PicoScope 2000

[MrW0lf]

PicoScope 2408B (100MHz, 4CH, 1GSa/s RTS, 20GSa/s ETS, 128MB memory, AWG) USB oscilloscope rolling review + general 2000 series thread

Even as a long time PicoScope owner (2205 (2010my), 2205 MSO (2013my)), was quite surprised by my new 2408B on several fronts. So decided to do "rolling review" on it and learn how to properly max out it's capabilities in progress. Currently planned: bandwidth & sampling rate limits; vertical / horizontal / automated measurements accuracy; advanced mask testing; FFT; decoding; bode plot (incl. phase), some math trickery. If any additional interests, please ask..



First task/test was to actually find out specs. Included here are those I see as most important and ones missing / different in official specs (emphasized in green). Scope performance relies heavily on PC software - so guess they forgot to update specs, while software got better. PC itself also plays large part, mine is Lenovo W520 laptop, Intel i7-2720QM, 8GB RAM, 64-bit Windows 7. So reasonably fast, but probably not enough to totally max out scope.

Disclaimer: review/test is done in non-controlled environment by non-professional. Use this info at own risk & responsibility

2408B SPECIFICATIONS

Oscilloscope - vertical
Analog bandwidth: 100MHz
Rise time (calculated): 3.5ns
Vertical resolution: 8bit (up to 12bit averaging)
Input coupling: AC/DC
Input ranges: +-20mV...+-20V (4mV/div...4V/div)
Vertical DC offset range: ±250mV (±20mV to ±200mV ranges); ±2.5V (±500mV to ±2V ranges); ±25V (±5V to ±20V ranges)
DC zero level calibration: on-the-fly
DC accuracy: ±3% of full scale ±200uV
Offset control accuracy: ±1% of offset setting, additional to basic DC accuracy
Overvoltage protection: ±100V (DC + AC peak)

Oscilloscope - horizontal
Real time sampling rate (RTS): any 1 ch 1GSa/s; any 2 ch 500MSa/s; any 3-4 ch 250MSa/s
Equivalent time sampling rate (ETS): any 1-4 ch 20GSa/s
Timebases (RTS): 1ns/div...5000s/div
Timebases (ETS): 50ps/div...200s/div
Max timebase at max sampling rate (RTS, Single trigger, fast sampling): 10ms/div
Max timebase at max sampling rate (RTS, None|Auto|Repeat triggers, fast sampling): 5ms/div
Max timebase at max sampling rate (RTS, Rapid trigger, fast sampling): 1ms/div
Max timebase at max sampling rate (ETS): 50ns/div
Max memory (RTS, Single trigger, fast sampling):
any 1ch 1x120MB; any 2ch 2x60MB; any 3ch 3x30MB; 4ch 4x30MB
Max memory (RTS, None|Auto|Repeat triggers, fast sampling):
any 1ch 1x60MB; any 2ch 2x30MB; any 3ch 3x15MB; 4ch 4x15MB
Max memory (RTS, Rapid trigger, fast sampling):
any 1ch 1x13x9.5MB; any 2ch 2x13x4.75MB; any 3ch 3x13x2.375MB; 4ch 4x13x2.375MB
Max memory (RTS, None|Auto|Repeat|Single triggers, slow sampling):
any 1ch 1x95MB; any 2ch 2x48MB; any 3ch 3x32MB; 4ch 4x24MB
Max memory (ETS):
any 1-4ch 1-4x16,380kB
Waveform buffers: 10,000
Maximum waveforms per second (RTS, persistence mode): 80,000
Maximum waveforms per second (RTS, Rapid trigger): >1,000,000
Initial timebase accuracy: ±50ppm
Timebase drift: ±5ppm / year
Sample jitter: 3ps RMS typical
ADC sampling: Simultaneous

Dynamic performance
Crosstalk (full bandwidth, equal ranges): Better than 300:1
Harmonic distortion: <–50dB at 100kHz, full-scale input, typical
SFDR (100 kHz, full-scale input, typical): ±20mV range: >44 dB; ±50mV range and higher: >52 dB
Noise: <30uV RMS (±20mV range)
Bandwidth flatness: (+0.3dB, –3dB) from DC to full bandwidth

