Recent Posts

Pages: 1 2 [3] 4 5 ... 10
Trying to make a 5A diode from two 3A ones is problematic.  I have a small welder with a number of small leaded diodes in parallel that doesn't look like it would last, hasn't failed yet. It is always a race  between higher temp lowering the forward voltage and attracting more current and internal resistance increasing forward voltage with more current. Definitely you need to keep them in close thermal contact.

The world has gone to small inductors and higher frequencies for size and weight savings. If you have the space larger inductors still work and lower frequencies make FET driving easy.  The micro was actually 30 feet away from this board. This used to be mounted on the wall next to the bedroom.  I would wake up to the drone of power. Now it is in a shed and adding the audio effect with changes in power just adds to the mood.

It is wired up a little better now.  That board is solid cherry.
Steve / Re: My Scratch Pad
« Last post by MadScientist267 on May 21, 2017, 03:48:25 AM »
Haha All good questions... I'll do my best...

First, the older doubler method results in a no load voltage of about 320V across the caps, which dips quite a bit under heavy load, just as any unregulated supply does, only worse because of the "half wave nature" of a doubler fed by sine.

In active PFC, the 380V point however is a set point (regulated), and is typically chosen to be above the older doubler scheme, but presumably only by the necessary amount to accommodate the nature of boost converters, as this would make the PFC more or less a "drop in" replacement for the doublers in tried and true designs (which is exactly what happened in this case). If someone knows something to the contrary, by all means, as that's just speculation on my end.

The exact voltage isn't critical when it's used in this way, provided the MOSFETs used both for the boost converter and the main converter that follows are adequately rated/snubbed. It's usually in the 360-380 range (the charger actually reads something like 376V or so, sagging only slightly under load, down to around 372-374). Many times, the reservoir caps used (electrolytic) are rated for 400WV, so it will of course be below this.

The margin above 320 is to accommodate the 240V mains countries, without needing to manually select the voltage with a physical switch like on the old supplies. A boost circuit is not capable of producing a regulated output that is a lower voltage than it's input, because the DC path from input to output has no switching components in series in between (only an inductor separates them). To accomplish high PFC, it must always be regulating, so the target voltage is set higher than the peak AC voltage would normally reach (240 * 1.404 = 336.96V). Why is it called "320"? Nominal? Anyone? LOL Anyhow...

The boost converter is set up slightly differently for PFC than for typical DC usage, but for the most part, this is mostly at the cap used at the input. Instead of a relatively large electrolytic or tantalum, usually a much smaller value metal film cap of some variety is found in its place, immediately following the mains rectifier bridge. This is because it needs to be able to track the 100/120Hz rectified sine rather closely for the PFC to reach as close to unity PF as possible. Ideally, for phase distortion considerations (which counteract the active PFC with capacitance on an otherwise "resistive" load), it wouldn't be there at all, but the boost topology needs it to work correctly.

To accomplish the PFC, the boost converter more or less functions just like it's DC counterparts that you're (hopefully) already familiar with. It targets an output voltage, varying it's PWM as required to maintain regulation, as the input voltage rises and falls over the rectified sine, and load on the output changes. The overall pattern this results in is near maximum duty at the beginning and end of each sine pulse, with the lowest duty being seen at the crests of the waves. Load at the output causes the overall duty percentages to increase, with most of the difference being seen toward the crests.

The result is the entire wave produces output power, and not just the crests as in the doubler scheme. The loading effects this has, as I'll get to next... In a nutshell, the upstream source doesn't experience the disturbances that cause distortion of the sine.

Why is this important? Well, the problems that a simple doubler scheme cause are a little bit different in part from the voltage/current phase issues seen by things like directly mains powered transformers and motors... the main difference being that it introduces wild harmonics back into the source, that can't be controlled and compensated for by legacy L/C correction schemes. This is largely a "grid phenomenon" (meaning it is easier to "overlook" on smaller scales such as local generator feeds and the like). As SMPS gained popularity and presence on the grid, the distortions increased, and began posing problems to the infrastructure that again, passive PFC (within the grid) could not correct.

