help reading circuit for welding machine (negative switching supply)

[coppercone2]

I am hoping that someone good at reading circuit pathes can answer a simple question for me

https://forum.millerwelds.com/filedata/fetch?id=615167&d=1622257766

In the middle of this circuit diagram, you can see the output for -15V . The description says its regulated from the 30V

If you just feed 30V into the board, should that -15V rail start up? The -15V is generated by Q21 and inductor L1 I think.

Does it have any other input requirements for starting to generate that rail?


I bought the pomona banana to round pin adapters (great purchase BTW, the kit of them allows you to connect banana cables to adapters that go into pin sockets or over pins, and they have necks that are flexible) so I can plug a supply into the board. When I add 30V, the 15V is made, but the -15V shows +0.5V.

I need to know if I am missing some input that enables the negative voltage rail to start working. I don't think so, but its a big circuit and I thought I might be missing something. 

I circled the area in red which I am confused about

[Andy Chee]

Your circled area appears to be an inverting buck-boost regulator, driven by an oscillator made from a schmitt trigger NAND gate.

But I can't work out how the regulation feedback functions, suffice to say, it looks like the designer has had some major difficulty in converting a negative voltage feedback into a positive voltage feedback signal.  The entire regulation circuit is non-intuitive!

[coppercone2]

yeah I loathe this blasted board

[coppercone2]

Around the area, all the 15, -15 and 28 points are connected to the pins labeled as such.

I ignore the Q11 and R13 area, because that is biased at 15V, so its just a 15V diff between30 and 15, aka its between two low impedance points, so that Q11 can't do anything to the +15 V rail, and it measures at +15V when its on.


Then you have the R64 tap and the R55 path, which go into some annoying bullshit that can mess up the gate some how. I don't know what that area is doing, it looks like its on because the difference is the 3 ohms on the 28v RAIL??? ?

[coppercone2]

what is supposed to oscillate this circuit.

The transistor driver is biased between +30V and +15V to get +15V full scale. That is correct because I measured it.

The transistor driver is controlled by Q25. All the oscillator signal must go through R55 to power the transistor base.
R55 is connected to the NAND gate.

THen it gets fucking confusing.

[coppercone2]

So every one of these circuits needs a feed back I think. I don't think they do open loop stuff.

So really your R19 resistor connected to -15V is the feedback from the output of Q21.

So the feedback path is mega weird to me.

[Andy Chee]

The oscillator is  the easy bit!  The 4093, C18 & R38 form a schmitt trigger oscillator.



Being a NAND gate allows the oscillator to be "enabled" and "disabled" by the signal on pin 1 of the 4093.

What I can't figure out is whether this pin 1 is the primary regulation mechanism (via Q16), or the regulation is through R22 & Q10 (which changes the feedback resistor, thus oscillator frequency).

[coppercone2]

is this what is happening?



I don't understand what is going on there

What the hell is R64. Is that current something?

waht in the shit is Q12 doing ?

[coppercone2]

and the main thing, did i miss something, because I don't see what could disable the -15V oscillator ? it has no other inputs there as far as I can tell?

I checked and there is continuity between all the +15, -15 and +28 pins in that diagram I re-drew. I thought maybe there was a missing external 15.

[Andy Chee]

What the hell is R64. Is that current something?

waht in the shit is R30 doing ?

R64 in conjunction with R53 & Q18, form some sort of current limiting circuit.

R30 looks is just a pull-up resistor for Q12 (to keep Q12 off when there's no signal on the base)

[Andy Chee]

and the main thing, did i miss something, because I don't see what could disable the -15V oscillator ? it has no other inputs there as far as I can tell?

When Q15 is on, this shorts out C18 and disables the oscillator.  This is probably related to the current limiting.

[coppercone2]

Ok but to step back from the complexities, the only "nodes" that might 'control' this thing are +15, -15V and +30V.


I measured every single node to the pins. All +15, -15 and +28V go to the pins on the bottom of the PCB. And the -15 is all feedback.

so I am thinking that something is wrong with the board that its not oscillating when I plug it into a power supply ?


**********
IT Is just a board, laying on a ESD mat, connected to a +28V power supply (linear). As far as the circuit shows, it should oscillate and make the -15V from just that ?

To me it looks like all the other pins are basically isolated from the function of the -15V rail.


I don't see any kind of lockout, disable, power down, etc functionality linked to the -15V rail.


