Why are there no 24-bit dso's?

[scatterandfocus]

In the world of audio, amplitude resolution is equally as important as frequency resolution.  And the generally accepted useful frequency range for audio is 20hz to 20khz (the audible range), although some audio circuit designers argue that there is human sensory benefit in capturing analog signals well above the audible range.  I won't go there, but there is undisputed benefit in digital audio sampling for having a significantly higher sampling rate outside of the audible range for pushing anti-aliasing filtering way above the audible range, which allows for less steep filtering.  It isn't uncommon for digital audio recording/playback interfaces to have available sampling rates of 96khz, 192khz, or even 384khz.  In terms of oscilloscope frequency range, even a very high 384khz audio sampling rate is very lowly, and some of the lowest cost 'toy' scopes far exceed that reaching into the multiple Mhz range.  But on the vertical resolution side of oscilloscopes, 8-bit sampling is still the most common, which is horribly low in the audio world.  And even 14-bit is very lowly in terms of audio amplitude, giving a dynamic range of 84 dB, where everyone today is making 24-bit audio devices (144 dB).  Given the upper limits  of analog frontends (~115 dB, give or take ~6 db or 1 bit)  24-bits aren't fully utilized, but it exceeds the range of the analog frontend by enough that it isn't the limiting factor.  In other words, 24-bits well captures the full dynamic range of the cleanest available analog audio signals.

So then, why isn't vertical resolution more of a concern for dso's?  I realize that testing audio circuits is only one of many uses for dso's, but is quantization noise not a major concern in other areas of electronics?  Is there a tradeoff to be had for more amplitude resolution and less frequency resolution?  Why are there no dso's which are weighted toward audio and other uses which involve concern for dynamic range?  It seems to me that at least for audio, a dso with a much lower frequency resolution in trade for a much higher vertical resolution could be very useful.  Also, with a 24-bit depth, the full range doesn't necessarily need to be utilized in cases where frequency resolution is much more important.  In the audio world, with a 24-bit device we can record at 16-bit or lower. 

[Kleinstein]

There is a natural trade of between amplitude resolution and bandwidth through noise. A slight step in the direction of high vertical resolution, but lower speed are some of the modern DMMs, like the Keithley DAQ6500 or DMM7510: not quite 24 Bits, but at least some 16-18 Bits and 1 MHz. There are slightly different functions, but the function of a slow high vertical resolution is there.

The normal scope has limited graphical resolution.

For higher dynamic range at high frequency there are spectrum analyzers.

[David Hess]

There *are* 24-bit DSOs but they are called audio analyzers, FFT spectrum analyzers (one channel), or dynamic signal analyzers (two channels) and are more specialized instruments than a DSO.  Because of the tradeoff between resolution and bandwidth, they are limited to 100s of kHz.

[Tom45]

24 bits would need 16 million pixels on the vertical axis. There just aren't any displays that come close to that.

I suppose some kind of Y-axis zooming could be used, but what would that really tell us? What would you hope to see with 24 bits that can't be seen with 8 or 10?

As David Hess points out, specialized tools have been developed to fulfill the needs of audio analysis.

[edigi]

I've never checked so far, but I'm surprised if no dozens of such project existed that tried to convert a sound card to an oscilloscope's front end.
(Just did a quick search, first result shows immediately such scope, albeit just 16 bit resolution...)
I think you're looking in the wrong place for solution.
The demand for audio DSOs would be far too low for any serious player to care but I'd expect quite some open source projects for this.

[Fungus]

So then, why isn't vertical resolution more of a concern for dso's?

a) Because we're mostly interested in the shape of signals. We're not trying to measure anything accurately.
b) Because it would be very expensive to make a 24-bit ADC that samples in the GHz range.
c) Because it would need a lot more RAM and processing power (wider lanes in the FPGAs) - more cost.

and

d) Because more bits only reduces the noise floor. DSO input noise is high enough that adding more bits won't achieve anything.

