SG3525 PWM controller OCP cell

[ym58]

There appears to be a band after the black?

Silver.

It's very unlikely that a faulty resistor would drop from 33Ω to 3.3Ω.

3.7Ω not 3.3Ω !
Unlikely but not impossible, apparently 

[xavier60]

This calculator says it's 3.3Ω for a 5 band resistor.

https://www.calculator.net/resistor-calculator.html?bandnum=5&band1=orange&band2=orange&band3=black&multiplier=silver&tolerance=brown&temperatureCoefficient=brown&type=c&x=56&y=20

[ym58]

This calculator says it's 3.3Ω for a 5 band resistor.
https://www.calculator.net/resistor-calculator.html?bandnum=5&band1=orange&band2=orange&band3=black&multiplier=silver&tolerance=brown&temperatureCoefficient=brown&type=c&x=56&y=20

Damned, you were right !
How come Meishile has chosen such low values for R7 whereas they could have changed the winding turns of T3 instead !
With such a low value, how can you possibly bias Q1 with precision and hence adjust the current limit precisely ?
To me the current protection cell made with T3 + rectifier + R7 (3.3R) + Q1 + R13 + R8+ R2 is a piece of s...
---
By the way, I just cooked my two last Mosfets, trying different values for R7.
I ordered 10 pieces in China but I will have to postpone my further tests.
That's not my day ...

[xavier60]

What were the operating conditions when the MOSFETs failed? Id expect that the charge current can be increased gradually by starting at a lower output voltage.
When MOSFETs fail violently, it's common for the Gate resistors and other drive components to be damaged which if isnt fixed will result in a short life for the replacement MOSFETs.
Even if the current limiting isnt working, no damage should occur until the PSU is actually overloaded.

[ym58]

Actually, I assumed (mistakenly, now we know !) that R7 was a 33R at the first place.
So to decrease the current limit, I decided to increase the R7 value and I placed a 100R variable resistor in series with it.
The initial test value was about 100R (33R + 77R) and the current limit decreased noticeably : an inverter connected to the PSU with a 25A input load was still working properly, but no longer worked on a 50A input load (using water heating resistances, one is 250-300W, the other one 500W).
Good point, I thought, but I don't know exactly where the current limit is actually triggering, so I decided to lower R7 again but 'ON-LINE' and connected a 173Ah/12V LFP4 pack load in place of the inverter.
The charge current settled to 20-25A or so (way better than the 100A++ that occured yesterday and which led to cook the Mosfets) but increasing the 100R variable resistor did not show any variation of the ouput current to the LFP4 pack ... strangely enough 
So at some point, I decided to remove the 33R resistor and left the 100R variable alone, but I must have screwed up with which way I turned the trimpot (multiturn) and the output current built up again and the MOSFETs fried ... 
Unfortunately, I had forgotten to place two fuses in the sources of the Mosfets, maybe it'd have prevented the mishap (should the fuses be fast enough ...)

[xavier60]

It's all getting messy now, too many unknowns.
With R7 at 100Ω. the current limit should have been a few amps.
If I was working on it, I would be using my DSO to get a clear understanding of what's happening.
Also Id be checking that the Gate drives are  good. Those are very rugged MOSFETs capable of handling very high currents.
Poor drive usually results in MOSFETs getting hotter than expected.
The  signal on R7 has such a low impedance, I dont understand why Q1 is needed.

[ym58]

Also Id be checking that the Gate drives are  good. Those are very rugged MOSFETs capable of handling very high currents.
Poor drive usually results in MOSFETs getting hotter than expected.

No real drives PER SE.
The SG3525 outputA (11) and outputB (12) drive the gates of the Mosfets thru a set of diodes and a small dual transformer T1, then two 220R resistors and that's it !
I will continue to reverse-engineer the PSU but it's a bit tough coz all the PWM and feedback section stand on a quite small piggy-back board.
There is still another circuit (8327B, on the left of the pic) that I have to investigate on as I don't know what it does ...

[Njk]

Good idea ... done !

Well, the theory is quite simple. When the SMPS is operating normally (no overcurrent), T20 and T21 are both closed (not conducting current), so the "dis" output signal is not asserted (is at GND level). In the case of overcurrent, the current sensor voltage at "lim" input became high enough to open T21. It opens and that, in turn, causes T20 to open. Because of that, the "dis" output became asserted (5.1V) causing the SMPS to shut down. Meanwhile, C20 starts charging and that keeps T21 opened even if "lim" signal is no more active. So the C20 charging time determines the pause time. When the charging has finished, T21 closes, T20 closes, "dis" de-asserts enabling the SMPS to start again. Meanwhile C20 starts discharging. D20 speed-ups the discharging to make the OCP cell ready for the next cycle more quickly

[mikerj]

3.7R means the resistor is FAULTY for it should be a 33R (orange-orange-black) !
It means that the current measurement was definitely underestimated ... I will replace it with a valid 33R first then increase it step by step until I get my targeted lower current limit ...
I will re-post to report the testings.

Orange, orange, black, silver, gold is a 3.3 Ohm resistor.

