Our hacker [haas] is at it again with the Haasoscope Pro, a full redesign of the original Haasoscope, which was a successful Crowd Supply campaign back in 2018.
This new Pro version was funded on Crowd Supply in April this year and increases the bandwidth from 60 MHz to 2 GHz, the vertical resolution from 8 to 12 bits, and the sample rate from 125 MS/s to 3.2 GS/s. Selling for $999 it claims to be the first open-everything, affordable, high-bandwidth, real-time sampling USB oscilloscope.
The firmware and software are under active development and a new version was released yesterday.
The hardware has an impressive array of features packed into a slick aluminum case with quiet 40 mm internal fan and 220 x 165 x 35 mm (8.66 x 6.5 x 1.38 in) form-factor weighing in at 0.9 kg (1.98 lbs). Also available is an active probe supporting up to 2 GHz analog bandwidth.
The Haasoscope Pro is miles ahead of alternatives such as this USB oscilloscope from back in 2010 and you can find a bunch of support material on GitHub: drandyhaas/HaasoscopePro.
In what universe is a 3.2 GS/s scope a 2 GHz scope?
Credibility spent.
Nyquist to the head.
https://www.eevblog.com/forum/crowd-funded-projects/haasoscopepro-open-source-2-ghz-oscilloscope!/msg5722779/#msg5722779
Ha – yes! Got there 3.2×10^-9 seconds before me.
It allows you to daisy-chain two together via a CAT5 to get 6.4, so maybe that’s the reason behind the claim? Still a lot less than the “equivalent” Siglent. Otherwise, totally agree :)
A lot less $$, I meant to say
When “ you can flexibly combine and sync multiple Haasoscope Pros to interleave ADCs for a sample rate of 6.4 GS/s”.
HaHa never fails to disappoint. At least he’s consistent.
Consistently right.
No.
Doesn’t matter what you can do with two of them.
That’s not a qualified claim.
It’s BS.
Whoops, accidentally clicked report instead of reply.
Agreed. Saw this a few days ago, and while it’s a really cool project/product, click baiting capability runs me the wrong way.
What the heck are you talking about? You can verify the analog bandwidth of a scope beyond Nyquist. The signal just aliases down, but it’s still there.
The analog bandwidth is higher, BNC is typically rated to 4.
Makes no difference.
You don’t report a digital scopes analog bandwidth.
That’s like rating a speaker at peak to peak, clear scammer territory.
Also Nyquist is theoretical maximum.
At 2x frequency sample rate you are getting 2 points per cycle (duh).
Could be square, could be sawtooth, you have no idea.
You don’t know the peak to peak voltage for sure, enough samples should get you there, unless syncpulse or similar.
In the real world practical scope bandwidth is sample rate / 3.
Some will argue for 2.6.
Nobody thinks 2 samples/cycle is a good rate.
Yeah. No. That’s not at all right.
First: every scope I own and every ADC I’ve ever worked with reports both analog bandwidth and sampling rate. This is totally normal.
It may seem odd to you to have a scope with a bandwidth higher than Nyquist, but it’s totally normal. Scopes normally are heavily oversampled to be user friendly, but it does limit their functionality (and I have ripped out filters to get closer to the analog chain’s max bandwidth). The ADC I’m using now has 4 GHz 3 dB analog bandwidth and 5 GSa/s sampling., which allows you to easily work in the second Nyquist zone of the converter.
You’re clearly used to working with scopes and not signal converters directly. Saying things like “you don’t know if it’s square or sawtooth” – yeah, if I have a signal that only has frequency components up to Nyquist in analog bandwidth it by definition is not a square wave.
Analog bandwidth is reported, but not first line, large font.
First line is useable bandwidth.
Sampling oscilloscopes have been around for more than half a century. They require repetitive signals, so you can’t record a one-time burst at the claimed frequency response.
The same universe in which a 40MS/s HP54100D can have a bandwidth of 1GHz. Or a Tek1502 ~30kS/s with a bandwidth of ~4GHz.
Hint: you don’t understand Nyquist’s theorem.
