New scope - LeCroy HDO9000

[Wuerstchenhund]

Just got news that LeCroy came out with a new scope:

LeCroy HDO9000




- BW 1/2/3/4GHz
- up to 40GSa/s
- 10bit ADC
- 128Mpts sample memory
- 16ch MSO optional, plus a High-speed Digital Analyzer option
- 15.4" 1280x800 touchscreen with multi-touch and gestures

Datasheet:
http://cdn.teledynelecroy.com/files/pdf/hdo9000-oscilloscope-datasheet.pdf

List prices from $21k for the 1GHz version to $37k fpr the 4GHz variant.

The HDO9000 is clearly positioned against the Keysight Infiniium S which is Keysight's 10bit proposition with BW ranges from 500Mhz to 8GHz, sample rates up to 20GSa/s, up to 800Mpts memory and a 15" XGA (1024x768) touchscreen display. Prices for the DSO-S start at $18k for the 500Mhz variant, at S21k for the 1GHz variant, at $37k for the 4Ghz variant and go up to $69k for the 8GHz variant. The prices are similar to the HDO900 which suggests that LeCroy has stopped to undercut Keysight on price.

Remarkable is that, like with the WaveRunner 8000 that came out earlier this year, LeCroy has again settled for what today in this class can only be considered as very moderate sample memory sizes (128M vs up to 800M on the Keysight).

On the other side, all the typical LeCroy features are there, i.e. WaveScan which is free (while the inferior Keysight InfiniiScan costs another $1200 or so), the UI has been modernized with multi-touch, gestures and tabs, and the list of available options is pretty long as one would expect.

Still, the small sample memory of the HD9000 is very disappointing, especially when considering that this time the distance to the competition is much smaller than ever before. The DSO-S is really a very attractive proposition.

Would be interesting to see both scopes in a trial.


Update: stupid thought error removed

[pascal_sweden]

The design looks similar to the Siglent SDS3000

[Wuerstchenhund]

The design looks similar to the Siglent SDS3000

That is because the SDS3000 is partially a LeCroy design and LeCroy specified the housing and front panel (so that the optics fit to their HDO Series).

It seems they have now standardized on that format for their new scopes. I'm not a huge fan of it, though (I liked previous formats like the one of the WR6zi or the "suitcase" style WRXi much better).

[pascal_sweden]

With some of these WaveRunner scopes you can change the screen from landscape to portrait:
http://www.avalonsellslecroy.com/lecroy_waverunner_640zi.htm

First time I see this actually

[nfmax]

Remarkable is that, like with the WaveRunner 8000 that came out earlier this year, LeCroy has again settled for what today in this class can only be considered as very moderate sample memory sizes (128M vs up to 800M on the Keysight). Which is even more surprising when considering the increased amount of data that the 10bit ADC is pushing into the sample memory, i.e. it requires more memory to capture the same period at the same sample rate as an 8bit ADC.

Surely not? The memory sizes quoted for both scopes are (mega) points, not bytes. So the extra word width called for by a 10 bit ADC is 'included in the price'. I believe the acquisition memory word is wider anyway, to allow for averaging and/or hi res modes (whatever they call them).

[Wuerstchenhund]

Remarkable is that, like with the WaveRunner 8000 that came out earlier this year, LeCroy has again settled for what today in this class can only be considered as very moderate sample memory sizes (128M vs up to 800M on the Keysight). Which is even more surprising when considering the increased amount of data that the 10bit ADC is pushing into the sample memory, i.e. it requires more memory to capture the same period at the same sample rate as an 8bit ADC.

Surely not? The memory sizes quoted for both scopes are (mega) points, not bytes. So the extra word width called for by a 10 bit ADC is 'included in the price'. I believe the acquisition memory word is wider anyway, to allow for averaging and/or hi res modes (whatever they call them).

You're absolutely right. My bad! 

[tautech]

The design looks similar to the Siglent SDS3000

Very much so.

When Mr W has one in his hot little hands he might like to tell us where they're made.

[Wuerstchenhund]

The design looks similar to the Siglent SDS3000

Very much so.

When Mr W has one in his hot little hands he might like to tell us where they're made.

