EEVblog #1298 - $70 2000W Power Supply?

[EEVblog]

A look at the $70 Juntek 2000W power supply with wireless display.
Also the 3000W model, but does it have a problem?
https://s.click.aliexpress.com/e/_dZVpwTI

[Kleinstein]

The constant power mode is likely mainly an issue with the electronic load. It start oscillating when the supply hit the CC limit.

For a pure switched mode regulator those oscillations are not so bad. Still it is some endurance testing for the CC/CV transitions. The supplies that don't like this are linear ones with a relay for transformer tap switching.

The reset of the higher power module is likely due to the input side power supply hitting the CC limit. So a drop in the input and thus a reset from there. Can't blame the regulator module for that - maybe except for an explanation in the manual for this probably not so uncommon situation.

The output ripple is really a little high, but this is kind of expected. The output side caps are relatively small for 20 A.
The 50 mV may be no load or a lower power version from the line. The higher power model may be better, as it may be a dual phase regulator - this at least would make sense with 2 equal power inductors and 2 power mosfets.

For the output loss, in a conventional buck converter the loss to a large part depends on the current, not so much the voltage level. So 20 V in an 5 V 20 A out is actually already causing quite a fraction of the maximum losses.  100 V in and 95 V out at 20 A would likely have less loss - it would be essentially the FET on nearly all the time.

The remote control unit looks really nice.

[jeremy]

Anyone else not seeing this video appear on YT?

[EEVblog]

Anyone else not seeing this video appear on YT?

It's been reuploaded, it's more betterer.

[johnlsenchak]

Dave

Wouldn't those units be (Buck)   "DC-DC converters" instead of calling  them   switch  mode  power supplies ?

[maukka]

Anyone else not seeing this video appear on YT?

It's been reuploaded, it's more betterer.

I see you did an overlay on the pro mille issue I commented on 

[r0d3z1]

Dave

Wouldn't those units be (Buck)   "DC-DC converters" instead of calling  them   switch  mode  power supplies ?

indeed I initially thought it was an AC/DC

[maginnovision]

I'd love for you to get a scope of theirs to look at.

I still don't know why individuals would want these. If you have a power supply that can feed this, assuming it's not a battery, then why would you not just use that as your supply. If you are using batteries as the supply I can see where these might be more useful as long as you don't mind the ripple.

[HwAoRrDk]

Anyone notice that the microcontroller is marked "STM32F" on the first line of text? This appears to be a telltale that they're not genuine ST devices, and probably a relabelled CKS32 clone. At the mentioned pricing, would not be too surprising - anything to save some pennies.

See the thread on fake STM32 chips in the Microcontrollers section of the forum:
https://www.eevblog.com/forum/microcontrollers/cheap-bluepill-very-likely-it-has-fake-stm32-right/

[filssavi]

The only way that unit is dissipating 160W is very briefly and while on fire, the heatsink while it seems to have a fair mass of aluminium has way to little fin surface, just as a wild guess I would say 50W dissipation is already quite optimistic with that puny fan (now with a proper delta 15krpm fan that might be different). Also since most of the heat will be dissipated in the single switching transistor, only part of the heatsink will be effective at cooling (even if low thermal resistance is a thing in metals too) cutting down the surface area even further

To put it into perspective a mid to high end CPU cooler will be capable of dissipating 160W from a single point source, however their design is much different, they have a ton more surface area and use heatpipes to evenly distribute the heat to all the fins

[maginnovision]

The only way that unit is dissipating 160W is very briefly and while on fire, the heatsink while it seems to have a fair mass of aluminium has way to little fin surface, just as a wild guess I would say 50W dissipation is already quite optimistic with that puny fan (now with a proper delta 15krpm fan that might be different). Also since most of the heat will be dissipated in the single switching transistor, only part of the heatsink will be effective at cooling (even if low thermal resistance is a thing in metals too) cutting down the surface area even further

