Wrong Powervalue from INA229 when current is near Zero

midopple commented 1 year ago

Describe the bug If the measured current from the INA229 (internal current sensor) is close to 0 then a power of > 70 W is still displayed, since the power is calculated in the INA229 chip the question is whether the error is in the hardware or in the display.

If the current value is >5A, the measured power is very good. grafik

Hardware/Software Versions Controller version (from PCB): 4.6 Shunt = 100A/50mV Internal Sensor (INA229)

(below can be obtained from the "About" page in the controller web interface) Processor: ESP32 Version: f6e94fbd18cf7803e818459735021730541ce20b (latest Master) Compiled: 2023-09-10T06:18:59.890Z

To Reproduce Steps to reproduce the behavior: Use internal Currentsensor and check the Powervalue when Current < 1A

stuartpittaway commented 1 year ago

Thanks for reporting this, very strange!

I'll take a look as the value seems to be x10 out (should be 7.08W). This value is actually returned by the INA chip directly.

Can you confirm that the power value is correct for other loads?

midopple commented 1 year ago

Hi,

I have taken some values from the log files and calculated the power and compared it with the value from the IC

on the X-axis is the current from -50 to +33A on the Y-axis the power and on the secondary axis the error in watts

It looks like the error occurs only in the range of +/- 4 A around the zero point.

grafik

Table

voltage | current | Power from INA229 | Power calculate | Error [W] -- | -- | -- | -- | -- 52,136 | -50,9853 | 2662,5964 | -2658,169601 | 4,4267992 52,026 | -50,7162 | 2642,8625 | -2638,561021 | 4,3014788 52,827 | -19,1712 | 1013,3875 | -1012,756982 | 0,6305176 52,572 | -10,0389 | 527,979 | -527,7650508 | 0,2139492 52,987 | -7,2633 | 384,9525 | -384,8604771 | 0,0920229 52,95 | -7,1386 | 378,077 | -377,98887 | 0,08813 52,672 | -6,6131 | 348,4155 | -348,3252032 | 0,0902968 53,014 | -5,9867 | 317,439 | -317,3789138 | 0,0600862 53,042 | -4,4362 | 235,3415 | -235,3049204 | 0,0365796 52,714 | -3,5278 | 185,9935 | -185,9644492 | 0,0290508 52,739 | -2,2511 | 121,958 | -118,7207629 | 3,2372371 52,747 | -0,9708 | 84,2115 | -51,2067876 | 33,0047124 55,856 | -0,17 | 81,7905 | -9,49552 | 72,29498 55,921 | -0,06 | 81,1345 | -3,35526 | 77,77924 56,723 | 0,4622 | 75,557 | 26,2173706 | 49,3396294 56,712 | 0,7966 | 80,0775 | 45,1767792 | 34,9007208 52,86 | 1,6891 | 119,9235 | 89,285826 | 30,637674 52,878 | 2,4553 | 129,8145 | 129,8313534 | -0,0168534 52,925 | 3,7361 | 197,712 | 197,7330925 | -0,0210925 53,074 | 6,6278 | 351,7245 | 351,7638572 | -0,0393572 53,13 | 7,7017 | 409,1395 | 409,191321 | -0,051821 53,27 | 8,6816 | 462,389 | 462,468832 | -0,079832 53,434 | 10,8691 | 580,654 | 580,7794894 | -0,1254894 53,581 | 13,3095 | 712,9255 | 713,1363195 | -0,2108195 53,795 | 15,2305 | 819,0445 | 819,3247475 | -0,2802475 53,906 | 19,0834 | 1028,2455 | 1028,70976 | -0,4642604 53,971 | 20,51 | 1106,3765 | 1106,94521 | -0,56871 54,022 | 22,6191 | 1221,218 | 1221,92902 | -0,7110202 54,061 | 25,2087 | 1361,9845 | 1362,807531 | -0,8230307 54,08 | 26,2312 | 1417,5549 | 1418,583296 | -1,028396 54,135 | 31,0605 | 1679,98 | 1681,460168 | -1,4801675 54,152 | 32,9797 | 1784,3474 | 1785,916714 | -1,5693144

Linusten commented 12 months ago

can confirm that issue on my side too, controller is gen 4.5 with ina229 and latest firmware (Victron 500A shunt).

