I need a new oscilloscope - 300€?
- Thread starter rock soderstrom
- Start date
Help Support GroupDIY Audio Forum:
I came to the same conclusion. The Analog Discovery 2 should be fine for frequency response, but not for THD nor noise measurements.Gold said:
For me, the absence of a frontend with calibrated attenuators and physical time base control is a deal-breaker.shot said:
Tubetec said:
The software is free to download. You can check it out and see if you like it.
I haven’t tried connecting both the dedicated interface and a soundcard simultaneously. It looks like that should work. You can have more than one AD2 connected to the software simultaneously.
jacomart
Well-known member
abbey road d enfer said:
who knows, maybe dithering techniques are used to reduce noise. In some cases SFDR improvements of over 20/25 dB can be achieved. So far I have not been able to find adequate information, we need a volunteer to do some tests! ;D
Cheers
JM
ok, quickly reading this thread on the Digilent forum I understand that there is no trace of dithering and that the measurable THD is about -70dB. Closed speech, I'm going to dinner, happy Sunday evening everyone!
https://forum.digilentinc.com/topic/19205-high-thd-at-analog-discovery-2-input/
Bo Deadly
Well-known member
Well you got me on this one jacomart. The datasheet does say 8 bits. I'm completely flummoxed by that. I guess to get up to MHz they had to limit the bit depth. It must be that the range setting adjusts the level going into the ADC so it can more fully utilize the very limited bit depth. That does make sense because when I adjust the range on my cheapo Rigol I can hear relays switching inside. And it explains why the FFT feature on digital storage scopes is completely useless. And the waveform in the display is presumably being oversampled / interpolated. At least it looks like it's a lot more than 256 steps.jacomart said:
If I ever get a new scope, I'm definitely going for 10 bit. That must make a huge difference in how the display looks.
jacomart
Well-known member
Surely the displayed waveforms take a sort of "makeup". A better resolution (10 or even 12 bits) helps in the exact reproduction of signals, e.g. glitches and their settling times. Instead 14 bits are too few for FFT, SNR and THD analysis as the theoretical SNR of a 14-bit ADC is about 86db but unfortunately the number of real bits is lower due to the ADC distortion (the infamous SINAD parameter ) and for this reason the real SNR can worsen significantly. For Analog Discovery 2 in the forum they say that it has a measured SNR of about 70dB, in this case the real number of bit (ENOB) is about 11dB instead of the theoretical 14dB. Oversampling and decimation are a different story but it seems that Analog Discovery doesn't use them.squarewave said:
Cheers
JM
Agreed. that's why we need to extract and measure separately noise and distortion from a signal. Magnifiers.Tubetec said:
rock soderstrom
Tour de France
Ok, the thread is developing splendidly, very good information in here now. Thanks!
I have looked at all the solutions mentioned, checked tons of PDFs, reviews and videos - my head hurts a bit now.
I must say I'm surprised what is possible
with today's devices, in positive as well as partly negative ways.
On the pro side are the really extensive possibilities especially of the computer based devices.
The "eLab in a box" fraction offers functions for almost all needs and that for a small dollar. That would have been unthinkable years ago. Great!
On the other hand I am surprised how many current devices still work with low bit resolution.
Another handicap, which I don't quite understand, is the way USB-based devices are presented. A buyer of the Analog Discovery 2 sums it up well in an Amazon review:
"What annoys me, however, are the fiddly connections via pin header. Sure, for the Arduino or Raspberry Pi hobbyists handy, because you can pin directly to a breadboard, but if you want to measure "normal" with a probe, it needs either the optionally available "BNC Adapter for Analog Discovery" board or you tinker something yourself. Here, a larger housing with already integrated BNC sockets would have been the much better choice, especially since it could quietly be included at the price."
I 100% agree. I understand the modular and open approach, but even the adapter solutions are not really suitable for everyday long term use.
I'm totally with Abbey on that, the front end is not good and yes, real buttons still have advantages.
What I didn't know is that Waveforms is freely available and also works with sound cards or USB audio interfaces. Thanks for this info! I'll give that a try today....
For those this is new territory, as it is for me, here's some good info on the subject.
https://www.nutsvolts.com/magazine/article/turn-your-computers-sound-card-into-a-scope
I have looked at all the solutions mentioned, checked tons of PDFs, reviews and videos - my head hurts a bit now.
I must say I'm surprised what is possible
with today's devices, in positive as well as partly negative ways.
On the pro side are the really extensive possibilities especially of the computer based devices.
The "eLab in a box" fraction offers functions for almost all needs and that for a small dollar. That would have been unthinkable years ago. Great!
On the other hand I am surprised how many current devices still work with low bit resolution.
Another handicap, which I don't quite understand, is the way USB-based devices are presented. A buyer of the Analog Discovery 2 sums it up well in an Amazon review:
"What annoys me, however, are the fiddly connections via pin header. Sure, for the Arduino or Raspberry Pi hobbyists handy, because you can pin directly to a breadboard, but if you want to measure "normal" with a probe, it needs either the optionally available "BNC Adapter for Analog Discovery" board or you tinker something yourself. Here, a larger housing with already integrated BNC sockets would have been the much better choice, especially since it could quietly be included at the price."