The other issue it causes, while different in originating nature, has the same end result - increased need for peak current handling. With SMPS (or countless, dozens at least) in every house on the planet pulling hard peaks at the crest of each wave, the impact on the grid is non-trivial. This is the more "visible" issue, in that the effects scale from single unit to the millions, with an even greater effect than the harmonics...

Let's take an SMPS that draws 1000W (for round number purposes) and ignore efficiency considerations for this example to keep the math cleaner, as it's a completely different concept.

It only *needs* 8.3A at 120V in doubler mode (or 4.15A at 240V in full bridge mode), but can, at the crest of the waves, draw current on the order of multiples of those during the peaks. Double, triple, quadruple even, and up. So let's play right down the middle and say the current spikes are hitting 3x average power used by the load... and rile the chip up... 1kW *average* power going in, 1kW coming out... but that's not what the upstream source needs to be capable of; it's seeing the need to provide a 3kW spike, even if only for a brief moment every cycle.

It is this spike of current multiplied by the instantaneous voltage within the "window" at the crest that becomes your "apparent power", and the upstream source, whether it be grid, generator,  or inverter, must be able to provide it while it's being drawn in order for the SMPS to feed it's load.

With PFC, the entire wave is supplying current to meet the actual power requirements of the load on the other side of the SMPS, the effective result is that the upstream source sees a mimicked resistive load, which has a perfect theoretical power factor (translates over to 100% "real power").

The harmonic distortion and behavior of the current spike are interdependent, in that the harmonic distortions induced in the upstream source are a result of the impedance of this source, and the distortions lead to further complexities that manifest as further disturbances in the current spikes, that result from the crests of the waves changing position, amplitude, and duration.

In the case of a smaller scale system, while the distortions still exist, the "round robin impact" is somewhat subdued, as the peak current demands can become the dominant problematic issue (by tripping fast acting overload protection schemes such as hiccup) before the distortions can cause problems, but not always. This is largely dependent on what types of other loads are also present within the system, and the impedance of the source. Generally speaking however, the main point is that "the grid will just keep on giving until something pops, and they'll just bill you for it, your generator and inverter will not".

(Ok, so generators and inverters can pop too, but that kind of popping is a massive PITA to "reset", and can come with a hefty "bill" of its own) :o

Hopefully, I've answered your questions with this, and didn't just add more confusion lol...

Haha nice indeed.

For years, my sole source of parts for everything I home brewed was from donor boards... then a brief window where there was some new part purchases going on, but the last few years, I've been right back at it again... PC power supplies, Car audio amps, UPSs, cheap MSW inverters, among a few others are absolute gold mines for experimentation parts... and as you pointed out with the UPS/battery thing, they have not only typical but also usually fairly obvious failures, so you almost know what you've got in hand before you ever even remove the first screw to take a closer look ;)

One thing... on paralleling diodes... it's actually very common in the SMPS world, particularly the TO220/247 Schottky "KAK packs" found in higher current lower voltage supplies (such as PC supplies)... The main thing is to balance each side with the traces they're connected to as closely as possible... and while that doesn't account for the minor internal differences, it's not generally a problem if they're properly sunk and not pushed all the way to the datasheet under warmer conditions. They're pretty tough little critters when they're done right.

Good post, glad to see I'm not the only one with projects like this that somehow just kinda stick to the wall/bench/table/ceiling haha... Efficiency is always a good thing, but there's a lot to be said for things that "just work" ;)

Steve / Re: My Scratch Pad
« Last post by lighthunter on May 20, 2017, 09:52:34 PM »
Ok now you have me curious about pfc. I've known about power factor for years but back when i learned what it was it referred to current out of phase with voltage causing something called apparant power and real power. A perfect power factor occurs with a resistive load where the voltage and current in load are in phase. A shift in the direction of capacitive or inductive involves apparant power such as connecting a capacitor directly across ac line. No real power is used, it is all apparant power therefore the power company will charge you for electricity used when you get no benefit. I can understand that. Now when it comes to an smps such as you have built...

Typically the 380v dc is arrived at by mains connected rectifier charging a bank of capacitors later used by a high frequency primary driver. So instead of mains current being sinusoidal over the full source voltage waveform, the mains current is more like a short duration pulse which the power company can charge you for the full duration and you are only using 50% of that.

