当前位置：首页 > news >正文

示波器traces采集

news 2026/4/29 18:50:52

业务上遇到一个示波器traces采集的问题，他们用的采集脚本是好几年前的开源脚本，一直没有更新和维护，跟新的示波器不匹配，第一步通讯就卡住了，然后把问题抛给我。我对示波器的了解仅限于当年在大学实验课上接触过，并没有过多的了解，这次就当学习了。

方案选择

他们原本是想让我在以前的脚本上做修改，但我打开那个开源仓库，上次更新时间是在7.8年前，对于示波器都还不了解，一上来就让我改工具，未免有点为难我，还是另寻他法。

首先弄清楚示波器的型号，是 PICO3000E系列的。
搜索一圈下来，发现 pico 官方是有提供软件和SDK的，这样可以直接用官方的接口，然后再改写脚本到我们希望看到的效果就好了。
具体怎么操作，官方有提供教程：PyPicoSDK: Getting Started

环境搭建

这里跟着官方的教程走就可以。

1. 安装 PicoSDK

下载地址：https://www.picotech.com/downloads
选择具体示波器型号和系统对应的安装包就可以

2. 安装需要的 python 库

pipinstallpypicosdk pipinstallmatplotlib pipinstallscipy pipinstallpandas

安装 pypicosdk ：
(可以看到结果，在这个过程 numpy 同时也被安装了）

安装 matplotlib :

安装 scipy :

安装 pandas :

编写脚本

在官方的仓库里，有提供挺多 example 可以参考的，可以在这个基础上加细节适配采集需求
官方仓库示例：examples

1. 验证示波器存在

用官方示例enumerate_units.py验证测试电脑能找到示波器，确保第一步成功

""" Copyright(C)2025-2025 Pico Technology Ltd. See LICENSEfileforterms. This example enumerates all PicoScopeunits(supported by pyPicoSDK), returns the number ofunitsand a list of serial numbers."""importpypicosdk as psdk# Enumerate unitsn_units,units=psdk.get_all_enumerated_units()# Print outputprint(f'Number of units: {n_units}')forunit, serialinunits.items(): print(f'{unit}: {serial}')

运行结果：

2. 适配采集需求

在需求的测试场景中，第一是要采集大量的traces用于后续分析，第二是要能够采集到触发的完整波形。
也就是需要解决波形问题，还有数据量的问题。

2.1 调整波形

首先是要配置好采集通道和触发条件。这个可以先打开上面安装的 PicoScope 7 T&M 这个工具，因为是图形界面，操作起来比脚本简单快速，也很直观。所以先用它调好参数，然后再把参数填到脚本中即可。
这是调好的波形：

