Siglent SP3050A probes marked “only for 2GHz model”

[bmjjr]

I have an SDS2104X plus which is upgraded to 500MHz. I wanted to upgrade the probes from the 200MHz PP215’s which I originally received. I bought two 500MHz SP3050A with autosense from tequipment.net, they were not in stock so I’ve been waiting about a month to receive them. They arrived today but I’m surprised to find they have a sticker attached to them that states “Only for 2GHz model”. Is this anything to worry about in using them with my 500MHz scope? Or have they been modified in some way which will be undesirable for my scope?  Thanks for your inputs.

[bdunham7]

The only thing I can think of is that the compensation range might not be compatible, or at least it is a close call.  The probes specify from 8 to 20pF, the SDS2104X+ says 17pF on the front.  Try compensating them and see if it works out.  I  have no idea why it would say '2 GHz only' unless that is referring to some issue with the non-export 6000HD models.  The SDS6k models here all come with this probe and can be upgraded to the 2GHz model.

[2N3055]

Explanation by Siglent in attachment.

Ideally, SP3050A without that label is meant to be used with SDS2000X+ or SDS5000X scopes.

tequipment.net got confused.

Since all SDS6000A are same hardware (500 MHz, 1GHz, 2GHz) these "2 GHz probes" are OK for all of these.
Marking is somewhat misleading. It should have been "Compensated for SDS6000A" or something like that.

They seem to be all the same, but these probes come with 3 compensation trimmers (2 in a box with BNC for high frequency trims, hidden under 2 little covers) and come set for a 2 GHz model front end.

Ideally, you should return them and ask for ones without tag.
They can be compensated for your scope, but you need more than standard 1kHz probe compensation signal on the scope and procedure is fiddly as trimming is iterative, i.e. trims are interactive, so you need to go back and forth until it looks OK.

[bmjjr]

Given the lead time I've already gone through to get them delivered, I can hardly stand to send them back.  In preliminary usage, they seem fine after basic LF adjustment. But I'm surprised to find the instructions state that the HF adjustment should be done on a 1MHz square wave with <0.7ns rise-time...?

I have an SDG2000X arb wave generator and I fed that1MHz square wave through the scope, it appeared pretty good but given those two adjustment knobs, I almost had to turn them to see if I could make it better. But, after double-checking the SDG2000X datasheet, I see the datasheet shows the square wave can only achieve 9ns rise-time.  Even the top of the line SDG6000X datasheet spec is 2ns rise time.  So I'm not sure what kind of equipment must be needed to hit that 0.7ns rise time on a 1MHz square wave?  I guess it's best to not make the adjustment on a square wave with 9ns rise time?

While the datasheet for the SDS2000X Plus oscilloscope states that "In interleaving, mode bandwidth is 500 MHz, rise time is 0.8 ns; in non-interleaving mode bandwidth is 350 MHz, rise time is 1 ns".  So even if I could get the 0.7ns rise time seems it would be beyond the limits of the scope.

[bmjjr]

Here’s what the 1MHz square wave currently looks like from my upgraded 2042x, connected directly by BNC, before any HF adjustment by myself. Opinions on whether or not the probe looks out of adjustment?

[bdunham7]

That's too much overshoot for a signal well below the scope's bandwidth.  Is that the best you can do with the LF adjustment? You can adjust it with a lower-bandwidth signal, but it won't be as accurate.  The best/cheapest way to do this is with the little Bodnar pulser.  1 MHz vs 10 MHz won't matter, it's the rise time that is important.  My entire workbench is disassembled at the moment or I'd send you a screenshot.

http://www.leobodnar.com/shop/index.php?main_page=product_info&products_id=295

[2N3055]

Well, I hate to be that guy, but "I told you so"...
You'll need equipment, and procedure.
There is LF, MF and HF trimmer and they are highly interactive.
It has to be 1 MHz, not because higher harmonics are not the same as 10Mhz but because you need that lower frequency so you can see signal settle on top..
And risetime should be 0,5-0,7 ns.
Slower than that, you're not seeing what you need. Faster, you're injecting too much high frequency energy into that edge and you'll get overshoot no matter what..

[bson]

Siglent should really sell these probes under a different code, like SP3051A or something, exactly to avoid these kinds of problems.  They can't reasonably expect a line worker at a reseller to know the difference, or even to know to look for it.  This practice of using a single code makes it the proverbial problem magnet.  Not to mention they may not have any idea what kind of scope the buyer has, to begin with!

