Professional Scope Purchasing Criteria

[TunerSandwich]

i really like what you put forth comparing the 3 scopes. i always thought it is impossible to see straight lines on scopes (unless maybe inside a RF shield chamber?) ... there has to be noise somewhere ... what did the lecroy register? 5mV of noise? what are all the 3 probes "measuring" from? what if you inject say a real white noise of 5mV and the rigol still shows a straight line LOL ... (add new bug to rigol list of bugs?)

like I said...I think the Rigol is using some post a/d trickery to put up a false representation.....I don't wholly trust that reading.  That line is only a few pixels in height.....seems fishy to me. 

As I said in the prior post, I highly doubt that the Rigol (below $2k new) pre- a/d amplifier could be anywhere near the noise floor of a $35k (new) WR64MXi......if so, LeCroy has some explaining to do 

all three units were referenced to ground

[TunerSandwich]

Hello,

the Rigol and many other DSO does not show the noise  at output from ADC if the input is set to ground. They shows only a useless straight line.

Set the input do DC and short the input, then you see the noise.

Regards
egonotto

I suspected the Rigol was not trustworthy, in this regard.  Mentioned it before.....

I checked again with the inputs terminated (against 50ohm) and ref to DC.....

This time I kept the 2225 in it's standard vertical mag....so it's 5mV/div....with a timebase of 1ms/div.....that SHOULD give us the full 50MHz bandwidth

I set the other two scopes accordingly....

So here is the result  2225 = 50MHz.........DS2102A-s = 100 MHz.........WR64MXi = 600MHz

The MXi obviously wasn't "nulling" the waveform, when referenced to ground 







 

[rx8pilot]

I was going through this thread again and appreciating the value of it even more. After TunerSandwich took the time to explain the considerations needed in scope choice, I did some reading. Lots of reading. Lots of learning.

The design of SMPS circuits is a delicate symphony of analog signals adjacent to MOSETS and inductors switching huge currents VERY fast. I have been learning the basics with a 100Mhz / 1GS/s scope that simply misses all the important details. The rise time of the signals and the resulting ringing is regularly well beyond 100Mhz for a DC/DC converter operating at 750khz. The high-current switching can and will damage the various control signals and the designer must be able to see what is happening. All the calculations and simulations in the world will not tell you that the PCB layout is killing your otherwise lovely design. I am counting on a tell-all scope to slash my design time and give me much better circuits in the end.

I was waiting for the Keysight announcement of the 3000T series before doing anything. In the meantime, I picked up a Tek TDS754 500Mhz/2GS/s off of eBay to get me by for a while. What is abundantly clear is that my original budget for an MSO of $5k-ish is completely ridiculous. I have decided that I need to spend $15k - $20k to get what I want out of a scope. While that is a lot of money, I feel it will pay for itself in a very short period considering that I cannot go any further without some sort of MAJOR upgrade.

Although it may be a little overkill for what I do today, I have settled on  a minimum of 1Ghz / 5GS/s / 4ch / MSO. I really want a good overall solution - not just what I would need to cover today's design challenges. I can only go by marketing bits and demos I have found on the internet so my view of these models is likely seen through rose colored glasses.
Top models being considered:
 
Keysight: MSOX-3000T
PROS: fast updates, hardware decode, touch/mouse control, triggering seems great, Arb Gen, lowest cost of the bunch
CONS: small memory, small screen

Keysight: MSOX-4000
PROS: fast updates, huge screen, hardware decode, touch/mouse control, triggering seems great, Arb Gen
CONS: small memory

LeCroy: HDO4000-MS
PROS: GIANT touch screen, 12bits, looong memory, power analysis
CONS: 2.5 GS/s (this conflicts with my hope of having 5 GS/s but gives 12 bit vertical resolution)
Not sure if trading sample rate for bit depth is worth it on these. 2.5 GS/s seems a little slow for a 1Ghz front end.

LeCroy: MXS-B
Still trying to add up the benefit of these. They have very fast sampling rates and lots of memory. Trying to figure out if they are near a model change. Does anyone have any opinion on these models.

LeCroy: Wavesurfer 10 with MS-500
PROS: 10 GS/s, 10Mpts, 10.4" touch screen, compact size
CONS: Not sure, nothing sticks out as marginal based on the marketing materials.

