ttrftech / nanovna Goto Github PK

Enabling and using cmd_scan keeps hanging the firmware. Can cmd_scan work?

Release 0.1.1 (commit d68ab20)

On screen keypad is broken if used via touchscreen. Bug shows after pressing the first key. It works if used through the jog switch.

1Gくらいの周波数があれば数10cmの解像度が得られるので実用的かも。
FFTはin-placeで実行できるはず。しかし、F072だと全くメモリの余裕がないのとCM0=FPUが無いので無理か。

Currently use pid of ChibiOS USB CDC and should replace it with unique one.

There are some candidates.

• pid.codes http://pid.codes/ (0x1209, ?) requires project is under a FOSS license.
• https://f055.io/ (0xf055, ?) requires project is under a FOSS license.
• STmicro might had provided pid sublicense service. see https://www.mikrocontroller.net/topic/285649

I updated last night from the original stock firmware that came with the device. It was successful however every now and again the traces crash, as well as the screen, becomes unresponsive for a few seconds. Also sometimes parts of the traces persist on the screen for a few seconds while other parts are redrawn etc.
Also before the update, I was able to use the multi-segment sweep with VNAsaver pc app, but now it seems to always crash the device too.
When I say crash, the system just locks up and freezes with the screen freezing.

Release 0.1.1 (commit d68ab20)

Bug found after updating my unit to this release. It worked fine with the stock firmware (GEN111.TAOBAO.COM Build time: Aug 2 2019 - 16:40:01). Double checked by flashing a backup.

After performing a full calibration (OPEN,SHORT,LOAD,ISO,THRU), the reference S21 shows a 0.5 to 1dB ripple. Ripple period depends on cable electrical length, but amplitude does not (tested with different quality cables).

If only ISO+THRU or just THRU calibration is done, the trace is flat.

To easily extend features, F072 is not enough to extend something valued.

Could you automatically switch the trace display to "real" after activating the domain transform?
Otherwise people have to deeply understand what to do.

After calibration you may use the device without the female-female adaptor or without the cables and you may want to have a negative edelay.
This implements the change

@@ -1211,6 +1211,7 @@ const keypads_t keypads_time[] = {
{ KP_X(2), KP_Y(0), 9 },
{ KP_X(3), KP_Y(2), KP_N },
{ KP_X(3), KP_Y(3), KP_P },

• { KP_X(3), KP_Y(1), KP_MINUS },
  { KP_X(2), KP_Y(3), KP_BS },
  { 0, 0, -1 }
  };

@@ -1712,6 +1713,9 @@ keypad_click(int key)
return KP_DONE;
} else if (c <= 9 && kp_index < NUMINPUT_LEN)
kp_buf[kp_index++] = '0' + c;

• else if (c == KP_MINUS && kp_index < NUMINPUT_LEN) {
• kp_buf[kp_index++] = '-';
• }
  else if (c == KP_PERIOD && kp_index < NUMINPUT_LEN) {
  // check period in former input
  int j;

There is exists trace format DELAY in the menu, but it doesn't works, because it is not implemented in the code.

I implemented group delay (GD) in this changeset: e81f4fe

Also it will be nice to add phase delay (PD) trace. But it requires continuous phase unwrap for entire sweep, so I'm not sure how to implement it with no use additional memory.

i can't upgrade my nanovna i'm on win10 x64
see Picture pls

https://pasteboard.co/IwBRO6Q.png

Do you have an instruction on building FW under Linux?

Is there an error in line 574 of plot.c
float zr = ((1+coeff[0])*(1-coeff[0]) + coeff[1]*coeff[1]) * d;

I think the right is
float zr = ((1+coeff[0])*(1-coeff[0]) - coeff[1]*coeff[1]) * d;

CALでやってる線形補完を、角度と振幅の補完に変更すべき。ステップ間の位相回転が小さいと誤差は小さいが、位相差が大きくなるとかなり効いてくる。

STR:

1. calibrate device: open,short,load,isoln,thru
2. make sure calibration is ok
3. open menu stimulus and change stop frequency
4. make sure calibration is unexpectedly reset

This bug appears after latest commits (scan command, etc)

Frequency error exists between LO and REF, which is around some numbers of 10Hz. It caused by fractional PLL which use inappropriate divisor.