Triggering
Trigger rearm time in rapid trigger mode: <1us on fastest timebase
Max. waveforms in rapid trigger mode: 10,000

Function generator
Standard signal frequency: 1MHz
AWG update rate: 20MHz
AWG buffer: 32kB
AWG Resolution: 12bit
AWG bandwidth: >1MHz
AWG rise time (10%...90%):  <120ns

Spectrum analyzer
Frequency: 0...100MHz
FFT points: 1M
FFT channels: all physical channels * viewports (same physical channel may be in different viewports, zoomed in different way) * math channels = many
Mask testing
Modes: YT, XY, FFT
Mask channels: estimated (12 viewports * 4 ch) = 48 masks simultaneous
Automatic measurements
Accuracy: 4-digit rounded at UI level, internally high precision floating point (can be accessed using "hacking")

Display
Interpolation: Linear or sin(x)/x
Persistence modes: Digital color, analog intensity, custom, fast or none
Zoom: 100Mx horizontal, 50x vertical in any timebase + offset + custom zoom

Connections
4xCH BNC; AWG; USB 2.0

Serial decoding
 (1-Wire, ARINC 429, CAN, CAN FD, DCC, DMX512, Ethernet 10Base-T, FlexRay, I²C, I²S, LIN, PS/2, SENT Fast, SENT Slow, SPI, UART/RS-232/RS-422/RS-485, USB (1.0/1.1))

Other features
math channels; event actions; supplied probes: 4xTA132 (1x/10x, 150MHz) incl. BNC adaptors

ANALOG BANDWIDTH TEST

VERTICAL SYSTEM INTEGRITY TEST

DC ACCURACY

FREQUENCY COUNTING ACCURACY

MICROPHONICS

CUSTOM PROBES

DIGITIZING ANALOG SIGNALS USING TRIGONOMETRY

DADC / VIRTUAL MSO USING EXTERNAL DAC

FREQUENCY RESPONSE ANALYSIS TOOLS

[JPortici]

It's dependent on usb load and on cpu/software load! notice that if you enable one or more decoders / math channels the waveform update rate drops down. ****

**** this, on the older pico with 48ks of memory. Probably on the new one it's using the additional memory for buffering and more wfms are transferred during CPU dead time?
i don't know. thinking loud.

+1 for math trickery

[MrW0lf]

PicoScope 2408B ANALOG BANDWIDTH

Officially it is 100MHz, 3.5ns rise. Bandwidth flatness (+0.3 dB, –3 dB) from DC to full bandwidth.

To test did set scope to 20ns/div, +-500mV (100mV/div), 1GSa/s RTS. Used good RF cables with 50ohm pass-thru (supplied probes were not tested).

First tried with sine wave from Siglent SDG2000X. Generator seems to drop ~-0.5dB@100MHz. Arrived at conclusion that scope drops <2dB@100MHz.

After that tried square from RF Explorer RFE6GEN. Gen is good to 6GHz. Outputs square wave. There should be no amplitude drop in 30...350MHz range. Could not choose ideal amplitude for scope vertical range due to fixed amplitude steps of generator. It is 592mVpp, so 59.2% of scope full range.





Compiled chart from both tests:



Since sine vs square results where quite different decided to go for classic BW(GHz)=0.35/RT[nsec] formula, set scope to 20GSa/s ETS:



0.35/2.521~=0.139 => Scope analog bandwidth is ~140MHz.

...but triggering system and auto-measurements work to at least 350Mhz.

[MrW0lf]

PicoScope 2408B VERTICAL INTEGRITY

Signal generator: Siglent SDG2000X         
Signal: sine wave, 100MHz         
Scope horiz. setting: 50ns/div
50 ohm system.
Generator amplitude drop at 100MHz is ~-0.5dB + drop due to scope analog bw limits.
1000x stats averaging applied.