[nctnico]

You don't consider getting a second hand high end scope? I did spot a great deal on a Keysight MSO7104A (4ch 1GHz/5Gs/s) the other day. It's already gone but if you have some time to wait I'm sure another great deal will come along.

[rx8pilot]

It's not off the table - the Keysight eBay store has some sweet deals on great stuff. I like the new bells/whistles though.

[Wuerstchenhund]

LeCroy: HDO4000-MS
PROS: GIANT touch screen, 12bits, looong memory, power analysis
CONS: 2.5 GS/s (this conflicts with my hope of having 5 GS/s but gives 12 bit vertical resolution)
Not sure if trading sample rate for bit depth is worth it on these. 2.5 GS/s seems a little slow for a 1Ghz front end.

Well, these are true 12bit scopes (not just 8bit scopes with oversampling that a certain unnamed company which recently changed its name sells as 12bit scopes).

LeCroy: MXS-B
Still trying to add up the benefit of these. They have very fast sampling rates and lots of memory. Trying to figure out if they are near a model change. Does anyone have any opinion on these models.

There's only one MXs-B which is the WaveSurfer MXs-B, which is now quite old (introduced around 2009). It's basically a WaveSurver MXs with faster processing, but it still runs Windowsxp. The successor scopes are the WaveSurfer 3000 for bandwidths from 200MHz to 750MHz and with the new WaveSurfer 10 for 1GHz.

The WSMXs-B is not a bad scope per se but unless the price is really good I wouldn't buy one anymore.

The new WaveSurfer 10 is a modernized version of the 1GHz WaveSurfer MXs-B, with faster processing and Windows 7 Embedded instead of XP. It starts at $10k which is pretty good for such a scope:
http://teledynelecroy.com/oscilloscope/oscilloscopeseries.aspx?mseries=470

However, there's also still the WaveRunner 6zi:
http://teledynelecroy.com/oscilloscope/oscilloscopeseries.aspx?mseries=352
The WaveRunner Series is more flexible than the WaveSurfer or HDO Series, there are much more options available for the WaveRunner, and those options or standard functionality that are also available on the WaveSurfer (i.e. WaveScan, Power Analysis PWA) are more flexible/more capable on the WaveRunner than on the WaveSurfer. As with the WaveSurfer MXs-B, the WaveRunner 6zi was introduced in 2009 but it's specs are pretty much still up to date. And the pivot-able 12.1" 1280x800 WXGA screen is very nice, too.

And since the WR6zi is an older model which will probably be replaced in the near future there should be very good deals available for that scope. In terms of analysis capabilities the WaveRunners are top-notch and much better than anything I've seen from Agilent/Keysight in their high-end scopes.

But as always, ask the manufacturers to give you a loaner unit for testing. Nothing beats real-life test, especially not when you're about to spend that much money.

[rx8pilot]

Did you work for LeCroy??? 

The question with the 12bits is this: Does 12bit vertical offer a significant practical advantage? Does sacrificing the sample rate negate the 12bit benefit? To me, it seems like a gain in vertical while significantly sacrificing the horizontal. 2.5 GS/s seems like absolute minimum for a 1Ghz front end.

[Wuerstchenhund]

Did you work for LeCroy??? 

No, I'm not affiliated with any T&M manufacturer. I'm just a user of LeCroy high end scopes scopes (sometimes, most of the times I use Agilent high end scopes scopes, though). And I tend to keep myself familiar with the highend scope offerings of the important vendors.

The question with the 12bits is this: Does 12bit vertical offer a significant practical advantage? Does sacrificing the sample rate negate the 12bit benefit? To me, it seems like a gain in vertical while significantly sacrificing the horizontal. 2.5 GS/s seems like absolute minimum for a 1Ghz front end.

It depends on what you want to do. 2.5GSa/s isn't that bad for 1GHz bw, and unless your signals are close to the upper bw limit then I#d go out on a limb and say that 2.5GSa/s should be sufficient.

But as I said, don't just rely on advice and specs, get some demo scopes and test them for what you need in your own setting.

[DJ]

What I would really like would be a kickbutt 2channel 2-4GHz scope without all the bells and whistles.

Just the front end box, with knobs for coupling, trigger, vert & horiz. Have it drive a tablet or LCD monitor directly.

[TunerSandwich]

I was going through this thread again and appreciating the value of it even more. After TunerSandwich took the time to explain the considerations needed in scope choice, I did some reading. Lots of reading. Lots of learning.