Watch this video:

[radiolistener]

amplitude resolution and time resolution are coupled together. If your scope has 8-bit 1 GHz sampling rate, it means within 20 kHz bandwidth it has dynamic range 6.02*8 + 1.76 + 10 * log10(1GHz/(2*20kHz)) = 94 dB. You can achieve such dynamic range just by applying Low Pass Filter with 20 kHz cut-off.

94 dB is about 15.3 bits of amplitude resolution. For most amateur cases this resolution (of low cost oscilloscope) is pretty enough.

15 bits is 32768 points. Did you seen display with 32k vertical resolution in order to display all details?

[scatterandfocus]

Very interesting, radiolistener.  If I'm calculating right, that should give modern entry level dso's that are 8-bit a dynamic range of around 84 to 87 dB @ 100 Mhz to 200 Mhz, which is far better than 48 dB without filtering.  At 12-bit 200 Mhz, that gives a dynamic range around 110 dB, which is approaching the edge of analog audio limits.  And 14-bit 100 Mhz gives 120 dB which I think is past the limits of analog audio dynamic range.  It slopes off pretty quick past around 1 Mhz.

And no, we aren't going to see any useful noise detail on a low resolution display without zooming in vertically.

[radiolistener]

scatterandfocus, yes, but there is no 14 bit ADC which can works at 1 GHz sample rate. Even 100-200 MHz 14 bit ADC has real resolution (equivalent resolution) about 12 bit. The best 16 bit high speed ADC has real resolution about 13 bits at 100 MHz sampling rate. If you're going above 1 GHz, then even 8 bits is unreachable resolution

So, in reality you cannot get even 140 dB for 20 kHz bandwidth

[scatterandfocus]

radiolistener, is that a limit of the type of adc being used in scopes, or something else?  On the analog side, I think I haven't seen any audio devices with a dynamic range up to 120 dB.  So yea, if 14-bit is effectively 12-bit, something like 12-bit 1G hz would just about cover it giving a dynamic range around 118 dB, which would be a fairly expensive scope for audio use.

[radiolistener]

scatterandfocus, this is limit of current technology for electronics. If you increase bandwidth you will lose dynamic range. If you increase dynamic range you will lose bandwidth. Both parameters are coupled together.

For example you can use averaging in order to increase precision of your measurement, but it requires more measurements, so it takes more time. You get precision, but lose measurement speed and vice versa. The same happens with ADC bits resoulution and bandwidth. More bandwidth = less bits resolution, because it is more hard to measure something with high precision at high measurement speed

ADC with wide bandwidth (which is used in modern oscilloscope) doesn't lose it's dynamic range, it's just spreading it over wide bandwidth and if you apply digital filter, you can get high dynamic range, but lose wide bandwidth. This is why high speed ADC has low bits resolution, because they have high time resolution.

[tv84]

Did you seen display with 32k vertical resolution in order to display all details?

That's 1.5x the resolution of the Dubai Mall's screen! 

[colorado.rob]

If you're going above 1 GHz, then even 8 bits is unreachable resolution

So, in reality you cannot get even 140 dB for 20 kHz bandwidth

Please see the Xilinx RFSoC chips...

https://www.xilinx.com/products/boards-and-kits/zcu111.html

Just imagine the sort of test instruments one could build with this thing.

 

[Tom45]

scatterandfocus, yes, but there is no 14 bit ADC which can works at 1 GHz sample rate.  ...

But the original query was about use with audio signals. A 1 GHz sample rate certainly isn't needed for audio.

[radiolistener]

Please see the Xilinx RFSoC chips...
Just imagine the sort of test instruments one could build with this thing.
 

according to their documentation it has SNR=58.33 dB at 3.93216GSPS. It means that effective resolution of their ADC at full bandwidth is just 9 bits.