[mawyatt]

Would be interesting to "see" how these probes are done with the 3 compensation trimmers. Maybe someone has a schematic??

Think 2N3055 has a good point, the usual single compensation trimmer isn't sensitive to calibration waveform frequency, only edge speed, and for proper calibration only requires an edge speed more than 3~4 times faster than the scope and probe combined response since these parameters are root summed squared. However, with the "extra" compensation likely to achieve some flatting in the overall frequency response, the calibration frequency may be somewhat important like achieving a very flat response across the waveform period.

Anyway, without a schematic or actual physical probe this is just speculation on my part.

Best,

[bdunham7]

Well, I hate to be that guy, but "I told you so"...
It has to be 1 MHz, not because higher harmonics are not the same as 10Mhz but because you need that lower frequency so you can see signal settle on top..
And risetime should be 0,5-0,7 ns.
Slower than that, you're not seeing what you need. Faster, you're injecting too much high frequency energy into that edge and you'll get overshoot no matter what..

I kinda sort have to disagree, especially the latter part.  First, the whole idea of a 10X/10M 500MHz probe is a bit sketchy in the first place, but they exist and I have one (not the SP3050A) and while setting them up may be tricky, so is finding an appropriate source.  I don't know the specs for the SP3050A, but typical similar probes specify something like -3dB with a 25 ohm source.  That's pretty tricky at 500MHz--I use the Bodnar pulser and an RF signal generator with a 50R pass-through terminator right at the source.  Not perfect, but things look OK to me.

If I was proposing 100MHz square waves, you might have a point, but I've found 10MHz to be OK enough.  And if you really want to see what happens further down that flat top, you can alway use a less-fast-edged 1 MHz or even 100kHz signal from a standard AWG.  Would I prefer a 1GHz AWG with a 0.5ns rise time?  Sure, but.... 

As for overshoot with 'too fast' a rise time I think this just accentuates that we probably aren't agreeing on what a properly compensated Hi-Z HF probe should do.  I suspect that the way you may be doing it will result in peaking the high-end to compensate for the combined system roll-off.  I would adjust for as perfect a step response as I can get with a very fast edge and then live with whatever the frequency response turns out to be.  After all that when you look at the phase response and reactive impedance of the probe at 500Mhz you might conclude that it really isn't all that worthwhile in the first place.

the usual single compensation trimmer isn't sensitive to calibration waveform frequency, only edge speed, and for proper calibration only requires an edge speed more than 3~4 times faster than the scope and probe combined response since these parameters are root summed squared.,

Actually it really shouldn't be that sensitive to edge speed either, since that primary adjustment is just matching the capacitive divider.  This is why you can get pretty close on compensation even with the relatively blunt 1kHz signal that the typical scope calibrator puts out.

[mawyatt]

the usual single compensation trimmer isn't sensitive to calibration waveform frequency, only edge speed, and for proper calibration only requires an edge speed more than 3~4 times faster than the scope and probe combined response since these parameters are root summed squared.,

Actually it really shouldn't be that sensitive to edge speed either, since that primary adjustment is just matching the capacitive divider.  This is why you can get pretty close on compensation even with the relatively blunt 1kHz signal that the typical scope calibrator puts out.

Probably true for the single-trim probes, but no so sure about the multi-trim ones. In the end you'll just use what you have and move on to the actual measurements.

Best,

[gf]

Actually it really shouldn't be that sensitive to edge speed either, since that primary adjustment is just matching the capacitive divider.

At lower frequencies, a capacitive divider is a good enough model. But at higher frequencies this model is IMO too trivial, due to parasitics.

Btw, nice paper on measuring probe performance:
https://www.picotest.com/downloads/DesignCon_2020_final_edit.pdf

EDIT:

This is why you can get pretty close on compensation even with the relatively blunt 1kHz signal that the typical scope calibrator puts out.

And 1kHz is in the region of the cross-over from the capacitive divider to the resistive 10MOhm : 1MOhm divider. If there is only a single trimmer, then only the (low-frequency) match between resistive and capacitive divider can be adjusted, hoping that the capacitive divider model is good enough up to the specified bandwidth.

[2N3055]

Well you can believe anything you want. By now you should know I don't just throw sentences out with no reason.

Setting a probe is about it's frequency response, phase response and those two combined will yield a pulse response.

