BK Precision 9130 Triple Power Supply repair.

[Kean]

Potential candidate for 2U14 is something like AT89S51, S52, or C55 - these use an 8051 core. 
These match the pin layout you've drawn up for power, reset, and crystal.  No I2C on pins 12 & 13, but that could be handled via GPIOs in software.
These apparently require access to almost all GPIOs for flash programming, so it probably wasn't programmed in circuit.
A pic of mine is attached in case you can decipher anything from it.

[Kean]

OK, I traced out the circuit from pin 18 of J1 on the analog board.  It appears that bridge rectifier 1D28 is used for temperature sensing of the heatsink.  It has no connections on the AC pins.

Edit: Schematic isn't quite right - see update below by jchw4

[fzabkar]

1D28 is reverse biased, so how does it do anything? :-?

[Kean]

1D28 is reverse biased, so how does it do anything? :-?

I traced the circuit late at night, but pretty sure I got it right.  Hopefully jchw4 can double check.

Reverse bias leakage saturation current of a diode can certainly be used for temperature sensing although typically only used when power consumption is an issue, but it also has a benefit in being non linear.  We don't need much in the way of accuracy here, just want to know when cooling is needed.

https://thecavepearlproject.org/2019/11/04/single-diode-temperature-sensor-with-arduino-icu-via-reverse-bias-leakage/

Edit:
I apparently should have said reverse saturation current, not reverse leakage current.
See page 4 of https://www.kennethkuhn.com/students/ee351/diode_characteristics.pdf
Still this seems a strange technique to use in a potentially electrically noisy environment...

[jchw4]

1D28 is reverse biased, so how does it do anything? :-?

I traced the circuit late at night, but pretty sure I got it right.  Hopefully jchw4 can double check.

I actually put it in exactly the same orientation as you did. But then there is fzabkar's comment.... Which seems reasonable. I double-checked everything but found nothing.
And only then I looked at the labels on the bridge rectifier itself 


Look at the photo, "+" is on the right and the key is on the right too. And KiCad's model has "+" on the first pin 
Which means that pins are counted from the right 

As you can see on the photos, it traced to "GND" and +12V.


So this is what I got:



Here is my schematics:

[Kean]

Oops, I missed that GND trace dissapearing under the heatsink.  I used some backlighting on the PCB and still missed it, but I can see it now I know it is there!
I traced the 1R5 pull up to 1IC12 pin 8, which I was thinking was going to be 5V but I now realise I didn't verify that.   

Thanks for your work in drawing up the schematics.  Hopefully my hints for 2U14 will help.

[fzabkar]

That makes sense.

The tempco of each diode pair is -4mV/C (-2mV per diode), so you should see the diode voltage drop as you heat it up.

Edit:

Would it be worth shorting the + and - terminals of the bridge rectifier to simulate a high temperature condition? Would the firmware see this is as a genuine high temperature event, in which case it should turn on the fan at full RPM, or would it detect the temperature reading as being out of bounds, ie a sensor fault?

According to the manual ...

https://bkpmedia.s3.amazonaws.com/downloads/manuals/en-us/9130_manual.pdf

... "If the internal temperature of the power supply exceeds 80 °C, the instrument will protect itself by automatically turning power OFF. When this happens you will hear a buzzer and the display will indicate the following: Over Temp".

[jchw4]

Hopefully my hints for 2U14 will help.

Sorry, I did not have time to dig into it. Let's solve the fan issue first 

[jchw4]

I don't understand how this circuit is supposed to work.
(I verified that pulling up base of 2Q2 to +5V turns on the fan though.)

2U8 has no way of talking to 2U14 except through the pin 10.
(It has 3 incoming digital lines thath seem to be CS/CLK/DATA but all of them come from the outputs of the level shifter 2U13 (SN74LVC07A), so these cannot be used for output).
Upd: this might not be true, see next message.

2Q2 controls the fan. It is attached to the pin 10 of the 2U8 (MSP430F1101A, https://www.ti.com/lit/ds/symlink/msp430f1101a.pdf)  which can be used as comparator output.
So it's reasonable to assume that it is indeed wired to the comparator output.

MSP430 has two comparator inputs: 11,12 which both could also be used for GPIO.
Pin 13 could be either auxilary clock output or GPIO.