The design of SMPS circuits is a delicate symphony of analog signals adjacent to MOSETS and inductors switching huge currents VERY fast. I have been learning the basics with a 100Mhz / 1GS/s scope that simply misses all the important details. The rise time of the signals and the resulting ringing is regularly well beyond 100Mhz for a DC/DC converter operating at 750khz. The high-current switching can and will damage the various control signals and the designer must be able to see what is happening. All the calculations and simulations in the world will not tell you that the PCB layout is killing your otherwise lovely design. I am counting on a tell-all scope to slash my design time and give me much better circuits in the end.

I was waiting for the Keysight announcement of the 3000T series before doing anything. In the meantime, I picked up a Tek TDS754 500Mhz/2GS/s off of eBay to get me by for a while. What is abundantly clear is that my original budget for an MSO of $5k-ish is completely ridiculous. I have decided that I need to spend $15k - $20k to get what I want out of a scope. While that is a lot of money, I feel it will pay for itself in a very short period considering that I cannot go any further without some sort of MAJOR upgrade.

Although it may be a little overkill for what I do today, I have settled on  a minimum of 1Ghz / 5GS/s / 4ch / MSO. I really want a good overall solution - not just what I would need to cover today's design challenges. I can only go by marketing bits and demos I have found on the internet so my view of these models is likely seen through rose colored glasses.
Top models being considered:
 
Keysight: MSOX-3000T
PROS: fast updates, hardware decode, touch/mouse control, triggering seems great, Arb Gen, lowest cost of the bunch
CONS: small memory, small screen

Keysight: MSOX-4000
PROS: fast updates, huge screen, hardware decode, touch/mouse control, triggering seems great, Arb Gen
CONS: small memory

LeCroy: HDO4000-MS
PROS: GIANT touch screen, 12bits, looong memory, power analysis
CONS: 2.5 GS/s (this conflicts with my hope of having 5 GS/s but gives 12 bit vertical resolution)
Not sure if trading sample rate for bit depth is worth it on these. 2.5 GS/s seems a little slow for a 1Ghz front end.

LeCroy: MXS-B
Still trying to add up the benefit of these. They have very fast sampling rates and lots of memory. Trying to figure out if they are near a model change. Does anyone have any opinion on these models.

LeCroy: Wavesurfer 10 with MS-500
PROS: 10 GS/s, 10Mpts, 10.4" touch screen, compact size
CONS: Not sure, nothing sticks out as marginal based on the marketing materials.

I had a feeling you would come around to this conclusion eventually.  It's why I posted all the data and scenarios. 

Since we work on the same systems, let me offer up a suggestion.  This is an ideal scenario, with an "unlimited budget".  HDO8000 from LeCroy.

If that kind of price isn't possible....then have a look at the 104MXi or MXiA.....you can score one used for roughly your projected budget....

HOWEVER, let me interject another critically important insight.

You NEED some in house standards.  I.E. some references that allow you to verify equipment accuracy, before making critical measurements.  I personally build all of my standards, and reference them against a known calibrated transfer standard.  This is as important, if not more, than any amount of bandwidth or features you may think you need.

The other consideration in the scope is going to be memory depth.  Remember that a minimum time window, to gather proper spectral results, is going to be greater than 1ms.  You are going to need a lot of memory to maintain the proper sample rate across that time division (otherwise it defeats the purpose of the bandwidth and bit depth). 

I HIGHLY recommend going with LeCroy.  Their statistical functionality and interface are head and shoulders above anything I have ever used or owned (i have had some of the fancy agilents you are considering).  Also the LeCroy triggering is unparalleled in these top tier models. 

I hate to sound like a fanboi here, but I have been through this all before...and took a leap of faith (recently) on a LeCroy midrange scope.....I wouldn't even consider going back to anything from agilent or tek....and I am damn picky about this kind of stuff. 

I wouldn't hold my breath on keysight, or any other company, catching up with the LeCroy philosophy, anytime soon.  That isn't saying that LeCroy makes the "best" scopes or that all agilent units are inferior etc.....they simply have different approaches, and cater to different specialties....and LeCroy happens to cater to OUR needs (you and I) above and beyond any other manufacturer....

P.S.  I sent you a reply to your PM a while back, but am not sure if it went through, as I never heard back from you....let me know (via PM) if you are still needing help, outside of the forum...

