New Rigol DS1054Z oscilloscope

[TheSteve]

Performance has been verified by many users. The four I have tested have all been good to 130 MHz + on all 4 channels.

So by all means UNLOCK your DS1054Z !

[Fungus]

has anybody posted the before and after results of a transient response test?  I have been watching for this but never found it.

Yes, many times. It's buried somewhere in the massive DS1054Z threads. Measured bandwidth on these scopes is about 140MHz.

Like I said, I specifically watched for this and never saw any results posted.  I just did another search and found nothing except for a depressing number of posts by me on the subject.

Did you use the forum search? It's bad.

I typed "eevblog ds1054z pulse rise time" into google and got this as the first hit.

https://www.eevblog.com/forum/testgear/pulse-generator-rise-time-and-rigol-ds1054z-bandwidth/

An entire thread on the subject! Complete with videos of tests!!

[David Hess]

A time domain test would be redundant to the frequency sweeps that have been made.  These scopes have no fancy DSP, so the time domain response can be inferred from the frequency domain response, which is known to be a smooth rolloff.

This is completely false.

Scalar frequency measurements do not reveal group delay.  A transient response test is more than just an independent way to confirm bandwidth.  And we do not know what kind of processing these oscilloscopes do which could affect the transient response which is why I suggested making both single shot and normal measurements.

Did you use the forum search? It's bad.

Are you kidding?  I went straight to Google.  The forum search is less than useless.

I will go through the discussion thread you linked ... again ... but I don't remember finding anything definitive.

Update: That did not take long.  We do not really know the conditions of the test but the upgraded MSO1074Z looks pretty horrible with a roughly 7ns something which is not quite dribble up after the leading edge.  It is a little difficult to tell without seeing the pulse on the 2467B at a slower timescale.  If I saw that on one of my 100 MHz oscilloscopes, I would start looking for what is broken.  I still want to see what it looks like without the display processing going on.

[Fungus]

I will go through the discussion thread you linked ... again ... but I don't remember finding anything definitive.

What about the video on the first page?



Start at 16:26

[paranoidsat]

Hi all

Got my new Rigol 1054z. Ha**ed, updated firmware and recalibrated it. Worked like a charm. Played the whole day with it 

Thank you guys for all your great work!! Keep on going!!

PS:
After recalibration the 500µV also worked fine. Without recal the 500µV where horizontal moved.

[David Hess]

What about the video on the first page?

Unfortunately the test conditions in the video are not ideal.  He has to use an external feedthrough termination on the Rigol but does *not* check it against the internal terminations on the other oscilloscopes so we do not know what effect it had.

The hacked Rigol's settling time seems reasonable but that 5ns ramp after the edge indicates a problem and doubly so since the ETS DSOs he used indicate that the pulse has real overshoot.  That is exactly the sort of thing I would expect on a DSO hacked for extra bandwidth because the transient response would look better at lower bandwidth.  Most users would never notice it however, it does *not* meet the specifications of a good 100 MHz oscilloscope.

It looks like his Tektronix 2467B is out of calibration; maybe the calibration data stored in non-volitile memory was lost.

I have an analog sampling oscilloscope for verifying the performance of my fast reference level pulse generator so that it can be used for oscilloscope calibrations.

It is fascinating how the displayed waveform changes his Agilent real time oscilloscopes when they operate in equivalent time sampling mode.

[Fungus]

What about the video on the first page?

Unfortunately the test conditions in the video are not ideal.  He has to use an external feedthrough termination on the Rigol but does *not* check it against the internal terminations on the other oscilloscopes so we do not know what effect it had.

The hacked Rigol's settling time seems reasonable but that 5ns ramp after the edge indicates a problem and doubly so since the ETS DSOs he used indicate that the pulse has real overshoot.  That is exactly the sort of thing I would expect on a DSO hacked for extra bandwidth because the transient response would look better at lower bandwidth.  Most users would never notice it however, it does *not* meet the specifications of a good 100 MHz oscilloscope.

That may be true but your question was this:

"has anybody posted the before and after results of a transient response test?"

We know how the 'scope works internally so we know there's no physical difference between a DS1054Z and a DS1104Z.

The video shows a rise time on a hacked DS1054Z that indicates 150MHz bandwidth.

What more do you want?  It's a $400 'scope...


