SMPS Power cycling issue

[Potato Cakes]

Hello, everyone,

I've discussed the issue I was having with a prototype build that uses Neve BA cards where the SMPS, which was rated for current draw eight times higher than the actual load of the circuit. The solution in that instance was to add a CLC filter, using 220uF/200mH/220uF respectively. Also, the reservoir caps had to be decreased in value. All of this made sense and then I applied this to the next iteration of this prototype build, which I set out to more cleanly having made all of the mistakes on the first one. I used a smaller open frame SMPS from the same manufacturer rated for 1.25A, still several times higher than what the circuit would draw (https://www.mouser.com/ProductDetail/490-VOF-60C-S48). The reservoir caps were reduced further and I used 470uF caps for the CLC filter. Now the problem is back, even with removing the caps at the power input to each card. When I disconnect power to one of the BA283AM modules (uses 2N3055 for output), the power supply powers on correctly and stays powered on when I reconnect VDC to that same card. I know I'm using bigger caps for filtering, but the total value of all of the electrolytics that are parallel in relation to the power supply is less than that of the working unit. I thought I had this problem solved forever more but alas, I once more must struggle with power supplies, the thing that will eventually be my demise.

Are the filter caps too big? I've seen larger values used with switching PSUs that have only a slightly higher current rating but were lower voltage.

Thanks!

Paul

 

[JohnRoberts]

how many caps in parallel? How much total capacitance... if a lot do you really need that much?

JR

 

[abbey road d enfer]

What happens if you disconnect the circuit, start the PSU and connect the circuit later?

 

[ruffrecords]

A lot depends on the current limit circuit of the SMPS. Simple ones have foldback limiting which means if the load initially exceeds its rated current it drops the current to a trickle which can never charge up all the electrolytics because their leakage is as much as the trickle current. We had exactly this problem with Coutant linear power supplies when I was at Neve in the 70s. They just could not charge up all the electrolytics in a big console. They had to be modified to increase the foldback current limit so it was high enough to slowly charge up the caps.

Things are different with an SMPS. There is no easy way to change the foldback current limit and it could be dangerous to do so. The solution is to look for an SMPS which has a hiccup mode current limit (most MeanWell SMPS have it). This will pmp the rated current into any load for a short period, turn off for a short period and repeat. I find them very good for firing up tube heaters rated at up to 50% of the SMPS output current.


Cheers

ian

 

[Potato Cakes]

how many caps in parallel? How much total capacitance... if a lot do you really need that much?

JR

The boards that I am using each have a reservoir cap that the BOM lists as 1000uF, which is way too much considering that there are 5 boards total. Right now I reduced all of those caps to 47uF except for one which has 100uF which wasn't changed because it was harder to remove once it was all built. A previous iteration of this design uses a 1.6A SMPS and has 100uF on each board's power input, so I thought I would be fine. I don't know how much total would be a good amount of capacitance for a class A preamp, EQ, and stereo summing amp. I do know that a previous issue I have encountered with SMPS's was not loading it enough which caused a similar issue to what I have now.

Thanks!

Paul

 

[Potato Cakes]

A lot depends on the current limit circuit of the SMPS. Simple ones have foldback limiting which means if the load initially exceeds its rated current it drops the current to a trickle which can never charge up all the electrolytics because their leakage is as much as the trickle current. We had exactly this problem with Coutant linear power supplies when I was at Neve in the 70s. They just could not charge up all the electrolytics in a big console. They had to be modified to increase the foldback current limit so it was high enough to slowly charge up the caps.

Things are different with an SMPS. There is no easy way to change the foldback current limit and it could be dangerous to do so. The solution is to look for an SMPS which has a hiccup mode current limit (most MeanWell SMPS have it). This will pmp the rated current into any load for a short period, turn off for a short period and repeat. I find them very good for firing up tube heaters rated at up to 50% of the SMPS output current.


Cheers

ian

I might have to look at the MeanWell option. I will see if I can get a response from CUI regarding this issue.

Thanks!

