Stange Heat Sink

[calzap]

I recently found a power supply on a back shelf that I bought as surplus a few years ago.  It's a Protek PX18-23 made in Taiwan.  120 VAC input; 5 VDC @ 2 A and 12 VDC @ 1.2 A output.  It's a large, rugged beast for its capacity.  Special attention was given to safety and robustness: cartridge fuse and NTC thermistor in on input; a several X2 caps and MOVs; an isolation transformer on the input; power resistors raised off the PCB with metal or ceramic stand-offs; output stage well isolated.   See the first pic (input is lower right; output is upper right).  Not surprising considering Protek's main business is supplying power supplies for medical equipment.

One thing puzzles me about the output stage.  There is what appears to be a large vertical heat sink with only a diode connected to it near its top.  The body of the diode is not in contact with it.  See the second pic. The heat sink is electrically continuous with the common output pin.  What's its function?  Hard to believe it could provide much cooling for the diode because only one lead of the diode is in contact with it.

Mike in California

[tautech]

You'd be surprised how much heat a 1mm copper lead can conduct.
Often there is a spec for heat conduction with a specified lead length in diode datasheets.

And of course this would have been cheaper than a TO-220. 

[BravoV]

Its common practice, and I found out its quite effective as well. Touched with fingers  when on, the heatsink was quite warm, meaning it dissipates the heat from that thick pin. 

Here shot from salvaged "branded" TV parts, front & rear shots of the heat sink with diode soldered on it.

[PA0PBZ]

...an isolation transformer on the input;

Are you sure? I'd bet on a common mode choke all day...

[wraper]

...an isolation transformer on the input;

Are you sure? I'd bet on a common mode choke all day...

+1. Isolation transformer  . If isolation transformer would be used, It would be not smaller in size as if just for linear psu of the same power. I don't see anything better than in average PSU.

[wraper]

BTW I don't see any MOVs. Yellow one near big electrolitic capacitor could be, but likely is not considering how it is placed.

[LukeW]

Looks like the same PCB layout is used for multiple different models, with an extra output rail on other models.

There is another rectifier diode on the output side, next to the large heatsink, which is not populated on this model - this is why the heatsink is so large. D12 is labelled on the silkscreen, next to D9. Also note that some extra LC output filter components have space on the board, but are not installed.

[amyk]

Cheap PC AT/ATX PSUs did this too.

[calzap]

Yup; you're right.  The two blue and one yellow discs to the left of the bottom of transformer are caps (C3, C4 and C20, in fact).  I should have looked more closely.

I've attached a pic of the bottom side.  Input is lower left; output section is upper left.   You're right about the input transformer too.   It may appear to be an isolation transformer, but continuity checking shows upper left solder joint continuous with lower right solder joint and vice versa.  There is also a copper shield around the coil of the other transformer.  The big green wire connects the shield to the input ground. 

Another puzzling thing is the vertical heat shrink to the left of the black heat sink.  It's around a ceramic capacitor.  To prevent it from being bent against the heat sink or some other reason?

Mike in California

[SeanB]

You often see axial diodes with a thermal spec calling for a heatsink typically 5-10mm ( 1/4 or 1/2in) from the body, as this is a common mounting method. That is why you find real power devices ( from a reputable manufacturer not the off brand same part number from a random vendor somewhere in the world) have such thick copper leads, with a generous tin plate or solder dip on them. I find genuine 1N540x will have a typically 3mm diameter lead, as opposed the clones which can come with 1mm leads. Motorola parts have even thicker leads, and these often have a minimum length before bending is allowed, so that you can use a needle nose plier to grip the lead without stressing the body itself.

The power supply obviously has a variant with higher current ( at lower voltage) so there is space for a second rectifier diode in parallel with the original to reduce voltage drop. The copper shield on the main transformer is a shield, used to reduce noise and to prevent the core from saturating as it damps any fields that are not identical in each outer core section.

The sleeve on the ceramic capacitor is there for both protecting the capacitor from touching, as the body cement is not classed as an insulator when hot, but also to contain any fragments in case the capacitor disintegrates, as it is run in a snubber and has a high peak current and high voltage stresses on it.

[T3sl4co1l]

Hmm, heatsink is at +HVDC, too.  They're using it as a jumper wire...

Not sure why they heatshrink caps some times.  Insulation I guess would figure.  Maybe something about temperature too?  The extra thermal insulation will raise its temperature some, which will tend to reduce its value (they're more than likely using a type II ceramic there).

