TPS2421-1 Problem

The TPS2421 output being low (4V)indicates that it is in current limit.  If it times out in overload, the -1 version should be latching off. So the "oscillating" symptom may be the result of an interaction between the TPS2421 current/power limit and the inverting converter startup.

You should measure the current, 2421's output voltage, timer capacitor, and -15V together to get more information. 

A few more things to consider are: 1) the TPS2421 has a power limit (V across x I thru) of 5W nominal, 2) the converter will have a startup inrush most likely greater than the steady-state, 3) the relatively small converter input capacitor may have the TPS2421 seeing the peak current (not just the average), 4) triggering the power limit is most likely to occur if the converter starts up as the 2421's output is rising.    Also make sure that the converter is not loaded with an active load.

If you find that once started, that the TPS2421 can carry the load, you might find that using its /PG to hold the converter off will allow a start (see item 4).

I have taken some scope measurements and will present them here to see if someone can spot anything abnormal. The reason I am particularly confused about  this, is because I protyped this same circuit on a Rev A PCB that we developed first, and it works fine. Now, on my Rev B board, I get this fault.

The first trace below shows the CT pin (yellow) on enable of device (purple). As you can see, it is not getting anywhere near the threshold voltage of 1.4V before starting to oscillate.

The image below shows the trace of Pin 7 or Vout (yellow) on enable (purple).

The image below shows Pin 8 or /PG pin (yellow) on enable (purple).

The below image is the -15V output from the MAX1846 agains the CT pin of the TPS2421.

Below is Pin 7 or Vout against the -15V output.

Below is again Pin 7 or Vout against the -15V output but at a smaller timebase.

Any feedback gratefully received.

Thanks.

The set of plots show a very large amount of "noise" on the Vout node.  The voltage overshoot shown (to almost 7V) would indicate that there was a fair amount of series inductance with insufficient capacitance to control the voltage.   The dc-dc converter switching is the obvious source, and the last picture shows that the converter load in some fashion pulls the TPS2421 output low to the point where the converter probably hits its UVLO and then does a restart once the Vout recovers.

Since this configuration worked at one time for you what changes were there from your rev A to rev B?

Some things to look at are:

1)  Check the ESR and layout of the C40, C44, C48, C52 to assure they are really doing their job in carrying the ripple current and controlling Vout
  As a note, input caps don't really form a filter (classically a two-element form), the supply ends up sourcing a lot of the pulse current/noise.  An LC input filter is really better for the system. 
2)  Check the Vin for similar noise.  Is the input inductance too high and does it need capacitance to control it.  I would expect the voltage overshoot on Vout to be from interrupted converter current (pulse-by pulse) through input inductance driving the voltage up.  Does the input supply droop as well as Vout?
3)  The converter current limit looks pretty high, perhaps R45 should be more like 48mOhm.
4)  Make sure that the inductor L5 is not saturating.  Looking at the pulse-by pulse current (or perhaps the voltage across R45) should tell.
5)  Is there more loading on the converter now than previously, perhaps causing the inductor current to run higher.
6)  Is the higher noise level leaking into the TPS2421 control pin (e.g. Iset) through proximity to L5, or some routing changes.

Thanks for your feedback Martin - very insightful.

FYI, I have been able to get the circuit to work by removing two of the MAX1846 output capacitors (C38 & C45). I'm not quite sure what, if any, the implications of this are yet, but I'll study your comments in more detail and see if there's anything I else I can try based on these.

Thank you.