I 100% agree. I understand the modular and open approach, but even the adapter solutions are not really suitable for everyday long term use.
I'm totally with Abbey on that, the front end is not good and yes, real buttons still have advantages.
What I didn't know is that Waveforms is freely available and also works with sound cards or USB audio interfaces. Thanks for this info! I'll give that a try today....
For those this is new territory, as it is for me, here's some good info on the subject.
https://www.nutsvolts.com/magazine/article/turn-your-computers-sound-card-into-a-scope
Bo Deadly
Well-known member
Unfortunately it seems to be limited to 16 bit. A USB audio interface cannot measure DC or high frequency signals like digital. So using WaveForms with a USB audio interface is basically useless for anything but a narrow range of low resolution low frequency AC-only measurements.rock soderstrom said:
jacomart
Well-known member
squarewave said:
Could one of the problems be that the ADC filtering cuts at something less than about 0.5*F(sampling) ?
A solution could be the use of a high speed 24 bit converter like the TI ADS127L01 of which there is an evaluation board which costs about 112 €. However, the software problem would remain ...
https://www.ti.com/tool/ADS127L01EVM
Cheers
JM
Bo Deadly
Well-known member
I was just talking about the "WaveForms" software. It's simply not designed to handle samples larger than 16 bit because it's oscilloscope software. Even if you could interface the WaveForms software with the chip you cite, it's bandwidth is limited to 100kHz so again, it would not be suitable for MHz digital signals.jacomart said:
Clearly there is an inverse relation between bit width and speed. If you want high res 24 bit like an audio spectrum analyzer, it's limited to 100kHz or so. Each sample is chewing up many clock cycles to do decimation and filtering and other such things and apparently you just can't do those things at higher speeds. It's just not possible to sample a tiny voltage once and get an accurate value.
I could imagine a technique where multiple faster chips sample a subset of the signal either divided up into frequency bands and / or signal levels. For example, maybe one could use overlapping filters for 10-100Hz, 100Hz-1KHz, 1KHz-10KHz and >10KHz and then feed that into 4 lowe-bit / faster ADCs. You could then limit the FFT window to those bands and average the low band 10x more than the next (meaning 1000x, 100x, 10x, 1x respectively). This could make for a really high res, realtime display that is high res at low frequencies when the display is log axis. Kind of a niche product but just thinking out loud.
jacomart
Well-known member
squarewave said:
Yes sure! in fact I just wanted to say that using low cost sound cards would make no sense and that the software problem would not be solved anyway.
Cheers
JM
squarewave said:
I don't think this is true. My Native Instruments Konnect2 will do 192K 32bit fixed or float according to the driver device settings in Waveforms. I haven't checked the measurements to see what the real world limitations are but the ASIO driver recognizes this high bit and sample rate interface.
Bo Deadly
Well-known member
And that all makes sense because the scope needs to be able to zoom in on a tiny voltage. So yes it will fully utilize the interface bit depth. But being an oscilloscope the WaveForms code only needs to manipulate and display a relatively narrow bit width. So it would also make sense that the code only uses 2 bytes per sample. Any more than that and it would just be wasting a potentially huge amount of memory trying to process samples at MHz frequencies.Gold said:
However, maybe the spectrum analzyer can switch to using 24 bit or maybe more. All of the screenshots of the spectrum analyzer window on the digilentinc.com website look like the noise floor bottoms out at around -96dB which would be the limit for 16 bits. For pro audio, you need to go quite a bit lower than that. It is not particularly difficult to make a mic pre with an SNR more than 120 dBFS. For that you need 24 bits. Maybe you can check and report back? Set up you USB interface in loopback, play a computer generated pure tone and see how low the noise floor is. Loopback should be able to get to -120 dBFS and triple zero THD no problem. That would prove for sure that the spectrum analyzer is using 24 bit samples in which case the could be interesting. Still no good for digital of coruse but I don't care about that. I have a regular scope for that. Most runtime SA displays are pretty chunky at low frequencies with a log axis. If WaveForms improves on that, I would want to try it for sure.
squarewave said:
Because they are using the 14 bit AD2 interface. The software will display any level and frequency range you set it up for. I just tried measuring a tone at -110dBfs with the Konnect2 and it worked and displayed fine.
Since the AD2 interface is 14bit if you want the features of the interface like attenuator and time base switching for oscilloscope functions you are limited to the 14 bit interface. If you use a 24/96 sound card you are only limited by its specs. You don't get the nice front end of the AD2 though.
rock soderstrom
Tour de France
squarewave said:
Sure, the sound card solution has limitations, especially for power users. For my primary purpose, signals in the audible range, the software already offers a lot!
I played around with it for about an hour after smooth installation. I immediately got results with a Steinberg UR22 MKII interface. Scope, signal generator and FFT I was able to test, my noise floor is clearly below -100dB.
It's definitely too early to make a qualified statement, but I think the soundcard WaveForms solution makes sense. Better than nothing.
I think the strongest limitation is the limited frontend.
For free, very promising so far, I will test further with it.
Edit:
PS: The combination of a sound card and Waveforms does not replace a real scope, but can still be useful.
