NEW Keysight HD3

[EEVblog]

in class ... Keysight slides compare HD3 against Tek MSO 4B and R&S MXO4 who are the leaders in the 4 class.   The "fastest update rate" seems to be just a false claim as R&S MXO4 does 4.5 M wfms/s.   Maybe Keysight will correct before intro. 

In the promotional comparison tables I've seen, the HD3 is shown as 1.6M vs the MXO4's 4M, but they use the words "*Uncompromised" for the Keysight and "*Conditionally" for the MXO4, which does sound correct in my testing of the MXO4 IIRC.

[pdenisowski]

In the promotional comparison tables I've seen, the HD3 is shown as 1.6M vs the MXO4's 4M, but they use the words "*Uncompromised" for the Keysight and "*Conditionally" for the MXO4, which does sound correct in my testing of the MXO4 IIRC.

I’m very curious as to what “uncompromised” and “conditionally” mean.  Are they claiming their waveform update rate is unaffected by other settings? 

And MXO is 4.5 million, for the record.

[EEVblog]

I’m very curious as to what “uncompromised” and “conditionally” mean.

New marketing wank words

Are they claiming their waveform update rate is unaffected by other settings?

Yes, as expected, just like the Megazoom IV scopes.

[maxwell3e10]

https://youtu.be/UmFTR4v7M1Y?t=333

This indicates 825 kHz update rate with all 4 channels on for MXO4

[maxwell3e10]

Just for the record, world's first 14-bit scope:

[electr_peter]

Just for the record, world's first 14-bit scope:

14-bit when moon and stars align - only 100MSa/s and 20MHz BW (while in 8-bit mode 1GSa/S and 200MHz BW). With such logic, one can consider first 14-bit (and 16-bit, and 24-bit) scopes were PC sound cards. Or even DMMs with data logging.

[hpw]

Just for your information as digital generated (none coherent) & dither signal as on 10...16Bits.

It shows up very clearly, better FFT Windows algorithm should be given (as currently seen), as the bell rises even on 14 bit and heavy on 16 bit even using BH7 FFT Window.

[maxwell3e10]

Just for the record, world's first 14-bit scope:

14-bit when moon and stars align - only 100MSa/s and 20MHz BW (while in 8-bit mode 1GSa/S and 200MHz BW). With such logic, one can consider first 14-bit (and 16-bit, and 24-bit) scopes were PC sound cards. Or even DMMs with data logging.

I would consider a scope is something with different voltage and timescale settings and a display. The first 12-bit scope or any digital scope was Nicolet 1090 from 1973 with a 1 MS/s rate. The first 16-bit scope as far as I know was Yokogawa DL708 from late 1990s with a 100kS/sec rate, I still use it sometimes.

[Someone]

In the promotional comparison tables I've seen, the HD3 is shown as 1.6M vs the MXO4's 4M, but they use the words "*Uncompromised" for the Keysight and "*Conditionally" for the MXO4, which does sound correct in my testing of the MXO4 IIRC.

I’m very curious as to what “uncompromised” and “conditionally” mean.  Are they claiming their waveform update rate is unaffected by other settings? 

And MXO is 4.5 million, for the record.

Well, it means almost nothing as there are various conditions under which it has been literally compromised:

[egonotto]

Just for the record, world's first 14-bit scope:

Hello,

no, Picoscope 5000 series were there earlier and Red Pitaya and Analog Discovery have also been around for a while

Best regards
egonotto

[maxwell3e10]

In the promotional comparison tables I've seen, the HD3 is shown as 1.6M vs the MXO4's 4M, but they use the words "*Uncompromised" for the Keysight and "*Conditionally" for the MXO4, which does sound correct in my testing of the MXO4 IIRC.

I’m very curious as to what “uncompromised” and “conditionally” mean.  Are they claiming their waveform update rate is unaffected by other settings? 

And MXO is 4.5 million, for the record.

Well, it means almost nothing as there are various conditions under which it has been literally compromised:

I would not say that a reduction of waveform update rate with the timebase setting is a compromise, it is just conservation of time. The percentage of time in data acquisition (which MXO4 displays) is a better metric.  On the other hand, if the waveform update rate drops for more than 1 channel active, it is a serious compromise.

[jusaca]

I would not say that a reduction of waveform update rate with the timebase setting is a compromise, it is just conservation of time. The percentage of time in data acquisition (which MXO4 displays) is a better metric.  On the other hand, if the waveform update rate drops for more than 1 channel active, it is a serious compromise.