[ym58]

Well, the theory is quite simple. When the SMPS is operating normally (no overcurrent), T20 and T21 are both closed (not conducting current), so the "dis" output signal is not asserted (is at GND level). In the case of overcurrent, the current sensor voltage at "lim" input became high enough to open T21. It opens and that, in turn, causes T20 to open. Because of that, the "dis" output became asserted (5.1V) causing the SMPS to shut down. Meanwhile, C20 starts charging and that keeps T21 opened even if "lim" signal is no more active. So the C20 charging time determines the pause time. When the charging has finished, T21 closes, T20 closes, "dis" de-asserts enabling the SMPS to start again. Meanwhile C20 starts discharging. D20 speed-ups the discharging to make the OCP cell ready for the next cycle more quickly

Wonderful, thanks 
It's now crystal clear for me !
Just one point though : what's C21 for, is it just there for decoupling HF spikes ?

[ym58]

Orange, orange, black, silver, gold is a 3.3 Ohm resistor.

Yeah, my mistake ... that's what I realized, afterwards.

[ym58]

It's all getting messy now, too many unknowns.

You bet !

The  signal on R7 has such a low impedance, I dont understand why Q1 is needed.

Maybe the trick is NOT to play with the R7 value to move the current limit threshold but with R8 and R2, hence why there is Q1.
You're right, I should use my scope & watch exactly what's going on at current limit.
The point is to be able to test the cell on several different high-current loads and this I don't have much : only two water heater resistors and nowadays, there's no more high-watt filament bulbs out there except maybe in the deep countrysides of Irak, Burma or Brazil (no offense intended for this guys ) !

[ym58]

Those are very rugged MOSFETs capable of handling very high currents.

[xavier60]

If you can adjust the output voltage? you can control the battery charge current.
Needing a mains isolation transformer, the low-side MOSFETs Gate waveforms should be checked first. Because the design uses no drive
buffering, the waveform is going to be normally messy.

[ym58]

If you can adjust the output voltage? you can control the battery charge current.

During the CC phase (constant current), I leave the battery dictate the PSU its output voltage and the PSU adjusts consequently (no choice !) while providing the battery the constant current that it's been set for (hence this thread !).
I never thought that I could **still** adjust (slightly) the voltage during this phase to modify the current, I need to test that as it's not a normal procedure for LFP4 technology.
On top of that, voltage range on LFP4 batteries is quite narrow (mostly from 12V min to 14.6V max, no linearity !).

Needing a mains isolation transformer, the low-side MOSFETs Gate waveforms should be checked first. Because the design uses no drive
buffering, the waveform is going to be normally messy.

I don't get it, why do I need a mains transformer ?
Can you clarify ?

[xavier60]

If you can adjust the output voltage? you can control the battery charge current.

During the CC phase (constant current), I leave the battery dictate the PSU its output voltage and the PSU adjusts consequently (no choice !) while providing the battery the constant current that it's been set for (hence this thread !).
I never thought that I could **still** adjust (slightly) the voltage during this phase to modify the current, I need to test that as it's not a normal procedure for LFP4 technology.
On top of that, voltage range on LFP4 batteries is quite narrow (mostly from 12V min to 14.6V max, no linearity !).

Needing a mains isolation transformer, the low-side MOSFETs Gate waveforms should be checked first. Because the design uses no drive
buffering, the waveform is going to be normally messy.

I don't get it, why do I need a mains transformer ?
Can you clarify ?

The suggestion of adjusting the voltage to vary the charge current is for testing reasons.
The PSU needs to be powered via an isolation transformer to be able to take oscilloscope readings on the live side.

[ym58]

The suggestion of adjusting the voltage to vary the charge current is for testing reasons.

Yes sure and that's what I had understood but, still, I have to make sure the BMS will like it or not !
Otherwise, I will have to disconnect it.

The PSU needs to be powered via an isolation transformer to be able to take oscilloscope readings on the live side.

Actually, since there is a fourth transformer which does not appear (yet) on my schematics, I have to wait until my reverse eng-ing the board is completed before clarifying how the FET gates drive is really handled ...

[Njk]

Just one point though : what's C21 for, is it just there for decoupling HF spikes ?

A sort of. "lim" input signal is of square wave. C21 shapes it to a triangle form to prevent false triggering at high (yet acceptable) output current levels

[ym58]

[ym58]

Well, the theory is quite simple. When the SMPS is operating normally (no overcurrent), T20 and T21 are both closed (not conducting current), so the "dis" output signal is not asserted (is at GND level). In the case of overcurrent, the current sensor voltage at "lim" input became high enough to open T21. It opens and that, in turn, causes T20 to open. Because of that, the "dis" output became asserted (5.1V) causing the SMPS to shut down. Meanwhile, C20 starts charging and that keeps T21 opened even if "lim" signal is no more active. So the C20 charging time determines the pause time. When the charging has finished, T21 closes, T20 closes, "dis" de-asserts enabling the SMPS to start again. Meanwhile C20 starts discharging. D20 speed-ups the discharging to make the OCP cell ready for the next cycle more quickly

@Njk
So, if I understood you well, that cell is more an 'OCP' cell than a 'CURR LIMITING' cell, is that right ?
By reducing C20, do you think I could make the controller recover FAST from the OCP state, at least fast enough to mimick a 'CURR LIMITING' behaviour that is to keep the current constant at a certain value (dictated by the choice of R9, e.g.) ?
Actually, my point is more to be able to control and adjust the CURRENT LIMIT than the OCP ...
(you will find the final PSU diagram in my previous post)