2.2 编写脚本

是在官方simple_block_capture.py这个脚本的基础上去改的。

importpypicosdk as psdkimportnumpy as npimportmatplotlib.pyplot as pltimporttimeimportos from datetimeimportdatetime# ============================================================# User-adjustable parameters (modify according to your chip signal)# ============================================================CHANNEL=psdk.CHANNEL.A# Channel where the probe is connectedVOLTAGE_RANGE=psdk.RANGE.mV500# Voltage range, match your chip's output levelCOUPLING=psdk.COUPLING.DC# Coupling mode (usually DC for digital circuits)TRIGGER_CHANNEL=psdk.CHANNEL.A# Trigger source channelTRIGGER_THRESHOLD=200# Trigger level in voltsTRIGGER_DIRECTION=psdk.TRIGGER_DIR.RISING# Trigger on rising edgeTIMEBASE=1000000# Sample interval in seconds (1e-6 = 1 us, 1 MS/s)# sample_rate = 9_765_625TOTAL_SAMPLES=5000# Total number of samples per captureTOTAL_CAPTURES=10# PRE_TRIG_PERCENT = 10 # Samples to keep before the trigger point# POST_TRIGGER_SAMPLES = NUM_SAMPLES - PRE_TRIGGER_SAMPLES # Samples after triggerSAVE_DIR="./captured_traces"os.makedirs(SAVE_DIR,exist_ok=True)# ============================================================# 1. Connect to the oscilloscope (PicoScope 3000E uses psospa class)# ============================================================print("Connecting to PicoScope...")scope=psdk.psospa()scope.open_unit()print("Scope connected.")# Optional: print some device informationtry: driver_info=scope.get_unit_serial()print(f"device info: {driver_info}")except: pass# ============================================================# 2. Configure the input channel# ============================================================print("Configuring channel...\n")scope.set_channel(CHANNEL,enabled=True,coupling=COUPLING,range=VOLTAGE_RANGE)print("set channel success...\n")# Disable other channels to save resources# for ch in ["B", "C", "D"]:# try:# scope.set_channel(ch, enabled=False)# except:# pass# ============================================================# 3. Set up the trigger (external chip signal, no auto-trigger)# ============================================================print("Setting trigger...")scope.set_simple_trigger(TRIGGER_CHANNEL,threshold=TRIGGER_THRESHOLD,direction=TRIGGER_DIRECTION,auto_trigger=0# Disable auto-trigger; wait for a real signal edge)print(f"Waiting for trigger on Channel {TRIGGER_CHANNEL} "f"({TRIGGER_DIRECTION}, {TRIGGER_THRESHOLD}mV)...")# Perform simple block capture via help function (inc. buffer setup, time axis, mV conversion etc.)# timebase = scope.sample_rate_to_timebase(sample_rate=sample_rate)# print(f"Calculated timebase: {timebase}")channel_buffer, time_axis=scope.run_simple_block_capture(TIMEBASE, TOTAL_SAMPLES,time_unit="ms")print(f"time_axis: {time_axis}")# channel_buffers, time_axis = scope.run_simple_rapid_block_capture(TIMEBASE, TOTAL_SAMPLES, TOTAL_CAPTURES)# all_traces = channel_buffers[psdk.CHANNEL.A]# timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")# filename = f"./captured_traces/traces_{TOTAL_CAPTURES}_{timestamp}.npy"# np.save(filename, all_traces)# print(f"Saved {TOTAL_CAPTURES} tarces to {filename}")# Release the device from the driverscope.close_unit()# # Use matplotlib to setup the graph and plot the dataplt.figure(figsize=(13,5))plt.plot(time_axis, channel_buffer[psdk.CHANNEL.A])# for trace in all_traces:# plt.plot(time_axis, trace)# plt.plot(time_axis, all_traces[0])plt.xlabel("Time (ms)")plt.ylabel("Amplitude (mV)")plt.grid(True)plt.show()

（因为后续是要采集大量traces的，这里只展示采集单个波形，所以脚本显得注释掉的很多，当时也是不断的在调，修修改改，不停的看效果，然后再优化）

看下没调对触发条件是怎样的：
只有噪声，没有触发事件。

调对了trigger ，但是波形只能采集到一半。

这里卡了挺久的，后面不断的调采集点samples 和采样间隔 timebase ，才悟出来，这两个东西反相关的。采集间隔小，那么采集点要足够大，才能采集到完整波形；采集间隔大，那采集点就可以不用那么多。

timebase: 采样间隔，用于控制采样率。 timebase 和采集的样本总数samples 共同决定总采集时间窗口：总采集时间=timebase * samples

