1 hour ago, Coulomb said:

The fact that the red point reads just over 12.0 V means that the power supply is fine

Yes, that is a dead giveaway the flyback converter's feedback is working correctly. 

1 hour ago, Coulomb said:

EDIT: Oops, that was before later posts. The fact that the red point reads just over 12.0 V means that the power supply is fine. The blue point reads less than 12 V because it's not a perfect square wave, and the power supply works on the peak, not the RMS or the average value. Notice on that reading only, it says "AC + DC True RMS".

Yep true rms 10.444v on broken one between blaack and blue.

But on a working one same True rms between black and blue is around 12.877v

 

On a working one between black and red is:

 

Edited December 27, 20223 yr by Abra

1 hour ago, BritishRacingGreen said:

What do you read on the 7915 and 7805 outputs? 

 

I can't find 7915, if you mean 7912 so:

7912 (U20) and 7805 (U5):

I'm very worried about the 10.444v value for high frequency power. All other values seem to be normal, similar to the values on a working inverter, but 10.444v is very low, and I'm afraid this is the reason for the non-starting of transistors on the CSS board.

1 hour ago, Abra said:

I'm very worried about the 10.444v value for high frequency power. All other values seem to be normal, similar to the values on a working inverter, but 10.444v is very low, and I'm afraid this is the reason for the non-starting of transistors on the CSS board.

As per Coulomb it can be excessive load. You can remove or partially lift  the diode that feeds the igbt driver transformer. It will disconnect the HFPWM feed to the igbt transformer. I think its called D20 or D40. Then you can check wether there is a rise in HFPWM feed. 

EDIT: never never ever  remove D54 for any testing purposes. Or any component on the SMPS for that matter.  Especially in the voltage feedback loop. 

Edited December 27, 20223 yr by BritishRacingGreen

10 minutes ago, BritishRacingGreen said:

As per Coulomb it can be excessive load. You can remove or partially lift  the diode that feeds the igbt driver transformer. It will disconnect the HFPWM feed to the igbt transformer. I think its called D20 or D40. Then you can check wether there is a rise in HFPWM feed. 

EDIT: never never ever  remove D54 for any testing purposes. Or any component on the SMPS for that matter.  Especially in the voltage feedback loop. 

Sometimes it happen when igbts fail, that some driver components are under stress. This may  cause  stress on the HFPWM feed.  I had such a problem, my luck was bad as I had to rewound the igbt drive transformer. 

1 hour ago, Abra said:

I'm very worried about the 10.444v value for high frequency power.

My initial reaction is that the readings on the good inverter make no sense, because the peak value can't be lower than the RMS value. But it's only 0.8 V lower, and you're measuring before the 22 Ω resistor, which might be compounding things. If D54 is a silicon diode, then that makes sense, if HFPW+ is a square wave. 

Sadly, there was an error with the position of that 22Ω resistor, R206 on the PIP-4048:

 

The fact that you still get 12 V out in the bad inverter suggests that the waveform of HFPW+ is bad/different/wrong. That could be because of an overload, or possibly due to a failure of some spike suppressing parts, e.g. the RC network between the drain and source of the power supply main transistor (R224/R211/C80 across Q36 in the PIP-4048MS).

Edit: I'll publish the revised schematic in a while, possibly after I've accumulated a few more changes.

Or it might be totally inconsequential, resulting from random differences between MOSFETs, transformers, or other factors. An oscilloscope would be handy here, to compare the good and bad inverters.

Edited December 27, 20223 yr by Coulomb

23 hours ago, BritishRacingGreen said:

As per Coulomb it can be excessive load. You can remove or partially lift  the diode that feeds the igbt driver transformer. It will disconnect the HFPWM feed to the igbt transformer. I think its called D20 or D40. Then you can check wether there is a rise in HFPWM feed.

Soldered D20, tried to turn it on, the same...

 

Edited December 28, 20223 yr by Abra

23 minutes ago, Coulomb said:

RC network between the drain and source of the power supply main transistor (R224/R211/C80 across Q36 in the PIP-4048MS).

Checked, and I think R224/R211/C80/Q36 working well

45 minutes ago, Coulomb said:

Or it might be totally inconsequential, resulting from random differences between MOSFETs, transformers, or other factors. An oscilloscope would be handy here, to compare the good and bad inverters.

Unfortunately, there is no oscilloscope, but in principle I was going to buy it sometime ... I'll see...

What brand/model would you recommend? The most versatile and accurate?

Edited December 27, 20223 yr by Abra

R215 is a possibility; it's the large wattage resistor in the source of the switching transistor. At 0.15Ω, it's very hard to measure; I'd use a current limited power supply set to about 3 A and measure the voltage drop. When the power is off, of course.

2 minutes ago, Abra said:

The most versatile and accurate?