(PS: there *are* other screenshots buried in the threads somewhere. It's a pity that the the resident Rigol-haters have made the threads ten times longer than they need to be).

[David Hess]

That may be true but your question was this:

"has anybody posted the before and after results of a transient response test?"

Maybe I should have been more specific.  Has anybody posted any rigorous transient response test results?

There have been lots of posts showing fast edges from various sources but most are not fast enough and have questionable or unknown characteristics.  That video is the only test I have seen with any credibility.

We know how the 'scope works internally so we know there's no physical difference between a DS1054Z and a DS1104Z.

The designs and production may be identical but Rigol could be grading them after production for the ones which will support 100 MHz with good transient response and the ones which will not.  As I recall, the design includes no adjustments so alternatively maybe they do board rework which takes time adding to the cost.

Some modern DSOs apply frequency and phase compensation through DSP.  The ones which can be field upgraded must have good transient response at their maximum upgraded bandwidth however Rigol does not do that so there is no such requirement for them.

While I have not seen it done at 100 MHz and I doubt Rigol did it, they could have relied on the stock passive probe to correct the transient response.  There are some aftermarket probes with high frequency compensation adjustments which could correct it but of course this does nothing when a feedthrough termination is used.  This is more common with higher bandwidth oscilloscopes.

The video shows a rise time on a hacked DS1054Z that indicates 150MHz bandwidth.

What more do you want?  It's a $400 'scope...

It *is* a great deal but if the results shown in the video are accurate, then the hacked model may not be a 100 (or 150 MHz) oscilloscope in the same sense that that the genuine model is which is what everybody is claiming.  An oscilloscope is a time domain instrument so I want the transient response to be as good as possible; if I saw the result shown in the video and it was accurate, then I would assume that the oscilloscope is broken or in need of calibration.

If I wanted just bandwidth, then I could tweak the high frequency compensation of on oscilloscope and get it at the expense of transient response.  Tektronix actually did this as an option on their old 7704A which came in 200 and 250 MHz versions.

There is a guy on Ebay who takes 150 MHz Tektronix 2445Bs, bypasses the fixed bandwidth limiting filter between the delay line and vertical CRT amplifier, reconfigures them, and sells them as 400 MHz 2465Bs.  Do they actually perform like real 2465Bs?  They do unless the user looks closely and most users lack a 50ps reference level pulse generator which would reveal what is going on and they would not notice when using standard passive probes with their high circuit loading except under very controlled conditions.  These hacked "2465Bs" have lower bandwidth and poorer transient response than real ones and can never be adjusted to meet the 2465B specifications.

[metrologist]

An oscilloscope is a time domain instrument so I want the transient response to be as good as possible; if I saw the result shown in the video and it was accurate, then I would assume that the oscilloscope is broken or in need of calibration.

David, what about the trace makes you assume that? The trace looks very similar on the following Tek 2467B with a little more than 1/3 the rise time. The following Tek and Agilent show overshoot. What should the trace actually look like?

[rstofer]

Maybe I should have been more specific.  Has anybody posted any rigorous transient response test results?

Apparently not, at least in terms you are willing to accept

We know how the 'scope works internally so we know there's no physical difference between a DS1054Z and a DS1104Z.

The designs and production may be identical but Rigol could be grading them after production for the ones which will support 100 MHz with good transient response and the ones which will not.  As I recall, the design includes no adjustments so alternatively maybe they do board rework which takes time adding to the cost.

Then how could I call up and get a secret squirrel code to implement the higher bandwidth?  All of the upgrades are just a matter of firmware passcodes.  Under your scenario, we could get an upgrade code but the scope wouldn't actually perform any better than the 50 MHz variant.  This is patently false.  It has been shown that the frequency domain bandwidth is in excess of 150 MHz and for all practical purposes, that is adequate.

The video shows a rise time on a hacked DS1054Z that indicates 150MHz bandwidth.

What more do you want?  It's a $400 'scope...

It *is* a great deal but if the results shown in the video are accurate, then the hacked model may not be a 100 (or 150 MHz) oscilloscope in the same sense that that the genuine model is which is what everybody is claiming.  An oscilloscope is a time domain instrument so I want the transient response to be as good as possible; if I saw the result shown in the video and it was accurate, then I would assume that the oscilloscope is broken or in need of calibration.