ITS CONFUSING THE HELL OUT OF ME BECAUSE IT LOOKS LIKE IT HAS 4 FEEDBACK PATHS!!!!!!!!!!

[Andy Chee]

There are two ways I can see the oscillator is disabled;

a) Q15 is on

and/or

b) pin 1 of U2 is low

[coppercone2]

but there are no pin that control that, its all integrated on the PCB.

That makes me think its broken.

I thought one of the other pins might be related. But I don't see anything. 


I can't believe it has 4 feed back paths. !!!!! who the hell does this !!!

[Andy Chee]

To check for oscillator function, I would remove Q15.

For safety, I would also remove Q21.

Doing this should enable the oscillator, with or without -15V.

[Andy Chee]

Where does -15V go on receptacle RC1?

[coppercone2]

You mean where on the welder? I am not sure. It goes into the mother board. I think part of it  goes to the LEM current sensor. That schematic no one has made yet. It might power the other PCB that is the high voltage cascode looking mosfet shit

I re-drew it again I think

[Andy Chee]

You mean where on the welder? I am not sure. It goes into the mother board.

Knowing where the -15V goes on receptacle RC1 is vitally important!  The current limiting of Q21 is there for a purpose!

[coppercone2]

I thought it did not matter since the card is on a bench.

This card had a explosion on it, which was repaired,on R2. All semiconductors other then diodes were replaced, the resistor was replaced, but it won't work. It happened because some idiot on a nother forum wrote the wrong transistor and i lost my diagram and trusted or possibly misunderstood somehting

For me the usual procedure for checking confusing things is to make power supplies run first. If those are good you can isolate.

[coppercone2]

ok so the answer in this test case its not connected to anything and reading 0.5V

I thought about putting a load on it. I guess I can try. Maybe it needs a minimum load.

I need to order more banana to pin adapters because if one more grabber hook slips off of something I am going to throw this machine into the woods. So the load test will wait a week. Then I can get 2 banana connectors and a resistor and have a nice load properly hooked up without getting acidosis

i know its simple but i am at my limit. i can give you the results of a loaded run in about a week. 

[Andy Chee]

Ok.  Try my suggestion, remove Q15 (and Q21 for safety).

This should make U2 pin 3 oscillate.

You do have an oscilloscope to view the oscillation, yes?

[coppercone2]

Ok I can try that tomorrow

I got pogo pin probes for the scope recently just for reducing stress level

they should come with scopes it makes it actually useful for repair because doing it with the normal way is like doing things with tree branches instead of tools  . I can't believe that they actually expected that people would poke inflexible probes into things for so long. i am really starting to think that its completely insane to not have gold plated pogo pin probes even if you have a harbor freight meter only.  its completely medieval

really i can't believe that I was jousting with circuits for like 2 decades before I finally listen to dave jones and got the gold pogo probes. its like your doing it wrong buddy. its like refusing to buy suspension for a car. we are so backwards



if you have a nice meter and you don't have pogo pin probes its like not having a suspension for a car


if you have a cheap meter but you don't have pogo pin probes its like not having a suspension for your bicycle

[coppercone2]

OK I got excited because I realized there is a new banana connector to make, that is you can get a regular banana end plug (mating end), and put a loop on it instead of a wire, so you get a series adapter that lets you hook up a oscilloscope clip to a banana cable nicely! very simple but it gave me the positive energy to remove the transistors.




this is the invention


the green should be a bare stiff copper wire loop, instead of a wire, so you can clip a scope ground probe on it nicely! with a series splice

how good is that for probing circuits you are powering with external leads during repair? instead of trying to clip to some random inconveniant metal or having to change probe ground leads etc. for low precision repair work.


however


U2 pin 3 is constant DC output  with the two transistors you thought should be removed. 
looks like the gate oscillator is stuck on at +15V.


That makes me think something is broken!


So this might isolate the problem to
r53, r44, c24, q18, r45, r46, q15, c18, d13, r38, and U2 ?

[Andy Chee]

U2 pin 3 is constant DC output  with the two transistors you thought should be removed. 
looks like the gate oscillator is stuck on.

With the two transistors removed, can you check U2 pin 1?  Is it high or low?

The oscillator will only work if pin 1 is high.

[coppercone2]

Pin #1 of U#2 is 1.8V, #2 = 15V, #3 = 15V

mean voltage on scope with no sign of any pulse activity

You know its weird I thought i change all the transistors but some of these solder joints look factory, I think I only changed the ones near the 'blast'. Well its worth trying a shotgun approach on 04's and 06's . I did change all the IC's though (including optos)


But the transistors all measure fine on diode test. Maybe there is another problem.


