Acquisition issues

I'm interested in buying a CS328A unit. I believe it has great performance. Also, i've been testing demo software from several vendors. By far, Cleverscope is the best. However, after reading user's manual downloaded from your web site, there are several things that remain unclear for me, mainly related to acquisition system.
User's manual says that there are several display modes. I don't understand what repetitive mode does exactly. Manual explains that several acquisitions of a repetitive signal are interlaced to get a higher resolution for HF signals. However to achieve this result, each acquisition must be slightly delayed from previous, otherwise, repetitive acquisitions of repetitive waveform will get always same data and therefore there will be no improvement at all. Is this the same feature referred as equivalent sampling rate by other vendors? If so, which is the maximum equivalent sampling rate that acquisition unit can achieve?
The second issue is related to the ""get frame"" feature since, obviously, can not be tested with demo software without actual acquisition unit. When this feature is enabled to transfer full buffer, is it performed in real time, so oscilloscope window can still be used normally? Maybe, i did not understand at all what is this feature intended for.

Best regards.

I'm interested in buying a CS328A unit. I believe it has great performance. Also, i've been testing demo software from several vendors. By far, Cleverscope is the best. However, after reading user's manual downloaded from your web site, there are several things that remain unclear for me, mainly related to acquisition system. User's manual says that there are several display modes. I don't understand what repetitive mode does exactly. Manual explains that several acquisitions of a repetitive signal are interlaced to get a higher resolution for HF signals. However to achieve this result, each acquisition must be slightly delayed from previous, otherwise, repetitive acquisitions of repetitive waveform will get always same data and therefore there will be no improvement at all. Is this the same feature referred as equivalent sampling rate by other vendors? If so, which is the maximum equivalent sampling rate that acquisition unit can achieve? The second issue is related to the ""get frame"" feature since, obviously, can not be tested with demo software without actual acquisition unit. When this feature is enabled to transfer full buffer, is it performed in real time, so oscilloscope window can still be used normally? Maybe, i did not understand at all what is this feature intended for. Best regards.

Hello Tiger,
Display modes: Repetitive mode is used to capture a signal whose bandwidth is higher than the instantaneous bandwidth of the scope, but inside the analog bandwidth. The scope samples at 100 MS/sec (ie 10ns resolution), and so has an instantaneous bandwidth of around 20 MHz. Above that there are insufficient samples per cycle to correctly see the waveform. A solution that works for stationary (that is unchanging) waveforms is to capture a whole bunch of them, based on a common trigger, and use the multiple sample points obtained to 'fill in the picture'. As you point out if the waveform is a low integer multiple of the sampling rate (ie 100 Mhz, 50 Mhz, 25 MHz) we have a problem, and you will not get access to all the points. However, it is rarely the case that the signal frequency is phase locked, and exactly the same frequency as the sampling system. So if we sample long enough (in our case 10ms), we can resolve down to frequency differences that are quite small (in our case we can handle all frequencies other than those within 100 Hz of the sampling frequency). Our sampling system, not being phase locked to the measured signal will sweep across it, generating points over the whole waveform. This is, as you say, equivalent time sampling. The maximum resolution is set by the time duration we sample for, and the number of time bins we resolve each sample into. In our case we offer a resolution of 200 ps.

The 'Get Frame' feature circumvents the usual display system (which decimates the 2M or 4M sample set into enough samples to fill the display, or generate an FFT), and decimates for either the full buffer, 50%, 10%, 5% or 2% of it. This sample set is then transferred to the PC, and is used from then on for waveform display. The oscilloscope window is used as normal. Doing a 'Save As' saves the transferred sample set, rather than the much smaller set used for updating the graphs. With 2M samples the files can be quite large (each values is a a real and occupies 8 bytes, and for an analog display only, this is 2M x 8 x 2 = 32M byte files).

So to summarize, usually only the samples required to fill a display are transferred, decimated out of the buffer, as required by the current graph setting (changing the graph timebase or time offset will result in a refresh from the buffer). We over-sample 8 times, so a 1000 pixel wide display will receive 8000 samples. This is done to keep things fast. We have a standard display update rate of 16 frames/sec.

Occasionally, when you want the whole buffer, you use 'get Buffer' to get the whole 2M samples. These are transferred as 2M x 32 words, or 4M x 32 words (depending on which Cleverscope you have), or 8M or 16 Mbytes. Even at full USB 2.0 speed of 30 Mbytes/sec this would take 260 msec, not taking into account other overheads, and so would result in a very sluggish system. We hope this helps.