To put it into perspective a mid to high end CPU cooler will be capable of dissipating 160W from a single point source, however their design is much different, they have a ton more surface area and use heatpipes to evenly distribute the heat to all the fins

Dave was definitely being nice. Like Dave though I'd be hard pressed to be able to sink(or source) the kind of current that'd be required to actually test it. My biggest power supply is 800W and if I put all of my supplies together, due to a voltage limit on a bunch of them, I'd only be able to supply roughly 55A @ 32V which is only 1760W on the input side. Probably enough to catch it on fire if I found a decent load though which would be the second problem although probably easier to solve. I guess the easiest supply side solution would be a decent 48V(or 2 in series) battery but what about a load? The simplest solution would probably be a decent inverter and just run some AC stuff with it.

[D. Head]

What do you mean by "NOTE the PSU integrated!"?
Isn't that allway the case, for scopes that the PSU is build in the device (execpt for the Tektronix TPS).

[firewalker]

The 5 volts on the communication port is probably for the optocouplers for the isolation.

So is it JUN-TEK or JUNK-TEK?

Alexander.

[Kleinstein]

For the worst case power loss one does not need the maximum power. The worst sould be somewhere with maximum current, but only some 50-70% of the maximum output voltage.

I would consider some water heater / cocking stove / electric grill (no electronic) / toaster / iron as a load. Some radiators also work. One may have to put a few in parallel.

It looks like the 5 V are together with the com port - maybe an interface to a remote control unit. This is actually not a bad idea, maybe just missing documentation.

[T3sl4co1l]

I wonder if they're even current mode controlled.

Wonder what the common mode (input to output) noise is.  As bad as the output, or worse?

Wonder how long they last at full power.  Guarantee it's not continuous, but it might be usable at a hefty derating.

Guessing the TO-220 at the end of the heatsink is the temp sensor.  Didn't get a good look at it.  Rather silly they didn't bolt it down, eh?

Oddly enough, I made a module with similar ratings last year; custom job for a custom product.  A bit higher max input voltage, two phase interleaved, and much smaller (and more efficient).  Digital control too (but not from an MCU, just an ADC/DAC onboard).  If anyone's interested I can provide contact info; the client might be interested in selling standalone modules.

Tim

[langwadt]

is that a fix for a ripped off track at ~16:44 ?

[Psi]

I reckon the chip marking removal was a service provided by the PCBA factory.
I think they put the boards back into the depanelization router stage after reflow and created a tool path to remove the tops of the chips. Looks like a goto x,y and plunge Z mm down/up then move to next x/y.

A PCBA factory already has all the tooling and equipment for doing that.
Also the size of the markings on the ICs matches the size of router bit you might find in a depanelization router machine.

Would also explain why it's only the SMT chips and not the TH ones.

[Nystemy]

I myself wouldn't refer to these as power supplies. Since that term tends to indicate that it runs off mains voltages.

Though, regardless of the price of these units, they clearly lack adequate output filtering.
Whacking in a pair of large caps do cut down on the ripple a bit, but doesn't really do much against the noise.

Though, the 3KW unit has two large inductors in it, two transistors, and two diode packages.
And the traces seems to indicate that it uses a dual phase converter in it. This would dramatically cut down on the ripple, and lessen the EMI just a bit.
Not to mention also make the power supply a little bit more responsive towards transients, not to mention increase efficiency at the higher end just a bit. (if they run 180 degrees out of phase with each other.)

Though, a good lab power supply should have a pass transistor for the output. A linear regulator one might say.
Then we can filter out all the noise much more efficiently before passing it through the transistor to drop another few mV before sending it out to the load. Though, this obviously adds to the cost of the device.

So, next time, can we see some actual power supplies?

[filssavi]

Though, a good lab power supply should have a pass transistor for the output. A linear regulator one might say.
Then we can filter out all the noise much more efficiently before passing it through the transistor to drop another few mV before sending it out to the load. Though, this obviously adds to the cost of the device.

