Commanda's design

[Opera House]

Madscientist,
Been gone for a few days but sensed my ears were burning.  Well that's a relief that I need not provide any further information.  My comments in the post were directed towards you only.
 
First. It really seems odd that you would design to protect the output caps from voltage but not set a maximum voltage the battery would see. Limiting drive can cause potential failure. I understand your reasoning but not the reluctance to attempt my suggestion.  I have admired your willingness to try out of the box.  Looking at things a new way is always good.  I spend 80% of my time designing for those rare 1/2 % conditions a circuit will see.

Second. 12%... I've looked at several data sheets for the 34063 and can' find any reference to minimum duty cycle. Frankly that seems way to high for any designed in function.  Data sheets have gotten a little sparse lately and often there are often things that happen which are not documented.  I was just working on a LAMBDA power supply that had ashut down 2845 regulator by clamping down the comparator feedback pin.  Nothing unusual there but that caused the chip VCC to drop to 8V (8.6V lockout) and the 5V ref to drop to 1V.  No direct literature reference indicating that would happen. But I ramble.  There is just no reason that you can't cleanly shut down the 34063 regulator chip and not have it over voltage. Please provide your reference source for this 12%.

Happen to have a number of 34063 boards from a product we used to make. Set one of them up on the bench nd tested it with no load.  Pulling the regulated output high by just a couple mv completely shuts off the pulse stream.  So the voltage snse pin can be used to cleanly control the input and output voltage

Third.  You remind me of someone I used to work with.  Every time I walked by his desk he would hide his schematics.  He was always afraid to let someone see his designs.  His stuff always worked but used a lot of parts. A year later he was found redundant. You got a camera.  Make a sketch so we can all figure out what you are doing.  This is your chance to learn new things. Don't be shy.

Fourth.  90-95%.....We all like to be optimistic, but the chances of a neophyte obtaining that with randomly selected components just isn't possible.  More than likely you are seeing a metering error or you are selecting best numbers.  Few meters can handle any kind of pulses accurately.  I remember these guys from an Air Force calibration group when a million dollar jet engine rebuild failed performance tests.  They were called in to "recalibrate" the instruments.  They would pick a high number here and low number there.  Suddenly the engine passed.  Fool who you want, but you can't fool me.  I've been in this business too long.  You are not likely to obtain those numbers, with that circuit, at those voltages.

I started building switching regulators in the 70's.  My comments on the board tend to be a little cryptic.  I don't have the patients any more to write long articles.   I prefer others to build these projects and give small gleanings of advice on how to improve them.  I have 40 years of design work behind me.  It's pretty unlikely someone will stump me here.  Good design is not a popularity contest.

Comanda,
I'm sure you were just having a bad day and with further thought would not have commented on minimum duty cycle.  It was irrelevant and only served to reinforce Madscientist's misunderstanding of the chips functioning.  This board needs both of you champion these solutions.

Never got around to adding a tag line to the signature.  Elsewhere I have used a line from Fog of War.  Seeing and believing are often both wrong.

[madlabs]

Commanda, the CNC uses Gerber files. Or rather, there is a bit of software that converts Gerber files to what is called G-code. The G-code is what actually tellls the CNC what to do.

What current level do you think the design could be pushed to? My luck with vero/proto went sideways after 5 amps or so. However, I could have been screwing all sorts of things up, it didn't have to be the vero board per se.

Mad, those all sound like things to keep and eye on/ work towards. I will test on a small panel I have kicking around before I try it on something bigger.

I am so hopeful I may be able to make a DIY MPPT!

Jonathan

[commanda]

Opera,

You must remember, we are running this chip in reverse logic. When the chip is off, our fet is on. And vice versa. To compare with your pre-made circuit, see if you can get 100% ON time.  It's a function of the assymetrical charge & discharge currents on the timing cap.  The typical values, from the datasheet, are 33uA and 200uA. The chip does a forced-off during that 200uA discharge. So the minimum off time (from the chips perspective) is 33/(200+33) = 14.16%.