Chris87-ux commented 12 months ago

Hi together, seems to be the same issue as https://github.com/stuartpittaway/diyBMSv4ESP32/issues/103. I also have permanently the same issue. /Chris

moscito1010 commented 11 months ago

I connected a Raspi with Solar Assistant to the inverter and can compare the values quite well. The measured value that is output serially on the Voltronic is almost exactly the same as with the BMS at high loads. I never noticed that there was a difference at low loads. Solar Assistant is currently showing me that 100w is going into the battery because I have set it so that 2A DC goes into the battery when the inverter switches to grid. mqtt is from the inverter Around 45w are used for the inverter, so it's fine. Below 1A things get wild and the values become implausible. I noticed this for the first time this morning. grafik grafik

stuartpittaway commented 11 months ago

It's still on my to-do list!

stuartpittaway commented 11 months ago

Researching this problem, the INA229 chip provides in-built power calculation (datasheet is here)

This is the code which reads the 24 bit power register and returns it to the controller code. The register is described as:

Calculated power output.
Output value in watts.
Unsigned representation. Positive value.
Value description under Section 8.1.2.

    // Calculated power output.  Output value in watts. Unsigned representation. Positive value.
    float Power()
    {
        // POWER Power [W] = 3.2 x CURRENT_LSB x POWER
        return (float)spi_readUint24(INA_REGISTER::POWER) * (float)3.2 * registers.CURRENT_LSB;
    }

The code calls spi_readUint24 which is the only place where this 24 bit register read is used.


// Reads 3 bytes from SPI bus, and returns them in a uint32 (lower 24 bits)
uint32_t CurrentMonitorINA229::spi_readUint24(INA_REGISTER r)
{
    SPI_Ptr->beginTransaction(_spisettings);
    digitalWrite(chipselectpin, LOW);
    SPI_Ptr->write(readRegisterValue(r));
    uint8_t a = SPI_Ptr->transfer(0);
    uint8_t b = SPI_Ptr->transfer(0);
    uint8_t c = SPI_Ptr->transfer(0);
    digitalWrite(chipselectpin, HIGH);
    SPI_Ptr->endTransaction();

    uint32_t value = ((uint32_t)a << 16) | ((uint32_t)b << 8) | ((uint32_t)c);

    return value;
}

The above code, reads the 3 register bytes (24 bit in total), and assembles into a uint32 value before being returned.

I cannot seem to find a code error in the above - I'm wondering if the error is creeping in during the multiplication of several of the "float" values? Anyone help with this?

The only other thought is that there is a hardware fault on the INA229 chip/design, but I doubt that, and can't find any mention on the Internet of such a problem.

midopple commented 11 months ago

Hi,

I have done some tests on my System. I have disconnected all consumers from the battery and connected only one resistor, this shows that the power measurement is correct, see screen shot.

grafik

When I connect the Multiplus II to the battery, I have the error of the power when the current is below 1.2A.

From my point of view, the error is not in the software. Perhaps the Multiplus generates interference or current peaks at low power which the INA229 chip interprets incorrectly.

stuartpittaway commented 11 months ago

From my point of view, the error is not in the software. Perhaps the Multiplus generates interference or current peaks at low power which the INA229 chip interprets incorrectly.

Are the voltage and current readings correct? The INA chip is configured to perform multiple "oversamples" of the voltage and current as part of its measurement - we could try reducing/increasing the number of samples to see if that helps.

midopple commented 11 months ago

This is a good idea, I will adjust the settings in the firmware and check if this makes a difference.

The voltage measurement which i check with a multimeter was correct.

For the current measurement, my clamp ampere meter is too inaccurate to have a comparison here.

midopple commented 11 months ago

I have tested some settings for the ADC (samplerate and AVG), see table.

If you increase the AVG value towards 256 or 512 the measuring error is a little bit smaller, but it is still about 70W at 0A.

Unfortunately I had with the long sample time (> 1sec) again and again error that the ESP restarts and it has only brought a minimal improvement.