I'm sure it'll be a while until I see one in real-life but I bet it's made in USA as all other HDO Series scopes are. Or pretty much all current LeCroy scopes with the exception of WaveJet (Iwatsu, Japan), WaveSurfer 3000 and WaveAce (both Siglent, China).

[Muxr]

Wonder how good its FFT is, there is also a "Spectrum Analyzer and Advanced FFT Option HDO9K-SPECTRUM" available.

[Wuerstchenhund]

Wonder how good its FFT is

FFT goes up to the full sample memory (128Mpts).

there is also a "Spectrum Analyzer and Advanced FFT Option HDO9K-SPECTRUM" available.

This is pretty much just a SA-like UI so that the FFT part can be operated like a SA. It also comes with markers and some other functionality.

Without the option you get the same resolution but you have to use the FFT UI (which is a bit more cumbersome).

[EEVblog]

Remarkable is that, like with the WaveRunner 8000 that came out earlier this year, LeCroy has again settled for what today in this class can only be considered as very moderate sample memory sizes (128M vs up to 800M on the Keysight). Which is even more surprising when considering the increased amount of data that the 10bit ADC is pushing into the sample memory, i.e. it requires more memory to capture the same period at the same sample rate as an 8bit ADC.

I don't think so. It's Mpts not MB, i.e. sample points. So doesn't matter what the sample size is.

[Wuerstchenhund]

I don't think so. It's Mpts not MB, i.e. sample points. So doesn't matter what the sample size is.

I guess you missed nfmax' post from yesterday then 

Plus my response

[richpike]

It doesn't appear to be a 10-bit ADC from reading the manual.  It sounds like it automatically adjusts resolution based on the number of channels and setting from 8-bits up to 10-bits.

From the manual - Page 49 (my emphasis):
*************************************
HD1024 high definition technology offers more quantization levels and greater precision when
digitizing waveforms. HD1024 dynamically reconfigures the ADC based on the number of active
channels and applies bandwidth filters in response to the timebase settings to prevent waveform
aliasing while maintaining high resolution. The HD Mode setting on the Timebase dialog
enables/disables HD1024. When HD Mode is Off, the oscilloscope operates in continuous 8-bit
resolution.


HD1024 is enabled by default on all HDO9000 oscilloscopes.

Note: HD Mode does not work with Sequence, Roll, or RIS Sampling Mode acquisitions; these
options are disabled when in HD Mode. Sample Rate is restricted to a minimum 500 ms/s and
the Time/div to a maximum 500 ?s. Pre-processor Enhanced Resolution (ERes) and userdefined
linear Interpolation are also disabled; interpolation is handled automatically.

HD1024 Methodology
The HD1024 technology utilizes two principal methods for maintaining high resolution: dynamic ADC
configuration (managing the number of active channels and resolution) and optimized bandwidth
filtering at the input.
Dynamic ADC Configuration
Interleaving allows the oscilloscope to put memory where it is needed to maintain the highest
possible sample rate, and the principal way to do this is to combine digital capabilities into a limited
number of active channels. There are three Active Channels options you may select from the
Timebase dialog. When in HD Mode, this selection affects the number of bits resolution at which any
channel may operate, as well as sample rate:

l Four leaves all channels active for acquisition, optimized for 9-bit resolution.
l Two leaves two channels active for acquisition, C2 and C3, optimized for 10-bit resolution.
l Auto, the default HD1024 selection, dynamically configures the number of active channels and
the resolution at which they operate depending on which channels are actually active and your
other Vertical and Timebase settings. You can enhance the performance of this selection by
whenever possible attaching your signals to input pairs C1 and C3, C1 and C4, C2 and C3, or C2
and C4. Doing so supports interleaving to 40 GS/s and maintaining 10-bit resolution.

Anti-Aliasing Filters
If the sample rate lowers to a point where waveform aliasing could occur, HD1024 will dynamically
apply bandwidth filters to the input according to the reduction in sample rate in order to ensure that
the Nyquist 2:1 ratio of at least two samples per cycle is met. Lowering bandwidth enables the
instrument to maintain or even raise the total resolution of that channel.
*************************************

[nfmax]

That sounds like it is distributing available ADC conversions across the number of channels in use (which almost all DSOs do anyway) but in a rather more flexible way; and using 'hi-res' or whatever it is called (we use the term 'fast time averaging' here) to gain an extra bit or two of resolution when you have more conversions available than the bandwidth requires. And also, adding extra filters, reducing the bandwidth, when it doesn't have enough ADC conversions to go round. Presumably this would kick in if you are using a lot of channels at fast sweep speeds. But I'm not sure I'd feel happy with a scope that changed bandwidth if I change the sweep speed!