The font size is very small and hard to read. There should be an option to choose bigger fonts. Thanks.

There are unstable measurement on frequency band boundaries (where reset pll is performed).

The root of cause is missing wait for si5351 PLL lock. Also there is minor affect from flashing LED during sweep, it leads to a small voltage deviations and as result higher measurement noise.

The bug can be noticed as a random spikes at frequency boundaries (at begin of sweep, at 300 MHz, etc).

Both issues are fixed in this changeset: 23dba7d

現在はメニュー項目だけがある。TRC_DELAYの実装を埋める。

Smithチャートで'R+C/Lコンポーネント'になっているのをR+Xjにしたいという要望がある
Displayフォーマット毎に、マーカー値のフォーマットが固定なのを可変にする

This is related with this issue: #70

The bug didn't fixed. The spikes are reduced but don't disappears.

Try rollback your changes and apply my small fix, this fix completely eliminates instability bug: qrp73@2ba09fb

This is how it looks with my changes (cal reset, S21 ref point=9, center 450, span 100):

Now let's check the latest code from your master branch with the same settings and exactly the same conditions:

As you can see there is a spike at 450 MHz with your change, it is still unstable.

My fix looks very easy, but it don't means that it uses random values. I tested hard different delays in different modes. I tested similar delays that is used in your changes in different configurations. But after all I found stable and fast configuration. After that I tested it very hard with different values. And after all I got these "magic values" 3 and 8 :)

It's long story to explain all details. Just in short, use minimum delay = 3 even for channel switch. For pll reset, band change and gain change use delay = 8. It works great and pretty fast

Downloaded and installed (2 attempts) firmware hangs on menu selection of any Trace.

本物にはこの表示機能がある(目的はよくわからない)。難しくはないので付けとく。

scaleからedelayの設定はできているが、画面に表示していないので、これを出す

Is it possible to add a new console command so that the NanoVNA-Saver app can turn off TDR if it is enabled?

chprintf doesn't supports lossless floating point formats such as "%.9g". It seems that it has very basic support for floating points. As result, the floating point value is truncated with data loss, especially for very small floating point numbers.

I propose to use snprintf instead of chprintf and then put the buffer to the output stream.

For example:

        char tmpbuf[32];
        int leng = snprintf(tmpbuf, sizeof(tmpbuf), "%.9g", s11[0]);
        for (int j=0; j < leng; j++) {
            streamPut(chp, (uint8_t)tmpbuf[j]); 
        }
        streamPut(chp, (uint8_t)'\t'); 
        leng = snprintf(tmpbuf, sizeof(tmpbuf), "%.9g", s11[1]);
        for (int j=0; j < leng; j++) {
            streamPut(chp, (uint8_t)tmpbuf[j]); 
        }
        streamPut(chp, (uint8_t)'\t'); 

I have seen some screen artifacts from traces since the first firmware version i have tested. Has anyone paid attention to this?

Hello,
Very impressed about the project.
I’ve seen it can be used to up to 900mhz but in practice can i measure the impedance at 868mhz?

There is some variation in the quality of the SI5351 being used. Going to 300MHz is formally always out of spec so you are never sure if the nanoVNA will be able to reach 300MHz
Would it be an idea to make the max SI5351 output frequency user settable (and stored in flash) so users experiencing stability issues can set the max SI5351 frequency to a lower value?

Hi,
I don't find a license file here, or on your CentSDR page. Is this an oversight?
Thanks for your good work!
73,
Rob
KL7NA

current scan command works incorrectly. It just duplicates sweep command functionality and don't returns data immediately. This is just a copy of sweep command. This is incorrect behavior.

The expected behavior of scan command is to get unlimited measurement point count. For example it should measure 1'000 points or 100'000 points. It should not store all points in the memory, because it will leads to out of memory error. So, it should return data immediately with no need for data command.

Another issue with current scan command is that it doesn't allow to select expected frequency step, because current implementation expects stop frequency instead of step frequency.

The expected scan command parameters is the following:
scan {start(Hz)} {step(Hz)} {count}

Note, it should use {step(Hz)} instead of {stop(Hz)}. Step may be negative, in such case scan command should perform sweep from higher frequency to lower frequency. The firmware should not calculate step, it should be provided in the command parameter.

scan command should NOT store all points in the memory, instead of that it should print measured data into console for each frequency step immediately.