But I did test all the transistors now in circuit with diode test, and all the diodes, it looks fine. The inductor too. I guess I can go through the resistors and then replace all the transistors again and maybe test the diodes with my new transistor tester.

Yeah, their transistors have the leads bent over in a certain way. I am sure now that I did not replace some of them because they tested fine and I was not thinking the damage can propagate there.

But holy crap could it really be one of those problems where they all look perfect on the diode test but one of em is off?  :'(


Unless you figure out what I missed.
I bet that maybe that oscillator has some other parameter that might be blocking its operation.



I keep thinking that in some of my repairs there is a bad transistor that measures good on the DMM. But then usually I find out its something else! its like paranoia.


Could it be a leaky transistor :O Could that day have finally come? Maybe it finally happened in a circuit because of the 800V power electronics nearby.


But if you have any more thoughts it would be welcome because i am extremely weary of working on this PCB

[Andy Chee]

Pin #1 of U#2 is 1.8V, #2 = 15V, #3 = 15V

Ok, with pin 1 @ 1.8V that suggests there's something wrong around Q12.  Don't forget to measure the collector of Q12, which is connected to pin 1, and should read the same because they should be connected together. 

I assume you've been supplying +28V during these tests?  With +28V & +15V connected, the voltage at U2 pin 1 should be 15V, not 1.8V.

With the schematic drawn as is, I'm not exactly sure how Q14 is supposed to turn on (thus turning on Q12 as well).  D14 & D12 look like zener diode symbols, so that should theoretically allow Q14 & Q12 to turn on. 

Do you know what voltage zener diode D14 & D12 are?  The voltage isn't marked on the schematic.  Note that if you inserted conventional diodes in these positions, the circuit will not work.

[coppercone2]

nah I left the diodes alone. I just measured continuity both ways.

OK to be extra  cautious I got my other power supply with is 30V on the mark. The last test actually turned out to be a little low (26.5V)


Now I do see a square wave!

I measured the two transistors on the tester and they look OK. I need to reattch them and check it again.

[Andy Chee]

Before you reinstall anything.  Connect oscilloscope to Q23,Q24 emitter.  There should be a squarewave here as well.

Reinstall Q15.  Leave Q21 out.  Connect oscilloscope to Q23,Q24 emitter again.  The squarewave should still be there.

[coppercone2]

The square wave on the gate of the removed power transistor (23 + 24 emitter) is a square wave from 19 to 30 volts,

so its floating on a DC

when I reinstalled the little transistor it looks the same

[coppercone2]

Ok so it actually works. Maybe the 26.5V test voltage was a little low or maybe there might have been a tiny short circuit somewhere because of conformal coating remains. The inductor sings and makes the -12V rail (I forgot to mention that it actually is/was -12V according to my previous measurements and other forum reports about this card)

There is a good chance its something on the motherboard then. Oh I loathe that Skiip IGBT construction. The 30V rail must be acting up.

At least that narrows it down by a huge amount.

Actually I can test to see what the UVLO is on this rail

It makes the power supply cranky at 27V turning on and off. I damaged the display on my 40V linear supply (thanks micro-gear-pump) (this is done with the 250V supply) and I thought I could get away using the 25V dual supply. So it was just a red herring.


BUT the noise that it makes, at 27V, where the supply is fluctuating/unstable is the noise the welding machine was making alot. I guess it means unstable negative rail, or unstable 30V rail, or low rail voltage, or fluctating rail voltage. It is a annoying CRT "dancing/squirming" noise. That is a good thing to know. I always suspected the machine was supposed to sound "more stable". Sort of like a angry fly trapped some where.



But holy crap that circuit for the negative rail is ludicrous IMO. I can't believe they did not try to integrate that a little more. It seems insane. Did they really need that much crap there? Its like 1/3 of the board.


It's like 40 parts, and it does not even feel like its built good. I can understand if there was protection parts and shit like that but it just seems like a big freaking mess. I mean if they had like filters and clamps and stuff I can see the parts count getting high, but this is just like wut?


Do you see some...  strength of this power supply? I am not buying that they were scared that a negative regulator IC is gonna be a supply chain limitation, they are too common.

