Test and Measurement Forum

Do any of your scopes support bit-depth over 8bits?

Post any questions you may have about our current range of oscilloscopes

Do any of your scopes support bit-depth over 8bits?

Postby Ben321 » Tue May 23, 2017 8:29 am

I noticed the specs on your scopes are all 8bits, and 12bits only with enhancement enabled. Do you actually sell any scopes that have bitdepth greater than 8 bits? I'm thinking 14 to 16 bits would be good actually. I know that even cheap soundcards have 16bit ADCs (analog to digital converters) in them, so it wouldn't seem to be too hard to take one of your 8bit scopes and replace its 8bit ADC with 16bit ADCs. Do any of your scopes in fact output 16bit samples? I hope so, because for my purposes 8bits is just not enough, and enhancements are unwanted. I want to be able to record at a sample rate of 20MS/S with 2 bytes per sample (that would be recording at a speed of 40million bytes per second, which is not outside the possibility of USB 2.0 to support such a high data rate, as USB 2.0 tops out at 60million bytes per second). I don't want any filtering either. I noticed for example that with Picoscope 2204A, it has a 100MS/S sample rate but only 10MHz bandwidth, which suggests additional filtering is being performed. An unfiltered ADC with a 100MS/S sample rate would have a 50MHz bandwidth (50MHz is the nyquist limit). The only filter that should exist is a very sharp one that cuts off at exactly 50MHz to prevent aliasing (frequencies higher than the nyquist frequency will appear to be lower than the nyquist frequency). However, for my experiment, I don't need even a 50MHz bandwidth. I only need a 10MHz bandwidth, which means I don't need a 100MS/S sample rate. I only need a 20MS/S sample rate. Anything with a higher sample rate only adds unnecessary cost to the price of the device. And of course (though I think all your scopes have this, I'm mentioning it because it's important) it should be DC-coupled

So to sumarize my question: Do you have any scopes that fit the following description?
DC-coupled
20MS/S sample rate
10MHz bandwidth (only filter is a very sharp one at the nyquist frequency)
At least 12bits per sample without enhancement (preferably 14 or 16 bits per sample)
Ben321
Newbie
 
Posts: 1
Joined: Tue May 23, 2017 8:09 am

Re: Do any of your scopes support bit-depth over 8bits?

Postby Martyn » Tue May 23, 2017 12:52 pm

Take a look at 4000 series which are 12 bit, or the 16 bit PicoScope 4262 which is designed for audio applications.

Additionally there are the flexible resolution 5000 series which can switch between 8/12/14/15/16 bits depending upon number of active channels and speed of sampling required.
Martyn
Technical Support Manager
Martyn
Site Admin
Site Admin
 
Posts: 2781
Joined: Fri Jun 10, 2011 8:15 am
Location: St. Neots

Re: Do any of your scopes support bit-depth over 8bits?

Postby Ben321 » Thu May 25, 2017 7:14 am

Martyn wrote:Take a look at 4000 series which are 12 bit, or the 16 bit PicoScope 4262 which is designed for audio applications.

Additionally there are the flexible resolution 5000 series which can switch between 8/12/14/15/16 bits depending upon number of active channels and speed of sampling required.


Thanks for the info.

There's still the other matter of sample rate versus bandwidth. Ideally, an ADC (analog to digital converter) would have a bandwidth exactly equal to half of its sample rate. So a 80MS/S ADC would have a 40MHz bandwidth. However, I notice all of your scopes have a much narrower bandwidth compared to their sample rate. For example, the Picoscope 4224 has a bandwidth of only 20MHz for a sample rate of 80MS/S. Why doesn't it have a 40MHz bandwidth? It seems that there's an extra filter somewhere in the signal path that significantly lowers the bandwidth below what it needs to. The only filtering needed for an ADC is one which limits the analog signal bandwidth to the nyquist frequency (half the sample rate). This is to prevent aliasing (a signal who's frequency is above the nyquist frequency will appear to have a frequency below the nyquist frequency). Any other filtering is unnecessary, and is actually bad if you want to be able to detect signals who's frequency goes all the way up to the nyquist limit.

