SMPSU

abbey road d enfer said:

Building high-power smps is enough to occupy a whole lifetime. I'm not that patient.
I know the guys at Powersoft. They spent their evenings at uni making sm power amps and PS. One was designing, the other replaced the exploded FET's!  ;D

Click to expand...

At Peavey we had a digital group with years (decades) of experience designing class D amplifiers while high power switching device technology was not quite up to that task back a few decades ago, like it is now. Switching PS are relatively easier, but not trivial. Chip makers often have detailed application notes including preferred PCB layouts, but I agree it is not worth the trouble for onesy-twosey projects.

Agreed for one-offs or as a small manufacturer using off the shelf OEM solutions makes sense. Back a couple decades ago (consulting after I quit Peavey) I inherited supporting a low volume DSP product using a custom switching supply. They cleverly tried to get two regulated rails from one switcher using a transformer. I killed a lot of brain cells trying to make that transformer quiet (it buzzed like an angry hornet).  I finally used the Gordian knot solution, simplifying the design to just a single regulated switching rail, and added a common pass regulator for the lower voltage rail. The small loss in efficiency was well worth it for the dead silence gained by using a cheap standard off the shelf inductor. I never even found a second source for the obscure transformer that buzzed.  (That was the same switcher that I piggy backed a small cap value, HF cap doubler/tripler to make a 48V phantom rail).

JR 

abbey road d enfer said:

That's correct. These smps have floating outputs, you can connect them any way you like. E.g. you can stack two of them for 48V.

Click to expand...

This implies a "flyback buck" topology.  Did it come with a schematic?  Or can you glean which chipset it's based on?

I'm wondering if there are standard feedback resistors that could be removed/changed to give a 48V output directly (provided the transformer can handle it).

Matador said:

This implies a "flyback buck" topology.  Did it come with a schematic?  Or can you glean which chipset it's based on?

I'm wondering if there are standard feedback resistors that could be removed/changed to give a 48V output directly (provided the transformer can handle it).

Click to expand...

If you can see the components  (not potted), search out the switching IC manufacturer for application notes. They should advise the capability of the design while 48V may require higher voltage switching devices and capacitors.

JR

Matador said:

This implies a "flyback buck" topology.  Did it come with a schematic?  Or can you glean which chipset it's based on?
I'm wondering if there are standard feedback resistors that could be removed/changed to give a 48V output directly (provided the transformer can handle it).

Click to expand...

IMO, it is based on standard TL431 IC as reference, so output voltage can be changed by changing resistors around it. But, it most probably cannot be changed 200%,  a different transformer's ratio is needed to achieve that (and some other parts should be changed).  Also, if it works as negative voltage source, one blue capacitor should be relocated.

moamps said:

IMO, it is based on standard TL431 IC as reference, so output voltage can be changed by changing resistors around it.

Click to expand...

The chip is AP8012H. The reference is integrated.

But, it most probably cannot be changed 200%,  a different transformer's ratio is needed to achieve that (and some other parts should be changed). 

Click to expand...

  Indeed. Typically smps can be trimmed by about 10%. Thaat's how I got my +/-15V rails for a 51X rack with nominal 15V smps.

Also, if it works as negative voltage source, one blue capacitor should be relocated.

Click to expand...

I believe you mean the CY capacitor between Neutral and ground of the secondary side...?
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwiStcmK0fvqAhVB8BQKHTa_B_QQFjAWegQICRAB&url=http%3A%2F%2Fclub.szlcsc.com%2Farticle%2FdownFile_756A0A8D8617647C.html&usg=AOvVaw1BVgt9o9M499xTN_oMcQrV
I have only academic knowledge of smps, never designed one, I always wondered what role exactly this cap plays. I believe it has to do with preventing the secondary side to emit RFI. BTW, any recent designs omit this cap.
It makes sense if the equipment it powers is not grounded.
But once the smps is integrated in a class 1 unit this becomes irrelevant since there is already a low impedance connection between "ground" and earth.
So instead of trying to move the CY cap, I would maybe just cross-connect the L and N inputs to the negative supply, but anyway I don't see how it could make a significant difference...
You seem to have a better understanding of these, so please elaborate.

abbey road d enfer said:

I believe you mean the CY capacitor between Neutral and ground of the secondary side..
I have only academic knowledge of smps, never designed one, I always wondered what role exactly this cap plays. I believe it has to do with preventing the secondary side to emit RFI. BTW, any recent designs omit this cap.

Click to expand...

I also have never designed one but I recall reading an article which said this cap is there to reduce/prevent common mode noise appearing on the output.

Cheers

Ian

abbey road d enfer said:

The chip is AP8012H. The reference is integrated.

Click to expand...

The voltage reference is U3 (TL431).

I believe you mean the CY capacitor between Neutral and ground of the secondary side...?

Click to expand...

Yes.

