HLv2 Beta3 current sensor test result

[Takashi K]

Hi Steve and all,

I have checked HLv2 beta3 current sensor. please refer to the attached pdf.

I used my homebrew constant current load.

I think that ADC raw data is more useful than Quisk readout.

If you want more data, please ask me. After soldering T3 transformer, I will never measure basically.

73, Taka  ji1udd

[in3otd]

Thanks for the measurements Taka,
the results look quite good for currents above 50 mA. Below, there a kind of offset which actually does not seem constant so we cannot simply subtract it.
Out of curiosity, did you check if the current sensor output offset is lower if you add a pull-down resistor at its output (let's say 10 kohm or less)? The datasheet seems to imply that the device output is nonlinear below 50 mV.

73 de Claudio, IN3OTD / DK1CG

[Takashi K]

Hi Claudio,

> Below, there a kind of offset which actually does not seem constant

The cause seems LMV324 offset.

While receiving,INA199 output is 0.045V to 0.056V, a little bit unstable. Quisk shows 35mA.

When add 3.3k ohm pull down resistor at INA199 output, INA199 output becomes 0V. But Quisk shows 25mA.

Remove 3.3k ohm pull down resistor. ADC AIN2 is 0.067V.

When add 100 ohm pull down resistor at AIN2, AIN2 becomes 0.004V. Quisk shows 2mA.

As Steve said before, we need accuracy around 100mA  and 200mA.

I think we need to fit Quisk readout to real current in this range and can ignore below 50mA.

73, Taka  ji1udd

[in3otd]

Hello,
I've also done some test on the current sensing circuitry and while the results at (relatively) low currents match Taka's measurements I saw something not so nice at high current. Above about 1.25 A the output of the LVM324 becomes nonlinear and saturates to 3.3V, even if the input from the current sensor is fine:

(measurement was done with an electronic load in place of the PA)
At first I thought that the LMV324 could be a counterfeit, even if it looks identical from the ones I bought from Digikey for the H-Lv2b2, but then I saw this comment about a very similar issue. Now I'm pretty sure the opamp is fine and just doing what it's supposed to do... looking closely at the datasheet it's clearly specified that the input common mode goes up to about Vcc-0.8 V - I've completely overlooked this before. So the opamp has rail-to-rail inputs and output but the common mode is limited.
I'm a bit puzzled since both Steve and Taka reported PA currents readings from Quisk around 1.4 A but here the readings jump quickly at around 1.6 A for actual currents above 1.25 A.

Note also that the opamp adds some small offset, visible at low currents, as pointed out already. I think we could get rid of the buffer completely and feed R108 directly from Vipa.

Same issue would affect the temperature sensing but the output for reasonable temperatures (< 200 C)is still within the allowed common mode so in practice this is not a problem; still, we could get rid of the opamp buffer also for this path.

73 de Claudio, IN3OTD / DK1CG

[Takashi K]

Hi Claudio,

I checked the linearity of current sensor up to 1A. But I didn't checked over 1A range because my current load equipment was overloaded beyond that.

Since I felt that the zero offset of ADC was not so big, direct connection seems to be good.

73, Taka  ji1udd

[Steve Haynal]

Hi Claudio,

We can add a modification to remove the opamp from the current and temperature paths on the beta3. For the next version, I can switch to a 2 channel op amp for the FWD and REV power. I was thinking of doing that for beta3 but was too lazy... I am a bit worried as the FWD and REV paths have never really been tested. We need a proper filter board to test with.

73,

Steve

KF7O

[in3otd]

Hello,
yes, I've modified my H-Lv2b3 by lifting the U18 pin 8 and to connect the nearby Vipa trace directly to the pad there, now the reading is fine up to the 1.5 A I've tested and there is also less offset, the reading is good from about 15 mA (even if we do not actually need to read so low).

I was also thinking at the FWD and REW power paths, is there a "recommended circuit" to connect there and try to see if everything works well?

73 de Claudio, IN3OTD / DK1CG

[Takashi K]

Hi Steve and Claudio,

Is your modification (next version)  filters (1k ohms + 0.1uF) remain ?

73, Taka  ji1udd

[Steve Haynal]

Hi Taka,

I haven't yet made this modification, but will try to leave the filters in place.

73,

Steve

KF7O

[Steve Haynal]

Hi Claudio,

The FWD and REV power circuits were lifted from the Alex design. See the Alex Manual. Schematics are in an appendix at the end of the document. In the past we discussed using a single binocular core for T1 and T2 in that schematic. That seemed like a good space saving way to go to me.