So i get it that its a power factor problem but not inductive or capacitive ????   

Also how would you correct it ???

And why would you care??? Does it in reality waste power or does it just cause inaccurate billing? And dont the new smart meters compensate?

Just crazy thoughts, not necessary to ans all of it. Thanks in advance!

Very nice! Obviously has stood the reliability test. Thanks for sharing that!
I was just reading on another thread about a 50V buck converter down to a 12V battery.
At camp a couple years ago I changed to a 36V array that gave about 50V at power point.
That higher voltage was powering some other stuff and still had lower voltage panels that
generally supplied most of the battery charging. The higher volt array was to help charge
the battery only on those desperate days with extreme overcast. I have resources at home,
but at camp only stuff I can scrounge at the town recycling center.  Fortunately they get
a lot of small UPS with dead batteries.  A lot of poor planning. Still, I manage to come
up with something out the garage when the need arises.

Should add that I think all electronics should be nearly free. There is a cornucopia of
consumer electronics with parts and sub assemblies that can be re purposed. While capable
of building some really nice electronics, If patient, nice commercial stuff can be had for
next to nothing that is not working. I just got a nice current model XANTREX 30V 10A power
supply for $46 shipped. These cost over $300 used. An hour later I had it working. Finally,
solar is becoming mainstream and broken converters will start flooding the market. Till then
I'm OK with hanging some ugly stuff on the wall for a while. It is a kick creating some of
this stuff out of the most improbable parts to show that a lot of money doesn't have to be
spent to have a working system.

Pictured is your textbook basic buck converter. This one only had to put out about 10A from
a vastly overpowered array. I have seen it put out 23A. The maximum it is now allowed to
produce is reduced by limiting the maximum PWM duty cycle. The FET came from an old UPS.
Three were riveted on the heat sink so I used three.  The high side driver is just an opto
isolator driving the gate with a 870 ohm resistor to the source. At these slow speeds the
optos work just fine.  Power for this high side driver is from a 12V wall wart. These happen
to be perfectly happy operating at reduced current with only 50V DC. The capacitor bank is a
bunch of 200V electrolytics out of PC power supplies.

I wanted a high current Schottky diode for the flyback.  PC supply ones are only good for
35V.  I did have a bunch of SB5100 diodes. These are 5A 100V and were like $3 shipped for
20. They say you shouldn't parallel diodes. You can if you are careful. I found some tin
plated steel used the shield a VCR. Two four inch squares were cut out for heat sinks. That
metal was so thin I only uses a sissors to cut it.  I think I used nine diodes with the leads
soldered close to the body on the heat sink. Leads were not cut so I could use them for other
projects at a later date. So that was my cheap $1.50 diode. A small RC snubber was placed across
them to catch any spikes.

The Inductor....

Suitable inductors are not easy to find in consumer electronics and spending money for one
is not in my vocabulary.  I rewound a generator when I was 17 and have been traumatized by
that ever since. Had a good size (400W) transformer out of a UPS.  The 24V winding looked like
it would work.  I know, it is going to saturate and should have an air gap on the iron. These
were desperate times so I used it. It worked pretty well on 480Hz. The iron core did heat up
as predicted, the windings stayed cool.  I rewrote the program to change the PWM port so I
could drop the frequency to 122Hz. That greatly reduced the iron core heating. While likely
not the best efficiency, nothing was heating up much. It has been running a couple years now.
A buck converter for next to nothing.  Just don't show it to people.

Steve / Re: My Scratch Pad
« Last post by MadScientist267 on May 20, 2017, 05:01:58 PM »
Hehe, yep that'll do it lol
then i think we can all agree that if we had a device called a "neutral grounding resistor monitor" connected up, which would then trip the breaker if there was a connection between earth and neutral while leaving 'neutral' floating and not bonded to earth, we would have the safest system going?

I believe that takes this full circle. Here, we call that a GFI/GFCI/RCD/PITA  ;D

I encountered this issue with the truck and it's umbilical cord connection, and wrestled back and forth with what was the best way to provision for every scenario I could think of... the answer? There wasn't one. Not that covered them all at least.

