I saw a couple of different problems. The slope of the pulse top on the RTB2000 looked like printed circuit board hook in the high impedance attenuator. The reverse of the top on the Tektronix looked like a mismatch in medium frequency compensation of the two path front end. However I would not rely on the flatness from your AWG without verifying it using a sampling oscilloscope. A reference level pulse generator is normally required for these sorts of measurements.
I have recently been corresponding with John Addis, the designer of the Tektronix 7A29 1 GigHz vertical amplifier, and principle engineer of the 7104 1 GigHz mainframe... which I own. Those guys had all these problems solved 50 years ago. The attenuator in the 7A29 actually works to 5 Gigs... the guys making this stuff today have no clue about how it was done, and are doomed to create glitchy stuff. My other scopes are: Tek R7844 400MHz dual beam rack mount Tek 7834 400MHz storage scope. Tek 2247A 100MHz portable with integrated digital co-processor. Non of these exhibit any of these issues. I use a Siglent SDS 2202X-E but mostly in FFT driven by residual out of a distortion analyzer, the Tek AA501. The degree of real science and engineering in the post 1970 to pre 1990 Tek products is unparalleled. No finer analog engineering exists.
For a good design the only residual pulse distortion will be from the calibration of the frequency compensation and calibration of the thermal balance. The specifications usually require aberrations to be within 0.2 divisions but performance is commonly much better than this. The results Dave got are probably considered within their specifications, but I sure would not put up with some of them. I have: 2 x 2230 2 X 2232 which are my most used DSOs 2247A 2246 7904A which is my favorite analog oscilloscope 7603 2 x 7834 7854 2440
Insult my wife or religion but not my dog or oscilloscope. Wish you had tried one of those old analog scopes you've repaired over the years, that would be interesting.
😂 was searching for that comment "It's schroedingers squarewave" But as others mentioned, how can we be sure the signal is generated totally perfect unless we measure? If we measure the exact same signal with several scopes and get a similar result, we might need to change the source, to check if the quirks come from there, right? 😯
Dave, it is just a pulse, representing 1 or 0 in digital world, it does not matter perfect or not. But we do like the perfect square wave oscilloscope, however. 😂
It's invariably true that any engineered product will have design choices made that involve achieving sufficient accuracy without pricing the product out of any reasonable market.
To jump further down the rabbit hole, now I want to try this on a fast scope! If I remember I'll try it on a Rohde RTP at the office...or maybe a 30GHz Tek 🤣
If that slope error is a result of error in the compensation network do you think some freezer spray would worsen it or is this more of a compensation network design/topology limitation?
That slope is more typical of circuit board hook in a high impedance attenuator and is not really suitable for compensation using a network. A measurement over a longer time period would make this more certain. The circuit board might have absorbed humidity after being manufactured causing this. Many differential high voltage probes suffer from this as well after months to years. Early oscilloscopes used plastic substrates (Polysulfone?) for their high impedance attenuators to avoid this problem before it was understood. It is not a mistake I would expect Rohde & Schwarz to make.
what if when using the same signal at the same time observed by an other scope for comparison (I would say a propper termination is not needed for this test)
you know what ? am ready to have this scope and pay for all costs to have it here in france next to my tek465 , by the way in my past engineers in rf experience R&S was already the best for ease of use reliability and support , way better in RF than HP....but COSTLY ! in this tests be sure of adaptation , the gen should have at least a 10 dB attenuator on output , coax and connectors should be rated to at least 4 gHz then we may talk about perf...this is why it is strange behavior IMO issue is SWR ...even at 1kHz at good gen as a rise time of gHz and so ...
The setup was good. You not need GHz generator for precison 1kHZ flat pulses. The key is here flatness, not rise/fall time. I have 3 scope, only the newest (DSO) have this problem. My combiscope and my fully analog scope made in 70s does not show any flatness error.
Analog oscilloscopes had to deal with all of the same problems and their design was very careful to avoid them. Knowledge of these problems has been lost over time and new engineers may not be aware of them.
This is tipical for shitty smd parts, but not because they are digital. My analog/digital VC-6145 combiscope does'nt have that problem at all, but my GDS-2104E is distorting like hell...