73,

Steve

KF7O

[Takashi K]

Hi Steve,

OK, I have modified my Beta3, only bypassed Op-Amp.

Hi Claudio,

 I got the following results.

While receiving, ADC 3pin is 0.055V. Quisk shows 29mA - 30 mA. OpAmp 8pin is 0.074V (not used now).

While spot 100%, ADC 3pin is 2.98V. Quisk shows 1510 mA. OpAmp 8pin is 2.66V (not used now).

Are these reasonable ?

You said "there is also less offset, the reading is good from about 15 mA"

Did you add pull-down resistor ?

73, Taka  ji1udd

[in3otd]

Hello Taka,
here I have
- on RX
    ADC pin 3: 95 mV, Quisk shows 48 mA
    OpAmp pin 8 : 3.24 V
- on TX with spot at 100 % at 14.2 MHz - about 39.2 dBm out:
    ADC pin 3: 3.44 V, Quisk shows 1610 mA - note that the ADC out is saturated at max
    OpAmp pin 8 : 3.24 V

Note that I lifted the OpAmp pin 8 so it has no negative feedback anymore, this is why it's saturated at 3.24 V.

I did not add any pull down; I saw that there is less offset now while doing the tests with the electronic load, without the PA. But I can also see this with the PA: if I set the PA bias to 0 for both devices I read about 5 mA of current.
The fact that we have a higher reading in RX is likely due to the INA199 not having any supply during receive, so it cannot pull down its output effectively.
I think we should add a pull down (R109), this will likely help in reducing the reading we have in RX (but that's not so important, we may simply blank the PA current reading during RX) and also allow to read higher currents, as currently the ADC is already at full scale with the PA in "standard" conditions and I think we may see more current drawn if the load has some SWR.

73 de Claudio, IN3OTD / DK1CG

[Takashi K]

Hi Claudio,

Thanks for your information.

> if I set the PA bias to 0 for both devices I read about 5 mA of current.

In my case, it is 8 mA, a bit worse but OK.

> Note that I lifted the OpAmp pin 8 so it has no negative feedback anymore,
> this is why it's saturated at 3.24 V.

Sorry, My beta3 modification is different from yours.
Once I removed R108, isolated its pad op amp side using kapton tape, reinstalled R108, made connection the pad to U14 6pin using jumper wire.

> I think we should add a pull down (R109), this will likely help in reducing
> the reading we have in RX (but that's not so important, we may simply blank
> the PA current reading during RX) and also allow to read higher currents,
> as currently the ADC is already at full scale with the PA in "standard"
> conditions and I think we may see more current drawn if the load has some
> SWR.

I agree with you. In this case we need to use low tolerance resistor for R108 and R109.

73, Taka  ji1udd

[Takashi K]

Hi Steve and all,

I have connected the tandem circuit (using two toroidal) on my filter board to HLv2 beta3.

1) While receiving, Quisk always shows 42Watts to 44Watts.

    Even if AIN1(CN7-6) is connected to GND, Quisk shows 40Watts. 

     I want to know how to calibrate watt meter.

2) While transmitting (14.2MHz), there is something strange.

      Spot     AIN1(CN7-6)    U13 1pin    Quisk RF Power / Temperature

     1.000      1.030V            0.704V      always 0.0W /  -50C

     0.985      1.011V            0.686V      always 0.0W /  -50C

     0.974      0.997V            0.672V      512W / 32C  , after few seconds, 0W / -50C

     0.969      0.992V            0.667V      512W / 32C  , after few seconds, 0W / 32C

     0.954      0.972V            0.651V      832W /  32C

    Why temperature is affected ?

    Watt meter is calibration issue ?

73, Taka  ji1udd

[Takashi K]

Hi all,

The temperature display on Quisk was improved a little by installing 0.1uF on B118.
When Spot 1.000, Quisk shows -50C before installing 0.1uF, after 31.5C on transmitting. 38C on receiving after transmission.
The cause seems RF interference.

73, Taka  ji1udd

[Steve Haynal]

Hi Taka,

Thanks for testing the tandem circuit. The correct Quisk power and SWR calculations have not been programmed in Quisk yet. If you would like to figure out the correction calculations, you can edit the file quisk_widgets.py:

Linux: quisk-4.1.10/hermes/quisk_widgets.py (Run from your local copy, not the system wide install.)