Tested:
+-20mV...+-200mV ranges (4mV/div...40mV/div);
ETS 20GSa/s, RTS 1GSa/s & 500MSa/s & 250MSa/s;
Sin(x)/x ON|OFF for RTS (ETS does not support/need Sin(x)/x)

Most boring test I've ever done. Amplitude error with sine seems to stay below ~5% across all modes. At 250MSa/s it's of course quite wobbly visually, but this does not bother stats reading. Over time sample points pile up and give correct reading.

[Fungus]

Prices?

[lukier]

Mini-teardown?

[MrW0lf]

About prices... Mine is most expensive 2000 - 949€ in EU Farnell. Looking at a bit excessive analog bw - cheaper 70MHz maybe more suitable if budget is concern, especially 2+16 MSO version. Just today discovered in "fine print" that MSO digital part is up to 500MSa/s! Was thinking it's only 200MSa/s like my old 2205.
Teardown... not a single screw in sight. Will not risk the warranty anyway. That's not my cup of tea. As mostly "software guy" I like to mess with their brains, while Dave dissects the bodies 

[rstofer]

Prices?

About $1000
http://www.saelig.com/product/picoscope-2408b.htm

[JPortici]

freaking brexit! (i suppose)
prices went up 20% since some months ago

[cv007]

Has anyone used the usb1.1 decode feature, or how far up the Picoscope line does one have to go to have a usable usb decode (for full speed usb). 

[MrW0lf]

Has anyone used the usb1.1 decode feature, or how far up the Picoscope line does one have to go to have a usable usb decode (for full speed usb).

It's 12Mbit/s? Their modern stuff (large memory) start with 2206B. Maybe ask/search in their support forum to be sure? https://www.picotech.com/support/
This here is just my hobby thing and I'm much more into analog stuff, so can only guess.
However I planned to try decoding sometime, so I might try if find suitable gadget to sniff on. Arduino Due might be full speed USB?

[tautech]

Exact model SDG2**2X ?
3 to choose from, state yours please.

[MrW0lf]

Exact model SDG2**2X ?

Might find subtle clues on the pic

[tautech]

Exact model SDG2**2X ?

Might find subtle clues on the pic

Subtle ?   

SDG2042X hacked to SDS2122X   

Please carry on.

[Messtechniker]

Naughty boy....

[irakandjii]

I am very grateful for this review,  It will help me with my final selection of a diagnostic tool.  I am looking at the 3405D Mso or the 5442B or the 2208B if it can do the job.

I have been investigating these scopes for a few weeks.  My needs are heavily biased towards protocol investigations and analysis.

This puts serial decoding at the top of my list of priorities.  But closely followed by bandwidth and frequency analysis.

It turns out that ALL of the standard bench scopes I looked at below $2000.00 US have significant short falls for this purpose, leaving Picoscope as a sole contender.  I did find Keysight scopes >6k $ US that would work but that is a whole other universe for me.

Problem:  I have long path lengths for multiple serial busses, some upwards of 4 meters.  To make things worse these busses are vulnerable to flexing, stretching, vibration and significant thermal variations.  Think sensors or remote devices on the ends of a robot arm working outdoors in the Canadian sub-arctic (+45 degC to -50 degC) in high winds. p.s. Its why I want a robot.  A human standing in that is subject to pain.

Standard scopes typically save no more than 500 mSec or so of signal and in many cases far less.  Allowing for very short exchanges to be diagnosed.  The math follows:
 
Total Time of Acquired sample =  (Time scale per division X Number of divisions per screen X the number of Screens in memory) / Sample Rate

I need very long diagnostic samples (ideally 10-20 sec.) that can be reviewed and analyzed for faults back in the lab.  I require the samples to include protocol decode info plus the analog signals (for fault isolation).  To handle "the very long" worst case diagnostic scenarios I came to the conclusion that streaming the data to an external PC would be optimum.  The picoscopes can stream approx. 12 MS/sec for USB 2.0 models and 125 MS/sec for USB 3 models. ( I would write my own application using the SDK).  I am now trying to figure out if there are dead spots in the stream analogous to the screen refresh dead spots on regular scopes.  Regular scopes just don't stream.

For best case, 128-256 MS memory would be sufficient and cover a lot of use cases.  This scale of memory is generally missing in regular scopes.

