Grounding a power supply and filtering.

[si2030]

Hi MrAl, Ron and others,

He thanks again for taking the time to help.

Well I was initially stumped by this last night and I couldn't for the life of me figure out why there was just under a volts difference between the input to the LT1083s and the output... it struck me this morning though, that it was infact working perfectly and exactly as MrAl had indicated. The fact was, with the way I had configured the Dead Time soft start off the TL494 it was limiting the voltage of the switcher section to just 24 volts. I had mistook the 950mV gap as an indication there needed to be a bigger gap between the input and output where infact the LT1083s had been adjusted to rise past 24 volts and couldnt... so they sat 950mV below the input voltage.

By adding the appropriate resistor values that MrAl had suggested and changed the voltage controlled resistor managing the different levels of output from the LT1083s it worked perfectly.... I also then configured the current sense to trip and engage at intervals based on a sin wave load resistance which ranged from 401.4Ω to 1.57Ω

View attachment 62351

So I think this is working now how its suppose to... and it has full isolation from the low voltage section... I now intend to bench construct this to see if it actually does work - it should!

I do have a question concerning the comments made by Ron earlier. He indicated that when the current sense detects a value over the set level the transistor cuts in... causing a large voltage across the LT1083s... will this be acceptable or will I have to mitigate this? I am guessing I need a voltage differential for the LT1083s that never goes over 30volts as per the datasheet...

Thanks for the fantastic help so far.

Cheers

Simon

 

[MrAl]

Hi,

There are at least two situations where this could happen, but it is a transitory thing not a constant condition.

If we have the unit up and running into a normal load and say 6v output that means we have 9v input to the LT1083. If we suddenly short the output we have almost instantly 0v at the output but we still have 9v at the input. That certainly means the LT has 9 volts across it all of a sudden.
But then we analyze the feedback mechanism which states (basically) that if we have more than 3v across the LT we will have large feedback which will cut the PWM pattern and thus cut the voltage down aiming again for 3v across the LT1083.
So it doesnt matter HOW we get a lower voltage on the output because the feedback will always try to maintain 3v across the LT.

Thus, we have a temporary condition in which we have more voltage than planned across the LT1083 but it only lasts for the time it takes for the feedback to act and so shouldnt be too long. It depends on the response of the error amplifier (it's a bit of an integrator) and the response of the output LC filter. There's going to be a lag because of one or both of these, but it shouldnt be long enough to develop any really significant extra heating.
Repeated shorting and unshorting could possibly be a problem so you could check for that if you think it will actually happen.

The other time this will happen is during turn on. The input to the LT will rise and the output will rise, but eventually the output stops rising so temporarily the LT becomes a higher impedance so the input still goes up while the output stays constant. There could be overshoot because of the delay in the feedback system so we end up with a similar condition here too. But after the small delay the feedback kicks in and limits the differential voltage yet again.

There is another issue though. That's about the current transfer ratio of the opto coupler. If it changes too much it could cause the voltage across the LT to rise accordingly. The best bet is to keep the diode series resistance as small as possible yet still provide protection for the overshoot condition above so the opto coupler doesnt get too much of a current surge during that short time period. Current degrades the opto coupler's LED brightness over time, so we do want to limit this, but also keep the resistor as low as possible so the main control law stays as non linear as possible (a series resistor linearizes the feedback to some degree).
Also, temperature change affects the opto current transfer ratio too, so a good final test is in order where the opto is heated somewhat to see how much the voltage across the LT becomes. If it changes too much we may have to add an amplifier to the circuit.
If we keep the current lower or equal to 1/2 normal operating current we wont get too much degradation over time.

 

[si2030]

Hi All,

Ok phase two... breadboarding this circuit to make sure it works in the real world.

..and I hit a snag.

I have no problem with the configuring of the TL494 and in this all I wanted was a simple signal with no adjustment from it 45kHz output... I was feeding this into the IR2183 and here is where I have come to grief... The IR2183 is overheating and I get no signal on the output. I am pretty sure I have this hooked up correctly as per..

View attachment 62386

..and it appears to be pretty easy to hook up.

This is the breadboard.

View attachment 62387 View attachment 62388

Power is coming from a 20volt fixed supply.