Having said all that, now let me contradict myself. In the simulator, I put a photocoupler in series with a zener across the output. The output of the photocoupler pulls the feedback pin low via a resistor. This drops the output current to zero, even though the fet drive doesn't drop to zero.
I'll try & get this in the next iteration of the pcb.



Blue trace is the battery voltage (sine wave with a dc offset).
Green trace is the current into the battery.
Red trace is the fet gate drive.

I've also been running your suggestion in the simulator, using a P-ch high side switch and the same totem pole driver. Managed to make that work as well.

It is possible to guesstimate efficiency as good or not-good, based on temperature rise, although the finger test is notoriously unreliable. I personally use an IR thermometer.

I agree about datasheets. There are often obscure side-effects from the way we "think" the chip should work.  Really need to build a prototype. Been caught out a few times myself with this.

I'm learning stuff all the time, and having a ball with LTspice.

Amanda

[Madscientist267]

It really seems odd that you would design to protect the output caps from voltage but not set a maximum voltage the battery would see

You must understand that there IS no way for the chip to do BOTH.

Because of the way the output gets inverted, the minimum duty it will provide the MOSFET with is (a confirmed) 12%. In the datasheet, you will find this listed as MAXIMUM duty cycle. The OUTPUT IS INVERTED.

Protecting the caps is TRIVIAL because there IS NO LOAD when it goes into shutdown. There is no negative effect of this protection, whatsoever. The instant a load is connected, it immediately falls out of range for the shutdown to stay engaged, and goes FULL DRIVE. NO in between. It is either ALL THE WAY ON, or ALL THE WAY OFF.

Set one of them up on the bench nd tested it with no load.  Pulling the regulated output high by just a couple mv completely shuts off the pulse stream.

Again, clearly you do not understand how the circuit works. IT IS NOT TYPICAL USE CASE FOR THIS CHIP. The datasheet and any implementation you may have them in means NOTHING.

Make a sketch so we can all figure out what you are doing.

I have nothing to hide. THERE IS NO SCHEMATIC other than the one I received from Amanda. I don't design off of paper, I play with breadboard until I get close enough to put it in a more permanent arrangement. Sometimes this is messy and its too 'beta' to even worry about a schematic. If you want to see what I built this around (more or less), see page one of this thread.

I'm not intentionally 'depriving' anyone on here of anything. It's how I work through the creative process, and unless its insanely complicated, I don't start with a schemo. This is particularly true when I have one handed to me. You already HAVE the schematic.

90-95%.....We all like to be optimistic, but the chances of a neophyte obtaining that with randomly selected components just isn't possible.

I'm sick of arguing this point. It works very well, and is very efficient. Heat (or lack of) doesn't lie. And neophyte? hahaha That's pretty funny. Very inaccurate, but funny.

I started building switching regulators in the 70's

Ok LAST TIME:    IT IS NOT A SWITCHING REGULATOR!!! IT DOES NOT CARE WHAT THE VOLTAGE AT THE BATTERY IS! THAT IS NOT ITS JOB! THIS IS INTENTIONAL!!!

The ONLY thing it 'regulates' is the load placed on the panels in an effort to keep the voltage as close to the Pmax power curve peak! Beside that, the only thing it gives two $#!+s about at the output is protecting itself if the load gets DISCONNECTED!

It was irrelevant and only served to reinforce Madscientist's misunderstanding

Don't you mean "reinforce Opera House's misunderstanding? It's clear you don't have a clue what this circuit DOES, nevermind how it WORKS!

READ the SCHEMATIC. THEN post.

Steve

[Madscientist267]

I've scoured the datasheets (I have two different ones) for the duty cycle numbers, and I concede they do not exist.