This may not be what is actually going on, but it is difficult to tell from the description. The language is a bit too much marketing speak for a user manual. A table showing the achieveable bandwidth & resolution for each combination of input channels and sweep speeds would be helpful.

[salviador]

LeCroy HDO9000


My dream .......... but the price is impossible! 

[Wuerstchenhund]

That sounds like it is distributing available ADC conversions across the number of channels in use (which almost all DSOs do anyway) but in a rather more flexible way;

That appears to be the case.

and using 'hi-res' or whatever it is called (we use the term 'fast time averaging' here) to gain an extra bit or two of resolution when you have more conversions available than the bandwidth requires.

I'm not sure it uses oversampling, details as of yet are sparse but it's supposed to be a hardware 10bit ADC hybrid.

But if not and if that is indeed a 8bit ADC using oversampling then I'd assume it doesn't use simple boxcar averaging (which is what 'HiRes' on other scopes do) but a FIR filter similar to ERES.

And also, adding extra filters, reducing the bandwidth, when it doesn't have enough ADC conversions to go round. Presumably this would kick in if you are using a lot of channels at fast sweep speeds. But I'm not sure I'd feel happy with a scope that changed bandwidth if I change the sweep speed!

This may not be what is actually going on, but it is difficult to tell from the description. The language is a bit too much marketing speak for a user manual. A table showing the achieveable bandwidth & resolution for each combination of input channels and sweep speeds would be helpful.

Yes, more information are definitely needed. I seriously hope it doesn't turn out to be an oversampled 8bit ADC, which considering that the DSO-S is true 10bit wouldn't look good at all. Especially when considering how poor the Agilent Infiniium 9000H (which also uses oversampling to gain more than 8bits) performed against the HDO6000 (which is true 12bit).

So far I can't say that this new LeCroy scope got me overly excited 

[robert_]

As far as i heard, they use 4 8bit ADC per channel with a small DC offset between them (1/4LSB), and sample those at the same time to obtain higher resolution in 10bit mode, and use them with no offset but with time interleaving to obtain high sample rate at 8bit. 9bit mode uses 2 time interleaved groups of 2 1/2LSB offset ADCs.
If the ADCs are very linear this should work fairly well.

Not too interested in that series either, i have little need for that kind of sample rate/BW. I do however really like their 12bit/2(2.5)Gs/s systems which do offer significant advantage over 8bits, at least for my applications.

[The Electrician]

Especially when considering how poor the Agilent Infiniium 9000H (which also uses oversampling to gain more than 8bits) performed against the HDO6000 (which is true 12bit).

Could you provide a link to the comparison, please?

[nfmax]

As far as i heard, they use 4 8bit ADC per channel with a small DC offset between them (1/4LSB), and sample those at the same time to obtain higher resolution in 10bit mode, and use them with no offset but with time interleaving to obtain high sample rate at 8bit. 9bit mode uses 2 time interleaved groups of 2 1/2LSB offset ADCs.
If the ADCs are very linear this should work fairly well.

At this sample rate, all the ADCs I know of are interleaved, multi-phase designs anyway. So it sounds like they have different ways of operating the interleaving, adjusting clock phases and offset voltages as appropriate. Clever stuff! Assuming you have the linearity (compensation method) to make it work, and given it's LeCroy I think we can assume they do.

Does anyone know if they have published anything on this architecture?

[Wuerstchenhund]

Especially when considering how poor the Agilent Infiniium 9000H (which also uses oversampling to gain more than 8bits) performed against the HDO6000 (which is true 12bit).

Could you provide a link to the comparison, please?

No, because these were internal tests incorporating some of our projects so nothing that can be released (which shouldn't be surprising, though, an 8bit ADC oversampled to 10bit or 11bit will never reach the same performance of a true 12bit ADC).

I'm not aware that anyone has done a public comparison that's really independent, though.

[Wuerstchenhund]

Does anyone know if they have published anything on this architecture?