I don't even think it could be considered reliable, because they share the NAND gate with other circuits. So they are sharing a NAND gate connected to a big inductor power supply, and basically linking the power supply to a whole bunch of other control circuits. I think that gate in the feed back path should be its own gate, not piggybacked, if you wanted this design to be 'good'. I mean there is a signal path in that chip to the 810VDC rail for christ sakes. Its sharing a NAND gate between a -15V instrumentation supply and the mains connected 810V down regulator. that seems like cascade failure to me. And its freaking controlling 810VDC through a header connector and that is for a iGBT control. I think this thing is built like the death star for self destruct. Why the hell is that regulator not placed locally. to me its bananas. I think thats fitting like a freaking   primary nuclear reactor control system in the mail room of a building

[Andy Chee]

But holy crap that circuit for the negative rail is ludicrous IMO. I can't believe they did not try to integrate that a little more. It seems insane. Did they really need that much crap there? Its like 1/3 of the board.

Any idea when this board was designed?  If this welder was from the 70s or 80s, integrated chips for dedicated SMPS may have been relatively expensive and hard to source back then.

Another explanation is that the power supply was first initially designed by an undergraduate student doing work experience for the company, and the design was just copied from generation to generation without any thought of revision.  R&D costs money.

Other than that, I agree, what a mess of a circuit design!

[coppercone2]

the miller dynasty is the 2000's flagship state of the art unit from miller at a price tag of 4k+ USD

[coppercone2]

See I thought maybe it had something to do with CMRR or immunity but

I don't see any components I associate with 'protected'

and I see a giant 'circuit'

I think that means its worse. 

Its like really freaking bootleg for a switcher. The output ripple looks crappy too. You would think it would have like filters

This reminds me of the circuit you learn how a DC DC converter works on a bread board in electronics 102 class with some extras, and given the gigantic size, it might have similar performance lol

and it sounds annoying too

[Andy Chee]

I don't see any components I associate with 'protected'

I mentioned several posts ago that R64, Q18 & Q15 form a MOSFET current limiting circuit.  It's pretty rudimentary, but functional.  As current increases through R64, the voltage increases across R64.  The increased voltage turns on Q18, which turns on Q15, which shorts out C18 and causes the oscillator to shutdown, which turns off the MOSFET.

I asked before what was on the -15V receptacle RC1?  Maybe this rudimentary protection is sufficient for whatever is connected to RC1?

[coppercone2]

I still have to investigate that. I think it provides power to the other PCB that is very complicated with lots of CD gates and op amps etc.

The card its connected to is called the inverter control card. this card is called the boost control card

By protection I mean ESD, RFI, awesome parts etc. Current limiting is a basic function of most controllers. Its not 'guilding the lilly"

I just mean you don't get any 'extra' feature from the way they built it, related to the parts count. You would think there should be a good reason behind that circuit being the way it is! I don't think it offers better performance or durability compared to a controller based one.


You know I mean like in the sense of its not  "they had to use this topology because they used extra high reliability hermetic 3904's in a proven mil-spec design with whatever whizz bang cool tech they have in nasa that helps justify the price and complexity over using a unreliable switching integrated circuit with generally superior performance"

Because there are sometimes discrete designs that have a justification. I don't think this is one of them lol


I can't talk too much because I don't work with designing switching power supplies and I am not too interested in them

[Andy Chee]

I just mean you don't get any 'extra' feature from the way they built it, related to the parts count. You would think there should be a good reason behind that circuit being the way it is! I don't think it offers better performance or durability compared to a controller based one.

As I suggested earlier, it's likely their R&D is being performed by an unpaid engineering student intern.

As far as cutting edge technology goes, welding machine control boards are a far cry from designing automotive engine management computers, for example.

[coppercone2]

the price says otherwise 

I am just being careful not to accidentally trash some nice 'analog' or 'discrete' design because it looks 'different'


I think they really could have made it a little better for the market segment they thought it should be in. it seems insane they handed the design of this machine control electronics to a intern.... its a real dilbert's boss type decision.

[Andy Chee]

the price says otherwise 

One shouldn't be surprised just how high product mark ups can be, particularly if there's name recognition behind the brand!

[coppercone2]

the price says otherwise 

One shouldn't be surprised just how high product mark ups can be, particularly if there's name recognition behind the brand!

its supposed to high light american industrial manufacturing might 

B2 bomber. GPS satellite. Space Shuttle. Miller dynasty.

Marketing put it up like a space shuttle . I guess its more like a "ford"