Hello Tiger, Display modes: Repetitive mode is used to capture a signal whose bandwidth is higher than the instantaneous bandwidth of the scope, but inside the analog bandwidth. The scope samples at 100 MS/sec (ie 10ns resolution), and so has an instantaneous bandwidth of around 20 MHz. Above that there are insufficient samples per cycle to correctly see the waveform. A solution that works for stationary (that is unchanging) waveforms is to capture a whole bunch of them, based on a common trigger, and use the multiple sample points obtained to 'fill in the picture'. As you point out if the waveform is a low integer multiple of the sampling rate (ie 100 Mhz, 50 Mhz, 25 MHz) we have a problem, and you will not get access to all the points. However, it is rarely the case that the signal frequency is phase locked, and exactly the same frequency as the sampling system. So if we sample long enough (in our case 10ms), we can resolve down to frequency differences that are quite small (in our case we can handle all frequencies other than those within 100 Hz of the sampling frequency). Our sampling system, not being phase locked to the measured signal will sweep across it, generating points over the whole waveform. This is, as you say, equivalent time sampling. The maximum resolution is set by the time duration we sample for, and the number of time bins we resolve each sample into. In our case we offer a resolution of 200 ps. The 'Get Frame' feature circumvents the usual display system (which decimates the 2M or 4M sample set into enough samples to fill the display, or generate an FFT), and decimates for either the full buffer, 50%, 10%, 5% or 2% of it. This sample set is then transferred to the PC, and is used from then on for waveform display. The oscilloscope window is used as normal. Doing a 'Save As' saves the transferred sample set, rather than the much smaller set used for updating the graphs. With 2M samples the files can be quite large (each values is a a real and occupies 8 bytes, and for an analog display only, this is 2M x 8 x 2 = 32M byte files). So to summarize, usually only the samples required to fill a display are transferred, decimated out of the buffer, as required by the current graph setting (changing the graph timebase or time offset will result in a refresh from the buffer). We over-sample 8 times, so a 1000 pixel wide display will receive 8000 samples. This is done to keep things fast. We have a standard display update rate of 16 frames/sec. Occasionally, when you want the whole buffer, you use 'get Buffer' to get the whole 2M samples. These are transferred as 2M x 32 words, or 4M x 32 words (depending on which Cleverscope you have), or 8M or 16 Mbytes. Even at full USB 2.0 speed of 30 Mbytes/sec this would take 260 msec, not taking into account other overheads, and so would result in a very sluggish system. We hope this helps.

Thanks for your response. That explains most of my issues. When you say ""instantenous bandwith"", what do you mean? Is this the frequency of a sine wave that can be displayed without absolutely destroying the waveform (about 5 or 6 samples per cycle). A good view of a sine wave requiere much more samples per cycle (say 15 or 20).

You say that refresh rate is 16 frames/sec (ie, one screen refresh every 60msec). I suppose this refresh rate is limited mainly by data transfer over USB as well as graphic processing (since decimation is done by A.U.), so it should depend on actual pixel resolution of your scope window. However, acquisition unit is able to capture a new set of data every 10 msec, 6 times faster than screen refreshing. Of course, refreshing screen 6 times faster is not very useful, however, when working in repetitive mode, acquisition unit may put together up to 6 captures to resolve frequencies closer to sampling frequency (up tp 15 hz). Is it posible in any way?

My last question about acquisiton system is related to digital channels, a very useful feature of this scope. While datasheet says that analog channels can be sampled simultaneously at 100 MSa/sec, it is not clear if digital inputs are also sampled simultaneously at 100 MSa/sec (ie, there are 8 samplers) or, on the other hand, there is a single sampler shared by all digital inputs. In other words, what is the shortest pulse that cleverscope can resolve? Does it depend on how many channels are being used? Also, i would like to know how much sampling memory is available for digital channels. I suppose that resolution of both analog and digital channels should be the same.

Thanks.