Honestly, for inverter work, you usually don't need top shelf equipment. I think I'd go a step above the pocket ones, or ones that you connect to your PC, but the Asian ones seem reasonable value for money, Rigol seems to be a good brand. But they make top shelf gear as well as more affordable ones, and of course the price varies accordingly.

About C80

When i desolder it and measure it's ok, around 2.4 nF i chek it on 2 boards. 1st, what i repair now, and 2nd broken one.

But when I measure it when it without desolder - it's around 44uF on 2nd, and around 99uF on 1st

Just an observation. I don't know where the correct value is, but they are just different. Perhaps this has something to do with my problem.

Edited December 27, 20223 yr by Abra

1 hour ago, Coulomb said:

Honestly, for inverter work, you usually don't need top shelf equipment. I think I'd go a step above the pocket ones, or ones that you connect to your PC, but the Asian ones seem reasonable value for money, Rigol seems to be a good brand. But they make top shelf gear as well as more affordable ones, and of course the price varies accordingly.

Thank you, I'll take a look, just maybe in the future I'll be doing something similar. For me, it’s like a hobby, I really like to do this, so I think with a gap for the future, that’s why I bought this model of the Uni-t multimeter

1 hour ago, Coulomb said:

I'd use a current limited power supply set to about 3 A and measure the voltage drop

It will probably be difficult... I think it will be easier to solder this resistor from another board and see if anything changes. I'll try to do it tomorrow, because the sun has already set  

45 minutes ago, Abra said:

because the sun has already set

Your multimeter is solar powered? 🙃

Edit: of course, you ultimately want the fault code 52 gone, and for that you'll need solar power.

Edited December 27, 20223 yr by Coulomb

13 minutes ago, Coulomb said:

Your multimeter is solar powered? 🙃

 

If globally, then it is The multimeter has a battery that is charged from an outlet that receives solar energy

 

In any case, I need to wait for the sun to see if a resistor soldered from another board will work. I checked the resistor with a multimeter on a broken board, it shows ~ 0.08 ohm.

I studied all this a little bit and if I understood correctly, the problem could be, for example, in the PWM generator or driver, if I understood correctly, then this is U10 on the diagram.

As I understand the frequency may be correct, but the on-time may be less than it should be, because of this, the voltage may be incorrect? Or am I thinking wrong?

For example, this is how it should be:

 

 

And this is how it really is:

 

Edited December 27, 20223 yr by Abra

18 hours ago, Coulomb said:

R215 is a possibility

I put R215 from another board, the same...

10 hours ago, Abra said:

As I understand the frequency may be correct, but the on-time may be less than it should be, because of this, the voltage may be incorrect? Or am I thinking wrong?

Hi @Abra you are on the right path here , however the actual duty cycle , or pulse width of that signal will vary automatically according to the needs on the +12V rail .  The feedback provided by this +12V rail will determine just how much duty cycle will be required. When the 12v sense voltage drops somewhat due to load requirements, this lower feedback voltage will instruct U10 to slightly increase the width of the duty cycle , and as a result the +12V rail voltage will rise again.  Al  other rail feeds that is derived from the TX9 transformer is secondary to this cause. This means their output is not fed back for regulation , but the circuit is designed in such a manner  that these voltages (-12V,+5V,HFPWM) are within their design limits when the +12V rail is regulated  with normal load requirements.  

So the voltage of the 12V rail is always a clear indicator  whether the SMPS is performing correcly.

In my opinion , the accurate RMS content of the HFPWM feed is not  the important issue . It should just be in certain limits in order to provide adequate minimum power to its loads. As an example , the IGBT drive transformer Tx7 is carefully designed to produce 20V outputs on its secondaries  when the HFPWM is at its lowest . It does not depend on the accuracy of HFPWM , as it will regulate its dc output via 5V6 and 18V Zeners .  And I think that all other isolated power supply supply transformers that use the HFPWM works on the same basis.

So here is a question for you . As an example to prove my point above ,what are the positive going and negative going dc drive voltages of the IGBT transformer outputs, e.g. what are  the DC voltages that you read across ZD7,ZD8      ZD1,ZD2   ZD3,ZD4    , under the conditions where your RMS of HFPWM is only as 'low' as 10.2V ?   If you typically read 5.6V and 12.8v across those zeners , its an indication that the drives are operating within their specs , and that HFPWM is 'good enough'.

 

 

10 hours ago, Abra said:

As I understand the frequency may be correct, but the on-time may be less than it should be, because of this, the voltage may be incorrect?

I've done some more reading on flyback converters. If I understand correctly, the output should really be pretty close to a square wave, varying only because the voltage on the output capacitor will dip a little when not being charged, and at the beginning of the cycle the current is higher, so the voltage drop across the diode will be a little higher.

Intriguingly, the  duty cycle should be fixed given a fixed voltage ratio and transformer turns ratio. As the load current increases, the primary of the "transformer" has to ramp up for longer, to reach a higher peak current, so that there is enough average current in the secondary to supply the load. So the duty cycle should stay the same, but the frequency should decrease.