Comparator is explained in the chaptor 16 of the user guide https://www.ti.com/lit/ug/slau049f/slau049f.pdf .
It can compare either its two external inputs, or compare one of them to the internal reference (selectable [0.25 x Vcc], [0.5 x Vcc], or "or a fixed transistor threshold voltage of ~ 0.55 V").
Vcc in our case is 2.5V, so it should be 0.625V, 1.25V, and 0.55V.

I tried to simulate the circuit for the temperatures 25C and 85C as two possible edge cases:


For all pins 11,12,13 input voltages decreases when temperature rises.
Threshold 0.5 x Vcc = 1.25V is too large for any simulated input.
Threshold 0.25 x Vcc = 625mV is barely crossed by pin 13 input around 25C (which is not a comparator input anyway) and pin 12 input, but now close to 85C.
I think this does not work, because by the time this sensor reaches 85C other parts will be cooked.

Another question is "why make that complicated circuit to start the fan only at about 80C"?

MSP430 manual suggests doing "precision slope analog-to-digital conversions".
There is actually a ceramic capacitor 2C31 which potentially could be used for that purpose.

Playing with random pulses in the LTspice I could make it cross 550mV threshold at about 35C:


I used 10uF capacitor for simulation, but I later measured it in-circuit 100nF. (by DE5000 at 100kHz)

It is also possible to make it work with the 100nF cap:


But it looks weird. Pulse width 0.0002ms suggests system clock at least 5Mhz, but this MC has only 4.096MHz crystal.

Also I ccould not see any change in behavior when  setting pin 13 to either 2.5V or 0V. It does not seem to affect anything.

Ok. I powered the device that I am repairing and measured:
- Pin 11: 816mV
- Pin 12: 685mV
- Pin 13: 570mV

which is very close to the static simulation above of 900mV, 755mV, 624mV.
I did no see any activity on these pins.


Could somebody help me understand how this sensor circuit is supposed to work?

[jchw4]

Edit:

Would it be worth shorting the + and - terminals of the bridge rectifier to simulate a high temperature condition? Would the firmware see this is as a genuine high temperature event, in which case it should turn on the fan at full RPM, or would it detect the temperature reading as being out of bounds, ie a sensor fault?

According to the manual ...

https://bkpmedia.s3.amazonaws.com/downloads/manuals/en-us/9130_manual.pdf

... "If the internal temperature of the power supply exceeds 80 °C, the instrument will protect itself by automatically turning power OFF. When this happens you will hear a buzzer and the display will indicate the following: Over Temp".

This is a very nice idea! Yes, it beeps and starts the fan

Hm, I wonder how it detects simple "start the fan" vs "overheating, beep"? I am probably missing something in my tracing, because I thought that MSP430 has only one way of sending data to the CPU and its via 2Q2. Which only starts the fan but does not beep....

[Kean]

I won't have time to look at this again for the next few days, but another thing you can try is shorting out just one of the diodes instead of both, i.e. one AC pin to one DC pin instead of both DC pins.

[fzabkar]

I'm wondering about the function of 2Q1.

When 2Q1 is switched on, the voltage at pin 12 is 685mV. When it is switched off, the voltage rises to 695mV. I thought this may be some kind of hysteresis, but that doesn't make sense because hysteresis could easily be implemented in the firmware, could it not? It might be instructive to see whether 2Q1 switches off after the diodes are shorted.

What is your ambient temperature?

As I see it, the tempco at the diodes is -4mV/C, so the tempco at pins 11 and 12 would be -3mV/C and -2.5mV/C, respectively.

You are currently seeing 685mV at pin 12. Your two reference voltages are 550mV and 625mV. The differences are 135mV and 60mV. These represent temperature differentials of 45C and 20C above ambient.

Edit:

I see some sources which state that the tempco at low currents can be typically -2.4mV/K.

[jchw4]

I attached my HP 66311B in parallel with the diode. This old HP can sink current, which is very handy here.

So when device boots at my ambient 26C, sensor output sits at about 1.116V.
Pin voltages:
  - 2U8(11) 832mV
  - 2U8(12) 699mV
  - 2U8(13) 580mV

Fan starts to spin (PWM 25%) when sensor output is about ~982mV.
Fan goes full speed when thansitioning from 953.6mV to 949.5mV (the discrete step of my 66311).
It beeps when sensor reaches ~805mV.

[jchw4]

To really investigate how it works I had to go find a real scope

The first screenshot shows the fan start sequence:


Zooming in to the beginning:


The right part (fan still runs at 25% PWM):


Pin 13 seems to have software control:


When pin 12 fully goes under 650mV threshold fan goes full speed:


Overheating signal:


Overheating still puzzles me. Screenshots have all the noise filtered out, so it probably detects overheating when comparator "significantly agrees that PIN 11 is below Vcc x 0.25.