Plus, as others have pointed out, you can parallel two of the Haasocopes to increase the sampling rate.
The creator also experimented a bit with overdriving the ADC sample (and FPGA) and achieved 6(ish) GS/s…
Plenty of ADCs have analog bandwidth above Nyquist. There are lots of uses for it, so long as you can isolate the aliasing.
According to the crowdsupply page linked above, the 2 GHz is ‘unlocked’ when using two scopes:
“You can combine two Haasoscope Pro units to achieve 6.4 GS/s on a single channel (unlocking the full 2 GHz bandwidth) or distribute the bandwidth across multiple channels (two channels at 3.2 GS/s each, one at 3.2 GS/s and two at 1.6 GS/s, or four at 1.6 GS/s).”
Still blows my mind that a 3.2 GHz 12 bit ADC can exist.
But Digikey is listing them for $2k (onesies, retail). How can the whole ‘scope come in at half that?
I did a: git clone https://github.com/drandyhaas/HaasoscopePro.git
and looked at the pdf of the schematic (I am still using KiCad V8 so can’t open the schematic itself right now) ADC used is: ADC12DL2500
For 10+, Digikey lists them as USD 575, and mouser for USD494
https://www.digikey.com/en/products/detail/texas-instruments/ADC12DL2500ACF/22462685
https://eu.mouser.com/ProductDetail/Texas-Instruments/ADC12DL2500ACF?qs=dbcCsuKDzFVNM4pXaT3ETg%3D%3D
But for me, putting in an ADC12DL500 (Listed as USD 168 and USD 145) would already be overkill.
For the rest, I wanted to have a closer look at the frontend, and what’s actually in it, but the schematic is drawn completely atrocious, with relays presented as an 18 pin dip in the schematic and opamps as rectangular blocs with pins at weird locations. Relays seem to be around bypassable Chebychev filters, but at a glance I can’t even find any gain stage in the signal path.
I am not sure what the target market of this thing is. Best guess is that it’s not intended as an general purpose oscilloscope, but as a special purpose instrument for some specific purpose, but that is not mentioned in the documentation for as far as I could see it (just gave it around 10 minutes of reading).
But still, Mouser lists the whole thing for USD860, and that does seem quite low considering the cost of the ADC alone. There may be a market for this thing, (but it’s not for me).
https://eu.mouser.com/ProductDetail/Crowd-Supply/HAASOSCOPE-PRO-01?qs=bN4HulGPsrGuOKJv67xg8Q%3D%3D
I’m not sure why you think this isn’t a general purpose scope. That’s definitely the goal – although clearly specialized for high bandwidth.
I wrote lots of project logs here on Hackaday.io…
https://hackaday.io/project/200773/logs
There’s one in particular on the front end so you can understand the signal chain there…
https://hackaday.io/project/200773-haasoscope-pro/log/236963-walkthrough-of-the-front-end-design
Feel free to reach out on there if you have any questions!
Yes, this amazingly affordable ADC was key:
https://www.ti.com/product/ADC12DL1500
It would be nice if the video actually showed anything.
And again, still without a decent analog frontend (max 3V input in 1MOhm input mode, according to the hackaday project. And an 8mV/div sensitivity is also pretty abysmal.
I liked the part of the “Haasoscope Pro v29 Updates” video(Released today) where he shifts the data a bit to compensate for sampling inaccuracies, but then, when the oversampled sine is jumping around with around 30 degree phase jitter and he calls it “pretty acceptable” (@11:00) I lost my last bit of interest. His approximation algorithm is completely failing in that case. That can still be improved in software later, but the too simplistic analog frontend can not.
Overall, I would be more interested in a relatively simple scope myself, as long as both the hardware and software is “pretty good”. Something comparable to the Owon VDS1022I (EUR 110, 100Msps, 25MHz bandwidth, 8-bit), but with good Linux support, as a companion to my Siglent SDS1104X-E (Which is quite good, but has a far too loud fan). A few months ago I had a brief look at the picoscopes, but they have made a very weird user interface, and prices go up extremely quickly for anything but their simplest model.
