Skip to content

Latest commit

 

History

History
109 lines (81 loc) · 6.74 KB

readme.md

File metadata and controls

109 lines (81 loc) · 6.74 KB

OpenHantek developer documentation

This document explains the basic data flow and concepts used and should be amended whenever a new concept is added to the code base. Detailed information about an implementation are to be found in the specific directory readmes though.

Directory structure

The res folder contains mostly binary resource data that are bundled with the executable, like icons, images, firmwares and translations.

The res/translations folder contains translations in the Qt format. Use Qt linguist as a handy graphical interface to translate OpenHantek. A Translation HowTo is available.

The source code within src can be divided into a core, that is responsible for device communication, data acquisition and post processing and the graphical interface with several custom widgets, a configuration interface and an OpenGL renderer.

Core structure

The raw device communcation takes place in the src/usb directory, especially via the USBDevice class. To find suitable devices, the FindDevices class in the same folder is used. Firmware upload is realized via the ezusb helper methods and the UploadFirmware class.

The hantek protocol structures and constants are defined within src/hantekprotocol.

The heart of OpenHantek is the src/hantekdso folder and its hantekdsocontrol class. All supported models, based on the DsoModel class, are implemented within a subfolder src/hantekdso/models and automatically register themself to a ModelRegistry class. A model is based on (contains) a specification, the ControlSpecification class.

The hantekdsocontrol class keeps track of the devices current state (samplerate, selected gain, activated channels, etc) via the ControlSettings class and field. It outputs the channel separated unprocessed samples via a samplesAvailable(DSOsamples*) signal.

Before the data is presented to the GUI it arrives in the src/post/postprocessing class. Several post processing classes are to be found in this directory as well.

Graphical interface structure

The initial dialog for device selection is realized in src/selectdevice where several models and dialogs are implemented. This is basically a graphical wrapper around the src/usb/finddevices class.

You will find the configuration dialog pages to be implemented in src/configdialog.

Custom widgets like a LevelSlider with a unit suffix reside in src/widgets, the custom main window docks are in src/docks.

The code that is responsible for exporting data to images or to the printer is stored in src/exporting.

The main window itself doesn't do and shouldn't do much more than connecting signals/slots between the core part and the graphical part.

All OpenGL rendering takes place in the GlScope class. A helper class GlScopeGraph contains exactly one data sample snapshot including all channels for voltage and spectrum and a pointer to the respective GPU buffer. GlScope works normally for OpenGL 3.2+ and OpenGL ES 2.0+ but if it detects OpenGL 2.1+ and OpenGL ES 1.2+ on older platforms it switches to a legacy implementation. If both OpenGL and OpenGL Es are present, OpenGL will be prefered, but can be overwritten by the user via a command flag.

Export

All export related funtionality is within src/exporting.

The following exporters are implemented:

  • Export to comma separated value file (CSV): Write to a user selected file,
  • Export to an image/pdf: Writes an image/pdf to a user selected file,
  • Print exporter: Creates a printable document and opens the print dialog.

All export classes (exportcsv, exportimage, exportprint) implement the ExporterInterface and are registered to the ExporterRegistry in the main.cpp.

Some export classes are still using the legacyExportDrawer class to draw the grid and paint all the labels, values and graphs.

The plan is to retire this legacy class and replace the paint code with a GlScope class shared OpenGL drawing code for at least the grid and the scope graphs.

Data flow

  • Raw 8-bit ADC values are collected in HantekDsoControl::run() and converted in HantekDsoControl::convertRawDataToSamples() to real-world double samples (scaled with voltage and sample rate). The 10X..100X oversampling for slower sample rates is done here. Also overdriving of the inputs is detected. The conversion uses either the factory calibration values from EEPROM or from a user supplied config file. Read more about calibration.
    • swTrigger()
      • which checks if the signal is triggered and calculates the starting point for a stable display. The time distance to the following opposite slope is measured and displayed as pulse width in the top row.
    • triggering() handles the trigger mode:
      • If the trigger condition is false and the trigger mode is Normal then we reuse the last triggered samples so that voltage and spectrum traces as well as the measurement at the scope's bottom lines are frozen until the trigger condition becomes true again.
      • If the trigger condition is false and the trigger mode is not Normal then we display a free running trace and discard the last saved samples.
  • The converted samples are emitted to PostProcessing::input() via signal/slot.
  • PostProzessing calls all processors that were registered in main.cpp.
    • MathchannelGenerator::process()
      • which creates a third MATH channel as one of these data sample combinations: CH1 + CH2, CH1 - CH2, CH2 - CH1, CH1 * CH2, CH1 AC or CH2 AC.
    • SpectrumGenerator::process()
      • For each active channel:
        • Calculate the peak-to-peak, DC (average), AC (rms) and effective value ( sqrt( DC² + AC² ) ).
        • Apply a user selected window function and scale the result accordingly.
        • Calculate the autocorrelation to get the frequency of the signal. This is quite inaccurate at high frequencies. In these cases the first peak value of the spectrum is used.
        • Calculate the spectrum of the AC part of the signal scaled as dBV.
    • GraphGenerator::process()
      • which works either in TY mode and creates two types of traces:
        • voltage over time GraphGenerator::generateGraphsTYvoltage()
        • spectrum over frequency GraphGenerator::generateGraphsTYspectrum()
      • or in XY mode and creates a voltage over voltage trace GraphGenerator::generateGraphsXY().
      • GraphGenerator::generateGraphsTYvoltage() creates up to three (CH1, CH2, MATH) voltage traces. For fast sample rates where less than 100 samples are displayed on screen a bandlimited (sinc) 10x upsampling creates a smoother display.
      • GraphGenerator::generateGraphsTYspectrum() creates up to three (CH1, CH2, MATH) spectral traces.

t.b.c.