It does look like the 10Ω resistor from the high side is touching the 0.1μF little blue cap however its not...

The green hookup wire transfers an inverted signal to the IR2183.. the brown transfers a no inverted signal as per the requirement of the schematic.

Power enters via pin1.

Ground is attached to pin4.

There are three green hookup wires leaving the board... the middle one goes to the middle of the half bridge.

The other two go to the high side and low side mosfets - IRFP450.

The half bridge has 40 volts over it.



First of all the IR2183 is heating excessively and heats within a minute of turn on.

I can only read a signal from the TL494 on HIN... I cannot get a signal on LIN... it appears to be grounded out.

I did try another IR2183 and the same thing happens..


Is it possible someone might have a look at the pictures and tell me why its overheating and also not putting out a signal on its outputs?

Kind regards

Simon

 

[MrAl]

Hi Simon,

We'd have to do a couple of tests to determine what is wrong here. It's probably something simple but to find it might take a little testing with a different setup.

The best thing to do is to check the 2183 chip independently. It doesnt matter that you tried another one because that one could be bad too. There are several reasons this could have happened but the best thing to do is to check it and see if it is working apart from the normal circuit.

Pulling the chip out or testing the other one, you can hook it up with some resistors and check that it can do basic switching from input to output. You can look at the data sheet to see how to connect it. I recommend a quick test here because there's no logical reason why one input pin would stay low especially when it has an internal pull up resistor. You can switch the input with resistors to ground or to +Vcc or using a pulse generator.

So test 1 is to carefully check one or both of the 2183 chips and make sure they can operate as per their function independently from the breadboard you already have set up.

Many times these problems are from something wired slightly wrong like MOSFET source drain swapped or something simple like that, but what's strange about this is that the input seems permanently grounded, so it could be the driver chip(s). There's a chance (not definite but a chance) that when the device started up for the first time the ringing on the output blew out the chip (hopefully it's not that).

Also, breadboards like these plugboards are not recommended for these kinds of power supply circuits because they have lots of limitations and extra capacitance everywhere which can cause ringing etc.

Also, when you run these breadboards up you should start with a low level input power (line) and gradually bring it up to full line voltage. This prevents blowing out chips and transistors if there are spikes and ringing present. It gives you a chance to look for these problems before coming up to full line voltage where components would start to blow if something was wrong. You also would monitor the input current to check for excessive input current and back off if it goes too high with little or no load. Of course running for the first time should be with only light load or no load.

 

[si2030]

Hi MrAl,

Again many thanks for taking the time on this... really does make a difference.

Here is a composite of parts of the IR2183 datasheet.

View attachment 62396

I am pretty sure I might have found the problem.. as can be seen from the two diagrams the actual pin out of the chip is vastly different from the hook up diagram IR presented... I took it that IR would use the same pinout for this diagram as for the actual chip... so when I connected one of the TL outputs to pin 3 I was actually shorting the output of the TL494 through the ground pin - no wonder it was heating up pretty quickly... I haven't tried it yet as its pretty late here and I just worked this out but it does help if you use the right diagram for the correct pin connections... and of course there would be no output as I had wired up every pin in the wrong place! doh.

I figure it will work when I use the right connections. Will let you know.

I do have a quick question regarding the capacitors used in the half bridge... they work as a potential divider and form the other side of the half bridge.. what kind of capacitors should these be... I read that you need to use special caps with low ESR however I am using Metallised Polypropylene caps as thats all I can locate in a pinch.. what would be ideal capacitors for this job... they were sized to 3μ in the simulation..

Kind Regards

Simon

 

[MrAl]

Hi,


Oh, hooking up the pins in random order is not recommended except for New Years Eve and some
Chinese birthday celebrations
Lets hope the chips are not blown out.

The caps you have are probably a pretty good choice. You'd have to use larger electrolytics like the ATX power supplies if you wanted that type instead.

I cant wait now to hear about the first run up of this project.

Oh yeah another trick i've heard about is to use undersized caps for the first run up to keep max transfer energies lower than normal in case anything goes wrong. This provides at least some sanity check before going all the way. As i said before too, the input power is usually brought up slowly a few volts at a time as some key points are checked with the scope just to make sure everything is connected right.