Exactly. When time base is 1ms/div the complete capture takes 10ms. So more than 100 waveforms/s would be rather magical.
I think a value of the "dead time" in percent or rearm time or something like that is a better approach for comparison.

[Someone]

Well, it means almost nothing as there are various conditions under which it has been literally compromised:
[IMG]

I would not say that a reduction of waveform update rate with the timebase setting is a compromise, it is just conservation of time. The percentage of time in data acquisition (which MXO4 displays) is a better metric.  On the other hand, if the waveform update rate drops for more than 1 channel active, it is a serious compromise.

Look closely and see if the change is predictable/explainable. I think it isn't in those plots and it would be fair to say the 3000X had a compromised update rate compared to the 2000X (other brands with more "controls" over the parameter affecting would likely have been able to avoid/uncompromise that).

Compromises that are baked in and not left for the user to trade off are still compromises.

[Someone]

I would not say that a reduction of waveform update rate with the timebase setting is a compromise, it is just conservation of time. The percentage of time in data acquisition (which MXO4 displays) is a better metric.  On the other hand, if the waveform update rate drops for more than 1 channel active, it is a serious compromise.

Exactly. When time base is 1ms/div the complete capture takes 10ms. So more than 100 waveforms/s would be rather magical.
I think a value of the "dead time" in percent or rearm time or something like that is a better approach for comparison.

Indeed you could transform that same dataset into a different metric such as % blind time. Up to you.

[EEVblog]

I think a value of the "dead time" in percent or rearm time or something like that is a better approach for comparison.

I agree, I like the way the MXO4 presents that. IMO it shouldn't be hidden away in a menu, it should be on the main display.

[Anthocyanina]

Well, it means almost nothing as there are various conditions under which it has been literally compromised:
[IMG]

I would not say that a reduction of waveform update rate with the timebase setting is a compromise, it is just conservation of time. The percentage of time in data acquisition (which MXO4 displays) is a better metric.  On the other hand, if the waveform update rate drops for more than 1 channel active, it is a serious compromise.

Look closely and see if the change is predictable/explainable. I think it isn't in those plots and it would be fair to say the 3000X had a compromised update rate compared to the 2000X (other brands with more "controls" over the parameter affecting would likely have been able to avoid/uncompromise that).

Compromises that are baked in and not left for the user to trade off are still compromises.

i think that apparent lower update rate on the 3000x when it intersects with the 2000x might be due to the larger memory on the 3000x series. when on small enough timebases, the memory goes beyond what's on the screen, so at 2.5 times the max sample rate, and 4 times the memory, there will be timebases (as seen on the display) that will essentially be treated as longer timebases by the acquisition system(because there are more timebases on the 3000x that will lead to data captured outside the screen, which you then can zoom out to see) when comparing the 2000x vs the 3000x scopes at apparently the same settings. i think this could be checked by looking at what timebases the 2000x and 3000x scopes change sampling rate

[Someone]

Well, it means almost nothing as there are various conditions under which it has been literally compromised:
[IMG]

I would not say that a reduction of waveform update rate with the timebase setting is a compromise, it is just conservation of time. The percentage of time in data acquisition (which MXO4 displays) is a better metric.  On the other hand, if the waveform update rate drops for more than 1 channel active, it is a serious compromise.

Look closely and see if the change is predictable/explainable. I think it isn't in those plots and it would be fair to say the 3000X had a compromised update rate compared to the 2000X (other brands with more "controls" over the parameter affecting would likely have been able to avoid/uncompromise that).

Compromises that are baked in and not left for the user to trade off are still compromises.

i think that apparent lower update rate on the 3000x when it intersects with the 2000x might be due to the larger memory on the 3000x series. when on small enough timebases, the memory goes beyond what's on the screen, so at 2.5 times the max sample rate, and 4 times the memory, there will be timebases (as seen on the display) that will essentially be treated as longer timebases by the acquisition system(because there are more timebases on the 3000x that will lead to data captured outside the screen, which you then can zoom out to see) when comparing the 2000x vs the 3000x scopes at apparently the same settings. i think this could be checked by looking at what timebases the 2000x and 3000x scopes change sampling rate

a) those scopes do not capture outside the screen in run mode
b) even if they did capture outside the screen, that would be compromising the waveform rate