If I wanted just bandwidth, then I could tweak the high frequency compensation of on oscilloscope and get it at the expense of transient response.  Tektronix actually did this as an option on their old 7704A which came in 200 and 250 MHz versions.

There is a guy on Ebay who takes 150 MHz Tektronix 2445Bs, bypasses the fixed bandwidth limiting filter between the delay line and vertical CRT amplifier, reconfigures them, and sells them as 400 MHz 2465Bs.  Do they actually perform like real 2465Bs?  They do unless the user looks closely and most users lack a 50ps reference level pulse generator which would reveal what is going on and they would not notice when using standard passive probes with their high circuit loading except under very controlled conditions.  These hacked "2465Bs" have lower bandwidth and poorer transient response than real ones and can never be adjusted to meet the 2465B specifications.

I suspect you are going to have to do your own tests although I am not certain of the end game.  It's just a simple 100 MHz scope.  If anybody wants serious bandwidth, this isn't where they'll start.

[Fungus]

We know how the 'scope works internally so we know there's no physical difference between a DS1054Z and a DS1104Z.

The designs and production may be identical but Rigol could be grading them after production for the ones which will support 100 MHz with good transient response and the ones which will not.  As I recall, the design includes no adjustments so alternatively maybe they do board rework which takes time adding to the cost.

They could be doing that, sure, but it makes no sense at all. The only components that could make a difference are a handful of things like discrete transistors and capacitors. The proportion of them that could work at 50MHz but not 100MHz will be ridiculously small. Much more likely is that if they don't work at 100Mhz then they won't work at 50Mhz either.

The most critical part of the electronics is downstream of those components and in that pert of the scope the different versions must behave identically. No grading/sorting is possible there.

In short: It makes no sense to spend time and money sorting/grading them. The cost savings will probably be less than the extra cost of those transistors and capacitors used to select the bandwidth.

[RoGeorge]

This post is just to easily follow the subject.

[sandor626]

After recalibration the 500µV also worked fine. Without recal the 500µV where horizontal moved.

Hi, your ds1054s  long ago when it was taken ?

[David Hess]

An oscilloscope is a time domain instrument so I want the transient response to be as good as possible; if I saw the result shown in the video and it was accurate, then I would assume that the oscilloscope is broken or in need of calibration.

David, what about the trace makes you assume that? The trace looks very similar on the following Tek 2467B with a little more than 1/3 the rise time. The following Tek and Agilent show overshoot. What should the trace actually look like?

The Tektronix 2467B is running at 10 times the sweep speed (500ps/div) so it does not show the same extent of the pulse to determine this.  I also suspect it is also out of calibration because of comparison with the later images; it should be showing at least some of the overshoot displayed in the later tests.  The 2467B is a little unusual in that the corner transient response adjustment is done electronically and if the NVSRAM calibration data is, it is lost also and calibrating it is not an easy task.

All of the later DSO traces show overshoot.  The real time traces show preshoot (and more overshoot) which I think is an artifact caused by the sin(x)/x filtering.  The two equivalent time sampled traces show no preshoot and probably accurately reflect the overshoot.

Maybe I should have been more specific.  Has anybody posted any rigorous transient response test results?

Apparently not, at least in terms you are willing to accept

I explained why the tests in the video are not rigorous but they are the best we have.  If you have a link to some better tests, I will be happy to look them over.  I could do the tests myself in a couple minutes but lack a Rigol oscilloscope.

We know how the 'scope works internally so we know there's no physical difference between a DS1054Z and a DS1104Z.

The designs and production may be identical but Rigol could be grading them after production for the ones which will support 100 MHz with good transient response and the ones which will not.  As I recall, the design includes no adjustments so alternatively maybe they do board rework which takes time adding to the cost.

Then how could I call up and get a secret squirrel code to implement the higher bandwidth?  All of the upgrades are just a matter of firmware passcodes.  Under your scenario, we could get an upgrade code but the scope wouldn't actually perform any better than the 50 MHz variant.  This is patently false.  It has been shown that the frequency domain bandwidth is in excess of 150 MHz and for all practical purposes, that is adequate.

The passcodes adjust what restriction are imposed on the analog bandwidth filters and we even have schematics to show how they are implemented courtesy of Dave.  The test in the video shows higher bandwidth at the expense of transient response.  Are you suggesting that the passcodes should fix the transient response?  That is possible but the video does not show it.