Indeed, I want to maximize the bandwidth of the signal I'm recording, so having the signal being limited only by the antialias filter (a sharp lowpass filter, with its cutoff at the nyquist frequency) would be ideal for me.
Ben321
Newbie
 
Posts: 1
Joined: Tue May 23, 2017 8:09 am

Re: Do any of your scopes support bit-depth over 8bits?

Postby Gerry » Thu May 25, 2017 5:24 pm

Hi Ben321,

There is no extra filter in the signal path.

In the industry of Data Acquisition, the manufacturers goals are to provide a well specified, high quality instrument at a cost-effective price. When designing with ADC's you do indeed need to avoid aliasing and other phenomena that will degrade the performance of the product. However, multiple pre-digitisation anti-aliasing filters (for each channel) that track every selected sample rate, taking into account the resolution (detectable levels of aliasing), incorporate some form of switching for when ADC's are stacked, and other considerations are going to add significant cost to the product. Then there is the consideration of filter characteristics, and what you are forgetting is that with a filter cut-off at the bandwidth limit (-3dB), you still need some frequency range to actually roll-off the signal before it reaches the stop-band level (you can't get an analog brick-wall filter). So, Nyquist may theoretically be a desirable limit, but is not possible to achieve in practice in an analog front-end anti-aliasing filter. Also, with a filter you need to consider pass band ripple, and phase distortion that can affect signal quality (and in audio, which is a common application for our PicoScope 4262, it's important to avoid distortion that causes group phase delays for quality audio reproduction).

Then there is the matter of displaying the signal once you have captured the data, and in this respect limiting to just below the Nyquist frequency is not sufficient because you need more data points to be able to reconstruct the signal with any accuracy (there are numerous sources on the internet that explain this in more detail).

So what you will find is that manufacturers tend to prefer specifying a more restricted bandwidth in order to provide sufficient frequency range for the highest bandwidth signal to be naturally rolled off by the time the Nyquist frequency is reached, and to be able to correctly display the data. So we tend to aim for a sample rate that is about 5 times the maximum expected bandwidth, with some slight variations (other manufacturers may have different ratios). This allows our customers to achieve their bandwidth target with the correct selected PicoScope, while making a measurement that still meets the quality parameters mentioned in our specifications, without paying a small fortune for the privilege.

That said, we have customers with very specific needs that take it upon themselves to buy anti-aliasing filters to apply to their signal before measurement with a PicoScope in order to give them extreme quality sampling, for the measurements that require it. Of course, you can buy Oscilloscopes with anti-aliasing filters inside, but they tend to be much more expensive pieces of equipment.

Regards.

Gerry
Gerry
Technical Specialist
Gerry
PICO STAFF
PICO STAFF
 
Posts: 158
Joined: Mon Aug 11, 2014 11:14 am

Re: Do any of your scopes support bit-depth over 8bits?

Postby Ben321 » Sat Jun 03, 2017 2:44 am

Gerry wrote:Hi Ben321,

There is no extra filter in the signal path.

In the industry of Data Acquisition, the manufacturers goals are to provide a well specified, high quality instrument at a cost-effective price. When designing with ADC's you do indeed need to avoid aliasing and other phenomena that will degrade the performance of the product. However, multiple pre-digitisation anti-aliasing filters (for each channel) that track every selected sample rate, taking into account the resolution (detectable levels of aliasing), incorporate some form of switching for when ADC's are stacked, and other considerations are going to add significant cost to the product. Then there is the consideration of filter characteristics, and what you are forgetting is that with a filter cut-off at the bandwidth limit (-3dB), you still need some frequency range to actually roll-off the signal before it reaches the stop-band level (you can't get an analog brick-wall filter). So, Nyquist may theoretically be a desirable limit, but is not possible to achieve in practice in an analog front-end anti-aliasing filter. Also, with a filter you need to consider pass band ripple, and phase distortion that can affect signal quality (and in audio, which is a common application for our PicoScope 4262, it's important to avoid distortion that causes group phase delays for quality audio reproduction).

Then there is the matter of displaying the signal once you have captured the data, and in this respect limiting to just below the Nyquist frequency is not sufficient because you need more data points to be able to reconstruct the signal with any accuracy (there are numerous sources on the internet that explain this in more detail).