The band around the power transformer is for shielding.  Note it's not around the windings (which would short everything out!), but around the windings plus core, which acts to force a balance in flux between the sum of both core legs and the center leg (around which the windings go).  Without this, some flux will naturally leak away, causing radiation; the ground lead also helps keep electric as well as magnetic flux under control.

Topology appears to be... blocking oscillator, self excited, BJT switch!  Flyback mode.  The transformer construction apparently sucks (probably no interleaving), hence the large resistor to burn away the energy stored in leakage inductance (an RCD clamp snubber, without which, the transistor would be destroyed).

Tim

[PA0PBZ]

The second voltage is from a series SCR regulator, that must be quite noisy.

[T3sl4co1l]

The second voltage is from a series SCR regulator, that must be quite noisy.

Where???

Tim

[PA0PBZ]

The second voltage is from a series SCR regulator, that must be quite noisy.

Where???

Tim

Here:

[Yansi]

Isn't that SCR only for basic overvoltage protection? It simply triggers on overvoltage and shorts the output out.

[T3sl4co1l]

Yeah, that's a crowbar.  For the... 5V, I'm guessing?

It's only connected to R15 and ZD1, not the best way to do it (the zener doesn't turn on sharply, so the SCR can just sit there cooking in linear BJT mode when overvoltage occurs slowly), but it's better than nothing.

Tim

[PA0PBZ]

Isn't that SCR only for basic overvoltage protection? It simply triggers on overvoltage and shorts the output out.

Ah yes, my bad, it is connected to ground. I was wondering if they only regulate one voltage and depend on the other one to follow, seems that way.

[macboy]

Isn't that SCR only for basic overvoltage protection? It simply triggers on overvoltage and shorts the output out.

Ah yes, my bad, it is connected to ground. I was wondering if they only regulate one voltage and depend on the other one to follow, seems that way.

They almost always only regulate one voltage, and depend on the turns ratio in the transformer to produce correct voltages all around. Some topologies also have a single large toroidal inductor that has separate windings for each voltage. This inductor allows the filter caps to be shared between the different voltage rails since ripple current is inductively coupled among the rails.
Old PC supplies which could have a heavily loaded (>30A) 3.3V rail would often have ways to boost that rail.

[Yansi]

and depend on the turns ratio in the transformer to produce correct voltages - exactly... But leakage inductance is the big enemy of that principle. Once upon a time, I tried to build small power (~10W) blocking converter with outputs rails of 6 and 200V. It simply didn't want to work, because of the weak coupling between the windings. After loading tha 6V rail, the 200V become 350V.  After moving the feeedback closer to 200V rail, the 6V rail become weak as a cat.

PC supplies regulated the 3V3 rail with a saturable reactor, which took current from the 5V rail secondary.

[T3sl4co1l]

A somewhat unusual and contradictory application: I have a tube oscilloscope project, which is powered by a flyback switching supply.  Tubes need heat (as in, there's a tungsten filament that glows red hot), so the load on the heater supply is huge at startup.  HV soars to about +/-300V as there's no load until the cathodes get hot enough to conduct, and the heater supply droops a similar amount.  Once everything warms up, HV settles down to its nominal +/-200V, and the heater supply rests at a comfortable 6V.

Feedback is designed as the average between 6V and +200V, so when there's a heavy differential load, it will do a sausage effect (one drops while the other rises).  Better than trying to regulate just one and not getting anywhere (if +200V were held constant, the heaters would probably drop to 3V, hardly enough to even start conduction; or if +6V were held constant, the high voltage sections would cook their 600V rectifiers!).

Tim

[Yansi]

Tim, does your tube scope SMPS also have a tube on the primary side? 

[T3sl4co1l]

No... that's why I said "contradictory": the power section is a traditional UC3843 peak current mode flyback circuit.  (CRT supply (-2kV and 6.3VAC), and auxiliary power for the controller, is provided by a self excited oscillator, using a MJE18008 or something like that.)

I could've made a sweep tube based blocking oscillator, but it wouldn't be very powerful at 160VDC supply (saturation is ca. 30-100V depending on type and current level), and I'd rather avoid a doubler just because.  Needless to say, any iron cored transformers need not apply; I need that space!

On a separate occasion, I have made a class D tube amp.  Plate efficiency was pretty good, ca. 80%, though overall efficiency was piss-poor for a number of reasons.

Tim