cupoftea
Advanced Member level 6
Hi,
It is believed that finally the senseless addiction to offTheShelf
Switch Mode Power Supplies is starting to reduce.
This is necessary because Power Supply modules very often go obsolete and then
the product mechanics need re-doing.
But also, they all have noise issues. Inevitably they all bare datasheets bragging
high, eg mid-nineties efficiencies. This is due to the fact that their FETs are switched
extremely fast and so have low switching loss, but in return, they spew noise all over the place
since the FET drain node transition in any SMPS is the prime causation for EMC problems.
Especially the common mode problems that are the most difficult to solve.
If a company designs its own SMPS, out of discrete components, then the FETs' switching can be damped,
and the heatsinking/cooling matched to enable this...giving a much better and "quieter" EMC situation.
Many Switch Mode PSU modules are in big systems where there are multiple enclosures all interconnected
by big cable runs of comms and power.....having noisy power supplies in amongst all this lot can
cause an EMC nightmare...since all those cables will re-radiate and act as antennae for the PSU switching noise.
Many companies that make these big systems using offTheShelf power modules, end up with massive numbers of quirky
"bits and pieces" that they have to add to the system in order for its operation to not be destroyed by noise.
If they made their own power supplies, and damped the switching transitions, then they would not have to bother
with all these quirky "bits and pieces". "Bits and pieces" which often need time consuming "tweaking" anyway since
they are acting on the strays, which can be slightly different from one product assembly to the next.
Also, you can fully use common mode chokes if you design your own power supplies..
Take the following Power supply module...
24V 25A DCDC (PQ60 power supply)
www.synqor.com
..Page 13 shows that if 3 of the PQ60's are paralleled, then only one common mode choke can be used.
Whereas if you design your own, you can parallel them and arrange for use of a common mode choke
right at the input terminals of each Power supply. -And get much better common mode EMC performance.
Page 15 shows that any remote sense connections must be to the upstream side of the output common mode
choke of the PQ60. So the common mode choke therefore wont be able to be placed close to the power supply,
which is a bad thing for EMC. If you design your own, you can arrange for remote sensing but still
have some common mode choking right at the output terminals of each power supply that is paralleled.
Another point is that the PQ60 has facility to be heatsunk to a flat metal baseplate which would be resting against the
flat surfaces of the various ferrite inductors and transformers. Unfortunately the FETs would have to rely
on squidgy gap pad to thermally couple them to this metal baseplate.
Anyway, the metal baseplate would need to be capacitively coupled to the input ground of the PQ60. This
would need to be done with a Y Capacitor. The PQ60 should have a y capacitor on it which is connected to its input ground.
The other terminal of this Y capacitor should go to a screw hole, so that external connections can be made to Y capacitively
connect the input ground of the PQ60 to the metal base plate (chassis) for the purpose of common mode noise reduction.
The PQ60 makes no mention of this kind of facility so straight away any potential buyer has noise concerns.
Also, the PQ60 PCB, should really comprise one full , internal copper layer which is dedicated to being a "chassis plane".
This should come out to the bottom of the PCB at a screw hole restring, so that this "chassis plane" can
be directly electrically connected to the external metal base plate (heatsink). However, again, the PQ60 datasheet makes no mention
of this. Having a "chassis plane" in the PCB like this acts like a stray y capacitor between chassis and the nets
of the PCB, which helps to reduce common mode emissions. However, the PQ60 datasheet makes no mention of any "chassis plane".
Again, if you design and build your own power supply, you can see to it that all of these EMC measures are well catered for.
Probably the biggest point is that any power supply you design yourself can have the FET drain transitions well damped.
This gives a massive reduction in common mode emissions. Even just a small amount of damping, leading to insignificant
extra switching loss can really reduce common mode emissions.