Thanks for your response. That explains most of my issues. When you say ""instantenous bandwith"", what do you mean? Is this the frequency of a sine wave that can be displayed without absolutely destroying the waveform (about 5 or 6 samples per cycle). A good view of a sine wave requiere much more samples per cycle (say 15 or 20). You say that refresh rate is 16 frames/sec (ie, one screen refresh every 60msec). I suppose this refresh rate is limited mainly by data transfer over USB as well as graphic processing (since decimation is done by A.U.), so it should depend on actual pixel resolution of your scope window. However, acquisition unit is able to capture a new set of data every 10 msec, 6 times faster than screen refreshing. Of course, refreshing screen 6 times faster is not very useful, however, when working in repetitive mode, acquisition unit may put together up to 6 captures to resolve frequencies closer to sampling frequency (up tp 15 hz). Is it posible in any way? My last question about acquisiton system is related to digital channels, a very useful feature of this scope. While datasheet says that analog channels can be sampled simultaneously at 100 MSa/sec, it is not clear if digital inputs are also sampled simultaneously at 100 MSa/sec (ie, there are 8 samplers) or, on the other hand, there is a single sampler shared by all digital inputs. In other words, what is the shortest pulse that cleverscope can resolve? Does it depend on how many channels are being used? Also, i would like to know how much sampling memory is available for digital channels. I suppose that resolution of both analog and digital channels should be the same. Thanks.

Hello Tiger,
Taking your last question first - all channels are sampled simultaneously at 100 Msa/sec. The 8 digital channels are synchronous with the analog channels (so you should see the rising edge on Chan A being the same as a rising edge on In 1 with probes connected together). The shortest pulse that Cleverscope can guarantee to resolve is just a bit longer than 10ns. Shorter pulse might or might not be seen, dependant on where the sample instant is. As all channels are sampled at the same time, it doesn't matter how many are being used. All of them can capture a 10ns pulse at the same time.

Each channel has its own 4M or 8M sample buffer. The resolution of the analog channels is 10, 12 or 14 bits depending on which sampling module (CS710, CS712 or CS714) is fitted. The resolution of a digital channel is just 1 bit. The threshold can be set between 0 and 10V.

The instantaneous bandwidth depends on how you use it. If you are using the spectrum analyser, just two samples is enough, and so you can happily measure signals betwen 0 and 50 Mhz. In fact if you understand aliasing you can use the spectrum analyser to about 150 Mhz and still get useable results. If you are looking at a time representation of a waveform, then yes, you need a few samples.

You are right that the update rate is set by the graphic processing (which is PC dependant), the USB transfer rate, and the decimation time (which is done in the Acquisition Unit). The decimator must sometimes process a complete frame - 2M or 4M samples (eg when the display is 20ms wide with 4M samples, or 40ms wide with 8 M samples). The peak capture decimator takes 14ns per sample, so a 4M sample frame can take 56ms to process.

The issue of not being able to resolve signals close to the sampling frequency has turned out not be too important. We have a 50 ppm oscillator , and most other oscillators are also 50-200 ppm. So even if they are meant to be the same frequency, 100 Mhz +/-50 ppm is 100 MHz +/- 5000 Hz. You can see the probability of them being exactly the same is low.

We hope these notes have answered your questions.

Hello Tiger, Taking your last question first - all channels are sampled simultaneously at 100 Msa/sec. The 8 digital channels are synchronous with the analog channels (so you should see the rising edge on Chan A being the same as a rising edge on In 1 with probes connected together). The shortest pulse that Cleverscope can guarantee to resolve is just a bit longer than 10ns. Shorter pulse might or might not be seen, dependant on where the sample instant is. As all channels are sampled at the same time, it doesn't matter how many are being used. All of them can capture a 10ns pulse at the same time. Each channel has its own 4M or 8M sample buffer. The resolution of the analog channels is 10, 12 or 14 bits depending on which sampling module (CS710, CS712 or CS714) is fitted. The resolution of a digital channel is just 1 bit. The threshold can be set between 0 and 10V. The instantaneous bandwidth depends on how you use it. If you are using the spectrum analyser, just two samples is enough, and so you can happily measure signals betwen 0 and 50 Mhz. In fact if you understand aliasing you can use the spectrum analyser to about 150 Mhz and still get useable results. If you are looking at a time representation of a waveform, then yes, you need a few samples. You are right that the update rate is set by the graphic processing (which is PC dependant), the USB transfer rate, and the decimation time (which is done in the Acquisition Unit). The decimator must sometimes process a complete frame - 2M or 4M samples (eg when the display is 20ms wide with 4M samples, or 40ms wide with 8 M samples). The peak capture decimator takes 14ns per sample, so a 4M sample frame can take 56ms to process. The issue of not being able to resolve signals close to the sampling frequency has turned out not be too important. We have a 50 ppm oscillator , and most other oscillators are also 50-200 ppm. So even if they are meant to be the same frequency, 100 Mhz +/-50 ppm is 100 MHz +/- 5000 Hz. You can see the probability of them being exactly the same is low. We hope these notes have answered your questions.