Windows: C:\Python27\Lib\site-packages\quisk\hermes\quisk_widgets.py 

Around line 75 there is a method called UpdateText. At the end of that method, the values of self.hardware.hermes_fwd_pwr and self.hardware.hermes_rev_pwr are displayed. You can edit the scaling here. Please share if you are able to calibrate the readings.

The temperature is always reading correctly for me whether in TX or RX. There have been one or two times when the readings from the slow ADC lock up. Then the values stay constant. I have to cold reboot the HL2 for the readings to return. You are supplying voltages to the other ADC inputs, but I don't see how that would affect the temperature reading. The HL2 firmware is already reading all 4 values, and the voltages you report are in range. Maybe check with a scope to make sure the values are true DC and not some RF signal.

73,

Steve

KF7O

[Steve Haynal]

Hi Taka,

The temperature reading has always worked for me in beta2 and beta3. I don't think B118 is required or should help based on the datasheet. A local decoupling capacitor from power to ground of Q6 may help. Maybe there was a bad solder job on this component in your build, or RF is leaking to other parts of the circuit now that you have the tandem detector connected.

73,

Steve

KF7O

[Steve Haynal]

Hi Claudio and Group,

I have made this modification and documented it on the wiki. My measurements saturate at around 1610 mA too with full power. I hit saturation at between 85% and 90% max TX DAC output into a good load. We can increase the range by lowering the value of R123, the 0.04 Ohm sense resistor, but this will trade off range for precision. I thought the 0.04 would cover the entire range, but I underestimated. When I went through the calculations with the 0.04 resistor, I came up with a precision of about +/- 5mA. Since both LDMOS devices are measured with the same sensor, they should experience the same offset. Also, people may want to look at when current starts to increase dramatically rather than an exact value in mA. U14 has a fixed x50 amplification.

I think we should go with a lower value for R123, maybe 0.03 Ohms. It would be helpful if someone would pick a low enough value to include most expected current readings (maybe not for the worst SWR) but high enough to not lose too much precision. Also, it would be helpful if this value were a common value. I rate something as a common value if there are many suppliers of the value on Digi-Key or Mouser.

There are also 2 modes supported by the slow ADC: 3.3V reference and internal 2.5V reference. It may be possible to switch to the 2.5V internal reference for low current measurements and increase the precision.

73,

Steve

KF7O

[in3otd]

Thanks Steve,
regarding the PA current sensing, I think we should simply add R109 to reduce the input to the ADC; even if we use divide the voltage by two we will still have 1 mA per ADC count which I think is enough for sensing the PA current. The readings do not seem much noisy here. I'm not sure we should go to an even lower sense resistor, I wonder whether we will start to see - or already see - some effects from parasitic resistances in the path; the current readings are about 5 % high already, IIRC, and it's not clear to me where this scale error comes from.
I also wanted to try the 2.5 V ADC internal reference, but from a quick glance at the code here I did not understand where the Setup Byte for the ADC is sent.

73 de Claudio, IN3OTD / DK1CG

[Takashi K]

Hi Steve and Claudio,

I replaced R108 with 470 ohms and added 1.5k ohms on R109 pad.

PA am meter seems to work well.
When only PTT, Quisk shows 150ma. It means real 197mA  (= 150 * 1500 / (470 + 1500) ).

When Spot 100%, shows 1240ma, means real 1629mA ( = 1240  * 1500 / (470 + 1500) ).

But I cannot measure Quisk readout vs PA current by using constant current load. because T3 has been already installed.

73, Taka  ji1udd

[Takashi K]

Hi Steve,

This temperature display -50C issue was solved.

I did the following modification.

・MCP9700 output is disconnected to op-amp 12pin and op-amp 12pin is clipped to GND.

・MCP9700 output is connected to ADC AIN1 directly. R111 1k ohms is removed. B95 0.1uF remains.

Also

・INA199 output is disconnected to op-amp 10pin and op-amp 10pin is clipped to GND.

・INA199 output is connected to voltage divider that I mentioned on another thread.

It seems important to disconnect op-amp input and clip unused op-amp input.

73 Taka  ji1udd

[Takashi K]

I forgot to mention,
When FWD/REV sensors are not connected to HLv2 CN7, the opamp input pins are floating.

I think that it is better to add pull-down resistors because input pin's floating is prohibited as a general opamp usage.