[Anthocyanina]

a) those scopes do not capture outside the screen in run mode
b) even if they did capture outside the screen, that would be compromising the waveform rate

are the 2000x and 3000x series doing something different from the 1000x series? if not, then they do capture data outisde the display while the timebase*number of divisions on screen is smaller than the total amount of time it would take for the memory to fill at the max sample rate. i remember on my 1102g, even if i stopped the scope(pressing run/stop) at 2ns/ i could zoom out until about 25us/. it just captured only what was on the screen when the sample rate went below the max. i think if the 2000x and 3000x scopes use the same megazoom architecture, they are very likely to behave that way.

if they do capture beyond the screen under those conditions, that wouldnt be compromising since all the data is there

[electr_peter]

MegaZoom IV scopes capture only what's visible on the screen (plus few points extra) when in RUN mode. Only when STOP/SINGLE is pressed, full memory is in use. It has to do with clever automatic memory management.
Of cource, screen = full memory on very slow time bases.

[Someone]

a) those scopes do not capture outside the screen in run mode
b) even if they did capture outside the screen, that would be compromising the waveform rate

are the 2000x and 3000x series doing something different from the 1000x series? if not, then they do capture data outisde the display while the timebase*number of divisions on screen is smaller than the total amount of time it would take for the memory to fill at the max sample rate. i remember on my 1102g, even if i stopped the scope(pressing run/stop) at 2ns/ i could zoom out until about 25us/. it just captured only what was on the screen when the sample rate went below the max. i think if the 2000x and 3000x scopes use the same megazoom architecture, they are very likely to behave that way.

if they do capture beyond the screen under those conditions, that wouldnt be compromising since all the data is there

Even if they were capturing around the screen (which they do not) that is compromising the update rate as the information outside the screen is not being displayed despite being stored. Which scope when stopped will show an intensity graded histogram of the region outside the screen?

This comes back to the conflating of trigger rate vs display update rate.

[tooki]

In the promotional comparison tables I've seen, the HD3 is shown as 1.6M vs the MXO4's 4M, but they use the words "*Uncompromised" for the Keysight and "*Conditionally" for the MXO4, which does sound correct in my testing of the MXO4 IIRC.

I’m very curious as to what “uncompromised” and “conditionally” mean.  Are they claiming their waveform update rate is unaffected by other settings? 

And MXO is 4.5 million, for the record.

Well, it means almost nothing as there are various conditions under which it has been literally compromised:

The inherent maximum update rate based on time base is not “compromise”.

[tooki]

Well, it means almost nothing as there are various conditions under which it has been literally compromised:
[IMG]

I would not say that a reduction of waveform update rate with the timebase setting is a compromise, it is just conservation of time. The percentage of time in data acquisition (which MXO4 displays) is a better metric.  On the other hand, if the waveform update rate drops for more than 1 channel active, it is a serious compromise.

Look closely and see if the change is predictable/explainable. I think it isn't in those plots and it would be fair to say the 3000X had a compromised update rate compared to the 2000X (other brands with more "controls" over the parameter affecting would likely have been able to avoid/uncompromise that).

Compromises that are baked in and not left for the user to trade off are still compromises.

Keysight is actually pretty clear on what they mean by “uncompromised” because they talk about it all the time in their marketing materials and technical info. In a nutshell, they mean “unaffected by enabling other features, like measurements, digital channels, serial decoders, etc”. It doesn’t mean “entirely unaffected by anything ever”, it simply means unaffected by enabling other scope features, all else held equal.

a) those scopes do not capture outside the screen in run mode
b) even if they did capture outside the screen, that would be compromising the waveform rate

are the 2000x and 3000x series doing something different from the 1000x series? if not, then they do capture data outisde the display while the timebase*number of divisions on screen is smaller than the total amount of time it would take for the memory to fill at the max sample rate. i remember on my 1102g, even if i stopped the scope(pressing run/stop) at 2ns/ i could zoom out until about 25us/. it just captured only what was on the screen when the sample rate went below the max. i think if the 2000x and 3000x scopes use the same megazoom architecture, they are very likely to behave that way.

if they do capture beyond the screen under those conditions, that wouldnt be compromising since all the data is there

Even if they were capturing around the screen (which they do not) that is compromising the update rate as the information outside the screen is not being displayed despite being stored. Which scope when stopped will show an intensity graded histogram of the region outside the screen?

This comes back to the conflating of trigger rate vs display update rate.

See also the attached PDF, which goes into all the things discussed here so far, and explains the definitions of different metrics, including the math behind them.

[pdenisowski]

See also the attached PDF, which goes into all the things discussed here so far, and explains the definitions of different metrics, including the math behind them.