[jchw4]

I attached thermocouple next to the temperature sensor and loaded the device with 90W.

Ambient 27C. While device was heating I wrote down some measurements:

Sensor temperature 29.8C gives 1.1425V sensor voltage.
33C 1.132V
34C 1.129V
35C 1.125V
36C 1.121V
38C 1.111V - Stable tmperature (Here I switched the load off for some time and waited for the temperature to stabilize.)
40C 1.104V
40.6C 1.100V - Stable temperature
45.3C 1.082V - Stable temperature
50.1C 1.062V - Stable temperature

I did not risk heating it more because my thermocouple does not have good thermal connection to the heatsink.

So it seems to match the simulation above. Which means that all the parts function correctly.
I still don't understand the designers intention for chosing these thresholds.

[jchw4]

Now we need to do something to make it work.

Simulation says that changing 2R10 to something about 1.8k should bring down start temperature to about 31C.
It would be nice to add some resistor in parallel instead of replacing this SMD resistor.
(There is a very handy unpopulated 2C32 in parallel to the 2R10 that could be replaced with a resistor.)

Resulting 1.8k would require 5.4k. I had 6.8k and tried it.

And it did not work. Double-checking eberything revealed that Pin 11 is very close to Vcc.
In fact Pin 11 to Vcc measures 10Ohm for me now (when power is shut off). So I probably killed this pin somehow. 
Still don't know how I did it.

Ok, I cut the trace to this pin and it is alive!

Putting load on the device showed that fan started at about 27C (thermocouple losely attached to the heatsink next to the sensor), and it went full speed at 31C.
Putting 176W load on the device thermocouple temperature stabilized at 49.6 (26.1 ambient). Which is probably good enough.

Adding my 66311B as a sensor revealed that "overheating" never occurs, so it is indeed Pin 11 resiponsible for this

[jchw4]

Well, I don't really know what to do next. On the one hand it works fine and probably will never overheat.
On the other hand it would be nice to fully restore it.

Maybe I should check whether the two other MSP430 controllers have this feature, so I could just exchange the fautly MSP with the one from another channel.
But this will require good soldering tools that I don't have.

It would also be nice to dump the firmware and replace this chip with the new one.

Both of these options are probably too much work for me now, so I will probably leave it as it is.

[fzabkar]

I was going to suggest connecting a pot across the diodes to vary the voltage. It probably would have been a safer approach. :-(

[Kean]

Thanks for all the research into the fan control jchw4.  I'll try a similar mod on the fan temperature control.

Apart from this PSU being a pain to work from, the big problem people seem to have with it is the caps prematurely failing - presumably a factor of poor cooling.
Now I just need to work out why I now get mixed readout between channels on mine 

[jchw4]

I was going to suggest connecting a pot across the diodes to vary the voltage. It probably would have been a safer approach. :-(

The MSP430  input is protected by 1k resistor from the sensor. So I doubt it was an issue. I think I probably shorted the test wire that I soldered to the pin11 itself after I disconnected the scope.
But I did not notice anything.
Well, it happens 

[jchw4]

Thanks for all the research into the fan control jchw4.  I'll try a similar mod on the fan temperature control.

Apart from this PSU being a pain to work from, the big problem people seem to have with it is the caps prematurely failing - presumably a factor of poor cooling.
Now I just need to work out why I now get mixed readout between channels on mine 

I cannot promice to make a schematics of it, but it is connected like this:

ADC 2U5, 2U6, 2U7 are connected to the main IC 2U14. They share the same serial bus.

ADC pins are connected like this:
(5) - Serial clock in (shared)   
(6) - Serial data out (shared)
(7) - ~CS (not shared)
(19) - ~RESET (shared)
(20) - ~INTERRUPT out (shared)
(23) - Serial data in (shared)

2U13 controls communications with all MSP430 and RS232 TX.
2U14 controls communications with ADCs.

So in your case something could be broken in these CS lines.

I did not trace it too far (as this part works for me), but I hope my tracing image will help you.

[Kean]

So in your case something could be broken in these CS lines.

I did not trace it too far (as this part works for me), but I hope my tracing image will help you.

Yes, maybe a damaged CS signal for channel 2.  Great suggestion!  I hadn't really studied how it was all connected, so that info is much appreciated.