Many Power supply modules sit within big systems which compirse large numberss of enclosures all interconnected by large
runs of power and communications cabling. This is an absolute recipe for a common mode EMC nightmare....very often resulting
in the product not even working. -And often leading to multiple ridiculous assembly practices being followed because they reduce the strays
such that the diabolical common mode EMC issues dont raise their ugly heads.....if these people had designed their
own PSUs and damped the switching transisiton, then they wouldnt even need all the "crazy" little EMC fixes that they need to do.
-Because they would be stopping the common mode from getting out of control in the first place.
Some examples of common mode nightmares:
Many peole do not realise just how bad common mode can be if the switching transition of an SMPS is not damped sufficiently.
1...150W LED driver:
At one company, they had had a PCB designed comprising three one hundred Watt Buck converters (48Vin, ~36Vout) on a PCB.
When the Buck converters were turned ON, the lab radio immediately went OFF, (even though it was ON at full blast)
when the Buck converters were turned OFF, the lab radio immediately came back ON….and so on.
The lab radio was plugged into the mains, the Bucks were powered from a 600W OffTheShelf PSU (from a well known, global PSU company)
also plugged into the mains.
I changed the series resistors in the FET gate drives from 2R2 to 4R7 to solve this “problem”.
This just shows how much of a fantastic EMC reduction measure is achieved by simply damping the drain node switching transition.
2....1W Buck causes big conducted common mode EMC failure:
A 1W High voltage Buck converter (using LNK304) had had its switching node copper routed on to a large copper pour that was
directly over an earthed heatsink on which the PCB sat (on a thermal pad). This resulted in failure of conducted EMC.
The problem, was solved by reducing the area of switching node copper. The switching node bits that remained, were kept
off the bottom layer of the PCB (away from the metal earthed heatsink). Also, “quiet node” copper was used as a
shield between the remaining bits of switching
node copper and the earthed heatsink. This shielding was done on the PCB layers that were closer to the heatsink than the
switching node copper.
This just shows that its the switching edges, and not necessarily the power level, that makes an SMPS noisy.
It is believed that finally the senseless addiction to offTheShelf
Switch Mode Power Supplies is starting to reduce.
This is necessary because Power Supply modules very often go obsolete and then
the product mechanics need re-doing.
But also, they all have noise issues. Inevitably they all bare datasheets bragging
high, eg mid-nineties efficiencies. This is due to the fact that their FETs are switched
extremely fast and so have low switching loss, but in return, they spew noise all over the place
since the FET drain node transition in any SMPS is the prime causation for EMC problems.
Especially the common mode problems that are the most difficult to solve.
If a company designs its own SMPS, out of discrete components, then the FETs' switching can be damped,
and the heatsinking/cooling matched to enable this...giving a much better and "quieter" EMC situation.
Many Switch Mode PSU modules are in big systems where there are multiple enclosures all interconnected
by big cable runs of comms and power.....having noisy power supplies in amongst all this lot can
cause an EMC nightmare...since all those cables will re-radiate and act as antennae for the PSU switching noise.
Many companies that make these big systems using offTheShelf power modules, end up with massive numbers of quirky
"bits and pieces" that they have to add to the system in order for its operation to not be destroyed by noise.
If they made their own power supplies, and damped the switching transitions, then they would not have to bother
with all these quirky "bits and pieces". "Bits and pieces" which often need time consuming "tweaking" anyway since
they are acting on the strays, which can be slightly different from one product assembly to the next.
Also, you can fully use common mode chokes if you design your own power supplies..
Take the following Power supply module...
24V 25A DCDC (PQ60 power supply)
Telecom DC-DC Power Converter | PQ60240HZx25 | SynQor
View the information for PQ60240HZx25 here
www.synqor.com
..Page 13 shows that if 3 of the PQ60's are paralleled, then only one common mode choke can be used.
Whereas if you design your own, you can parallel them and arrange for use of a common mode choke
right at the input terminals of each Power supply. -And get much better common mode EMC performance.