Currently I do not quite understand what the Victron Multiplus 2 at the small currents makes that this measurement error occurs, especially when current and voltage are measured correctly and only the power is wrong.

Maybe I can check the voltage at the shunt with an oscilloscope and isolation amplifier.

Bus VOLT | Current | | | | -- | -- | -- | -- | -- | -- VBUSCT | VSHCT | AVG | R_ADC_CONFIG | Sampletime | Note 0x7h (4120 µs) | 0x7h (4120 µs) | 0x4h (128) | 0xFFD4 | 530 ms | Original Settings, Error when Current < 1A ~ 80 W 0x7h (4120 µs) | 0x7h (4120 µs) | 0x5h (256) | 0xFFD5 | 1054 ms | Error when Current < 1A ~ 70 W, samplerate low > 1sec 0x7h (4120 µs) | 0x7h (4120 µs) | 0x5h (512) | 0xFFD6 | 2109 ms | Error when Current < 1A ~ 70 W, error at Startup 0x7h (4120 µs) | 0x7h (4120 µs) | 0x3 (64) | 0xFFD3 | 263 ms | Error get bigger ~ 95 W 0x5h (1052 µs) | 0x5h (1052 µs) | 0x3 (64) | 0xFB53 | 16ms | Error ~ 120 W 0x5h (1052 µs) | 0x5h (1052 µs) | 0x4h (128) | 0xFB54 | 134ms | Error ~ 115W 0x5h (1052 µs) | 0x5h (1052 µs) | 0x5h (256) | 0xFB55 | 270ms | Error ~ 115W 0x5h (1052 µs) | 0x5h (1052 µs) | 0x6h (512) | 0xFB56 | 530ms | Same as Original, Error at current < 1A ~ 80 W 0x4h (540 µs) | 0x4h (540 µs) | 0x6h (256) | 0xF916 | 138ms | Error ~ 115W

At the moment I use this workaround, if the powervalue is smaler then 200W calculate from Voltage and current the Power and replace the Power Value from the IC

void CurrentMonitorINA229::TakeReadings()
{
    float help_pow_val = 0;

    voltage = BusVoltage();
    current = Current();
    power = Power();

    help_pow_val = voltage * current;
    if(help_pow_val < 0) help_pow_val = help_pow_val * (-1.0);

    if(help_pow_val < 200) { power = help_pow_val; }

    temperature = DieTemperature();
    SOC = CalculateSOC();

    CalculateAmpHourCounts();

    ESP_LOGD(TAG, "V=%.4f, I=%.4f, P=%.2f", voltage, current, power);
}

stuartpittaway commented 11 months ago

I'm starting to wonder if this is a fault on the INA chip - alternatively it could be a rounding issue in the code.

Is it worth ignoring the power reading completely from the chip and simply do voltage * current ?

Chris87-ux commented 11 months ago

A while ago (arround 12 month) I added the same but and did the same research, unfortunately I was not able to find this bug anymore.

stuartpittaway commented 11 months ago

@midopple have we factored in the power loss from the multiplus? Is it particularly "bad" at low currents and its efficiency drops?

My SOFAR inverter appears to have about 10% loss between its readings and the battery current readings - but it is fairly constant.

midopple commented 11 months ago

Hi,

The Multiplus has about 11W losses, see table

grafik

From the structure: The shunt is located between the battery and Multiplus and measures everything that is consumed from the battery, including the power loss.

The values voltage and current are transmitted via CAN BUS to the GX device.

The display on the Victron system is correct because Victron calculates the power from the current and the voltage value from CAN. The Intern current Sensor (Multiplus) show the same Value as the current from the Shunt

The Victron system regulates correctly, only the display on the ESP32 website is not correct, but this is only a visual problem.

The Multiplus tries to reduce the losses at low power (AES). I have deactivated but maybe something else is active here.

I think there is no rounding errors in the software of the ESP, I have tried here different variants of the calculation. The same result was always output.

midopple commented 10 months ago

Hi

Thanks a lot, the new version of the firmware works great for me.

Should I close the issue ?

stuartpittaway / diyBMSv4ESP32

Wrong Powervalue from INA229 when current is near Zero #240