* Are you seeing differences in the frequency between the good and bad inverters?

* Are you testing the two inverters with the same power source?

Assuming that the output wave is close to square, it's interesting that the RMS secondary voltage before the diode is well over 12.0 V on the good inverter. That means that there is significant negative voltage at the output of the transformer when the input is fluxing up ("charging" the primary with current). I guess that makes sense : when the primary is fluxing, there is still standard transformer action happening, which would produce a little over -12 V at the secondary. You don't want the secondary winding to draw any current at that point, because this would take away from energy that should be going into the primary.

So what if there was current during fluxing? It seems to me that this would reduce the efficiency of the converter, but I don't see how it would affect the output voltage. Maybe the output voltage is weak during fluxing, such that even a low load during that time would result in lower negative voltage at the output, and hence lower RMS voltage at the secondary, while still producing 12 V during the discharge part of the cycle. The converter frequency would have to lower to supply the wasted power.

Anyway, this suggests checking the rectifier (D54) and its snubber components (R206 and C82).

Edit: This also suggests checking all the HFPW+ rectifiers, e.g. on the comms, parallel, and SCC boards. Perhaps unplugging them one at a time may reveal the culprit(s).

Edit 2: And that also includes the gate driver rectifiers that operate on HFPW+, like D40.

Edited December 28, 20223 yr by Coulomb

13 minutes ago, BritishRacingGreen said:

When the 12v sense voltage drops somewhat due to load requirements, this lower feedback voltage will instruct U10 to slightly increase the width of the duty cycle , and as a result the +12V rail voltage will rise again.

It's my understanding that U10 will increase the time that the transistor conducts, but it also has to increase the time that the transistor is off, because the inductance of the secondary winding is fixed, and the voltage is fixed, so dI/dt in the secondary is fixed, so to get higher average current into the secondary, it has to continue the current for longer. So both on and off times will increase proportionally to the current, meaning that the duty cycle will remain constant, and the frequency will go down as the load current goes up. I'd expect the frequency to decrease by 1% for every 1% increase in load current, ignoring losses.

6 minutes ago, Coulomb said:

It's my understanding that U10 will increase the time that the transistor conducts, but it also has to increase the time that the transistor is off, because the inductance of the secondary winding is fixed, and the voltage is fixed, so dI/dt in the secondary is fixed, so to get higher average current into the secondary, it has to continue the current for longer. So both on and off times will increase proportionally to the current, meaning that the duty cycle will remain constant, and the frequency will go down as the load current goes up. I'd expect the frequency to decrease by 1% for every 1% increase in load current, ignoring losses.

@Coulomb that is interesting observation . Unfortunately i dont have any MAX's on the bench any more as they have been repaired, but I seem to think that the PWM was fixed at 100khz and that the on time is the only variable.  The uc3845 has a cap on duty cycle at 50% , regardless of feedback . I do know that i measured a pulse width of as low  as 800ns in order to get 12.3VDC , at 100kz. That was without control card , and obviously other circuits disabled.

The circuit is unusual in that the VFB pin on UC3845  is not actually used for feedback , instead the COMP (compensation ) pin only is used. But I think that is a result of the finicky dynamic range and linear operating region of the feedback opto isolator. 

16 minutes ago, BritishRacingGreen said:

@Coulomb that is interesting observation . Unfortunately i dont have any MAX's on the bench any more as they have been repaired, but I seem to think that the PWM was fixed at 100khz and that the on time is the only variable.  The uc3845 has a cap on duty cycle at 50% , regardless of feedback . I do know that i measured a pulse width of as low  as 800ns in order to get 12.3VDC , at 100kz. That was without control card , and obviously other circuits disabled.

The circuit is unusual in that the VFB pin on UC3845  is not actually used for feedback , instead the COMP (compensation ) pin only is used. But I think that is a result of the finicky dynamic range and linear operating region of the feedback opto isolator. 

I can confirm that the one max with the broken feedback circuit , had the uc3845 at 50%  at 100khz . That energy had the power to raise the 12V rail to nearly 50VDC! 

In addition , when i fixed the feedback , the pulse width was 800ns in order to produce 12VDC , that was still at 100kz. below is an image i still have on my phone.

 

The HFPWM feed looks something like this :

 

By the way my Fluke 177 had dificulty with this signal . After being rectified by D20 to the IGBT transformer , its reads 0VDC ! I am still to investigate why it does that ?

 

Edited December 28, 20223 yr by BritishRacingGreen

26 minutes ago, BritishRacingGreen said:

But I think that is a result of the finicky dynamic range and linear operating region of the feedback opto isolator. 

Although the arrangement of TLC431 and opto isolator is widely regarded as the de-facto standard for isolated feedback for flyback converters .