Those probes have 3 trimmers. One is standard that gets set at 1 kHz, low frequency response and base probe gain. Other two, are part of a filter that is scope side, on the other end of cable. Those two will have large influence to a first 30-40ns of the pulse response. So you need, say 5MHz or smaller frequency to have a piece of flat signal plateau after that first part of the edge to be able to set it right. In which case you go with factory recommended 1 MHz, instead of pulling some numbers out of the hat.

Also you don't set pulse response to a 25 Ohm source that has 33ps risetime. Probe is not designed for that. Higher part of the spectrum will shoot trough. Don't believe?

I have at my disposition those exact probes SP3050A. I have Leo's pulser. I have two 1 GHz scopes, one of which happens to be SPS6000. And also have SDG6000X (that can do 500MHz and 1 ns pulse risetimes).
And I do have 1GHz/ 500ps pulse risetime capable source.
So it is not an opinion nor theory on my side.

As usual Mike is right. Standard is to use 3x faster edge than probe response.

As for sensitivity to edge speed, you forgot that probes are complex impedances, HF probes even more so. They will have resonances etc. How we deal with them? We try to push them higher, over frequency probe is meant to work on. There will be peaking too. So you have first trimmer that shunts divider, than cable than a box with compensator. All of that is interacting with scope input, it's capacitance and transmission line properties.. all those things together can ring like a church on Sunday..

And then you tune them up best you can. Some will prefer signal to overshoot to get that clean edge because they need to measure risetimes. Some will tune it down, making probes have risetime of 1,5-1,8ns making them 200Mhz probes but with really nice pulse response without overshoots.

But none of them will be perfect. I always chuckle when I see Nico's signature.. That is so right..

[2N3055]

Actually it really shouldn't be that sensitive to edge speed either, since that primary adjustment is just matching the capacitive divider.

At lower frequencies, a capacitive divider is a good enough model. But at higher frequencies this model is IMO too trivial, due to parasitics.

Btw, nice paper on measuring probe performance:
https://www.picotest.com/downloads/DesignCon_2020_final_edit.pdf

EDIT:

This is why you can get pretty close on compensation even with the relatively blunt 1kHz signal that the typical scope calibrator puts out.

And 1kHz is in the region of the cross-over from the capacitive divider to the resistive 10MOhm : 1MOhm divider. If there is only a single trimmer, then only the (low-frequency) match between resistive and capacitive divider can be adjusted, hoping that the capacitive divider model is good enough up to the specified bandwidth.

That is a good paper, and Steve Sandler is a scary smart guy. I recommend reading anything he writes..
These SP3050A probes actually have 2 HF trims, so they are fiddly to set.
I don't have schematic for them, and you are correct, it's going to be physical modeling too to get it right.
 
So, when in doubt, follow the manufacturers procedure. If they say 1MHz /700ps edge, so be it.

[bdunham7]

As for sensitivity to edge speed, you forgot that probes are complex impedances, HF probes even more so. They will have resonances etc. How we deal with them? We try to push them higher, over frequency probe is meant to work on. There will be peaking too. So you have first trimmer that shunts divider, than cable than a box with compensator. All of that is interacting with scope input, it's capacitance and transmission line properties.. all those things together can ring like a church on Sunday..

And then you tune them up best you can. Some will prefer signal to overshoot to get that clean edge because they need to measure risetimes. Some will tune it down, making probes have risetime of 1,5-1,8ns making them 200Mhz probes but with really nice pulse response without overshoots.

But none of them will be perfect. I always chuckle when I see Nico's signature.. That is so right..

For me this whole discussion just reinforces the idea that using 10X/10M probes at these frequencies is highly problematic.  As I predicted, I think we simply have different notions about what a properly compensated probe is.  I I'm not claiming your version is 'wrong', but I'm not sure I see the value in doing it that way (overcompensated, IMO).  Your last screenshot looks nicely dialed in, but is it the truth? 

One minor nitpick--the SP3050A does not specify a 1MHz/700ps signal.  It specifies a 1MHz/<700ps signal, no implication AFAIK that there is a lower limit.

My workbench is torn down right now so I can't show a complete rundown on how I would characterize a probe using what I have, but I did get my hands on two 500MHz+ probes, the Bodnar pulser and the SDS2354X+.  The two probes--Tek P6139A and P6156-10X--don't work with traditional BNC adapters and I can't find the little ground collar wire at the moment, so the ground here is literally a straight cut section of a paper clip.  Also, since I typically use probes like this to minimize circuit loading, not display 500MHz or fast edges, I only now discovered that only 3 of the 4 trimmers in the P6139A move properly, so I didn't kill myself trying to dial in a pretty picture.  However, with these probes in their present setup I certainly do not have '200MHz probes' or anything remotely like that.  They display 500MHz signals and sub-ns rise times as advertised, without the grotesque overshoot shown in your first screenshot. 