P.P.S  we should really discuss references and standards, as they are going to become a VERY important part of your on-going discovery...and they are an art-form all to themselves.  Removing AC coupled line noise from the measurements is going to be HIGHLY critical, and you can do that with some proper stabilized standards.  You can then generate a mask (this is why LeCroy leads the way for us) to remove the measured AC/line components from your critical switching measurements.   This is really where my experience has taught me that your measurements are key factors in determining potential problems in an SMPS.  ESPECIALLY when dealing with DC-DC conversion....

P.P.P.S    another critical factor you need to look at, is the total dynamic range of your front end solutions (probes, i/o amplifiers etc).....this is equally important, as any other factor you might consider....you can't trust measurements that are clipping 

[EEVblog]

So here is the result  2225 = 50MHz.........DS2102A-s = 100 MHz.........WR64MXi = 600MHz

Beware comparing analog and digital scopes for noise, it's not apples vs apples.

[EEVblog]

What is abundantly clear is that my original budget for an MSO of $5k-ish is completely ridiculous. I have decided that I need to spend $15k - $20k to get what I want out of a scope.
Although it may be a little overkill for what I do today, I have settled on  a minimum of 1Ghz / 5GS/s / 4ch / MSO.

Don't forget the refurbished Keysight store:
http://stores.ebay.com/keysight

You can get a 2GHz 9000 series on there for $12K
http://www.ebay.com/itm/Keysight-DSO9204H-High-Definition-Oscilloscope-2-GHz-Agilent-DSO9204H-/181559422822?pt=BI_Oscilloscopes&hash=item2a45c8ef66

Or lesser 1GHz one for less than $8K
http://www.ebay.com/itm/Keysight-DSO9104H-High-Definition-Oscilloscope-1-GHz-Agilent-DSO9104H-/191454554954?pt=BI_Oscilloscopes&hash=item2c9394ab4a

Up to 12 bit vertical res and low noise!
And 11 bits at 1 GHz is pretty darn good. Sound like exactly want you want for under $8K
How much is competing 12bit Lecroy HDO?

Or the 4CH 1GHz MSO3000A series with arb is only $8K
http://www.ebay.com/itm/Keysight-Premium-Used-MSOX3104A-Oscilloscope-1-GHz-Agilent-MSOX3104A-/181616054054?pt=BI_Oscilloscopes&hash=item2a49290f26

[TunerSandwich]

So here is the result  2225 = 50MHz.........DS2102A-s = 100 MHz.........WR64MXi = 600MHz

Beware comparing analog and digital scopes for noise, it's not apples vs apples.

Although I wholly agree, it's not apple to apples, it is still a sanity check...in regards to the "trustworthiness" of what we see on a screen.  I may have "under-emphasized" the nuances involved in comparing (DSP, aliasing, pre/post filtering etc). 

I DO find a great deal of validity in making a comparison, on a "simplified" level....because a lot of posts here are mentioning that they "trust" what they see on their oscilloscope screens.  I was trying to emphasize that those assumptions should not be made. 

The biggest problem we are facing (which I mentioned), is that most manufacturers don't state what level or type of dithering is used to remove noise in the quantization step response. 

However, at the end of the day, I fear that a lot of folks are bypassing these points and trusting manufacturer "specs" as holly gospel....and that if YOUR specific scenario somehow deviates from those specs, that something is "wrong".  It's simply not true.  A proper measurement should be qualified against a known standard and that should be the basis of your uncertainty (SDEV), NOT some random spec., quoted on XYZ datasheet. 

I have a sneaking suspicion (always have) that Rigol is removing noise components, post a/d, with dithering....I simply don't trust that Rigol can produce a low noise front end (pre a/d) that can compete with a $30k scopes analog input section.  Not for under $2k.  So how trustworthy is the Rigol measurement, below 1mV/div.....?  Who the hell knows....I haven't taken the time to verify it against a known standard, as we get into some serious probing issues at 500uV/div.  Also the offset in the DS2000A is all over the place, when compared against my in-house standards.  Although I can't confirm the drift, and therefor can't confirm the accuracy of the SDEV/mean measurements in the Rigol unit I have, I DO trust the LeCroy MXi, as the measurements I did earlier in this thread shows a more linear SDEV across all standards.