ADDED:
The other thing to check is the RMS current rating. It should be able to handle the current without any problems like heating.

 

[si2030]

Hi There,

I have been tinkering with the setup (above) on the bench and the first time around it blew both IRPF450 Mosfets and the IR2183 with the IR2183 going quite spectacularly (I always wear safety glasses when working with high voltages on the bench)... This was after it worked well at 40 volts -another bench power supply.. but when I put in the much higher voltage across the half bridge it destroyed those components... I have looked closely at the circuit and I think its because I did not connect the ground of the Half bridge voltage to the ground the IR2183 was working on so they were floating with respect to each other... I am interested to know what other people think here... in any case three components junked and back to the drawing board... (used a variac for the higher voltage where as the 40volt test the grounds were linked)

In the circuit I have an IR2183. It requires a bootstrap capacitor... there are a few application notes on how to size this on the internet which I worked through and applied for the my circuit. I used this application note to size the capacitor.

View attachment 62546

I got ~112nF which I multiplied by 15 to get ~1.6μF... as per their directions.

What I am unsure about with the IR2183 and also how this works... is... given I have a half bridge rail voltage of 320Volts... what voltage rating should the 1.6μF bootstrap capacitor be... I am unsure what voltage is between Vb and Vs and so what the capacitor voltage rating should be here to handle this voltage.. does it see 320 volts between these two pins or (12 volts => Vcc)... I am not sure?

View attachment 62547

Kind regards

Simon

 

[MrAl]

Hi,


The bootstrap cap should have a relatively low voltage across it, but if VB can draw any current during startup it could have 160v across it. Im not sure if VB can draw much current during startup or not.

The thing to check is that diode. It has to have a high voltage rating at least 350 volts.

When you are running up for the first time you have to check for spikes across the transistors. The parallel diode of the opposite transistor catches the spike but the lead inductance prevents this from happening perfectly so some of that inductance causes a spike across the transistors. If this happens you have to use a snubber circuit across each transistor to absorb the spike temporarily while the small but not insignificant inductance gets a chance to conduct.
Snubbers are usually made with a fast high voltage diode, a capacitor, and a resistor in parallel to the cap. The idea is that the diode conducts quickly and the cap absorbs the spike, then when the transistor next turns on the resistor dissipates some of the energy in the capacitor, thus getting ready for the next cycle.
The connection is simple: diode to cap, resistor across cap, network across transistor drain to source.
Alternately (or in addition to) controlled turn off can be employed.

Obviously the connections to the entire circuit have to be absolutely correct or it will never work.

I forgot to say You're Welcome to your many thanks for helping out with this circuit and im happy to help when i can, and it's also nice to find that you are interested in these kinds of circuits. Can i ask what you intend to use it for?

 

[ronsimpson]

At some point you will need to move form that bread board to a real PCB. At this current, voltage and frequency you need a good PCB with short traces. Long wires will cause problems. As MrAl pointed out, spikes comes form long wire.

 

[si2030]

Hi MrAl

Actually I trained in accounting and IT but this stuff really interests me (I should have done engineering instead).. I am interested in High voltage generation in particular but at the root of this disipline are power supplies... I am working in HV at home but at the heart of this is knowledge about power supplies - all types. I blew up my crappy power supply here and contemplated buying one.. then made the decision to build one as I wanted to learn how they work better... and I wanted to have a piece of equipment that I knew how to fix if it died.. by doing this I have learnt a lot about switch mode power supplies and in particular magnetics... which is invaluable if you want to know about HV generation..

OK tomorrow I will design and implement a snubber system across the two mosfets...

Thanks again

Simon

 

[MrAl]

Hi again,


Oh ok so you want to have a power supply to work with again and you dont want to have to worry about it blowing up and not being able to fix it. That's really a good idea and something i like to do with various things myself. I also like to obtain schematics for things that i can get, and these days might be inclined to look for a schematic BEFORE buying the equipment just so i know i can fix it if i have to.

I forgot to mention that the snubber circuits have to be very very close to the physical transistors, with leads as short as humanly possible. They would be mounted on the heat sink right with the transistor perhaps on a few standoffs.

 

[si2030]

I thought about this and I have a number of quick questions that I need clarified if thats OK....