I don't think they have, and considering that so far there's still only sparse documentation of their other technologies like X-Stream (which has been out there for over a decade) I wouldn't hold my breath that there will be more in the future.

Unfortunately LeCroy is pretty poor when it comes to documentation.

[Wuerstchenhund]

Not too interested in that series either, i have little need for that kind of sample rate/BW. I do however really like their 12bit/2(2.5)Gs/s systems which do offer significant advantage over 8bits, at least for my applications.

For me it's exactly what I'd be looking for, as for the areas I work in 1-4Ghz is pretty much the 'bread and butter' range, and we've been waiting for better-than-8bit for a long time. The other HDO 12bit scopes are great but unfortunately lack the BW we need.

But based on the specs and what is currently known I'm just not convinced that the HDO9000 is the right scope, tbh. I'm actually pretty disappointed. From where I stand, the Keysight DSO-S looks a lot better, if it only supported all serial protocols that the HDO supports.

Which is a shame really as buying scopes in this class isn't as easy as buying entry-level scopes. At frequencies of 1Ghz and more you're looking into active probing, which is very costly, so being able to "stay" with a brand usually helps to protect that investment when scopes are refreshed. Because changing vendor not only means new scopes but new set of probes, too.

[robert_]

Does anyone know if they have published anything on this architecture?

I don't think they have, and considering that so far there's still only sparse documentation of their other technologies like X-Stream (which has been out there for over a decade) I wouldn't hold my breath that there will be more in the future.

Actually there is:
http://www.edn.com/electronics-products/electronic-product-reviews/other/4442666/10-bit-oscilloscopes-adjust-to-signal-conditions

They could really throw in some more memory, doesnt cost that much.
Looks like they reused some demux ASIC from the good old Waverunner 6 series, also being limited to 64M/Channel. Might not have been the smartest move to have that little memory, esp. when the slower and cheaper HDO4/6k have 256M/ch installed (sure, only the 6k can officially use all of it, but its there on both...)

[Muxr]

But based on the specs and what is currently known I'm just not convinced that the HDO9000 is the right scope, tbh. I'm actually pretty disappointed. From where I stand, the Keysight DSO-S looks a lot better, if it only supported all serial protocols that the HDO supports.

Does R&S have a competitive scope in this category? Remember you saying you didn't like RTO's UI implementation.

[Wuerstchenhund]

Actually there is:
http://www.edn.com/electronics-products/electronic-product-reviews/other/4442666/10-bit-oscilloscopes-adjust-to-signal-conditions

Thanks! Seems I missed that one.

But based on the specs and what is currently known I'm just not convinced that the HDO9000 is the right scope, tbh. I'm actually pretty disappointed. From where I stand, the Keysight DSO-S looks a lot better, if it only supported all serial protocols that the HDO supports.

Does R&S have a competitive scope in this category? Remember you saying you didn't like RTO's UI implementation.

No, they haven't. R&S has the RTO (recently refreshed as RTO2000), but that is a 8bit scope with an optional (paid for  )16bit oversampling mode.

The RTOs are good scopes with some nice featues, and the list of options is comparable to the DSO-S (and there's even an OCXO option which is nice). However the scope and everything else is very expensive, and the UI has always been received negatively in our evaluations (it's very flexible, but not fully thought through as often UI elements cover up important information, plus it's pretty cumbersome to setup).  Also, the probe selection is really small, and the probes they offer are also expensive.

Besides, I don't like how R&S appears to hide key information. A while ago I wanted to find out what the max FFT size is, especially since they make all this noise about their hardware-accelerated FFT. I couldn't find the figure in the specs, so I contacted R&S UK, got passed through R&S Germany, and ended with an engineer who went into length to tout their special architecture and how great it is, and that it works differently so there's no specific FFT memory size (yeah, sure), and so on. Unfortunately the scope itself doesn't seem to say how much memory it uses for FFT either so it could well use only a few Mpts where other scopes in that class use 25Mpts or even hundreds.

It also seems that despite being late to doing their own scopes (they started 2009)  they learned the game of bullshit marketing pretty fast.

So as it stands I can't really recommend the RTO (the smaller scopes, why not). The probe situation alone would make me wary of investing in that platform, but charging extra for something like 16bit HiRes is silly, really. Even Keysight doesn't do that, and they charge you where they can just because