NEW Keysight HD3

[EEVblog]

Tek is steering people away from the MDO, so there's going to be a market hole there.

"Market hole" implies that there is a market for something - I'm not sure that's the case

True analogue spectrum analyser input?
RF output for VNA?
Wavegen here would have been nice, but clearly they have other plans

That's what I speculated in my video. It's an N connector cutout.

[EEVblog]

Tek is steering people away from the MDO, so there's going to be a market hole there.

"Market hole" implies that there is a market for something - I'm not sure that's the case

Interesting point. Has anyone else done an MDO equivalent since Tek?

[Martin72]

Does not really make sense to have 14Bits and a channel separation of only 40dB, does it ?

The ENOB of the 1Ghz model is “only” 8.2 bits and is therefore on a par with the ENOB of a 12-bit scope in the same bandwidth.
One could now (rightly) argue, OK, the 14 bits show their strengths at lower bandwidths.
And indeed, at 20Mhz it is 11.3 bits and at 350Mhz it is still 10.3 bits....Impressive.
But these are not the values of the HD3.
With the HD3 it's 10.4 and 8.8 .
The values mentioned above are from the Batronix Magnova Scope....
(Both scopes 100mV/Div, 50 Ohm)

[EEVblog]

Still I do not see the point for MSOX3104T users to upgrade to a HD3, unless they are working in unicorn applications.
I mean, how important are 14bits for the EE today?

MSOX3000 users aren't going to rush out and buy one like a new model iPhone, but that was never the point.
The point it have a new option available when people are ready to update because their needs change, or probably more importantly, new customers and/or new needs customers come along.
If you have the budget required for this, are you seriously going to buy the 8bit 3000X over the 14bit HD3? Maybe if you absolutely need to save few $K (in which case you'd likely be considering other brand options anyway), but otherwise you are going for the HD3.

[mawyatt]

Tek is steering people away from the MDO, so there's going to be a market hole there.

"Market hole" implies that there is a market for something - I'm not sure that's the case

True analogue spectrum analyser input?
RF output for VNA?
Wavegen here would have been nice, but clearly they have other plans

That's what I speculated in my video. It's an N connector cutout.

With the latest semiconductor integrated technologies, even for Analog/RF/MW signals, with an integrated down converter this could cover a wide frequency range. One could envision a modest performing SA which uses the core ADCs for digitizing the IF??

Best,

[egonotto]

Does not really make sense to have 14Bits and a channel separation of only 40dB, does it ?

The ENOB of the 1Ghz model is “only” 8.2 bits and is therefore on a par with the ENOB of a 12-bit scope in the same bandwidth.
One could now (rightly) argue, OK, the 14 bits show their strengths at lower bandwidths.
And indeed, at 20Mhz it is 11.3 bits and at 350Mhz it is still 10.3 bits....Impressive.
But these are not the values of the HD3.
With the HD3 it's 10.4 and 8.8 .
The values mentioned above are from the Batronix Magnova Scope....
(Both scopes 100mV/Div, 50 Ohm)

Hello,

I'm afraid there is a mistake in the Magnova data sheet. Example 1 V/div 50 Ohm 20 MHz. The data sheet says 2.19 mVrms and ENOB 11.5 bit.

11.5 * 6.02 dB + 1.76 dB = 70.99 dB = SINAD.

To achieve this, the signal would have to be around 8 Vrms (at 1 V/div). I suspect that Vrms has been confused with Vpp.

Best regards
egonotto

[EEVblog]

With the latest semiconductor integrated technologies, even for Analog/RF/MW signals, with an integrated down converter this could cover a wide frequency range. One could envision a modest performing SA which uses the core ADCs for digitizing the IF??

I suspect that might have been the intention. Not a spectacular SA, but worthy of a MDO badge.

[EEVblog]

Front end. RIP 3 pads.

Relays under shield are the exact same model as those outside.

[Martin72]

But the workmanship is also superb, nothing else can be said.
The BNC socket.....Nice.

[jusaca]

Oh wow, these look really beautiful
What are the small green parts with text in the lower picture? Are these some sort of caps? They actually look like some small polyfuses, but that does not make sense.

// Ahh no, I assume they are some precision resistors.

[EEVblog]

What are the small green parts with text in the lower picture? Are these some sort of caps? They actually look like some small polyfuses, but that does not make sense.
// Ahh no, I assume they are some precision resistors.

Yes, 499k resistors for the 1M input impedance. Looks like they have two lots for the different divider taps. I don't think I've seen that before, usually it's just the one 1M input and then you relay select the attenuator.

[tautech]

Front end. RIP 3 pads.

Some one please mailbag Dave a decent big old soldering iron from yesteryear.
No, I'm keeping mine. 

[nctnico]

It also looks like they added the option to have a local low-noise power supply on the input module. With the compute module and the input modules it would not surprise me if Keysight respins the main board at some point so everything is on a single board.

[tautech]

It also looks like they added the option to have a local low-noise power supply on the input module. With the compute module and the input modules it would not surprise me if Keysight respins the main board at some point so everything is on a single board.

Yet many keep the processor on a separate PCB where it can easily be upgraded as HW availability changes or allow the same processor to be used in different models.
Each to their own on this and only the cost of an edge socket is additional.

[EEVblog]

With the compute module and the input modules it would not surprise me if Keysight respins the main board at some point so everything is on a single board.

That would change the entire physical design.

[EEVblog]

Yet many keep the processor on a separate PCB where it can easily be upgraded as HW availability changes or allow the same processor to be used in different models.
Each to their own on this and only the cost of an edge socket is additional.

They did this on the cheap 1000X series as well, so it's no-brainer to also do this on a much higher margin product.