Try this:

Grab a scope, take one of your ready-made working-to-your-satisfaction MC34063 based everyday bucks, and load the crap out of it (or twist up the reference however you wish with a pot or something), and watch the waveform at pins 1 or 2 (depending on it's configuration).

If you have a meter that does duty cycle, hit it with that as well. The highest you will see is ~85 - 90% duty.

But because of how we are using the chip, this translates to a MINIMUM duty of 10-15% instead. Mine happens to be 12.

Sorry to spring so far out of shape about this, but you're arguing points that prompt us to keep throwing back sound reasoning to you, but you're not hearing us. It gets old repeating yourself.

I have no issue with someone that doesn't get it and wants to understand what they're seeing, but flinging blind 'experience' and calling people 'neophytes' isn't how you get anything other than what you've been getting. I don't discount your knowledge of what you've done, I ask the same in return. Ok, so yours is 40 years, mine is 25. We do things differently, but that doesn't warrant all the BS that has come to the table.

Apparently Amanda has a little more patience when it comes to trying to keep things politically correct. You caught me earlier after a 3 hour drive and coming home to all hell breaking loose. Not that you could have known, but it wasn't the best time for the straw to break a camel's back.

So for the 'yelling', I apologize. Please just do the homework, and forget what you may have already seen this chip used for. This design is very different from anything else I've seen either.

Didn't completely grab ahold of everything that was going on either until I went to build the first one.

Steve

[commanda]

So today I wound a new inductor. Same E42 ferrite core. This time 5 strands of 1mm wire. Managed to get 15 or 16 turns on. Set the gap to 0.8 mm to give 50uH using the inductance meter. Fixed the gap using a cut piece of Press Pahn (elephantide). This is the cardboard stuff you sometimes see used to isolate mains wiring.

Calculated resistance should be 6 milliohms. Previous coil with one winding of 18 turns of 1.2 mm should have been a calculated 18 milliohms.

No sunshine today to really test it.

Also tested the opto coupler feedback idea to limit output voltage. Basically seems to work. Just need a higher gain opto. And because it connects to the feedback node, flying leads are a really bad idea.  I know I said I used a zener in the simulator, but the real one uses a TL431 precision shunt regulator.

On another note, I've sorted out my charge controller. Because of differences between the solar application and my vehicle applications, I realised I didn't need to reference it to the negative input at all. So now it's powered straight off the battery.

Amanda

[Madscientist267]

MOSFET locked up today (or maybe yesterday...?), had to replace it.

Not sure of the cause, but I suspect heat. The unit is completely enclosed with no vents.

Also, my early testing without a heatsink may have weakened it or something.

I'm going to run it without the cover on for a while, but supposedly, tomorrow is set up for rain. 

Stay tuned for updates.

Steve

[commanda]

Die cast aluminium enclosure. TO-220 devices laid flat on pcb. Aluminium angle sandwiched between  pcb and devices. The angle piece of said aluminium angle then bolts to inside of enclosure.

Amanda.

[commanda]

For those that are interested, here is the schematic, pretty much "as built" by myself.



Q1/Q2 are BC327/337, although BC547/557 should work just as well.
Q3, N-ch mosfet. I used an IRL3705.
Q4, pre-regulator, optional. I used a TIP31.
D1, I used an MBR2535.
D2, optional (stops leakage back through the panel at night). I used an LTC3457 ideal diode.
R17 is apparently redundant, but is required for the add-on mppt controller, assuming I ever get it working to my satisfaction.
P1 should give enough adjustment range for 12 or 24 volt panels.
L1 is 50uH, as I described earlier.  Adjust value to suit your exact application.

U3 can be 5 volt or 12 volt, depending on the gate voltage requirements of the main fet, Q3.
5 volts is also required by the current sensor I'm using on the output, for the mppt. So I kill 2 birds with one stone here.
E3 supplies the gate current at switch-on of the main fet. This is essential.
Please study the datasheet for the MC34063. Note that its output transistor is used to switch the fet off. So it operates in reverse to what you would normally expect.