Also had a look at the Labnation Smart scope. (Also Open source Software), but it’s relatively expensive for the hardware, also has a very limited front end, and there does not seem to be much development going on over there for quite a few years now, and that also made me hesitant.
Batronix Magnova baby!
Ah, yes, Impressive specifications, doesn’t have a fan so it’s silent and probably a decent performance to price ratio, Nice big screen, impressive user interface, both touch and buttons… But it’s much more scope then I can justify for the work I do. These things start at EUR 3500. I paid EUR 450 for my Siglent, and it’s already more capable then what I need. (But I would like a bigger screen and quieter fan).
…afforda… BWAHAHAHAHA! At one grand?!? I can get 3-4 VERY usable scopes for that. No matter how huge of an open-everything nutjob I might be, first I gotta EAT.
To clarify, it’s +-30V in 1 MOhm mode with a 10:1 probe. Also, since it’s 12-bits, that’s equivalent to a 16x better sensitivity than an 8-bit scope, so 8 mV/div is really more like 500 uV/div since you can zoom in digitally. These are pretty standard specs, I think.
Thanks for your thoughts about the software trigger stabilizer. I’ll consider updating it soon to be even more precise when sinc upsampling is being done.
Well, no, “digitally zoom in” does not cut it, sorry. The sparkle of 12B ADCs just isn’t there in an old 8B ADC. It’s not the same. While it’s OK to not support higher input voltages at the scope itself, as they are unviable at high bandwidths, I find it odd to pair this with a High-Z input. It’s either High-Z and High-Voltage or none of them. High-Z + Low-Voltage is just begging for people to break the input.
Also, there is no separate trigger circuit with time-to-digital conversion. So the sub-sample time of the trigger is lost, so the sample that matched the trigger will jump around, and so will the whole captured signal. That’s only OK in a streaming architecture or maybe in a single-shot architecture. Yet usually a scope overlays many captures to capture rare effects. You’ll have to bend over backwards to support this, and play tricks like correlate the captures to overlay them properly (yet, that’s just a guess about the signal and nothing real).
Also, I see USB 2.0 I/O – USB 4.x is the way to go. So that you can actually sample often or even stream samples into the computer. You can actually buy laptops with Thunderbolt 5 which offers 120 Gbps in one direction.
Finally, I still consider this schematic a rage bait. You got so much helpful in the EEVBlog forum.
According to the crowdsupply page linked above, the 2 GHz is ‘unlocked’ when using two scopes:
“You can combine two Haasoscope Pro units to achieve 6.4 GS/s on a single channel (unlocking the full 2 GHz bandwidth) or distribute the bandwidth across multiple channels (two channels at 3.2 GS/s each, one at 3.2 GS/s and two at 1.6 GS/s, or four at 1.6 GS/s).”
2Ghz scope as this price is a pretty nice achievement.
However I would backed funding if it had a better memory spec, or ability to stream samples to PC at high speed (as Thunderscope does)
Oh wow, those schematics are horiffic…. I’d be embarrassed to release something like that with my name on it.
I don’t have much confidence in the project with those schematics, doesn’t bode well for the rest of the design…. They need a complete redraw.
i keep looking at these USB scopes and every time i’m impressed by the price and then absolutely repelled by the fact that it doesn’t have a decent analog frontend. i need my test equipment to accept the largest voltage on my bench (roughly 20V) without complaint or pre-configuration, and i’d really like it to accept 240VAC just as easily. i.e., i want auto-ranging. a scope without auto-ranging would only be attractive if i didn’t have any choice, but the $500 cheapo scopes are great these days.
and this one isn’t even cheap. i’m sure someone will appreciate it on bandwidth alone but not me.
The Nyquist theorem says you must sample at Greater Than twice the highest frequency. If you sample at a little more than Nyquist, it will take a little less than an infinite number of samples to know the signal is present. If you sample at exactly twice you will get zero or a pair of oscillating constants. One sees various rules like sample at 3 times or at 5 times, etc. Realistically, oversample as much as you can. I like numbers that are both powers of 2 and have an integer square root. Like 4 and 16 and 64. This makes the arithmetic easier.