chassis gave huge ripple figures. While bonding clean ground of the device to the device chassis gave good ripple figures.
I hope this clears the things up a bit.