因为采集需求是要大量的traces ，所以这里牺牲采样时长来缓解内存压力，这样采样点就可以少，只是采样时间会变长。
最后采集出来的波长跟示波器捕捉到的一致。

下面这个脚本是采集多条traces 的，结果保存为 .trs 文件

importpypicosdk as psdkimportnumpy as npimportmatplotlib.pyplot as pltimportos from datetimeimportdatetimeimporttrsfile from trsfileimportTrace, SampleCoding, TracePadding, Header# ============================================================# User-adjustable parameters (modify according to your chip signal)# ============================================================CHANNEL=psdk.CHANNEL.A# Channel where the probe is connectedVOLTAGE_RANGE=psdk.RANGE.mV500# Voltage range, match your chip's output levelCOUPLING=psdk.COUPLING.DC# Coupling mode (usually DC for digital circuits)TRIGGER_CHANNEL=psdk.CHANNEL.A# Trigger source channelTRIGGER_THRESHOLD=200# Trigger level in voltsTRIGGER_DIRECTION=psdk.TRIGGER_DIR.RISING# Trigger on rising edgeTIMEBASE=10000000# Sample interval in seconds (1e-6 = 1 us, 1 MS/s)TOTAL_SAMPLES=500# Total number of samples per captureTOTAL_CAPTURES=100# PRE_TRIG_PERCENT = 10 # Samples to keep before the trigger point# POST_TRIGGER_SAMPLES = NUM_SAMPLES - PRE_TRIGGER_SAMPLES # Samples after triggerSAVE_DIR="./captured_traces"os.makedirs(SAVE_DIR,exist_ok=True)# ============================================================# 1. Connect to the oscilloscope (PicoScope 3000E uses psospa class)# ============================================================print("Connecting to PicoScope...")scope=psdk.psospa()scope.open_unit()print("Scope connected.")# Optional: print some device informationtry: driver_info=scope.get_unit_serial()print(f"device info: {driver_info}")except: pass# ============================================================# 2. Configure the input channel# ============================================================print("Configuring channel...\n")scope.set_channel(CHANNEL,enabled=True,coupling=COUPLING,range=VOLTAGE_RANGE)print("set channel success...\n")# Disable other channels to save resources# for ch in ["B", "C", "D"]:# try:# scope.set_channel(ch, enabled=False)# except:# pass# ============================================================# 3. Set up the trigger (external chip signal, no auto-trigger)# ============================================================print("Setting trigger...")scope.set_simple_trigger(TRIGGER_CHANNEL,threshold=TRIGGER_THRESHOLD,direction=TRIGGER_DIRECTION,auto_trigger=0# Disable auto-trigger; wait for a real signal edge)print(f"Waiting for trigger on Channel {TRIGGER_CHANNEL} "f"({TRIGGER_DIRECTION}, {TRIGGER_THRESHOLD}mV)...")# Perform simple block capture via help function (inc. buffer setup, time axis, mV conversion etc.)# channel_buffer, time_axis = scope.run_simple_block_capture(TIMEBASE, TOTAL_SAMPLES)channel_buffers, time_axis=scope.run_simple_rapid_block_capture(TIMEBASE, TOTAL_SAMPLES, TOTAL_CAPTURES,time_unit="ms")# print(f"time_axis: {time_axis}")all_traces=channel_buffers[psdk.CHANNEL.A]timestamp=datetime.now().strftime("%Y%m%d_%H%M%S")filename=f"./captured_traces/traces_{TOTAL_CAPTURES}_{timestamp}.trs"headers={Header.DESCRIPTION:'Encryption traces for side-channel analysis', Header.LABEL_X:'Time (ns)', Header.LABEL_Y:'Voltage (mV)', Header.SAMPLE_CODING: SampleCoding.FLOAT}with trsfile.open(filename,'w',headers=headers,TracePadding=TracePadding.AUTO)as trs_file:fori, trace_datainenumerate(all_traces): trace=Trace(SampleCoding.FLOAT, trace_data)trs_file.append(trace)if(i+1)%200==0: print(f"Writing trace {i+1}/{len(all_traces)}")print(f"Saved {len(all_traces)} traces to {filename}")# time_base, unit = scope.get_time_axis(...)# Release the device from the driverscope.close_unit()# # Use matplotlib to setup the graph and plot the data# plt.plot(time_axis, channel_buffer[psdk.CHANNEL.A])# below is to display the waveform, it's not necessary, can delete. Add here is to verify the waveform we fetched.plt.plot(time_axis, all_traces[0])plt.xlabel("Time (ms)")plt.ylabel("Amplitude (mV)")plt.grid(True)plt.show()print("traces acquisition completed")