STR:

1. power on device
2. during device initialization, touch at the center of screen with frequency 2-3 times per second
3. make sure device freezes and don't respond on any input anymore.

This bug is a little floating and may require 2-3 times to repeat in order to reproduce it.

There is strange bug. Sometimes the first measurement point has very high spike.

I found the following STR that allows to reproduce this spike with high probability:

1. power on
2. open CONFIG -> VERSION screen and wait 5-10 sec
3. tap on the center of screen to close version screen
4. note that first sweep has high spike at first measurement point

The same spike may happens when you use joystick to move cursor left or right.

Also it may happens during calibration, in such case calibration will be wrong for the low frequency.

There is a little imbalance of gain, which leads to some discontinuities. You can find it by reset calibration and setup DISPLAY TRACE S11 LOGMAG scale=1 with reference position about 27.

@hugen79 tried to fix it in NanoVNA-H, but not complete.

More precise fix is done in this commit: qrp73@675b5bb

I have installed your last Firmware,
hat trouble to save more than 3 Memories. At the 4th the cal procedure is fine but if i want to recall the Memory it contains Garbage.

Have Reset the Calibration Data before and try to erase the config.

Its not important on witch place i store the memory.

73s Holger, DF6DBF

Current code has some syntax tricks which have issue with C standard compliance. Please fix it. It's hard to merge all changes when this fix is not applied. Thanks.

The main issues are following:

1. implicit cast from void(*)() to void*
   Proposed solutions:

• explicit cast
• use unnamed union to store void pointer or function pointer at the same memory, but with different field name

2. struct field initializers are not compatible with C standard
   Proposed solution: use C standard struct field initializers:

config_t config = {
  .magic =              CONFIG_MAGIC,
  .dac_value =          1922,
  .grid_color =         0x1084,
 //...
};

3. Binary constants are not supported in C standard.
   Proposed solutions:

• use hex constants
• use header file with predefined constants starting from letter b to translate binary constants to hex: #define b10001010 0x8A

fix is already done in the following two changesets:

• e1e5ca2
• abdb025

USB serial 経由のデータ(など)が化けることがあります。

• https://github.com/edy555/ChibiOS/blob/fe0ba1049c38346ceb2a396fa560848ef8323dd1/os/hal/lib/streams/chprintf.c#L121 tmpbuf がグローバルになっている
• shellThread の USB 経由で chvprintf が呼ばれる
• sweep スレッド (Thread1) の draw_all_cells() 中で chvprintf が呼ばれる (ここで tmpbuf が上書きされてしまう)

が原因かと思っています (実際 tmpbuf を stack 変数にすると再現しない)。が、tmpbuf は意図的に global にされているようなので、issue にします。

逆のケース (sweep スレッドの chvprintf が shellThread で上書きされる) もあると思いますが扱う文字数が違うので観測するのが難しそうです。

Add choice of marker value format on smith chart

• re+im
• R+component value (current)
• R+jX
• G+jB

エラー係数(=校正状態)をPCとやりとりできるようにする

To find maximum/minimum and jump to there on trace.

Is there some dll's for C# program?
THX!
Donald

Hello
is it possible tot have Bigger Fonts, like in the AA version of 2/8
if not possibel, 2 traces can be enough
Since used for field work and older man, bigger fonts can be very usefull
Tnx for soft anyway

Have here "strange" Peaks after upgrading to v0.2.1. v0.2.2 shows same behaviour. Took Screenshots with nanoVNA-saver to show the Problem. The v0.2.0 from 27. Sept. seems ok.

Screenshot from 0.2.0 (27.Sept)

Screenshot from 0.2.2 (29.Sept)

For TDR , instead of doing a window function its better to extend the FFT to
twice the length with the complex conjugate and use a 256 point FFT

現在CALキットは理想的である前提での実装。(<300MHz以下だったから)
周波数範囲が上がったので、理想ではないCAL Kitのモデルを適用できるようにしたい。
オープンのC0-C3
ほか全てのZr, delay
定義を画面でさせるのはツライので、コマンドラインからのみサポート