The video shows a rise time on a hacked DS1054Z that indicates 150MHz bandwidth.

What more do you want?  It's a $400 'scope...

It *is* a great deal but if the results shown in the video are accurate, then the hacked model may not be a 100 (or 150 MHz) oscilloscope in the same sense that that the genuine model is which is what everybody is claiming.  An oscilloscope is a time domain instrument so I want the transient response to be as good as possible; if I saw the result shown in the video and it was accurate, then I would assume that the oscilloscope is broken or in need of calibration.

If I wanted just bandwidth, then I could tweak the high frequency compensation of on oscilloscope and get it at the expense of transient response.  Tektronix actually did this as an option on their old 7704A which came in 200 and 250 MHz versions.

There is a guy on Ebay who takes 150 MHz Tektronix 2445Bs, bypasses the fixed bandwidth limiting filter between the delay line and vertical CRT amplifier, reconfigures them, and sells them as 400 MHz 2465Bs.  Do they actually perform like real 2465Bs?  They do unless the user looks closely and most users lack a 50ps reference level pulse generator which would reveal what is going on and they would not notice when using standard passive probes with their high circuit loading except under very controlled conditions.  These hacked "2465Bs" have lower bandwidth and poorer transient response than real ones and can never be adjusted to meet the 2465B specifications.

I suspect you are going to have to do your own tests although I am not certain of the end game.  It's just a simple 100 MHz scope.  If anybody wants serious bandwidth, this isn't where they'll start.

I am not as concerned about the bandwidth as I am about the transient response; the former can always be pushed at the expense of the later and reverse.

You are right about testing and a couple years ago I bought and calibrated a pair of pocket sized 450ps reference level pulse generators for oscilloscope testing in the field.  They were intended for quick and dirty bandwidth and functional testing but with my sampling oscilloscope, I recalibrated them to produce a slightly slower but maximally flat pulse for transient response testing.  In practice they are not significantly faster than my PG506 and the later produces an even better pulse so that is why I usually use.

[metrologist]

The passcodes adjust what restriction are imposed on the analog bandwidth filters and we even have schematics to show how they are implemented courtesy of Dave.  The test in the video shows higher bandwidth at the expense of transient response.  Are you suggesting that the passcodes should fix the transient response?  That is possible but the video does not show it.

Thanks, David. Are you suggesting that the passcodes break the transient response, so an unhacked scope is preferable in this regard, or that the entire line is broken?

The pulse generator I have handy is not fast enough.

[metrologist]

I was just playing around with my scope. What does this mean? Does it say anything about the transient response?

[David Hess]

Thanks, David. Are you suggesting that the passcodes break the transient response, so an unhacked scope is preferable in this regard, or that the entire line is broken?

The pulse generator I have handy is not fast enough.

What I am saying is that the transient response of the Rigol tested in the video was not good enough to support the higher bandwidth unlocked by the passcode.  At a lower bandwidth, the transient response may have been adequate but we do not know because that test was not done.

I was just playing around with my scope. What does this mean? Does it say anything about the transient response?

That is more like what it should look like but it still is not very good; it looks like a ground lead with alligator clip was used with a x10 probe.  The test setup matters like how the oscilloscope input is connected to the source.  The pulse source needs to be calibrated; normal pulse generators are usually not suitable for this type of test.

[metrologist]

How about LiPo battery (perhaps with a cap in parallel) and a mercury switch, short leads, to 50 ohm pass through?

[David Hess]

How about LiPo battery (perhaps with a cap in parallel) and a mercury switch, short leads, to 50 ohm pass through?

The early highest performance reference level pulse generators with the Tektronix Type 109 being the prime example worked this way using a coaxial mercury switch.  The Type 109 has a specified rise time of less than 250 picoseconds and exceptionally clean settling.

The problems with this method are low repetition rate at 10s to 100s of Hz and low relay operating life at 100s of hours.  Low repetition rates are of course not a problem with digital storage oscilloscopes.  One big advantage of this method is good control of the pulse amplitude; both low and high voltages are not a problem.

Transistor based reference level pulse generators like the PG506 are good to 500 picoseconds.  Some fast logic like 74AC can work also.  The problem is verifying performance if you do not already have a way to accurately measure the characteristics of the pulse.