All it does is maintain the input voltage constant.

As usual, free for personal use. Commercial rights reserved.

Looking at the circuit like this, it's almost too simple.

I hope the image comes out alright. I didn't want to scale it down any further, lest it become illegible.

Amanda.

EDIT: as Steve pointed out earlier, the input and output filter caps probably aren't big enough.

[DamonHD]

Is U3 itself a small switching reg for efficiency (I can't quite read the part number)?

Rgds

Damon

[commanda]

Is U3 itself a small switching reg for efficiency (I can't quite read the part number)?

Linear regulator.  78L05. TO92 low power version of a 7805.  You can of course use a TO220 package if you so wish. Especially if it's already in the junkbox.  Power dissipation is not really an issue. Average current consumption is only a few milliamps.

Only use a 5 volt regulator if your main fet is a logic level type.

Amanda

[Madscientist267]

EDIT: as Steve pointed out earlier, the input and output filter caps probably aren't big enough.

Yup. And I'm thinking that even MINE doesn't have enough yet.

I am considering a bank of 8 or 10 1000uF caps with two 10000uF caps at the very beginnings of the output for the next rendition of this.

It takes no length at all in the output cables to dissipate what you've gained by the buck. Unless you're going to just about mount the buck right on top of your battery (or main terminal buss), you need the caps.

Probably going to stiffen up the input as well. Right now I have a 4700 and 1500uF, with a 1000uF right at the MOSFET, between the tranny and the coil. Not convinced it's enough either.

On mine, I didn't need quite as many caps for input because my input voltage hugs the 36V range, so current is lower there. But on the output, it is imperative. 

Steve
 

[Madscientist267]

Mine blowed up real good the other night...

Went into 'fake panel' test mode and was of course sure of the design, so didn't see any need for the ballasts...

My daughter commented starting with 'WOAH', later mentioning 'but it was pretty'...

Three 12V SLA's in series to hit my roughly target [reverse] voltage and ZAAPP!!! Totalled the connector for starters, then thought I had taken the MOSFET and Schottky with it, but all it ended up mangling was the 12V zener for the preregulator....

Wasn't sure if it took out anything else, so worked my way backwards, and replaced both the Zener and the TIP31.

Think I'm going to throw a reverse polarity protection diode on both the input and output with a fuse for each one.

Added a secondary coil to the toroid this time around to drive a fan during heavy load. Doesn't seem to steal much power, but keeps things running a bit cooler.

Also had to add a heatsink to the Schottky (probably should have anyway); it was getting quite a bit warmer than the MOSFET, which I already had to replace once due to an overheating failure.

Added a 'windowed' pot with a knob on it to easily find the sweet spot in differing conditions. The upper voltage cutoff (where the panels no longer produce power) is adjustable by a trimmer cermet right next to the main pot. Seems to work well.

Also found out last night that it can produce in excess of 10A of output (not sure what the input was), but a 10A fuse tapped into my 8D sure enough let go! Not from solar obviously, but proves it can convert some serious juice.

Pics to come of the newest revision as soon as everything dries, I finish the final touches, and I can get it to pose for the camera...

Steve
 

[commanda]

Mine blowed up real good the other night...

but all it ended up mangling was the 12V zener for the preregulator....

Steve
 

This is why I like the pre-regulator. Cheap sacrificial lamb.

And may I suggest, if you're gonna do the fake panel test, you leave some ballast in series with the source. As you have seen, output current can get very serious, very quickly.

And heating in the flyback diode is why this style of converter isn't used much in high current applications. Once you get to around 10 amps or so, a synchronous converter becomes much more attractive.

I have yet to find a synch controller, or even a driver, that isn't surface mount. And they're mostly designed for computer motherboard applications, so all low voltage.