I had a close look at the circuit etc on spice and across the transistor however how do I check for spikes on the real circuit.. do I just run this up to say 40 volts and attach, for the high side, the oscilloscope probe ground clip onto the source end and the probe onto the drain.... and check the signal..

I also put two 16 volt zeners back to back across the gate and source for both Mosfets as well... I have not placed any resistors in the gate leads from the IR2183.. should I do this as well?

Just to confirm... there will be two snubber networks each one across the drain/source or the Mosfet.

Ron... yep thats definitely next.. have everything ready.. just want to confirm the circuit first.

 

[ronsimpson]

si2030,

I do not use 16 volt zeners on the gates of MOSFETS unless the gate is a long distance form the driving IC.

There are formulas for gate resistance but for this application use 4.7 to 10 ohms. For home use it may not matter but if you are fighting EMI problems the resistor helps. If you PCB has long traces you should use 10 or 22 ohms to slow the MOSFETs down some.

When I am trying a new approach to power supplies; I may make two PCBs. One PCB for the MOSFETS, inductors and the high current items. A different PCB for the "brains" like the TL494. Then I can change one PCB and keep the other. To keep the price down I only get one PCB but put many different projects on it. When it comes I cut the big board into many small ones. Where I get PCBs from the charge for a small board is about the same as a big board.

 

[MrAl]

Hi,

I assume that you are not using that little plug board anymore.

Yes that's right, you check right across the transistor, but of course with a probe made for the voltage not 1:1 or it might be a strain on the scope input.

Usually there are resistors in the gate leads for the purpose of reducing the ringing and to slow down the turn off slightly as that reduces the spikes too. 10 to 20 ohms is usual.

Yes one snubber per transistor.

 

[si2030]

Hi again,

Well I am experiencing some real problems understanding the nuances on how the IR2183 half bridge driver works. I have faithfully hooked up the chip (after the earlier problems). I used the pin diagram as per the datasheet.

As per the spice model and IR2183_test.asc the inputs should be HIN and LIN (INVERTED) and I should get the following wave forms....

View attachment 62649

I hooked up the TL494 with one channel working as an emitter follower (taking the signal from below the output transistor) representing a non inverting signal and the other channel with an inverting signal.. (output from above the output transistor).

This provided the following traces...

View attachment 62652

View attachment 62653

Checking the outputs of the IR2183 I found that LIN (INVERTED) input signal was inverted as expected... however the HIN signal was also inverted...

I clearly dont understand how this IC is suppose to work as I am now turning the half bridge Mosfets on at the same time... no wonder they blew.. Having 4 of these driver chips I tried another with exactly the same outcome...

A clue to this lack of understanding is this diagram from the datasheet...

View attachment 62654

Initially the HIN and LIN (INVERTED) inputs are in sync and the outputs are alternating (LIN has been inverted and HIN has not)... but then I dont understand how the rest works...

In the bigger scheme of things though HIN is being inverted and I do not know why?

Hoping someone might identify the cause of this for me..


Kind Regards

Simon

 

[MrAl]

Hi Simon,


You can't look at a single pin and claim that it is being inverted, like saying that Hin is being inverted. You have to look at both signals at the same time and apply the logical connective for the outputs.

The truth table is such that with both inputs high you get HO turned on, and with both inputs low you get LO turned on, and anything else turns them both off.

Looking at it another way, with Lin high and Hin low both outputs are off, then if Lin goes low LO turns on, but if Hin goes high then HO turns on.

What this means from the standpoint of the TL494 is that Hin has to connect to the emitter and Lin has to connect to the collector. The schematic you have posted shows them reversed.

Taking the logic right from the data sheet this way should work.

 

[si2030]

Hi MrAl

Feeling a bit obtuse here but I seem incapable of getting this IC (IR2183) to work as intended.

Here is a picture of the traces LIN and HIN. As you mentioned HIN is connected to the emitter and LIN is connected to the collector of the TL494.

View attachment 62671

They overlap where HO should turn on for the full length of HIN being on.... however.

Here is the HIN and HO together. Whats indicated from the datasheet is that while both LIN and HIN are both on then HO should be on... by comparing HIN with HO instead of HO being on at the same time as HIN its instead inverted... at least thats how I am reading it...