In the end, I went with looking at the truck as an "appliance" that has most of its threat potential inside of it, and gave it its own internal N-E bond, complete with a set of downstream GFIs. Once someone is inside, there's no more threat of getting zapped than in a normal everyday dwelling wired for AC.

The caveat to this is that because of the tires, it is for all intents and purposes inherently isolated from earth, as the "single appliance". The problem then became how to deal with bonding the frame to my upstream source's ground to keep it at earth potential, while keeping an upstream GFI happy under no-fault conditions. Put simply, it wasn't possible to do *both* internal and external safety protection to my satisfaction at the same time.

If the upstream connection is to a non-GFI protected source, the earth bond is more important than ever, because if I'm parked on grass or the asphalt is wet, there's a much higher chance that anyone with one foot on the ground and the other on the rear bumper will be part of a lower resistance path at both, than if it is parked on pavement and everything is dry.

Having the internal N-E bond however, this means that the normal current return path would share with the earth lead.

If there *IS* an upstream GFI (a scenario I only envisioned as likely to increase in probability of encountering), this results in an immediate trip of the GFI with nothing more than the *capacitive reactance alone* of the internal wiring connected as a "load".

As a side note, this was a bit baffling until I realized (and subsequently confirmed) that this was what was happening... the fluke was showing completely open resistance measurements, but that is done with DC. AC had snuck parasitic capacitance in under my nose. This however is ultimately moot, because even without it, the moment a real load gets applied, the same result happens.

What I ended up doing was placing a "jumper" in that could leave the truck frame and umbilical disconnect/breaker box bonded to the umbilical's earth, still leaving the internal bond in place to provide the protection on the inside. This of course then ultimately means that an open (or other) neutral fault between the truck frame and upstream source's N-E bond relies on the upstream GFI to prevent shock hazard conditions. Again, I weighed this against the idea that it's then essentially the equivalent of grabbing a toaster chassis in an up-to-code kitchen in terms of threat. GFIs fail too, after all, but that failure would need to be combined with the open neutral or other fault to become a problem.

Whether or not to have the jumper in place of course requires knowledge of the upstream system I'm connecting to... which leads me to a final point...

The thing I keep seeing here, and it's either overlooked, ignored, or rationalized away, is something that was ingrained in me early on: "Treat every wire as if it can kill you". You can rationalize and argue against the idea that its possible all day long, but get into an argument with physics, and (I really hope) I don't need to justify my reasoning for betting on my choice as to which one will win, right?

As for isolation transformers: Under the wrong circumstances, they *will NOT* save your neck. They're designed in general to isolate power and ground leakage to levels that *equipment* can safely tolerate, not people. Don't fool yourself into thinking otherwise. If you don't believe me, connect one with only the hot side connected to the appropriate primary terminal, and leave the neutral and ground disconnected... then check how much AC current is available between one of the secondary leads and earth...

You no longer have a transformer... you have a *capacitor*, a non-trivial one at that, which passes AC rather freely. Still think it'll save you?

</rant> </2cents>


Yes I agree it would be very safe, however I believe such a system would be unworkable for general distribution for at least 3 reasons.

1. Cost - the monitoring devices are expensive, at least I'm not aware of any easily available inexpensive ones.

2. Lack of discrimination - A fault anywhere on the system has to bring down the entire supply. So your immersion heater for example could go faulty and that leaves you with no power until you track down the fault (which would take a while) and repair it. The more extensive your distribution the more complex this becomes.

3. False alarms - In order to be useful the system would be need to be set at a quite sensitive setpoint. Since you have to monitor the entire system as one, a long term deterioration in one or many devices could combine to lower the resistance to earth close enough that it could trip unnecessarily during periods of high humidity for example.

Hope that makes sense!

Steve / Re: My Scratch Pad
« Last post by lighthunter on May 20, 2017, 04:19:41 PM »
You are right! i didnt think there was enough inductance in the wiring to make a difference. From the 1 minute i spent trying to come up with inductance of 200 feet of parallel 12awg conductor i came up with 460uH. While that may not be exact, i put it in circuit simulator and came up with these two screenshots. With/without snubber.
Pages: 1 2 [3] 4 5 ... 10