Paul

 

[Potato Cakes]

What happens if you disconnect the circuit, start the PSU and connect the circuit later?

It's totally fine then. I as I posted, all I really need to do is disconnect one of the BA283AM cards while the PSU is powered on then reconnect power and everything is right as rain. This led me to consider using a soft start option but I'm running out of room in the chassis and also this is one more thing outside my electronics know-how. Plus, I have a working unit without needing one and I'm using the same boards in this current build, so I should have this version's power working properly.

Thanks!

Paul

 

[abbey road d enfer]

The boards that I am using each have a reservoir cap that the BOM lists as 1000uF, which is way too much considering that there are 5 boards total. Right now I reduced all of those caps to 47uF except for one which has 100uF which wasn't changed because it was harder to remove once it was all built. A previous iteration of this design uses a 1.6A SMPS and has 100uF on each board's power input, so I thought I would be fine. I don't know how much total would be a good amount of capacitance for a class A preamp, EQ, and stereo summing amp. I do know that a previous issue I have encountered with SMPS's was not loading it enough which caused a similar issue to what I have now.

Are these caps directly on the incoming rails? I always have a resistor inserted there. Good operation of a circuit should not rely on the PSU stiffness, but rather on that of the decoupling caps. Particularly in class A, where the average current is constant. That reduces significantly the turn-off current surge, in addition to reducing X-talk and increasing PSRR.

 

[Bo Deadly]

I used a smaller open frame SMPS from the same manufacturer rated for 1.25A, still several times higher than what the circuit would draw (https://www.mouser.com/ProductDetail/490-VOF-60C-S48). The reservoir caps were reduced further and I used 470uF caps for the CLC filter. Now the problem is back

Read the datasheet. It says the maximum capacitive load of the 48V output model is 1000uF.

One thing about SMPS that folks using linear supplies seem to have trouble understanding is that SMPS do NOT need large capacitor filters like linear supplies and in fact they will see large amounts of capacitance as a short circuit and kick on the overcurrent protection. SMPS are inherently less noisy than linear. Linear supplies use bridge rectifiers to brute force the AC into a very noisy DC that requires massive amounts of capacitance to smooth (and it never quite does because there's almost always some mains frequency bleed). The output of an SMPS is already pretty good DC that requires only a small amount of filtering.

So your 470uF / 200mH / 470uF filter is far too much of a load at DC for that little 1.25A supply. The capacitors alone are almost at the limit of the SMPS and the presence of the inductor might not be completely insignificant either. The CLC filter is only meant to remove the high frequency switching noise so it could be 1/10th the size for that supply.

As I have mentioned before many times, the proper solution is a capacitance multiplier which can be as few as 4 parts plus only one large capacitor. The two transistor version is 5 parts. If you can afford the voltage drop of a mosfet, it's only 3 parts and a cap.

Another thing that linear supply people have trouble with about SMPS is that SMPS work BETTER under load. So you don't want a supply that's more than what you need. In fact, I have recently considered that it might be better to actually pull a constant current supply into the CC region using a shunt regulator. Then you can also pick your exact voltage.

 

[JohnRoberts]

Reservoir caps fed by a dc-dc switching supply are redundant, reservoir caps are not intended to provide low impedance but to supply current during ripple sags. However when a bunch are in parallel their (low) impedance proves problematic for the regulated supply trying to quickly charge them.

An easy answer already suggested is a small series resistor that would current limit the initial surge. Some switchers use a soft start that will keep trying, but fold back current limiting may get stuck folded back. I suspect some combination of smaller caps and build out resistors could get this happy.

JR

 

[ruffrecords]

The thing to rememeber about vintage Neve class A circuits is that their power supply rejection ration (PSRR) is very poor by today's op amp standards. They do need significant decoupling in order to achieve good noise performance. However, none of these caps is placed directly across the supply, they are always fed via a resistor.

