Help needed after capacitor replacement, now MORE noise??

[TERRA Operative]

I'm currently poking away at a HP 437B Power Meter here and running through some tests.
One of the tests was to check the noise on the various voltage rails and if it was more than 10mV, the manual states that filtering caps probably need to be changed.
It was a little over 10mV, so I pulled the old capacitors out and replaced them with new, but I am now getting MORE noise, like 30mA+ etc.

I have measured the capacitors and the only obvious thing I can see is the old capacitors on the output of the vregs have a much higher ESR.

Attached is the power supply schematic, it all looks like a pretty basic design, the HP schematic is basically a copy of the data sheet reference designs it seems.
I have replaced C3, C4, C5, C6, C7, and C11. C12 is a Kemet T110 Hermetically sealed Tantalum which I didn't replace.
U1 is a LM2935.
U2 is a LM325H and isn't anything special either. There is no mention about ESR that I can see in the data sheets.

The ripple on the unregulated side after the capacitor swap are as follows (I didn't measure before):
+5V: 496mV Peak to Peak
+15V: 776mV Peak to Peak
-15V: 648mV Peak to Peak

Here are the measurements I took with my Der-ee DE-5000 LCR meter, it doesn't have great resolution for low ESR, but I can fire up my HP 4276A if more accurate measurements are needed.

C3/C4
Old - 470uF, 50V Sprague 674D
Measured 652uF/481uF, 0.0ohm ESR

New - 470uF, 100V United Chemicon KMG
Measured 419uF/420uF, 0.1ohm ESR


C7
Old - 4700uF, 25V Sprague 674D
Measured 4610uF, 0.0ohm ESR

New - 4700uF, 35V United Chemicon KMG
Measured 4620uF, 0.0ohm ESR

C5/C6/C11
Old - 10uF/10uF/10uF, 25V United Chemicon KM Axial (Equivalent to KMG)
Measured 8.5uF/8.4uF/8.4uF, 18.3ohm/16.2ohm/19.4ohm ESR

New - 10uF/10uF/10uF, 50V Nichicon VX
Measured 9.8uF/9.8uf/10uF, 5.7ohm/5.5ohm/5.2ohm ESR

C12 Tantalum
100uF 10V Kemet T110
Measured 100.7uF, 0.2ohm ESR




Any ideas what to try to get this noise on the output back down under 10mV???
Should I up the capacitance on C3, C4 and C7 or try low ESR caps? Or maybe the new caps on the output side are too low ESR?

[Tomorokoshi]

From "HP 437B +5V Noise Check - New Caps.png" it seems the period of the half-cycle is 12.5 ms (80 Hz), for a full-cycle period of 25 ms (40 Hz)? However, "HP 437B +5V Unreg - New Caps.jpg" shows a clean half-cycle of 10 ms (100 Hz) with clean full cycle of 20 ms (50 Hz) as I would expect.

With the switching noise on top of that strange frequency, are you getting into some kind of motorboating phenomenon?

Can you measure Z with the HP 4276A and check for high series inductance with that bulk capacitor?

Can you zoom into the noisy waveform and see the switching frequency?

[TERRA Operative]

Here's the measurements I took with the 4276A, I averaged the readings where I had multiple caps to measure.
I'll have to pull a new 470uF cap from the unit to test later.

Old 10uF
8.4uF
ESR - 19.3ohm
Dissipation - 0.1222
Q - 8.2
Z - 159.64ohm, -83.06deg

New 10uF
10.04uF
ESR - 6.28ohm
Dissipation - 0.48
Q - 20.9
Z - 132.25ohm, -87.3deg

--------------------

Old 4700uf
4670uF
ESR - 0.017ohm
Dissipation - 0.06
Q - 17
Z - 0.284ohm, -87deg

New 4700uF
4620uF
ESR - 0.029ohm
Dissipation - 0.104
Q - 10
Z - 0.289ohm, -84deg

--------------------

Old 470uF
566.5uF
ESR - 0.051ohm
Dissipation - 0.0235
Q - 45
Z - 2.399ohm, -88.7deg

[duak]

I would expect to see ripple voltage like that on a +5V rail because it provides power to some digital circuitry.  I'm sure the 437B has a microproccessor and some digital logic that are generating all sorts of current pulses.

What probing method are you using to observe the supply rails?  A 10X scope probe and ground connection will easily pick up EMI radiated from the digital section.  You may have to make up a super short ground clip to the barrel of the probe to minimize the loop area.  When I have to look at low level signals in the midst of digital circuitry and they're not brought out on coax connectors, I use a coax cable with a 1k0 isolating resistor in series with the center conductor to make a simple 1X probe.

[capt bullshot]

Too low of an ESR can cause the regulators to get unstable (would be rather obvious on the scope), or if it's an edge case ...
So check your replacement caps at the LDO outputs if their nominal ESR is within the specified range (sometimes there's a lower limit for ESR too).

I second duak's hint about the probing techniques. Use a short and bold GND connection and avoid any kind of loop area in your probing circuit, otherwise you'll pick up noise that doesn't exist. In the past I've wasted a day hunting noise, because the probe was intended to be set 10:1 but was accidentally set to 1:1.

[TERRA Operative]

I'm using a 1x probe, connected directly to the test pins as per the HP manual.
When I need to loop to reach adjacent pins, I'm using a short bit of wire.

I think it might be ESR related, seeing as the old 10uF capacitors have much higher resistance than the new ones. I might try reinstalling them to see what happens.