So what you will find is that manufacturers tend to prefer specifying a more restricted bandwidth in order to provide sufficient frequency range for the highest bandwidth signal to be naturally rolled off by the time the Nyquist frequency is reached, and to be able to correctly display the data. So we tend to aim for a sample rate that is about 5 times the maximum expected bandwidth, with some slight variations (other manufacturers may have different ratios). This allows our customers to achieve their bandwidth target with the correct selected PicoScope, while making a measurement that still meets the quality parameters mentioned in our specifications, without paying a small fortune for the privilege.

That said, we have customers with very specific needs that take it upon themselves to buy anti-aliasing filters to apply to their signal before measurement with a PicoScope in order to give them extreme quality sampling, for the measurements that require it. Of course, you can buy Oscilloscopes with anti-aliasing filters inside, but they tend to be much more expensive pieces of equipment.

Regards.

Gerry
Ben321
Newbie
 
Posts: 1
Joined: Tue May 23, 2017 8:09 am

Re: Do any of your scopes support bit-depth over 8bits?

Postby Ben321 » Sun Jun 04, 2017 10:01 am

For some reason, when I made the previous post, the quoted post remained, but my own content disappeared. So I'm posting again, this time with JUST the content, which was supposed to have been in the previous post.

Thanks for the info. However, that still doesn't take into account my intended use. I intend to be able to record at a bandwidth of 6MHz, to capture an NTSC color video signal, from the composite video output of a camcorder (maybe 1 second of recording, for a total of 30 frames). I then intend to write my own software which will load the raw recorded signal (a file that is saved by Picoscope, or at least I hope the Picoscope software can save the raw wave stream to a file) and then after loading this file, my program will decode it (taking into account the v-sync, h-sync, and color burst) and generate a series of picture files (one for each frame). However, in order to capture this video signal (which has a 6MHz bandwidth), I'll need a 12MS/S sample rate scope, but can't find any Picoscope hardware that fits these specs. In fact, if your rule of thumb were followed (samplerate = bandwidth * 5), I'd be buying a 30MS/S scope. Of course a 30MS/S scope costs more than a 12MS/S scope. Since I only need a 12MS/S scope for my experiment, I shouldn't have to pay for a 30MS/S scope. Of course, for such a wide bandwidth to be possible (going almost up to the nyquist frequency of 6MHz), the rolloff width for the antialias lowpass filter must be kept as small as possible. How close should it be to the nyquist frequency? Maybe within 10kHz of the nyquist frequency. Such a sharp lowpass filter, isn't impossible. It just requires a large number of passive components. It should be able to implement this as an analog IIR filter, using dozens of resistors and capacitors, and maybe inductors.
Ben321
Newbie
 
Posts: 1
Joined: Tue May 23, 2017 8:09 am

Re: Do any of your scopes support bit-depth over 8bits?

Postby bennog » Mon Jun 05, 2017 8:43 am

bennog
Advanced User
Advanced User
 
Posts: 91
Joined: Mon Nov 26, 2012 9:16 am
Location: Netherlands

Re: Do any of your scopes support bit-depth over 8bits?

Postby Ben321 » Tue Jun 06, 2017 4:35 am



8 bits scopes don't work too well, because the brightness range for an ntsc signal includes levels that are "darker than black" which are used for blanking (turning off the electron gun) and for sync (both horizontal and vertical). If you want 8 bits for brightness (displayable light levels) in the image portion of the signal you'll still need more bits to take into account the sync levels. The scope must have at least 9 bits per sample for my intended use. And to save money, it should have a sample rate no higher than twice the bandwidth.
Ben321
Newbie
 
Posts: 1
Joined: Tue May 23, 2017 8:09 am

Re: Do any of your scopes support bit-depth over 8bits?

Postby Martyn » Tue Jun 06, 2017 8:41 am

Unfortunately you are looking for a device that is tuned to your specific application, and it is very unlikely that you will find this in an off the shelf product. And at some point you will need to make a compromise, and this will either be in the specifications or the price.
Martyn
Technical Support Manager
Martyn
Site Admin
Site Admin
 
Posts: 2781
Joined: Fri Jun 10, 2011 8:15 am
Location: St. Neots


Return to USB PC Oscilloscopes

Who is online

Users browsing this forum: No registered users and 0 guests