73, Taka  ji1udd

[Takashi K]

Hi Steve,

> Please share if you are able to calibrate the readings.

here is my result.

    fwd = (fwd-40.0)*6.0/840.0

There might be 10 % error or so.

Also for swr, I described,

    rev = (rev-40.0)*6.0/840.0
    swr = (1+math.sqrt(rev/fwd))/(1-math.sqrt(rev/fwd))

but phython error occurs around math.sqrt.

Do you have any advice ?

73, Taka  ji1udd

[Steve Haynal]

Hi Taka,

What is the error reported? You'll see an error for a negative argument to math.sqrt. You'll see an error if fwd is zero due to the division by zero. This may work:

fwd = (fwd-40.0)*6.0/840.0

fwd = fwd if fwd > 0 else 0.0001

rev = (rev-40.0)*6.0/840.0

rev = rev if rev > 0 else 0

swr = (1+math.sqrt(rev/fwd))/(1-math.sqrt(rev/fwd))

73,

Steve

KF7O

[Takashi K]

Hi Steve,

Thank you for your advice.

Another error is shown. It seems that some definition is needed.
division by zero will be error as you said. so I fixed script.

73, Taka  ji1udd

[Graeme Jury]

Hello Taka san,

There is a further source of ambiguous readings which happen when reading very low swr. Typically when the swr is very low and the reverse voltage is less than the forward conduction of the diode in the Stockton bridge you get a small series of zero reverse readings which all produce an swr of 1. This leaves the reading ambiguous as can be seen in the table below in the swr1 column. The fix is to make the reverse_voltage = 1 if it is zero and an unambiguous result can be obtained as shown in column swr2. The downside is that swr2 is telling tiny fibs about the actual swr but when you are measuring voltages with a diode down in that part of the curve it will be very non-linear anyway and the error is negligible.

fwd    rev    swr1    swr2
388    3    1.016    1.016
388    0    1.000    1.00517
390    0    1.000    1.00514
392    0    1.000    1.00512 *** Actual minimum swr point
387    0    1.000    1.00518
386    4    1.021    1.021

Sorry this is in C. I really need to get on top of Python :-(

void measure_swr()

if (forward_voltage > reverse_voltage) {
  if (reverse_voltage == 0) reverse_voltage = 1;
  swr = float(forward_voltage + reverse_voltage) / float(forward_voltage - reverse_voltage);
}
else {
  swr = 100;
}

73, Graeme zl2apv

[in3otd]

Hello Taka,
did you add import math at the beginning of hermes/quisk_widgets.py ? From the error it seems that the math module is not known there.

73 de Claudio, IN3OTD / DK1CG

[Takashi K]

Hi Claudio,

Thanks.

I added "import math", then another error ...

73, Taka  ji1udd

[Takashi K]

Hi Graeme-san,

Thank you for your professional advice : )

Someone, please convert to python script.

73, Taka  ji1udd

[in3otd]

Hello Taka,
I see now that in hermes/quisk_widgets.py the fwd and rev variables are redefined as strings.
Around line 86 in the file please change these lines

    fwd = self.hardware.hermes_fwd_power
    fwd = " %3.1f W" % float(fwd)
    self.text_fwd_power.SetLabel(fwd)
    rev = self.hardware.hermes_rev_power

to

    fwd = float(self.hardware.hermes_fwd_power)
    fwds = " %3.1f W" % fwd
    self.text_fwd_power.SetLabel(fwds)
    rev = float(self.hardware.hermes_rev_power)

Hope this works :),

73 de Claudio, IN3OTD / DK1CG

[Takashi K]

Hi Claudio,

Now FWD power and SWR display work well. Thanks.
I needed to fix another point because SWR is calculating during reception.
It's better that SWR is displayed only on transmission.

73, Taka  ji1udd

[Takashi K]

Hi Steve and HLv2 testers,

FYI,

I attached the schematic of my slowADC modification.

73, Taka  ji1udd

[in3otd]

Hello,
I did some experiments to see if the noise on the ADC readings could be reduced. I've disconnected U18 pin 14 and connected its pad (which goes to the ADC AIN1 after an RC filter) to an heavily filtered DC voltage coming from an additional regulator supplied by the 12 V going to the H-L.
I've then collected about 10000 readings from the ADC under various conditions: the default configuration, with an additional 10 uF tantalum capacitor right on the ADC supply pin and by adding a 4.7 ohm series resistor on the ADC supply, as suggested on the datasheet.