P.S.  I think the point was missed....I was trying to show that the noise on an analog scope input is NOT necessarily lower than that of a DSO/MSO.  That was why I included the theoretical bandwidth figures.  2mV pk-pk @ 600MHz (MXi) is damn impressive IMO, and simply outclasses the Tek 2225 @ 50MHz (and likewise the Rigol @ 100 MHz). 

I did this comparison BECAUSE, someone here claimed that "analog scopes have lower noise"....and by that token I also have to assume "better" linearity....and I was trying to show that is simply not true.  I hope the tests I provided reflect this. 

[DJ]

"I personally build all of my standards, and reference them against a known calibrated transfer standard."

Can you share some standards you have put together for those who might like to try their hand at it?

[TunerSandwich]

"I personally build all of my standards, and reference them against a known calibrated transfer standard."

Can you share some standards you have put together for those who might like to try their hand at it?

What kind of standard?  Voltage?  AC/DC?  Impedance?  Null? CC?  Theoretical range etc....

I can use the "standard" I built for this specific thread (the MAX6350) to emphasize some points.  Remember I built it with a specific "error" in it's stability, to illustrate the problem of resonant tanks, and how you need bandwidth to discover these problems...in relation to DC-DC SMPS/POL

I have two discreet nodes on that standard.  One is a ballasted direct output, and the other is a low pass filtered output.  The standard nodes also employ a variety of capacitance scenarios, which illustrate potential issues in stability (remember the MAX6350 excels at driving capacitive loads).  Also it's construction is 100% point to point, as I wanted to illustrate these resonant tank issues, without bringing a PCB substrate into the equation.  If I wanted to quantify the PCB interactions, I would need to measure past the 1GHz mark, and that is going to bring up a whole other set of issues.  Primarily the effects of available probing solutions. 

P.S.  we should start by discussing artifact based standards (PWM) vs your more traditional "linear" standards.  One might confuse a "linear" standard with being more "linear"....which is not at all the case.  PWM based "artifact" standardizations can be much more linear, free of noise and relevant to calibrating "digital" equipment.....especially DC on a DVM

Another relevant standard for designing multi-phase SMPS would be a +/- voltage standard....which can be very useful for characterizing op-amp performance.  In a highly critical, multi phase architecture one might be forced to hand select (bin) precision op-amps.  You could consider this a "null" standard, but it's a bit more specific than that.  "Nulling" op-amps with the comp pins isn't necessarily useful....if your measurement standards can't resolve the TRUE offset.  2 discreet DVM is hardly a good way of doing this....so we need a null RMS measurement, and to acquire that we need a known "null" reference.....

I think the problem of a true "null" standard is often overlooked in production designs.....however if you do some reading into metrology grade devices, you will quickly discover how important null readings can be, in defining baselines. 

[Wuerstchenhund]

You can get a 2GHz 9000 series on there for $12K
http://www.ebay.com/itm/Keysight-DSO9204H-High-Definition-Oscilloscope-2-GHz-Agilent-DSO9204H-/181559422822?pt=BI_Oscilloscopes&hash=item2a45c8ef66

Wow. $12k and only 30 days warranty? That's not a lot of trust they're putting in their product, and when spending that much money I'd want my purchase underpinned by a bit more assurance that at least for a while the investment will be a good one.

Or lesser 1GHz one for less than $8K
http://www.ebay.com/itm/Keysight-DSO9104H-High-Definition-Oscilloscope-1-GHz-Agilent-DSO9104H-/191454554954?pt=BI_Oscilloscopes&hash=item2c9394ab4a

Again with just 30 days warranty.

Up to 12 bit vertical res and low noise!
And 11 bits at 1 GHz is pretty darn good. Sound like exactly want you want for under $8K
How much is competing 12bit Lecroy HDO?

The 1GHz HDO4k is roughly $17k but then it's a true 12bit scope (not 8bit oversampled like these Agilent's) and it's brand new not used (these are not even remanufactured units) and comes with 3yrs warranty.

Or the 4CH 1GHz MSO3000A series with arb is only $8K
http://www.ebay.com/itm/Keysight-Premium-Used-MSOX3104A-Oscilloscope-1-GHz-Agilent-MSOX3104A-/181616054054?pt=BI_Oscilloscopes&hash=item2a49290f26

Not a bad deal if you primarily need a MSO but as a 1GHz scope I'm not sure if $8k for a refurbished 1GHz MSOX3k (which is rather simple scope) with 2Mpts (4Mpts optional) is that great of a bargain when a brand new WaveSurfer 10 (1GHz 10Mpts, MSO-capable) costs only $2k more (and gives you a much more advanced scope).

