Gameduino Oscilloscope

[lawrie]

I have started building my Gameduino 3 oscilloscope. It uses some of the techniques discussed in recent posts. It uses an interrupt on button2 for the FPGA to tell the STM32 when data is ready and then uses QSPI to fetch the samples so they can be displayed on the Gameduino screen.

My biggest problem is power. I had to use spacers on the Arduino headers so that I could take a lead from the Vin 5v feed to the AN108 AD device, which needs 5v. I already had to have both USBs connected to give the Gameduino enough power. With both the Gameduino and the AN108 taking power, it is unreliable. It works for a while and then stops. I then have to remove the USBs and plug them in again. The spacers for the Gameduino are a good idea anyway as it leaves room to get to the buttons, RPI header etc.

I am experimenting with user interfaces. I am not sure whether to use dials on the touchscreen or physical dials.

[lawrie]

This is a (more or less) working version of my oscilloscope. The AN108 device seems to draw less power and work more reliably at lower bit rates. I currently have it at 8Mhz. The Gameduino seems a bit temperamental. It gives me a blue screen coprocessor error if I try to write too many pixels (points). At least I think that is what the problem is.

The user interface involves touching the relevant parameter and then turning the touchscreen dial to change its value. It is a bit fiddly.

The FPGA interrupts the STM32 when data is available and the STM32 using QSPI to send the current parameters and receive back the latest samples.

Here is the code:

https://github.com/lawrie/verilog_examples/blob/master/fpga/oscilloscope/oscilloscope/oscilloscope.ino

[Folknology]

Nice work @lawrie great all around use of BlackIce features, one of the best I have seen using the STM32 in addition to the Ice40

[afiskon]

@lawrie You don't stop to surprise me How do you find so much time for these projects?

[lawrie]

I am retired and have quite a lot of spare time (at least when my family leave me alone). Usually, I am very lazy but I am a bit inspired by FPGAs at the moment and want to explore the limits of what you can do with them.

[ken.boak]

Lawrie,

Great to see this useful project come together. Have you given any updates to James Bowman or the wider Gameduino community?

Incidentley, at OSHCamp 2017, an attendee showed me a commercially available pocket oscilloscope that contained both an ICE40 and an STM32.

Using the FTDI EVE chip to control the display and GUI, allows these projects to come together fairly quickly with impressive results.

With a BlackIce board and a Gameduino III, a student of electronic engineering could create a useful, low cost multifunction test equipment, scope, DVM, logic analyser, FFT analyser, arbitrary waveform generator etc.- as well as learning a lot about how these instruments function.

Keep up the good work!

[lawrie]

Hi Ken,

I just posted the project on the Gameduino 2/3 forum so James Bowman and the other Gameduino users can see it.

I know very little about electronic test equipment so these projects are teaching me about it. I didn't own an oscilloscope until about 6 months ago. I did start looking at waveform generation using the DAC on the AN108 device, so will probably add that to the project. A voltage display would be simple to add but I assume that for accurate voltage measurement, I would need a more accurate ADC such as https://store.digilentinc.com/pmod-ad5-4-channel-4-8-khz-24-bit-a-d-converter/.

I tried out @hoglet's port of the Open WorkBench Logic Sniffer recently - https://forum.mystorm.uk/t/open-bench-logic-sniffer-on-blackice-ii/417/4. It should not be too hard to run the Sump client on the Gameduino 3.

I played about with spectrum analyser's for radio projects a while ago, so I might look into implementing fourier transforms and spectrum analysis.

[lightcalamar]

Hello, my name is Tomas from Spain.

Excuse my bad English.

A great project for oscilloscope, I have screens of FTDI 5 and 7 inches. I have New Haven Display, EVE2 FTDI screen with chip FT813Q where it has support for QSPI

Currently using Nucleo-STM32F767, trying to export this project with modified GD23ZU library by James Boswman for my friend and me.

Thanks!

[lawrie]

TinyFPGA as a sound source for the BlackIce oscilloscope:

[lawrie]

My oscilloscope has made it to Hackaday, thanks to @jamesbowman:

Gameduino + Mystorm = Oscilloscope!