I have only academic knowledge of smps, never designed one, I always wondered what role exactly this cap plays. I believe it has to do with preventing the secondary side to emit RFI. BTW, any recent designs omit this cap.
It makes sense if the equipment it powers is not grounded.

Click to expand...

It is connected usually between primary and secondary 0V reference, it helps to lower jitter in control signal and overall lower RFI.  If you leave it in -V power supply as is, it can dump garbage in your -V line.

But once the smps is integrated in a class 1 unit this becomes irrelevant since there is already a low impedance connection between "ground" and earth.
So instead of trying to move the CY cap, I would maybe just cross-connect the L and N inputs to the negative supply, but anyway I don't see how it could make a significant difference...

Click to expand...

I don't follow you here, you can't connect 0V of primary side to safety earth; in class1 environment you can add Y caps to earth.

I found one ok explanation here:
https://electronics.stackexchange.com/questions/216959/what-does-the-y-capacitor-in-a-smps-do

"Switched mode power supplies use what is known as a "flyback converter" to provide voltage conversion and galvanic isolation. A core component of this converter is a high frequency transformer.

Practical transformers have some stray capacitance between primary and secondary windings. This capacitance interacts with the switching operation of the converter. If there is no other connection between input and output this will result in a high frequency voltage between the output and input.

This is really bad from an EMC perspective. The cables from the power brick are now essentially acting as an antenna transmitting the high frequency generated by the switching process.

To suppress the high frequency common mode is is necessary to put capacitors between the input and output side of the power supply with a capacitance substantially higher than the capacitance in the flyback transformer. This effectively shorts out the high frequency and prevents it escaping from the device.

When desinging a class 2 (unearthed) PSU we have no choice but to connect these capacitors to the input "live" and/or "neutral". Since most of the world doesn't enforce polarity on unearthed sockets we have to assume that either or both of the "live" and "neutral" terminals may be at a sinificant voltage relative to earth and we usually end up with a symmetrical design as a "least bad option". That is why if you measure the output of a class 2 PSU relative to mains earth with a high impedance meter you will usually see around half the mains voltage.

That means on a class 2 PSU we have a difficult tradeoff between safety and EMC. Making the capacitors bigger improves EMC but also results in higher "touch current" (the current that will flow through someone or something who touches the output of the PSU and mains earth). This tradeoff becomes more problematic as the PSU gets bigger (and hence the stray capacitance in the transformer gets bigger).

On a class 1 (earthed) PSU we can use the mains earth as a barrier between input and output either by connecting the output to mains earth (as is common in desktop PC PSUs) or by using two capacitors, one from the output to mains earth and one from mains earth to the input (this is what most laptop power bricks do). This avoids the touch current problem while still providing a high frequency path to control EMC.

Short circuit failure of these capacitors would be very bad. In a class 1 PSU failure of the capacitor between the mains supply and mains earth would mean a short to earth, (equivalent to a failure of "basic" insulation). This is bad but if the earthing system is functional it shouldn't be a major direct hazard to users. In a class 2 PSU a failure of the capacitor is much worse, it would mean a direct and serious safety hazard to the user (equivilent to a failure or "double" or "reinforced" insulation). To prevent hazards to the user the capacitors must be designed so that short circuit failure is very unlikely.

So special capacitors are used for this purpose. These capacitors are known as "Y capacitors" (X capacitors on the other hand are used between mains live and mains neutral). There are two main subtypes of "Y capacitor", "Y1" and "Y2" (with Y1 being the higher rated type). In general Y1 capacitors are used in class 2 equipment while Y2 capacitors are used in class 1 equipment."

moamps said:

If you leave it in -V power supply as is, it can dump garbage in your -V line.

Click to expand...

I would have thought that the impedance of the output at EMI frequencies would be extremely low, so it wouldn't really matter. In practice, as far as I can see, it doesn't. If I was making medical instrumentation, taht would be different maybe.

I don't follow you here, you can't connect 0V of primary side to safety earth

Click to expand...

I didn't suggest that; what I meant is the output is connected to the audio "ground", which is in turn earthed, so the CY cap is actually bypassed by a hard link.
As to this smps used as a negative supply, the hard link between its positive output and ground is dominant over the CY cap, IMO.
Indeed, class 2 equipment is another story.

EDIT: Actually, reading the whole thread is very enlightening.
"A new engineering tradeoff for the use of Y capacitors has emerged within the last decade or so, due to NEC (US National Electric Code) requirements for use of GFCI and AFCI circuit breakers. "

"In a class 1 (earthed) PSU we can use the mains earth as a barrier between input and output either by connecting the output to mains earth (as is common in desktop PC PSUs) or by using two capacitors, one from the output to mains earth and one from mains earth to the input (this is what most laptop power bricks do). "

I understand this as a proper justification to simply getting rid of the CY caps, since the secondary sides are properly earthed.
I see further justification in the app notes of Power Integration. Their non-isolated PS have no CY cap, which makes sense to me.