What I haven't tried, is a winding on the coil to drive the gate of a fet used for the flyback diode. I guess shoot-through, where both fets are on at the same time, may become a problem. Might have to discrete logic my own synch controller.

What I did do though, was pick the best diode I could find, in terms of forward voltage drop.

If you have a diode test on your multimeter, this will show the forward voltage drop at low current. Mine is 570mV at 25 amps, from the datasheet, and 110mV on the diode test.  MBR2535 (I think from memory)

Next thing to optimise is the main fet. You want a low RDSon. I'm using an IRL3705, with 8 milliOhms on resistance.

Next on the list is the inductor. I detailed previously what I'm using now.

But work on this has temporarily come to a halt. I'm tidy'ing up the control panel, and managed to short my dual shunt amplifier to battery positive.
'but it was pretty'...
and it stunk for days.

Never was happy with the performance of the thing anyway, so I'm rebuilding it using some INA213's.  Size of these things, or rather lack of, is just about at the limits of what I can print as a printed circuit board.

Amanda

[DamonHD]

I'm too much of a coward and a cheapskate to toast that much hardware...  It's why I mainly do software!

Rgds

Damon

[Madscientist267]

Damon -

I'm more afraid of scrambling something in a winbloze registry than I am of shorting something with a soldering iron. hahaha

Good thing I use Linux now and like fireworks!  

Amanda -

sacrificial lamb

Insert baaaaaa-d joke here? LMAO

Been looking into this whole sync/inter/poly thing to get my feet wet (yes I killed the main breaker first!) and it really seems rather simple when it's all put together.

From what I can gather, the interleaving is the PITA with that, everything else works as if it were all in parallel (well, ok it is).

No fear, I'm not looking into taking it on quite yet, but having gathered almost all the other basics from this project, I'd say the higher power units would be a breeze, since really it's just more of the same thing. Rather than making one huge coil and driving the crap out of it, drive a bunch of smaller ones in a 'ring' and get a much smoother output to boot!

Guess I should really just tackle a syncro version next, and then from there it would really just be about slaving a bunch of single syncs with a chip designed for the purpose, right?

Steve

[commanda]

Funny, I gave up on windoze and became a hard core Linux user years ago.
Only use Win XP now for running Protel to do schematics & pcbs.
It will run under dosbox, but the mouse synchronisation is broken. At least it was last time I tried.

Going beyond what we have here is not as simple as you may think.
There are plenty of specialist chips out there.  Many are low voltage only for PC motherboard applications.
Many have fixed frequency oscillators, anywhere from 500KHz to several MHz.
And only come in surface mount packages.
To use them, you have to be able to do your own pcb layouts, and make one off boards.
Layout at this level is not trivial. Matrix board is totally out of the question.
I think "fools rush in where angels fear to tread" is apt here.

Amanda

[Tight Yorky]

Hi Steve & Amanda,

I have read through this thread and can see that there has been a lot of effort and time put into your circuits. I, for one, would like to compliment you for publishing the details for us all to see.

I see that you have been investigating the efficiencies of your devices. I note that a blocking diode, D2, has been added to the original schematic. I thought I would make a suggestion that may contribute. And that is to reduce the losses across this D2 component it could be moved to the PV supply leg of the Inductor/D1 node. Although the losses would still be present during the Q3 conducting cycle, the diode would be out of circuit for the flyback part of the cycle with no voltage drop, no power loss. I take it from the thread that D2 has been added due to discharging of the batteries. Are you experiencing this because of the diode built into the MOSFET being forward biased by the batteries when the PV is not generating a voltage?

Hope the above is relevant and helps you progress.
Regards.
Clive.

[commanda]

Clive,

My panels don't have built-in blocking diodes. I did note that D2 was optional.
Also, I used an LTC3457 ideal diode. This has a forward voltage drop of 20mV. But it is surface mount. And very tiny.
Your idea of re-locating D2 sounds fair enough. It would then be more like a built-in blocking diode.