[EEVblog]

Front end. RIP 3 pads.

Some one please mailbag Dave a decent big old soldering iron from yesteryear.
No, I'm keeping mine. 

The can and joint heated up in a second no problems, looks like not that great a laminate adhesive combiend with the solder pads not beign part of the main ground plane, just tiny pads.
I also physically ripped a large connector ground pad off the main board with some excessive force.

[Kleinstein]

The separate PCB for the inputs make absolute sense. It is easier to respin the relatively small PCB. The analog front end may just need more revisions to really get good performance. Modern software can simulate a lot, but there is still some RF magic left with possible unpleasant surprises. For really high performance they may even want better than standard FR4.
The extra connector can also reduce mechincal stress from the BNC input.
The separate PCBs may also allow a repair if needed.

I find it a bit odd that the shielding cans are not connected more often, even if just with spring contacts. There seem to be provisions, but not used.

The extra compute PCB also saves on area for the main PCB. The overall scope is already rather large for the moderate size screen.

[mawyatt]

What are the small green parts with text in the lower picture? Are these some sort of caps? They actually look like some small polyfuses, but that does not make sense.
// Ahh no, I assume they are some precision resistors.

Yes, 499k resistors for the 1M input impedance. Looks like they have two lots for the different divider taps. I don't think I've seen that before, usually it's just the one 1M input and then you relay select the attenuator.

Splitting the 1M Resistor makes sense as ideally you split the parasitic SMD capacitance. Also note the small (0201 ?) component in series with the 2 499K Resistors.

At first thought this was a small inductor to help compensate for the SMD capacitance, but makes no sense as the impedances are too high, so likely a PTC or NTC resistor to compensate the TC of the two large 499K SMD resistors.

BTW thanks for the reviews and ongoing discussions, very interesting indeed!! Wish we could find out about the ADC and it's design heritage 

Anyway, speculating which we shouldn't do

Best,

[nctnico]

With the compute module and the input modules it would not surprise me if Keysight respins the main board at some point so everything is on a single board.

That would change the entire physical design.

Still, having modules means extra assembly steps (expensive) and potential contact problems (angry customers). In my experience, the connectors Keysight uses for the modules in the HD3 are not easy to clean while these connectors also like to collect flux residue. And it wouldn't be the first time you tore a scope down with lots of modules inside while later versions consist of a single board.

I agree with Kleinstein that having modules makes it easier to do faster design iterations to bring a product to market quicker in small numbers. However, when scaling up, modules become a nuisance.

[gslick]

Front end. RIP 3 pads.

Relays under shield are the exact same model as those outside.

With the front end modules removed, where do the traces up and to the left of J302 go?

[Zucca]

Hello,

I'm afraid there is a mistake in the Magnova data sheet. Example 1 V/div 50 Ohm 20 MHz. The data sheet says 2.19 mVrms and ENOB 11.5 bit.

11.5 * 6.02 dB + 1.76 dB = 70.99 dB = SINAD.

To achieve this, the signal would have to be around 8 Vrms (at 1 V/div). I suspect that Vrms has been confused with Vpp.

Best regards
egonotto

In case you find to have a monkey face (happened to me) after reading the above...
This will help and also this

[nimish]

You say there is no limit, so go ahead and write us a software implementation that does it. Others of us can calculate the computational task and know it is non-trivial and there are limits.

I don't have to, the authors of libraries like OpenGL have already done that. Or do you think upscaling and applying a simple moving average shader is "non-trivial"? There's no requirement that your DSP engine and your display engine share resource constraints!

There are no hard _limits_ to this, just market and product _choices_. Which was my point. That Keysight could have, but chose not to, use a higher-res screen even if the "intensive" calculations happen elsewhere and are simply rendered at screen-res. They could have even kept the hw rendering and have it composited with a nicer UI (straightforward to do on a Zynq with its GPU)

[egonotto]

Hello,

I'm afraid there is a mistake in the Magnova data sheet. Example 1 V/div 50 Ohm 20 MHz. The data sheet says 2.19 mVrms and ENOB 11.5 bit.

11.5 * 6.02 dB + 1.76 dB = 70.99 dB = SINAD.

To achieve this, the signal would have to be around 8 Vrms (at 1 V/div). I suspect that Vrms has been confused with Vpp.

Best regards
egonotto

In case you find to have a monkey face (happened to me) after reading the above...
This will help and also this

Hello,

I don't understand what you mean. Do you think my calculation is wrong?

From Walt Kester:
"Signal-to-Noise-and-Distortion (SINAD, or S/(N + D)
is the ratio of the rms signal amplitude to the mean value of the root-sum-square (rss) of all other
spectral components, including harmonics, but excluding dc"


Best regards
egonotto

[nctnico]

Hello,

I'm afraid there is a mistake in the Magnova data sheet. Example 1 V/div 50 Ohm 20 MHz. The data sheet says 2.19 mVrms and ENOB 11.5 bit.

11.5 * 6.02 dB + 1.76 dB = 70.99 dB = SINAD.

To achieve this, the signal would have to be around 8 Vrms (at 1 V/div). I suspect that Vrms has been confused with Vpp.

Best regards
egonotto

In case you find to have a monkey face (happened to me) after reading the above...
This will help and also this

Hello,

I don't understand what you mean. Do you think my calculation is wrong?

From Walt Kester:
"Signal-to-Noise-and-Distortion (SINAD, or S/(N + D)
is the ratio of the rms signal amplitude to the mean value of the root-sum-square (rss) of all other
spectral components, including harmonics, but excluding dc"

Read the R&S appnote Zucca linked to. You have to compensate for the full range of the ADC (peak-to-peak). So yes, I think your calculation is wrong