[lawrie]

Currently I am adding a couple of things to the oscilloscope. I have added playing the samples as audio through the Gameduino speaker output, which works if they are audio or in the right frequency range. I have also added a Fast Fourier Transform for spectrum analysis, which works, but more work is needed on displaying it. I found Arduino STM32 ARM assembler code that did this. It does 1024 samples in two or three milliseconds.

I am planning to use these things on a voice or music recorder application that works in a similar way to the oscilloscope, but uses a Digilent microphone Pmod to capture the sound.

Unfortunately the oscilloscope still has power problems. I only connect the 5v source to the AN108 for a short period as after a while it crashes the device. I have ordered a Flashy ADC board similar to the one that @tbalon is using as they seem to use less power, only need a 3.3v source and can capture samples faster.

[tbalon]

Are you using fixed or floating point for the FFT. I found the STM DSP lib works very well although I'm not sure if you can link with it and fit the resulting file in the stm flash.

Nice display, are you using BlackSoc on this project ?

[lawrie]

I am using this code -

rogerclarkmelbourne/Arduino_STM32/blob/master/STM32F1/libraries/stm_fft/cr4_fft_1024_stm32.asm

I think it uses fixed point. The input and output values are integers.

No, this project doesn't use BlackSoc. I have BlackSoc doing QSPI, but I haven't worked out how to do it at full speed with BlackSoc yet. The code I run on the FPGA is fairly simple - it just takes samples and sends them over QSPI. It accepts a few parameters from the STM32 including a prescaler to set the sampling speed, and the trigger type and value.

The way I am currently using QSPI, it is slower than your use of SPI. I need to change the Verilog QSPI slave code to support larger chunk sizes.

[lawrie]

Is the source of your project(s) available? I would be interested in how you are using SRAM. My samples are small enough that I am using BRAM at the moment, but might want to use SRAM for my audio project.

[tbalon]

Looks like your using software floating point ( softvfp ) the SM32L433 support hardware floating point and DSP instructions that will make the FFT run much quicker if you need it.

The Arm ® Cortex ® -M4 with FPU processor is the latest generation of Arm ® processors for embedded systems.

The processor supports a set of DSP instructions which allow efficient signal processing and
complex algorithm execution.

Its single precision FPU speeds up software development by using metalanguage
development tools, while avoiding saturation.

[tbalon]

For SRAM storage I am using the example SRAM controller. I'm sampling every 10ns and write to SRAM every other clock cycle thereby following its requirement to write on every other clock. Since I'm using an 8-bit ADC and the SRAM is 16-bit, it works out very nicely.

Source is not available, I'm working on a C API from the STM32 that can program things like buffer size, trigger over SPI then once it's working I can post a copy for you.

[tbalon]

Actually that optimized version looks pretty good considering the clock speed... For comparison, I'm using the M4 DSP to perform basic HP and LP digital filtering on the input. I do an FFT on the signal, apply the LP and HP and the an inverse FFT back before display.

On my STM32F429 board, using 512 point FFT and the ARM DSP lib, I can do the filtering in ~1.2ms.

[lawrie]

Actually looking at the timing again, it is taking 14 milliseconds to do 1024 points. What speed is our STM32? I know the crystal frequency is 12Mhz but I do not know what the multiplier is. I saw somewhere that the L4 has a maximum of 80Mhz.

UPDATE: Ignore that. I looked at the wrong timing in my output. the 14ms is the time it has QSPI to transfer 4kb, 1-byte at a time. The FFT varies between just under and just over 3ms.

[tbalon]

STM is running at 80Mhz, There may be an optimized version of the code you have for the STM32L433 that leverages the FPU and DSP instruction set. This would be good to have, cubemx may be able to pull it in as a library.

[evan.bovie]

You've got a very cool product.

I highly recommend these rotary encoders from Sparkfun for the main control interface, supplemented by touchscreen for more complex configurations and interactions.

The OP-1 has accomplished so much with just four primary knobs.