[EEVblog]

Again with just 30 days warranty.

But what are you paying? something like half the price?
Fair compromise I think if you want performance on a budget.

The 1GHz HDO4k is roughly $17k but then it's a true 12bit scope (not 8bit oversampled like these Agilent's) and it's brand new not used (these are not even remanufactured units) and comes with 3yrs warranty.

Again, double the price, your call.
But you aren't the OP, he may have an entirely different opinion. Indeed he seems to be constrained by price, and is considering the MDO3000T, so the "almost" 12bit performance of the 9000 is likely good enough.

[TunerSandwich]

Not a bad deal if you primarily need a MSO but as a 1GHz scope I'm not sure if $8k for a refurbished 1GHz MSOX3k (which is rather simple scope) with 2Mpts (4Mpts optional) is that great of a bargain when a brand new WaveSurfer 10 (1GHz 10Mpts, MSO-capable) costs only $2k more (and gives you a much more advanced scope).

I was recently offered the same scope, fully loaded (all licenses) for $3500....with all original boxes, documentation, software and accessories. 

I don't think the 30 day warranty period is worth $4.5k 

I also think a used MXi (MX, MXiA etc) simply decimates the Agilent, for the same money (used).

If the HDO4k is only a few grand more, BNIB with full manufacturer warranty and support....hmmmm.....although the HDO4k might not have the memory depth that OP needs. 

The Keysight/Agilent refurb stuff is a good price point for used Agilent, however I am not sure that means it's the best value for dollars spent.  Agilent/HP/Keysight/etc.. has always been a bit "overvalued" IMO.  However nothing like Tek (snickers).....

[tautech]

You can get a 2GHz 9000 series on there for $12K
http://www.ebay.com/itm/Keysight-DSO9204H-High-Definition-Oscilloscope-2-GHz-Agilent-DSO9204H-/181559422822?pt=BI_Oscilloscopes&hash=item2a45c8ef66

Wow. $12k and only 30 days warranty? That's not a lot of trust they're putting in their product, and when spending that much money I'd want my purchase underpinned by a bit more assurance that at least for a while the investment will be a good one.

I'd be very concerned if that was the best warranty offered and for that reason alone I would give it a wide berth.
Hell if I can offer the remaining factory warranty on an 1 year old Siglent demo unit so should Agilent.
And chuck in new probes to boot.
Piss poor IMO

[Wuerstchenhund]

Again with just 30 days warranty.

But what are you paying? something like half the price?
Fair compromise I think if you want performance on a budget.

Is it really? That's alot of money and I'd say in any terms accounts as a serious investment. 30 days means that in the worst case you're out of money and left with non-working kit after a month.

If your fine with 30 days lifetime then you could save a whole lot of money by just renting the kit you need.

This aside, I can buy some used later-model (i.e. 2004 vintage and later) LeCroy scope from anyone and anywhere (you, me, ebay, the trunk of a car at a lonely car park at night, garage sale, Craig's List, pawn shop, whatever), send it in to LeCroy for calibration, and as long as it passes and the original factory seal is still in place I can buy full manufacturer warranty for that scope for one, two or three years, sometimes (depending on the model) even more. That's how confident they are in the stuff they produce.

And Agilent/Keysight can't even provide a simple year of warranty for a pretty young scope they claim has been checked by them and which is sold through their official store? That's ridiculous.

30 days is what I usually get when buying 2nd hand kit from private sellers on ebay.

The 1GHz HDO4k is roughly $17k but then it's a true 12bit scope (not 8bit oversampled like these Agilent's) and it's brand new not used (these are not even remanufactured units) and comes with 3yrs warranty.

Again, double the price, your call.
But you aren't the OP, he may have an entirely different opinion. Indeed he seems to be constrained by price, and is considering the MDO3000T, so the "almost" 12bit performance of the 9000 is likely good enough.

Of course I'm not the OP, and the oversampled 8bit scope may well be good enough, but especially on a tight budget there are other criteria than just technical specs. In regard to these used Agilent scopes, I'm sure the OP wouldn't feel confident to pay for a potentially very expensive repair say 2 months after splashing out $8k or more.