[metrologist]

I may be naive, but so far I am rather pleased with my measurement. You said it (presumably the transient response of the scope) was not very good. I understand that we (I) cannot verify the actual signal response, and my test was not highly repeatable (usually would completely fail to capture an edge and I'd get flat or diagonal trace), but I did get several very similar signals ranging from 1.9 to 2.1 ns.

I'm going to keep looking for a more reliable signal. I do not like the avalanche diode as I want a longer pulse width. I do have some microwave stuff around. Maybe I'll find something with an SRD or tunnel diode, and I also read that there is some kind of driver in HDD's. I do want to know how my scope actually performs.

Thanks.

[David Hess]

I may be naive, but so far I am rather pleased with my measurement. You said it (presumably the transient response of the scope) was not very good. I understand that we (I) cannot verify the actual signal response, and my test was not highly repeatable (usually would completely fail to capture an edge and I'd get flat or diagonal trace), but I did get several very similar signals ranging from 1.9 to 2.1 ns.

The test edge needs to be faster, say about 4 times, than the rise and fall time of the oscilloscope to effectively test it.  A PG506 with a transition time of about 500ps is good for testing up to 200 MHz.  To test the 400 MHz 2465B which was shown in the video, a tunnel diode pulser was normally used.

I'm going to keep looking for a more reliable signal. I do not like the avalanche diode as I want a longer pulse width. I do have some microwave stuff around. Maybe I'll find something with an SRD or tunnel diode, and I also read that there is some kind of driver in HDD's. I do want to know how my scope actually performs.

A charge line can be used with an avalanche pulse generator to extend the pulse width to a usable length.  The problem with avalanche pulsers is that they need to be adjusted to get a clean edge with a minimum of aberrations which becomes a chicken in the egg problem if you do not have access to a fast oscilloscope which is already calibrated.

The easiest solution now is probably the output from some fast CMOS logic but again, the output needs to be verified somehow.

Below are two examples of good transient response on 100 MHz oscilloscopes.  Note that DSOs that have a sharp non-gaussian rolloff have a different transient response and aliasing in a DSO without equivalent time sampling causes problems as well; in the last part of the video, the transient response changes (and improves) when ETS is used.  I think that may explain the preshoot bump shown in the Rigol in the video.  The 2467B in the video has preshoot for a different reason which is unrelated to DSOs.

[Fungus]

I was just playing around with my scope. What does this mean?

It means the DS1054Z is a total bargain for $400. 

I may be naive, but so far I am rather pleased with my measurement.

Yep. Sub 2ns is pretty good.

What was requested is the difference before/after the unlock. If you get an ethernet cable and telnet into the 'scope (port 5555) you can make it into a 50Mhz 'scope by typing this:

:SYSTem:OPTion:UNINSTall

That removes all the unlockable options. To re-enable them you can do:

:SYSTem:OPTion:INSTall RDJxxxxxxxxxxxxxxxxxxxxxxxxx

Where xxxxxxxxxxxxxx is one of the unlock codes generated by Riglol.

[RoGeorge]

I'm going to keep looking for a more reliable signal. I do not like the avalanche diode as I want a longer pulse width.

A very simple, but well tested circuit for sharp edges:
#88: Cheap and simple TDR using an oscilloscope and 74AC14 Schmitt Trigger Inverter

[David Hess]

I'm going to keep looking for a more reliable signal. I do not like the avalanche diode as I want a longer pulse width.

A very simple, but well tested circuit for sharp edges:

It is well tested for sharp edges and horrible transient response.  If your Rigol produced the results shown in that video with a perfect test signal, would that be acceptable?

W2AEW points out in his video that transient response does not matter for his TDR application because the step will have settled before the reflection returns.  Some small additions to that circuit will fix the transient response.

[RoGeorge]

Thank you for pointing me to the ugly transient response, I was not aware about it.

I saw that video some years ago, and I was remembering the circuit for what strike me most at that time: a very simple circuit that can generate 500ps edges using only common parts, and without any tricks like avalanche junctions or so. Your question made me watch again both videos (there is another one where w2aew is measuring the rise and falling time of a kit with the same circuit: "#203: Schmitt Trigger Oscillator revisited").

Indeed, that circuit not only that it has horrible ringing and ugly asymmetric transient response, but it also has 500 ps only for the falling edge. The rising edge is about 2 ns, nothing spectacular.