Anyway, here's the Bodnar pulser connected directly to the scope at 50R and 1M, then through the two probes at 500R/50R and 10M/1M.  In the next few weeks if I get my stuff back together maybe I'll try and illustrate the loading effect that the various probes have.  You can see that the direct-coupled signal into the 50R scope input is pretty faithfully rendered through the probes. 

[2N3055]

Did you measure risetimes? That would be interesting. Note the combined risetime of 1ns.
Take into account that scope I used has twice the bandwidth of SD2000X+ so it will show spikes 2000X+ won't.

P6156 doesn't fit into this discussion. It is a special Lo-Z probe high speed probe (1pF, 3.5 GHZ) that is bound to have excellent frequency/step response in this test, and any imperfections won't show on a 500MHz scope. SP3050A is a Hi-Z 10X probe..

As for 1MHz/<700ps signal, yeah I agree it is not mathematically rigorous. But until Leo made his pulser few years ago, it wasn't trivial having 40ps risetime signal source.. Pretty much you had to make one, and then have fast enough equipment to characterize it.. so they specify 1MHz/<700ps signal, not expecting you'll get much faster than few 100s ps.
And it is not critical if it is 400 or 500 or 600ps. But when you jump from 1ns to 40ns you can see the difference, if your scope is fast enough... And controls on probe might not be able to suppress all of it. Some of it will shoot trough.

One thing we agree on wholeheartedly is that passive probes are fiddly and not to be trusted at very high speeds. And not only because of some of the stuff show here, but what you see will be highly dependent on source impedance... Add quite large capacitance that will detune many circuits and you have to be careful what you're probing and be skeptical at what you see..

[nctnico]

One thing we agree on wholeheartedly is that passive probes are fiddly and not to be trusted at very high speeds. And not only because of some of the stuff show here, but what you see will be highly dependent on source impedance... Add quite large capacitance that will detune many circuits and you have to be careful what you're probing and be skeptical at what you see..

I agree and IMHO you shouldn't even bother to use a passive high impedance probe on signals >100MHz if you expect something on the screen that resembles the signal you are measuring.

[mawyatt]

Basically anything and everything you measure is Wrong, just a question on how wrong.

Heisenberg Uncertainty Principle 

Best,

[bdunham7]

Did you measure risetimes? That would be interesting. Note the combined risetime of 1ns.
Take into account that scope I used has twice the bandwidth of SD2000X+ so it will show spikes 2000X+ won't.

P6156 doesn't fit into this discussion. It is a special Lo-Z probe high speed probe (1pF, 3.5 GHZ) that is bound to have excellent frequency/step response in this test, and any imperfections won't show on a 500MHz scope. SP3050A is a Hi-Z 10X probe..

I mention it and demonstrate it because at 500MHz 'Lo-Z' and 'Hi-Z' get turned on their heads. 

One thing we agree on wholeheartedly is that passive probes are fiddly and not to be trusted at very high speeds. And not only because of some of the stuff show here, but what you see will be highly dependent on source impedance... Add quite large capacitance that will detune many circuits and you have to be careful what you're probing and be skeptical at what you see..

I agree and IMHO you shouldn't even bother to use a passive high impedance probe on signals >100MHz if you expect something on the screen that resembles the signal you are measuring.

I think you can think about it by dividing the probing errors into two elements--the actual effect the probe has on the circuit (as measured by a hypothetical second 'perfect' probe) and the difference between what the signal actually is with the probe in circuit and what you see on the screen.  Look at my example of connecting the Bodnar pulser directly to the scope with a 10M/15pF input--it clearly has an extended rise time.    If the scopes 15pF affects the rise time that much, the 11pF + complex impedance of the probe likely affects it a similar amount as well, although not as cleanly due to inductance.  Then you 'compensate' your probe to give you a picture that represents what you think should be there without the probe in circuit, but in reality you are compensating for the probe itself, the actual effect the probe has on the signal, and perhaps the scope as well.