I am very interested in the impact of the PC hardware, my plan is to use an original Surface Pro as the PC interface.  It is unclear to me if this will be sufficient.

I look forward to seeing this post progress

[MrW0lf]

PicoScope 2408B DC ACCURACY

While checking my AD584kH, AD588 voltage standards against stuff I got, developed technique to substantially exceed specified DC accuracy of this scope (±3% of full scale ±200uV), making use of math channels.

General idea: High sampling rate will generate noise, which will induce "artificial accuracy" when averaged.

Setup with some example values:
Scope horizontal setting: max RTS sampling rate, large amount of points (1GSa/s, 1M)
Vertical setting: good enough for measured (DC) voltage, make use of DC offset if single polarity (+-10V, -5V for ~12V stuff)
Make math channel formula that irons out DC error ((A+4.984)*1.0056)
Apply DC averaging auto-measurement to math channel.
Configure stats reading count that will give stable reading (64 in test, but much less ok for regular use).
Calibrate using high accuracy DMM for 0 and max volts. Watch out for ground loops!

After initial effort all this can be saved as settings file and reloaded any time.
Do not use "Zero offset", because this does not get saved ("DC offset" does).

Just as "sanity check" included pretty ok 4-digit DMM for comparison - UNI-T UT210E.

There is some (temperature related) drift with scope, but for hours it seems to hover in the couple mV vicinity for +-10V range. Looking at test results, this technique results in DC error mostly far <1%, without even utilizing full scale of 8bit ADC. Could be used to monitor battery voltage etc on dedicated channel no worse than average DMM.

[MrW0lf]

I am very grateful for this review,  It will help me with my final selection of a diagnostic tool.  I am looking at the 3405D Mso or the 5442B or the 2208B if it can do the job.

Think 3405D MSO is everything mine is + some on top of that + MSO. 5000 series is interesting with flexible bit resolution up to 16bit... but according to my findings even 8bit one do not exactly suffer from inaccuracy... 100MHz analog seems just about right, balances with MSO part which is max 500MSa/s = 100MHz/200Mbit also.

Edit: So much going on already forgot. You do not get cut out of 16bit game with MSO! Just pick any external ADC, feed into digitals and if your ADC voodoo is good its up to 16bit extra channel. For simple analog sensor listening youre already in with Arduino Uno/Due:
https://www.eevblog.com/forum/testgear/what-you-do-with-math-channels-on-oscilloscope/msg1106353/#msg1106353
Heres final version with faster scripts etc:
https://www.eevblog.com/forum/testgear/what-you-do-with-math-channels-on-oscilloscope/msg1130997/#msg1130997
Managed 59kSa/s@8bit with Uno,  687kSa/s@12bit with Due.



This works in standard software using math channels. Developed on my old 2205 MSO.
So with 3405D you get 4x8bit analog channels + 1x12bit sensor channel + 4x1bit digitals to spare. Or any other combination with digitals... Of course if you fire it all up at once it bogs down sampling rates making 50MHz max scope (Nyquist limit for 125MSa/s).

So you could calibrate formulas in standard GUI and then insert into your custom app. If you ever do this would be excellent to get some feedback about how it worked in real life.

Its why I want a robot.  A human standing in that is subject to pain

We here in Estonia know all about that Robot building is quite popular 

I am now trying to figure out if there are dead spots in the stream analogous to the screen refresh dead spots on regular scopes.  Regular scopes just don't stream.

Think no spots. Stream is stream, it goes into similar mode on ultra slow timebases. Going for USB 3 good idea also. I can look into that a bit on coming days. Doing this review-thing is very interesting because I learn new stuff every day.

I am very interested in the impact of the PC hardware, my plan is to use an original Surface Pro as the PC interface.  It is unclear to me if this will be sufficient.

My second PC is similar - Lenovo Yoga 14 - folds into (a bit heavy) tablet. Runs good there. Unless you go completely crazy with math channels processing load is very little.

I look forward to seeing this post progress

Think next part will be something for timing nuts (femtoseconds  ). After that will look into streaming and processing load then. Thanks for good feedback! 