Is there no agreed spec for 20MHz LP filters on scopes? This all looks like a bunch of different impulse responses to me. But I can't explain why it changes with vertical scale.
It changes with the volts/div setting because of switching in and out of the high impedance attenuators, and operating point changes in the high impedance buffer. It is a difficult task to design each to avoid these problems.
@@Alamagosa Yes, I was thinking about that. A scope attenuator is composed of capacitors. There are resistors too, but they only serve to allow a response down to DC. Clearly as ranges are switched reactive loads change. Given enough DSP this could be equalised but it really isn't worth it. Nobody expects that degree of accuracy from a scope.
The high impedance attenuator sections are designed to be constant input and output impedance so switching them in and out should not affect anything except the attenuation. Some DSOs implement corrections using DSP, but I do not know that any modern low end DSOs do. The displayed results are likely within the specifications, but do reveal some real flaws. I would not tolerate that slope shown on the on the Rohde & Schwarz RTB2000 on any of my oscilloscopes. Some of the other results are questionable also.
It's strange how the more expensive scopes have that annoying glossy look to them. It Siglent and UNI-T can put a proper low reflection screen on even their low-end scopes, why can't R&A or Tek? Annoying. My Siglent SDS2504XP scope does not have this problem from a 10mV input up to a 10V input. Not even slightly. My Siglent SDS1204X-E has it, but to less extent than the ones shown here. No hint of it on my old Tek 465 either - made when Tektronix care more about the quality of their products than their executive bonuses. What's up? Just because a lot of scopes do this, does not excuse it. Especially when those scopes have names like R&S, Tektronix and Agilent on them. Should we not expect more for the higher price we pay?
Dave, please don't call your oscilloscope videos a full review if they don't include a DC coupled 10Hz/100Hz square wave flatness measurement using a coax cable! In December 2021 I wrote here "EEVblog Electronics Community Forum »Products »Test Equipment »GW Instek GDS-2000E teardown /short review, comparison with Rigol DS2000(A)" under the name "Krisztián" that the GDS-2104E has serious problems with e.g. the 10Hz square wave flatness. The same signal is perfectly straight on the excellent JFET input Hitachi VC-6145 combiscope.
If the bandwidth varies with volts/div, then the design is flawed. Bandwidth and pulse shape should not change with the position and volt/div setting. Modern DSOs which only use one high impedance input attenuator have to operate their high impedance buffer over a much larger signal range, and may be slew rate limited causing variation in their full power bandwidth.
Which is likely a problem, since the front end 1 meg input has part of the probe compensation in it, as I recall, so a 50 Ohm cable has little chance of matching the probe cable and compensation?
@@derkarhu5079 It does not work like that. Like a x1 probe which is 1 megohm, a 50 ohm direct connection requires no compensation. What is required is a 50 ohm termination at higher frequencies.
My SDS2504X+ doesn't have this problem. My SDS1204X-E is also fairly consistent except on the 200mV/div range which actually shows a sharper edge. Either way, it's a non-issue and neither are anywhere close to as bad as the Rhode & Schwarz.
I saw a couple of different problems. The slope of the pulse top on the RTB2000 looked like printed circuit board hook in the high impedance attenuator. The reverse of the top on the Tektronix looked like a mismatch in medium frequency compensation of the two path front end.
However I would not rely on the flatness from your AWG without verifying it using a sampling oscilloscope. A reference level pulse generator is normally required for these sorts of measurements.
I have recently been corresponding with
John Addis, the designer of the Tektronix 7A29 1 GigHz vertical amplifier, and principle engineer of the 7104 1 GigHz mainframe... which I own. Those guys had all these problems solved 50 years ago. The attenuator in the 7A29 actually works to 5 Gigs... the guys making this stuff today have no clue about how it was done, and are doomed to create glitchy stuff. My other scopes are:
Tek R7844 400MHz dual beam rack mount
Tek 7834 400MHz storage scope.
Tek 2247A 100MHz portable with integrated digital co-processor.
Non of these exhibit any of these issues.
I use a Siglent SDS 2202X-E but mostly in FFT driven by residual out of a distortion analyzer, the Tek AA501. The degree of real science and engineering in the post 1970 to pre 1990 Tek products is unparalleled. No finer analog engineering exists.