And if you apply low pass filter with 20 kHz cut-off, you can achieve just 58.33 + 10*log10( 3.93216GSPS / (2*20kHz) ) = 108 dB dynamic range for 20 kHz bandwidth on this super duper ADC

108 dB is about 17.6 bits. So, this super duper ADC allows to provide just 17-18 bits for 20 kHz bandwidth

You will use a lot of resources of very expensive FPGA which working at crazy speed and has high power consumption just to get the same result which can be provided with cheap 18 bits ADC which works at 40 kHz and has very small power consumption.

[ogden]

But the original query was about use with audio signals. A 1 GHz sample rate certainly isn't needed for audio.

There's serious problem with assumption that audio signals shall be analyzed with DSO.

hint: https://www.rohde-schwarz.com/my/product/upv-productstartpage_63493-7558.html

[thm_w]

There's serious problem with assumption that audio signals shall be analyzed with DSO.

hint: https://www.rohde-schwarz.com/my/product/upv-productstartpage_63493-7558.html

THD of -110dB (0.0003%), dynamic range 120dB and noise floor -140dB is crazy. It does cost $10k+ though..

[ogden]

THD of -110dB (0.0003%), dynamic range 120dB and noise floor -140dB is crazy. It does cost $10k+ though..

Yes. That's why sane solution is to stop whining and start thinking - maybe 24-bit USB audio card is all you need.
[edit] I did not address you but OP obviously   

[colorado.rob]

Please see the Xilinx RFSoC chips...
Just imagine the sort of test instruments one could build with this thing.
 

according to their documentation it has SNR=58.33 dB at 3.93216GSPS. It means that effective resolution of their ADC at full bandwidth is just 9 bits.

And if you apply low pass filter with 20 kHz cut-off, you can achieve just 58.33 + 10*log10( 3.93216GSPS / (2*20kHz) ) = 108 dB dynamic range for 20 kHz bandwidth on this super duper ADC

108 dB is about 17.6 bits. So, this super duper ADC allows to provide just 17-18 bits for 20 kHz bandwidth

Oh, no doubt it's stupid to use this for audio.  I was just responding to your comment that 8 bits at 1GHz was unreachable.

[radiolistener]

colorado.rob, well... but everything changes and progress does not stand still 

So, let's say 8 bit is not reachable for 10-20 GHz sampling rate 

[Someone]

colorado.rob, well... but everything changes and progress does not stand still 

So, let's say 8 bit is not reachable for 10-20 GHz sampling rate 

Lecroy disagree:
http://cdn.teledynelecroy.com/files/whitepapers/comparing-high%20resolution-oscilloscope-design%20approaches-wp.pdf
Add enough parallel ADCs and you can get to the performance you desire, brute force but works well with Moore's law.

[Rerouter]

It is not particularly easy to make a true 24 bit ADC even at a few KHz sample rate, (says the guy who is trying to improve a 24 bit ADC with only a 50Hz sample rate..) the quantisation noise goes down in most cases as you increase sample rate, but the wider your system bandwidth, the faster the op amps and resistors end up swamping out any chance at 24 bit worth of signal to noise.

There are fun things like oversampling to shape noise out of your bandwidth of interest, but there are limits to this, In reality, I think you will find any audio sample rate "24 bit" ADC will have a noise free number of bits closer to 16-18

[rodpp]

 This NI daq board for audio and vibration signals has 24bits ADC, but only 102.4Khz sample rate:

http://www.ni.com/pdf/manuals/373861d.pdf

[Frex]

Hello,

Maybe you can look about a new open source project i started to design,
a 2 channels analyzer based on LTC2380-24 high performance 24bits 1.6MSPS SAR ADC.
I started a thread about it on the DIY audio forum here :
https://www.diyaudio.com/forums/equipment-and-tools/335005-osva-source-versatile-analyzer.html

I started this project after previous prototype described in the other topic here:
https://www.diyaudio.com/forums/equipment-and-tools/292107-sar-adc-performance-audio-adc-project-ltc2380-24-a.html


Frex

[scatterandfocus]

Frex, it looks like an interesting project, but it is way above my head right now.