Amanda

[Madscientist267]

Even with blocking diodes in the panels, there is still backfeed from the battery.

Mine is on the order of between 10 and 20mA, which doesn't seem like much, but every little bit counts... 

The levels are pretty much accounted for with the LED and driver chip operating currents.

So far, since my whole system is nowhere near a 'permanent' installation, I just disconnect the battery when I know there is no usable output. This is of course not an ideal solution for someone who wants a 'set it and forget it' approach to using this.

Steve

PS - Amanda - How does the 'ideal' work (in terms of principle)? Is the end effect essentially a MOSFET with current sense for one direction that only turns on the gate when the current flows one way, and off for the other?

[commanda]

PS - Amanda - How does the 'ideal' work (in terms of principle)? Is the end effect essentially a MOSFET with current sense for one direction that only turns on the gate when the current flows one way, and off for the other?

Essentially a mosfet with voltage differential sense. The driver chip turns on the fet to maintain  20mV from input to output.
You could always Google the part number and find the datasheet. There was a guy on ES that was selling made up units.

Amanda

[boB]

Even with blocking diodes in the panels, there is still backfeed from the battery.

What PV panels have actual blocking diodes ??   Do you maybe mean, bypass diodes ?

All blocking diodes I know of are added externally.

boB

[Madscientist267]

Well, ok. They're technically 'external', but they're usually within the junction box on the back of a commercial panel.

Typically a single Schottky of the 3A-ish variety.

Bypass, blocking, tomato, tomato.

That doesn't work as well with text... 

The diode that prevents battery drain into a darkened panel.

Steve

[ghurd]

Blocking diodes are in series to prevent reverse current. 
Only very small panels include them from the factory, maybe 5W and under.

The diodes in most commercial panels are bypass diodes.
Typically they are in parallel with 6V sections of the panel, and a 12V panel will have 2.
They allow current to bypass shaded sections in higher voltage strings.
They do not stop reverse current.

The connection diagram for the bypass diodes is shown in this old Siemens J-box.
G-

[Madscientist267]

Interesting... guess they come in varying flavors... ?

On both of mine, they're just blocking diodes, preventing reverse flow. It's the only provision for them in the J box. The panels are sets of 2 x 6V in series, with a single blocking diode in series with the leads going into the box.

Granted, I rewired them, but only in the sense of physically replacing the copper. I didn't change the configuration.

Then again, these things appear to have possibly powered the motor that might have lowered the bridge had the red sea not been 'partable'... LOL

Steve

[DamonHD]

G, my Topray panels (~12W or 15W), sold in the US by HF I suspect, have blocking diodes within the panel.

Rgds

Damon

[ghurd]

The 15W HF are different.  They have the diode in the box too.
I think the Topray may be sold by Northern Tool (same type of retailer as HF) and a few other names.  I think they always have a bulkier (and plastic) frame.

I have 1W, 3W, 5W multi and mono that do not have the blocking diode built in.  The factory supplied them with a blocking diode in a separate little heat shrink harness.
Pretty sure the only panels I have with the blocking diode built in are the $10 red 1.5W HF PVs, and a few of the no-charge-controller-in-the-plug VW PVs.
G-

[Madscientist267]

Confirmation that this circuit is worth the time and effort -

I saw (although briefly) a 4A hit on the meter today. This worked out to about 54 watts of output, the panels being rated at 60 (collectively, with 2 cells bypassed).

I'd say that's damn efficient!

Works out to 90%. Probably more since I am not taking into account the two bypassed cells in this equation.

The limitation is in the panels; today is partly cloudy, and the spikes are being seen when the sun can peek through at full brightness for brief moments at a time. The temperature of the panels is staying relatively low during all of this, so full power comes through when the panels see full light.

I have a smaller panel that I may use to either circulate water or run a fan for heatsinks. The logistics of that are a whole different thread however.

Thanks again Amanda! 

Steve