For a good design the only residual pulse distortion will be from the calibration of the frequency compensation and calibration of the thermal balance. The specifications usually require aberrations to be within 0.2 divisions but performance is commonly much better than this. The results Dave got are probably considered within their specifications, but I sure would not put up with some of them.
I have:
2 x 2230
2 X 2232 which are my most used DSOs
2247A
2246
7904A which is my favorite analog oscilloscope
7603
2 x 7834
7854
2440
Dave, tell us again why those "Super Precision" Big $$ scopes are soooo much better, more accurate, etc. etc..... 😂
Insult my wife or religion but not my dog or oscilloscope. Wish you had tried one of those old analog scopes you've repaired over the years, that would be interesting.
Dave you fool !!!
Its schroedingers signal.
Its perfect until observed. 🤣
A man with one oscilloscope knows the pulse shape. A man with two or more oscilloscopes is never sure.
😂 was searching for that comment
"It's schroedingers squarewave"
But as others mentioned, how can we be sure the signal is generated totally perfect unless we measure?
If we measure the exact same signal with several scopes and get a similar result, we might need to change the source, to check if the quirks come from there, right? 😯
Dave, it is just a pulse, representing 1 or 0 in digital world, it does not matter perfect or not. But we do like the perfect square wave oscilloscope, however. 😂
It's invariably true that any engineered product will have design choices made that involve achieving sufficient accuracy without pricing the product out of any reasonable market.
check your signal source
Thanks for showing it on all the different brands, really interesting. Too bad no Lecroy one there :)
To jump further down the rabbit hole, now I want to try this on a fast scope! If I remember I'll try it on a Rohde RTP at the office...or maybe a 30GHz Tek 🤣
Like what you said : " no scope is perfect".
Imagine a review that includes this kind of stuff ... I guess it would be like 8h long.
I don't know what you're driving from, but doing 50 ohm into a 1 Mohm input... cable capacitance, awg.
Without using probe and probe compensation ...
If that slope error is a result of error in the compensation network do you think some freezer spray would worsen it or is this more of a compensation network design/topology limitation?
That slope is more typical of circuit board hook in a high impedance attenuator and is not really suitable for compensation using a network. A measurement over a longer time period would make this more certain.
The circuit board might have absorbed humidity after being manufactured causing this. Many differential high voltage probes suffer from this as well after months to years. Early oscilloscopes used plastic substrates (Polysulfone?) for their high impedance attenuators to avoid this problem before it was understood. It is not a mistake I would expect Rohde & Schwarz to make.
I have the Rohde & Schwarz HMO3004 500Mhz scope. It would be interesting to see if it exhibits the same thing.
Well? Did it?
what if when using the same signal at the same time observed by an other scope for comparison (I would say a propper termination is not needed for this test)
Interesting, one should always be aware of the uncertainty of a measurement.
why is this on channel 2
Are you sure the ringing not from the pulse source?
The pulse source definitely has some aberrations, however the different DSOs Dave tested showed different results.
Use Leo Bodnar's lovely pulse device.
Is that the same on analog scopes?
It is. Analog oscilloscope design went to great lengths to remove these types of errors.
No! Most of analog scopes made in the 80s have been ruled out these errors simply buy using bigger transistors or using jfets in the input stage.
you know what ? am ready to have this scope and pay for all costs to have it here in france next to my tek465 , by the way in my past engineers in rf experience R&S was already the best for ease of use reliability and support , way better in RF than HP....but COSTLY ! in this tests be sure of adaptation , the gen should have at least a 10 dB attenuator on output , coax and connectors should be rated to at least 4 gHz then we may talk about perf...this is why it is strange behavior IMO issue is SWR ...even at 1kHz at good gen as a rise time of gHz and so ...
The setup was good. You not need GHz generator for precison 1kHZ flat pulses. The key is here flatness, not rise/fall time. I have 3 scope, only the newest (DSO) have this problem. My combiscope and my fully analog scope made in 70s does not show any flatness error.
Is this typically for digital scopes? Or have analog scopes the same issue? Thanks.
Analog oscilloscopes had to deal with all of the same problems and their design was very careful to avoid them. Knowledge of these problems has been lost over time and new engineers may not be aware of them.