Thanks!  We have a similar whitepaper (written last year - Keysight paper is from 2018, I believe)

https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/pdm/cl_brochures_and_datasheets/white_paper/3609_0543_52/Selecting--Why-fast-update-rate-matters_wp_en_3609-0543-52_v0100.pdf

[Someone]

In the promotional comparison tables I've seen, the HD3 is shown as 1.6M vs the MXO4's 4M, but they use the words "*Uncompromised" for the Keysight and "*Conditionally" for the MXO4, which does sound correct in my testing of the MXO4 IIRC.

I’m very curious as to what “uncompromised” and “conditionally” mean.  Are they claiming their waveform update rate is unaffected by other settings? 

And MXO is 4.5 million, for the record.

Well, it means almost nothing as there are various conditions under which it has been literally compromised:
[IMG]

The inherent maximum update rate based on time base is not “compromise”.

As already replied to above, the plot shows not just limitation on some simple characteristic (trigger re-arm, or pixel rate, or some other obvious limitation) but a rather complex mix of constraints on all those scopes.

Well, it means almost nothing as there are various conditions under which it has been literally compromised:
[IMG]

I would not say that a reduction of waveform update rate with the timebase setting is a compromise, it is just conservation of time. The percentage of time in data acquisition (which MXO4 displays) is a better metric.  On the other hand, if the waveform update rate drops for more than 1 channel active, it is a serious compromise.

Look closely and see if the change is predictable/explainable. I think it isn't in those plots and it would be fair to say the 3000X had a compromised update rate compared to the 2000X (other brands with more "controls" over the parameter affecting would likely have been able to avoid/uncompromise that).

Compromises that are baked in and not left for the user to trade off are still compromises.

Keysight is actually pretty clear on what they mean by “uncompromised” because they talk about it all the time in their marketing materials and technical info. In a nutshell, they mean “unaffected by enabling other features, like measurements, digital channels, serial decoders, etc”. It doesn’t mean “entirely unaffected by anything ever”, it simply means unaffected by enabling other scope features, all else held equal.

a) those scopes do not capture outside the screen in run mode
b) even if they did capture outside the screen, that would be compromising the waveform rate

are the 2000x and 3000x series doing something different from the 1000x series? if not, then they do capture data outisde the display while the timebase*number of divisions on screen is smaller than the total amount of time it would take for the memory to fill at the max sample rate. i remember on my 1102g, even if i stopped the scope(pressing run/stop) at 2ns/ i could zoom out until about 25us/. it just captured only what was on the screen when the sample rate went below the max. i think if the 2000x and 3000x scopes use the same megazoom architecture, they are very likely to behave that way.

if they do capture beyond the screen under those conditions, that wouldnt be compromising since all the data is there

Even if they were capturing around the screen (which they do not) that is compromising the update rate as the information outside the screen is not being displayed despite being stored. Which scope when stopped will show an intensity graded histogram of the region outside the screen?

This comes back to the conflating of trigger rate vs display update rate.

See also the attached PDF, which goes into all the things discussed here so far, and explains the definitions of different metrics, including the math behind them.

That PDF does not explain uncompromised, as you expand on. Which if you want to be specific for an Agilent DSO-X 3000:

Turning on math, can increase or decrease the trigger rate
Turning on more analog channels, can reduce the trigger rate (without changing memory depth)
Hiding channels, can increase the trigger rate (without changing memory depth)
Adding digital channels, can reduce or (when causing a change in sample rate due to memory sharing) increase the trigger rate
Turning on serial decode, can (when causing a change in sample rate due to memory sharing) increase the trigger rate

Those were your choices of parameters that would cause the update rate to be "unaffected". I can readily reproduce them significantly affecting the update rate. Claims of uncompromised on that measure are false.

Fixed operation that have little or no user control makes it appear/feel like there is no compromise. But when you look past the (I feel obnoxious) marketing smoke and mirrors there are some ways to push/prod the underlying mechanisms and expose the compromises. Reducing the configuration space and possibility for user errors is not 100% bad or 100% good, it can suit different users or situation. However those fixed choices are compromises, and the physical hardware is plainly/evidently/known to be capable of more, but HPAK decided they would NOT expose that.

Compromised: "cause to function less effectively or become vulnerable"

[EEVblog]

a) those scopes do not capture outside the screen in run mode
b) even if they did capture outside the screen, that would be compromising the waveform rate

are the 2000x and 3000x series doing something different from the 1000x series?

The 1000X uses the exact same Megazoom IV ASIC.