Here are the histograms of the ADC reading for the 3 configurations:

adding the 4.7 ohm resistors ahead of the local filtering capacitors brings some improvement, which was not seen with the additional capacitor alone. I did not (yet) try with a standard ferrite bead to see if it brings a similar improvement. It may be interesting to see if the internal reference brings an even lower noise.

73 de Claudio, IN3OTD / DK1CG

[Takashi K]

Hi Claudio,

It is very interesting.
I would like to check my ADC data with the same way as you.
Could you provide your software (modified Quisk?) to acquire it ?

73, Taka  ji1udd

[Steve Haynal]

Hi Claudio and Group,

Thanks for looking into this. I will add a footprint for either a FB or resistor in series with the U18 power supply. The Versa clock IC uses a 2.2 Ohm resistor in similar fashion, so to not add another BOM line I'd like to use either that 2.2 Ohm resistor or a FB for U18.

I've been working on the current sensing changes this weekend. I am a little concerned with a voltage divider here and possible interactions with the driver's source resistance, but am trying out 191 Ohms for R108, 1K for R109 and 100 pF for B93. This keeps the impedance lower than the 1.5K recommended by the datasheet. It also includes the 100pF recommended by the datasheet for when input impedance is on the high side.

This combination allows for just over 1.9A of current to be measured. I presented my HL2 with some mismatches with bad antenna tuner settings and saw it only hit 1.9A occasionally. With a good load, I see max of 1.82A. Do you think this is enough range? I don't want to have too much wasted range above the maximum expected operating current. 

With the next firmware release, I want to enable some simple firmware protection of the PA. If the current reading pegs, or the temperature hits 60C, the HL2 will disable the PA. A cold reboot will be necessary to reenable the PA. We can refine this in the future, but do you think this a reasonable starting point?

73,

Steve

KF7O

[in3otd]

Hello Steve,
I'll try with a lower resistance and a FB to see if these also help in reducing the ADC noise.
Regarding the source impedance, my understanding was that it's important only if you want to have a large system bandwidth, if we keep B93 to 100 nF that could help to keep the input voltage stable when using higher impedance.
Regarding the max measurable current, a couple of amps may be enough; anyway even if we had 4 A maximum we still would have 1 mA/LSB which should be enough even for setting the bias - worst case we can do some averaging in HW or SW.
Ok for the PA FW protections; don't know how easy it would be, but a visual feedback on the H-L to signal an error condition like this, e.g. all four leds blinking, would also be nice, IMHO.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

From your ADC measurements, I'm not sure we can rely on the full 12 bits. There are other sources of error >1mA too, so maybe it is not as bad as I think to extend the range a bit more.

73,

Steve

KF7O

[in3otd]

Hello,
I've tried a FB (MPZ1608S601ATA00, like the others on the board) on the slow ADC supply, still with an additional 10 uF afterwards, and the results were even slightly better than with the 4.7 ohm resistor:

(I've just noticed that all the histogram titles have a typo - should read H-Lv2b3)
I'll try later to lower the added capacitance also.

[in3otd]

Hello Taka,
a simplified version of the python script I used for testing the ADC is enclosed. Copy this file in the same directory where Quisk 4.1.10 is and run it with the same config file you use normally for Quisk, e.g. python test_slow_ADC.py -c hermes/quisk_conf2.py
It will connect to the H-Lv2b3 and start printing on the terminal lines with a timestamp and the ADC code read every 100 ms; the same data will be saved to a file. The file name is defined towards the end of the python script. To select the ADC channel which will be logged change the line 190 to use one of the variables temp_raw, ain1_raw, ain2_raw or current_raw. The script will run indefinitely, to stop it press Ctrl-C (just one time, it may need a little to actually stop). The log file can be processed by the enclosed GNU Octave script to generate the related histogram.
Let me know if it works for you, I've tested it only with my (slightly) modified Quisk.

73 de Claudio, IN3OTD / DK1CG

[Takashi K]

Hello Claudio,

Thank you for providing your software and kind explanation.

Unfortunately an error occurred when I started the software.

Please refer to the attached file.

Of course, quisk.py runs well instead of test_slow_ADC.py

73, Taka  ji1udd

[in3otd]

...works practically equally well with the FB alone (and the default 100 nF bypass on the ADC supply), no need to add further caps:

73 de Claudio, IN3OTD / DK1CG