And as to list prices, only the lazy ones pay them. I'm pretty sure LeCroy would give him the HDO4104 for a lot less, and even might throw in some of the options ne wants. Keysight does so, too, especially on new scopes (and sometimes even on their CertiPrime remanufactured scopes). Just Tek seems to sit a bit tight on their arses.

[Wuerstchenhund]

I was recently offered the same scope, fully loaded (all licenses) for $3500....with all original boxes, documentation, software and accessories. 

I don't think the 30 day warranty period is worth $4.5k 

If it was from the same seller we got our LeCroy scopes from then it would have come with 30 days as well 

[nctnico]

Look at it this way: if you pay less then half the 'new' price you can buy two and still save money.
In my experience with used equipment the age is a very important factor. The older a piece of equipment is the bigger the chance of serious problems.

[EEVblog]

If your fine with 30 days lifetime then you could save a whole lot of money by just renting the kit you need.

Oh come on, be reailistic, you know very well the odds of a (proven working "burned in" so the speak demo) Agilent/Keysight scope failing within a few months is not far off zero.
And if it's half price of a new one, it's actually not a bad calculated risk. i.e if the used one does fail after the 3 months and before the 3 years you would have got on a new one, then you can buy another one and have not paid any more than a new one.
And of course the odds of a 2nd one failing in the same way are lower again.
In fact you would be ahead even if it did fail because the failed unit is actually worth something on ebay to someone, or perhaps you could get it repaired for once again a lot less than a new one would have cost.
But odds are very high indeed that it will be fine and you are going to save half your money.
Fear of lack of warranty often clouds people's opinion of the real economics and statistics of the matter.

[Wuerstchenhund]

Oh come on, be reailistic, you know very well the odds of a (proven working "burned in" so the speak demo) Agilent/Keysight scope failing within a few months is not far off zero.

Most of the enivironments I work use Agilent (heck, very few use LeCroy), and while it doesn't happen very often it's not unheard of that young kit gives up. I myself had a 9 months old DSO 90604A dying on me during work. The overall failure rate is still low, but certainly not close to zero.

And quite frankly, if the failure quote would be as close to zero as you say, then why doesn't have Agilent/Keysight more confidence in their product and offer more than a measly 30 days of warranty?

And if it's half price of a new one, it's actually not a bad calculated risk. i.e if the used one does fail after the 3 months and before the 3 years you would have got on a new one, then you can buy another one and have not paid any more than a new one.
And of course the odds of a 2nd one failing in the same way are lower again.
In fact you would be ahead even if it did fail because the failed unit is actually worth something on ebay to someone, or perhaps you could get it repaired for once again a lot less than a new one would have cost.
But odds are very high indeed that it will be fine and you are going to save half your money.

So what if the first one fails after 3 months, and the second after four (I've actually seen that happening)? It means in less than a year you spend roughly 2x 50% of the new price and are still left with non-working kit. That's playing the lottery.

This aside, you forgot the whole software option issue. Once you bought options for one scope you can't just transfer them to the 2nd one if the first one fails. That means you'd have to buy a 2nd set of options, which makes the whole thing even less attractive.

What you describe can make sense for a hobbyist (although I still can't see why he couldn't just buy the same scope somewhere else for cheaper). But for a professional setting I would want to be able to rely on my kit and the support behind it. Having to deal with broken kit is already a hassle, and because it means I spend the time on dealing with the issue I can't spend that time on making money it's also causing some loss. Warranty and support are there to minimize the impact of such occurrences.

But that's just me, and of course I appreciate YMMV.

Fear of lack of warranty often clouds people's opinion of the real economics and statistics of the matter.

Statistics is fine unless you're finding yourself inside that low probability figure, which can happen sooner than expected. Don't forget that statistics is a calculation following a standardized mathmatical model and is based on certain defined parameters. It's no way a gurantee that your real-life situation will follow that model.

Playing lottery by buying unwarranted/low-warranty kit in the hope it will survive for a certain period is bonkers in a commercial setting. In most cases the more sensible thing is to buy supported/warranted kit and if that fails just get a loaner (either from the manufacturer, which is pretty common for high end scopes, or from a rental provider) and continue working.

'Real economics' is taking all subsequent costs into account, which is much more than the initial purchase price (which often turns out to be peanuts anyways).

[EEVblog]

But that's just me, and of course I appreciate YMMV.

Yep, and I'm not going to bother arguing the point furher. If I want to see people argue the worst case scenario I'll go watch Doomsday Preppers, far more entertaining