Page 15 shows that any remote sense connections must be to the upstream side of the output common mode
choke of the PQ60. So the common mode choke therefore wont be able to be placed close to the power supply,
which is a bad thing for EMC. If you design your own, you can arrange for remote sensing but still
have some common mode choking right at the output terminals of each power supply that is paralleled.
Another point is that the PQ60 has facility to be heatsunk to a flat metal baseplate which would be resting against the
flat surfaces of the various ferrite inductors and transformers. Unfortunately the FETs would have to rely
on squidgy gap pad to thermally couple them to this metal baseplate.
Anyway, the metal baseplate would need to be capacitively coupled to the input ground of the PQ60. This
would need to be done with a Y Capacitor. The PQ60 should have a y capacitor on it which is connected to its input ground.
The other terminal of this Y capacitor should go to a screw hole, so that external connections can be made to Y capacitively
connect the input ground of the PQ60 to the metal base plate (chassis) for the purpose of common mode noise reduction.
The PQ60 makes no mention of this kind of facility so straight away any potential buyer has noise concerns.
Also, the PQ60 PCB, should really comprise one full , internal copper layer which is dedicated to being a "chassis plane".
This should come out to the bottom of the PCB at a screw hole restring, so that this "chassis plane" can
be directly electrically connected to the external metal base plate (heatsink). However, again, the PQ60 datasheet makes no mention
of this. Having a "chassis plane" in the PCB like this acts like a stray y capacitor between chassis and the nets
of the PCB, which helps to reduce common mode emissions. However, the PQ60 datasheet makes no mention of any "chassis plane".
Again, if you design and build your own power supply, you can see to it that all of these EMC measures are well catered for.
Probably the biggest point is that any power supply you design yourself can have the FET drain transitions well damped.
This gives a massive reduction in common mode emissions. Even just a small amount of damping, leading to insignificant
extra switching loss can really reduce common mode emissions.
Many Power supply modules sit within big systems which compirse large numberss of enclosures all interconnected by large
runs of power and communications cabling. This is an absolute recipe for a common mode EMC nightmare....very often resulting
in the product not even working. -And often leading to multiple ridiculous assembly practices being followed because they reduce the strays
such that the diabolical common mode EMC issues dont raise their ugly heads.....if these people had designed their
own PSUs and damped the switching transisiton, then they wouldnt even need all the "crazy" little EMC fixes that they need to do.
-Because they would be stopping the common mode from getting out of control in the first place.
Some examples of common mode nightmares:
Many peole do not realise just how bad common mode can be if the switching transition of an SMPS is not damped sufficiently.
1...150W LED driver:
At one company, they had had a PCB designed comprising three one hundred Watt Buck converters (48Vin, ~36Vout) on a PCB.
When the Buck converters were turned ON, the lab radio immediately went OFF, (even though it was ON at full blast)
when the Buck converters were turned OFF, the lab radio immediately came back ON….and so on.
The lab radio was plugged into the mains, the Bucks were powered from a 600W OffTheShelf PSU (from a well known, global PSU company)
also plugged into the mains.
I changed the series resistors in the FET gate drives from 2R2 to 4R7 to solve this “problem”.
This just shows how much of a fantastic EMC reduction measure is achieved by simply damping the drain node switching transition.
2....1W Buck causes big conducted common mode EMC failure:
A 1W High voltage Buck converter (using LNK304) had had its switching node copper routed on to a large copper pour that was
directly over an earthed heatsink on which the PCB sat (on a thermal pad). This resulted in failure of conducted EMC.
The problem, was solved by reducing the area of switching node copper. The switching node bits that remained, were kept
off the bottom layer of the PCB (away from the metal earthed heatsink). Also, “quiet node” copper was used as a
shield between the remaining bits of switching
node copper and the earthed heatsink. This shielding was done on the PCB layers that were closer to the heatsink than the
switching node copper.
This just shows that its the switching edges, and not necessarily the power level, that makes an SMPS noisy.