[MrW0lf]

this technique results in DC error mostly far <1%

Did fire up same settings today. After "cold startup" zero point was drifted about -25mV vs U1282A. Waited for ~0.5h. Results with 64x averaging (values in volts, single +-10V (20Vpp) range as yesterday):

U1282A	2408B	diff	% of range
0.0045	-0.005	-0.0095	0.0475
4.9209	4.915	-0.0059	0.0295
9.938	9.927	-0.011	0.055
14.909	14.9	-0.009	0.045

After writing this 0-level is drifting between -2...0mV.

This cries for built in cal-tool, could cal all ranges easily with PSU/DMM at hand.
      

[irakandjii]

Fantastic! 

I will follow up on the MSO + ADC once I get the unit.  My requirement for 12 bit+ is a future need.  I want to play with some ultra-sonic gear and see if I can get some improved object or location sensing.  Your proposed solution will work wonders for that and give me more flexibility. 

Looks like the 3405 MSO is the way to go.  As far as I can see the differences are: memory amount, memory architecture HAL 3, more analog channels, USB 3.  I wanted both MSO and 4 analog channels.  I wanted 4 analog channels so I can decode any of the available protocols in Analog mode.

Thanks again and the new data looks great!

[MrW0lf]

PicoScope 2408B FREQUENCY COUNTING ACCURACY

This scope has sample jitter spec of 3ps RMS typical and supports 20GSa/s ETS. So I started getting ideas about pushing it much further than default 4-digit frequency reading.

After initial tests it was clear that in MHz range ETS is best option. You need both very high sampling rate, and at least 10k points for stats. Accessing extra accuracy goes like this:

1) apply standard counter for "sanity check"
2) make math channel counter "freq(A)", apply DC Average to it.

Suppose reading is 119.6MHz on both counters.

3) make second counter "freq(A) - 119E+6", apply DC Average to it.

Then you can read MHz part of reading from 1/2 and kHz part from 3.

Sadly no femtosecond-accuracy and guaranteed 6-digit reading. Worst error 1.165ps in this test making it typically 5-digit accurate. It's not a FemtoScope after all  Will do separate test for sub-MHz range and get my missing digits from there...

Think math channel (or even standard reading) digit count should made configurable in settings, so volt/timing nuts could access it without so much trouble...

Note that "standard error" means error for reading using standard counter func:

[MrW0lf]

PicoScope 2408B MICROPHONICS

Dave brought up microphonics issue with some scopes:


Did turn on ChA, ChB. Channels unterminated. Tuned trigger to barely rise from noise.

Tapping on table: no result
Tapping box by hand: no result
Tapping box with pen: no result
Strong blow directly above channels with pen: result

So for any practical intents and purposes - think no problem. Seems that effect has nothing to do with caps in this case. Purely "electron inertia" in conductive parts.

Now I have buy it new probe or something... Poor little thing never understood why such beatings 

[MrW0lf]

Had no time to roll review further but find interesting stuff time to time... So far been using 12Mbit/s USB isolator when working on "nasties" but this or similar could prove useful as well:

https://www.meilhaus.de/en/promo-ranger2201.htm

[alank2]

Excellent review MrW0lf.  I was going to do a picoscope review as well, and I may still someday, but you outdid me in your testing by far.  Good job!  I have a 2205A and a 2207B, and while the 2205A exceeds it specs, the 2207B exceeds them much further like you 2208B does.  I'm going to sell some equipment and get a 2407B as well so I have a 4 channel model.  One thing I really like about the picoscope software is the custom probe option.  I can setup a custom probe for my small PDI 60A two range clamp or the pintek DP-25 differential probe and you can even define the ranges so you can know when you are over range for the probe even if the scope isn't over range.

[MrW0lf]

Thanks Feel free to add some chapter here if like. I decided to check hardware first because if thats no good - little point to dig into software features. But luckily hardware is well engineered so software feature testing has green light. Had some other matters that stalled review progress but think time to resume.  Currently found offline postprocessing feature on saved data very useful. Will post couple words about it soon.