View attachment 62674

Here is the way I connected it up... simple and exactly (unless I have screwed up) to the datasheets requirements... This shows that I am using both probes one on HIN (pin1) and one on the HO (pin7).

View attachment 62675

So the output of the IC as I am observing has HO inverted which then creates the situation of both HO and LO being on at the same time....

I would be interested in yours and anybody else's comments and apologise in advance if I have made a stupid mistake...

Kind Regards

Simon

 

[MrAl]

Hi again,


Going by the data sheet HO should be on when both inputs are high, and LO should be on when both inputs are low. If this is not the case then either that is not the right part or they made a bad mistake on the data sheet.
This isnt one of the parts that was involved in one of the blow ups was it? It is a new part never used, or never involved in a blow up failure? If it was involved in a blow up it could be damaged. A new part would have to be tried. There are other problems that could crop up if the chip is only partially blown.

It might also help to show Hin and Lin pins on one scope picture, then Lin and LO on another scope picture.
If you can show both LO and HO on another scope picture too that would be good, perhaps with the trigger input
on the Lin pin.

 

[si2030]

Hi again MrAl

Thanks again for taking the time on this... I had not anticipated this little part would cause me so much grief..

With regard to the IR2183s, I originally got 4 of these when Ron suggested them as a replacement... and they do indeed say IR2183 on the top (I got reasonable pictures with super macro)... I blew one up... actually it vaporised the ground or com leg.. so I am down to three.. I have one still in its plastic casing. I checked the two I have out and they are giving identical results...

OK the traces as requested from the IR2183:

1). Hin and Lin with Lin being the top trace:

View attachment 62684

Here are the probes settings for the previous picture... Channel one is the channel being triggered from and its on pin 2 or Lin - top trace.

View attachment 62686

2). Lin and LO (triggering off Lin)

View attachment 62687

..and the way I connected the probes for this... Channel one is connected to Lin and is the channel used for triggering.

View attachment 62688

2a). I added this in so you can see that the output HO for HIN is actually inverted which is, from what I can gather different from what I should be expecting...

View attachment 62692

Here is the way the probes were connected.

View attachment 62693

3). Finally LO and HO. I left the channel 2 probe on LO and moved the channel 1 probe from Lin to HO so in this picture its triggering off HO.

View attachment 62689

..and the way the probes were hooked up.

View attachment 62690


So to summarize this.... Pin 1 is HIN and pin 2 is LIN. The inputs for these are coming from the TL494 emmitter and collector respectively. This provides a situation where the input for HIN and LIN are non inverted and inverted as per datasheet.

Output for Lin is inverted.

The problem I note in 2a and 3 above is that HO is inverting along with LO. HO should not be inverting but it is and since we have an input where one signal is inverted we need an output were again only one signal inverts... both signals invert again and so they are synchronous... not good for a half bridge.

This is how I see it... ultimately HO is being inverted from its input... when the datasheet says otherwise.

Let me know if there is anything else you need or whether I have made a simple mistake in the way I have hooked up or.... interpreted the detail..

Again kind regards

Simon

 

[MrAl]

Hi Simon,


Before we go any further the first scope pic shows Lin on the top and Hin on the bottom. That means pin 2 on the top and pin 1 on the bottom. That's reversed from what it should be.

Take the lead out of pin 2 and place it into pin 1, then take the original lead out of pin 1 and place it into pin 2. The idea here is that Lin should have a long top and short bottom, and Hin should have a short top and long bottom. Until the inputs are correct you'll see strange stuff. Before you do this however verify that pic 1 is really correct in that it shows pin 2 Lin on top and pin 1 Hin on bottom.

I'll take a look at the other pics next, but once you do that you can then take another look at HO and LO and be sure to mention which one is LO (like top or bottom) and which one is HO. Lets see the scope pic after the swap of pins 1 and 2.

Taking a look at pics 1 and 2a, i see Hin with long on the top and short on the bottom, but in pic 2a i see Hin with short on the top and long on the bottom. It cant be both ways, so one pic is not correct in that it is not Hin.

What you can do is mark probe 1 with a small piece of tape so it shows up in the pics, and make sure that scope trace on the top is ALWAYS from probe 1. You have to be very careful about this or we'll be here forever

Nice pics BTW.