@@Alamagosa Thank you :)
This is tipical for shitty smd parts, but not because they are digital. My analog/digital VC-6145 combiscope does'nt have that problem at all, but my GDS-2104E is distorting like hell...
Is there no agreed spec for 20MHz LP filters on scopes? This all looks like a bunch of different impulse responses to me. But I can't explain why it changes with vertical scale.
It changes with the volts/div setting because of switching in and out of the high impedance attenuators, and operating point changes in the high impedance buffer. It is a difficult task to design each to avoid these problems.
@@Alamagosa Yes, I was thinking about that. A scope attenuator is composed of capacitors. There are resistors too, but they only serve to allow a response down to DC. Clearly as ranges are switched reactive loads change. Given enough DSP this could be equalised but it really isn't worth it. Nobody expects that degree of accuracy from a scope.
The high impedance attenuator sections are designed to be constant input and output impedance so switching them in and out should not affect anything except the attenuation. Some DSOs implement corrections using DSP, but I do not know that any modern low end DSOs do.
The displayed results are likely within the specifications, but do reveal some real flaws. I would not tolerate that slope shown on the on the Rohde & Schwarz RTB2000 on any of my oscilloscopes. Some of the other results are questionable also.
@@donepearce I would expect that precision! Its easier to redesign the analog frontend than equalise that with a DSP.
It's strange how the more expensive scopes have that annoying glossy look to them. It Siglent and UNI-T can put a proper low reflection screen on even their low-end scopes, why can't R&A or Tek? Annoying.
My Siglent SDS2504XP scope does not have this problem from a 10mV input up to a 10V input. Not even slightly. My Siglent SDS1204X-E has it, but to less extent than the ones shown here. No hint of it on my old Tek 465 either - made when Tektronix care more about the quality of their products than their executive bonuses.
What's up? Just because a lot of scopes do this, does not excuse it. Especially when those scopes have names like R&S, Tektronix and Agilent on them. Should we not expect more for the higher price we pay?
Dave, please don't call your oscilloscope videos a full review if they don't include a DC coupled 10Hz/100Hz square wave flatness measurement using a coax cable! In December 2021 I wrote here "EEVblog Electronics Community Forum »Products »Test Equipment »GW Instek GDS-2000E teardown /short review, comparison with Rigol DS2000(A)" under the name "Krisztián" that the GDS-2104E has serious problems with e.g. the 10Hz square wave flatness. The same signal is perfectly straight on the excellent JFET input Hitachi VC-6145 combiscope.
Does bandwidth depends to the selected front end attenuation?
Bandwidth should not vary with the volt/div setting.
@@Alamagosa yep, should not, but what if does?
If the bandwidth varies with volts/div, then the design is flawed. Bandwidth and pulse shape should not change with the position and volt/div setting.
Modern DSOs which only use one high impedance input attenuator have to operate their high impedance buffer over a much larger signal range, and may be slew rate limited causing variation in their full power bandwidth.
So what parts in the frontend do we need to change
Will a sweep find out the dip in the response ?
It depends on exactly what the cause is. Many of these aberrations can be caused by non-linear operation which a simple sweep would miss.
No teardown
:D OMG
I would like to see you repeat this after calibrating the probe.
No probe is being used.
Which is likely a problem, since the front end 1 meg input has part of the probe compensation in it, as I recall, so a 50 Ohm cable has little chance of matching the probe cable and compensation?
@@derkarhu5079 It does not work like that. Like a x1 probe which is 1 megohm, a 50 ohm direct connection requires no compensation. What is required is a 50 ohm termination at higher frequencies.
@@derkarhu5079 No. A straight, properly terminated good quality coax will perform well into the 10's of GHz and requires no compensation.
Why is it so damn hard to get a decent and cheap oscilloscope in a modern world! Everything else electronic gets cheaper way faster than these tools.
first
Say Hi to your mum for me.
pointless
This scope has a low quality front-end just like my GDS-2104E.
My SDS2504X+ doesn't have this problem. My SDS1204X-E is also fairly consistent except on the 200mV/div range which actually shows a sharper edge. Either way, it's a non-issue and neither are anywhere close to as bad as the Rhode & Schwarz.
"it's a non-issue" Wrong.
@@aklankrisz Explain...