So, next time, can we see some actual power supplies?

When we are talking regular signal electronics you can’t have a linear supply, however when the power is high unless you have a very specific need, and you are willing to pay dozens of k$ (I have seen a 30 kW class A linear power amplifier/function generator/four quadrant power supply)

Moreover Switching mode power supplies can be as low noise as linear supplies, when designed correctly, however you will not find that in a el cheapo unit from AliExpress

[T3sl4co1l]

A lowpass filter will be cheaper, more efficient and more effective than a linear pass regulator.

I don't get why people fantasize about those things...

Tim

[Nystemy]

A linear pass transistor on the actual output allows us to correctly go into current limit without worrying about output capacitance to the same degree.
It can also help cutting off the last bit of noise, but that is honestly a secondary concern most of the time.

A low pass filter most often also brings along a bunch of output capacitance. Depending on the load, this might not matter, but in cases where it does matter, the transistor is the better solution.

[Nystemy]

A weren't talking about linear supplies, nor class A amplifiers.

I just stated that you can have a transistor on the output to do final regulation. Ie, cut out both noise and ripple, and ensure proper constant current behavior.
We can still have a switch mode solution providing the actual power.

And the transistor is a superior solution when it comes to constant current applications.
But if voltage is all one needs, a good old low pass filter is usually sufficient.

[T3sl4co1l]

I should've written "post regulator" to be more specific.

Yes, fast response current limiting is one useful feature.

Even with "capacitor multiplier" style circuits, it's difficult to get extreme attenuation, and impossible at high frequencies where device gain is low and parasitics are dominant.  Common mode noise is completely unaddressed.  Whereas a passive filter, just keep tacking on stages and shielding until it's done.  (Yes, you can stack linear stages as well, but that degrades efficiency proportionally.)

A couple years ago I made a low voltage, multiple output module with <= 0.1mV output noise, from 1MHz to well over 100MHz (noise <1MHz was specified differently).  Easy enough with a modest switching frequency, a few stages of filtering, and good layout.  Single board, no shields needed.  Probably, to do better, shielding would be mandatory.

Linear circuits can still be attractive around extreme corner cases, like where a passive filter would be too bulky, the efficiency hit is unimportant, and attenuation at relatively low frequencies is required.  There are also active power filter circuits, mostly for low frequencies because again, device/loop gain is degraded at high frequency.

Tim

[Nystemy]

Yes, low pass filtering should still be there. Passive filters are usually superior at higher frequencies, especially if we want to handle a fair bit of power.
The pass transistor would mostly be there on the final output, it could handle the low frequency filtering like taking out ripple.
The higher frequencies will most likely be out of reach for it.

Though, I wouldn't drive the transistor as a capacitance multiplier, that seems honestly a bit crude.
Since then we wouldn't be able to easily control the output current.

Another advantage of using a pass transistor in this fashion is that we can more quickly react to load changes if we have remote voltage sensing. Since we can characterize the resistance over the power cables, and take that into consideration. Ensuring that the voltage dropped by the pass transistor under no load is roughly equal to the voltage drop of the cables under full load. Meaning that we don't need to rapidly force the low pass filtering and bulk capacitance supplying power to the pass transistor to shift up in voltage to handle the higher load.

Ie, a lab bench power supply just becomes better with a pass transistor on the output. Though we still will need actual filtering before it. The transistor is there to regulate and filter out the ripple. So we don't need large passive filters for the low frequency stuff.

[Muny]

I posted this as a comment on the YouTube video, but I doubt it'll be seen; copying here:

The weirdness you were having seems like it could be from the voltage drop across your input wires. You have your main supply set to 20V, but the voltage that ends up at the input is likely less than the minimum rated input voltage of 20V. And I doubt it's very stable near the low end of its probably-immodest input rating. How about re-doing some of your tests, but feeding it say 30V or more from your supply?