I'll revisit this when and if Grainger manages to deliver my new workbench and I get all set up again.  I hope the OP has managed to sort out his probes.  One interesting thing that I noticed looking at scope specs is that the higher-bandwidth Siglent scopes actually have more input capacitance--SDS1000X-E is 15pF, SDS2000X+ is 17pF and the SDS6000A is 20pF.  So if there is an issue, the OP's probes may be overcompensated for the SDS2000X+.

[mawyatt]

One of the important requirements of RF/MW/MMW measurements, especially at the chip level is to de embed the measurement artifacts caused by the test instrument, probes and test fixture. Here known "standards" are employed, and corrections applied to the measurements from these standards measurements.

In the time domain (TD) the unit step function is one of these "standards", and why most scopes have this feature built-in. Just like frequency domain (FD) standards, as the frequency/speed goes up the uncertainty of the standards becomes greater, and thus better standards are required.

When the speed becomes high enough to engage system level parameters that are not normally observed at lower speeds, then the measurement system "model" begins to deviate from the 1st order, then 2nd order, 3rd and beyond which requires more complex compensation to "de embed" the probe and instrument (scope in this case, and thus maybe helps explain the attentional compensation trims on the high frequency probes under discussion). Unlike the FD where software de embedding techniques are employed post data capture, the TD is done in real time literally on the measured signal of interest.

Which begs the question if at some point the TD takes a page from the FD measurements and employs post processing beyond the simple offset and gain correction?? Maybe the upper end 100GHz scopes do some of this, dunno, but jumping back and forth between TD and FD isn't quite as simple as doing an FFT or IFT, especially when non-linearities are involved.

Anyway, just thinking out loud...sorry for the ramblings.

Best,

[2N3055]

One of the important requirements of RF/MW/MMW measurements, especially at the chip level is to de embed the measurement artifacts caused by the test instrument, probes and test fixture. Here known "standards" are employed, and corrections applied to the measurements from these standards measurements.

In the time domain (TD) the unit step function is one of these "standards", and why most scopes have this feature built-in. Just like frequency domain (FD) standards, as the frequency/speed goes up the uncertainty of the standards becomes greater, and thus better standards are required.

When the speed becomes high enough to engage system level parameters that are not normally observed at lower speeds, then the measurement system "model" begins to deviate from the 1st order, then 2nd order, 3rd and beyond which requires more complex compensation to "de embed" the probe and instrument (scope in this case, and thus maybe helps explain the attentional compensation trims on the high frequency probes under discussion). Unlike the FD where software de embedding techniques are employed post data capture, the TD is done in real time literally on the measured signal of interest.

Which begs the question if at some point the TD takes a page from the FD measurements and employs post processing beyond the simple offset and gain correction?? Maybe the upper end 100GHz scopes do some of this, dunno, but jumping back and forth between TD and FD isn't quite as simple as doing an FFT or IFT, especially when non-linearities are involved.

Anyway, just thinking out loud...sorry for the ramblings.

Best,

Please , more ramblings...:-)

Higher end scopes do dembedding  in software and hardware (DSP). Some have it as a basic part of acquisition block and do linearizations and all kind of DSP to make response better despite front end /ADC imperfections...
All kinds of phase/gain response EQ are in question...
And 100GHz scope from LeCroy divides bandwidth into 3 bands 33GHz wide, downconverts upper 2 bands down and samples them separately and then assemble signal back together in digital domain. There is math, DSP, and all kinds of calibration involved to make it work...

On The Signal Path, explanation by the designer of the beast...

[mawyatt]

@2N3055,

Thanks so much for the video, fascinating stuff!! Although not finished yet!!

Remember when LeCroy was using the IBM8HP SiGe BiCMOS process way back sometime ago, we were early access partners with IBM with 7HP and 9HP but not 8HP (were waiting for 9HP), Tektronix was also involved back then.

Best,

[Hydron]

My RTB2004 came with similar probes, needing both LF and HF (x2) adjustment. The probe comp wizard guides you through setting both of these, and the second HF step uses a 1MHz square wave (LF is 1kHz). If you don't want to use the wizard, you can setup the pattern generator manually to spit out the appropriate square wave, though that might need the right licence to enable access

The key to this working is that the probe comp (/pattern gen) output has a risetime of ~1ns or so (fast enough that it's hard to accurately measure without another better scope), and a design such that you can ground the probe using the conductive ring near the tip for good signal integrity, rather than using a normal ground clip lead, which would totally wreck things at that edge speed.

I gave away my Bodnar pulser to a friend who makes better use of it than I do, though I'm tempted to get another!