Cheers

Ian

 

[Potato Cakes]

Each of BA boards have a power input section that goes +VDC - resistor - capacitor (between +VDC and 0V) - circuit. I looked at my second version of this build and the SMPS being used is 1.12A using CLC filter of 220uF/220mH/220uF but the reservoir caps on each board are 220uF with one being 1000uF, so I'm not sure why the current version I'm working on is having such a problem besides the size of the filtering caps. There does not seem to be a capacitance load listed on the data sheet of this particular SMPS (https://www.mouser.com/ProductDetail/TRACO-Power/TXM-050-148?qs=ckJk83FOD0WnaoJ1CQyw/Q==) so I don't know if I'm comparing similar things.

I've also managed to recreate the motor boating issue I had a while back, so now I get to rummage through old posts to re-remember how to fix it.

The biggest frustration right now is that I have two working prototypes, but when I try to combine what I learned from both builds into a more finalized version and now I'm running into old problems. I've been at this for about 2 years now and the PSU issue keeps kicking me in the teeth. Everyone here has been extremely helpful navigating me through this one issue so I don't know why at this point I can't complete this particular build and have it work correctly the first time.

I'm taking a break from this for a couple of hours then I'll tinker with it some more and report back.

Thanks!

Paul

 

[ruffrecords]

I think your CLC filter is a little on the heavy side and it is possible that it can make the feedback loop unstable which could cause all sorts of problems. I would recommend changing the 220uF to 47uF and your 220mH to 220 micro Henries.

It is worth putting a scope or meter across the output so you can see if the hiccup is working or not when you power up. The data sheet say the SMPS you are using has hiccup / constant power so it should work. I have done this on 12V heaters powered by an SMPS. Typically you can see the voltage rise to 5 or 5 volts then drop, the rise to about 11V and drop and on the third try it reaches 12V and stays there.

Cheers

Ian

 

[Potato Cakes]

I guess I have a misunderstanding of selecting values for a CLC filter. From what I gathered it was better to have a higher capacitor value to further slow the delay of the pulsing DC, lowering rippling.

Right now I've had to delay looking into this matter as I found the filter board I'm using (not my design) is creating a permanent low pass filter down to about 400Hz, so I have to spend some time debugging that first.

Thanks!

Paul

 

[john12ax7]

SMPS usually have a max capacitor load specification, you should find out what this is for your particular supply. Bigger is not always better.

 

[ruffrecords]

I guess I have a misunderstanding of selecting values for a CLC filter. From what I gathered it was better to have a higher capacitor value to further slow the delay of the pulsing DC, lowering rippling.

In general this is true but an SMPS is a rather special case because you are adding a network to the negative feedback network that regulates the output voltage. There are plenty of articles on this on the web.

Cheers

Ian

 

[Khron]

In general this is true but an SMPS is a rather special case because you are adding a network to the negative feedback network that regulates the output voltage. There are plenty of articles on this on the web.

Cheers

Ian

Wouldn't that only apply if one added parts BEFORE the point the feedback is taken from? Usually that's taken from the first smoothing cap after the diodes, and whatever follows downstream is less relevant.

 

[ruffrecords]

Wouldn't that only apply if one added parts BEFORE the point the feedback is taken from? Usually that's taken from the first smoothing cap after the diodes, and whatever follows downstream is less relevant.

Anything attached at the point the feedback is taken from can affect the loop response and hence the stability and generally the output of the SMPS is that point.

Cheers

Ian

 

[Potato Cakes]

Managed to do some more testing on the issue at hand and adding a 12R resistor in series with the +VDC fixed the issues. Changing the filter caps valued or even removing the CLC all together had no effect. It's a little confusing as each of the BA circuits have a resistor in series with the power input before it goes to the rest of the circuit, but maybe it needed a little more resistance overall. I have no idea but the issue is now resolved, so now I get to move on to the ground hum/buzz issue which is not present in the other two prototypes. The pulsating noise is also gone but I think that

As always, I'm very grateful to everyone's help and suggested remedies. I would be completely lost without the generous, knowledgeable, and helpful individuals who make up this community.

Thanks!

Paul