Author Topic: Frequency response of your DSO  (Read 37297 times)

0 Members and 1 Guest are viewing this topic.

Offline EV

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: fi
Re: Frequency response of your DSO
« Reply #75 on: April 17, 2014, 02:14:41 pm »
Maybe the risetime can be corrected as follows: Rt^2 = 400^2 -190^2 and Rt = 352 ps.
Then BW is 350 / 352 = 994 MHz.
 

Offline Neganur

  • Supporter
  • ****
  • Posts: 1114
  • Country: fi
Re: Frequency response of your DSO
« Reply #76 on: April 17, 2014, 02:18:04 pm »
The Agilent spec sheet is for a 10-90% rise time of 450psecs giving the product as 0.45.

You're right, the datasheet lists 1 GHz BW with a calculated 10%-90% rise time of ~450 ns, so that would not be Gaussian. The user manual manual however, mentions the Gaussian response of 0.35/f_BW for 10%-90% so maybe the manual is wrong.

I did more tests with a Tek 284 Pulse generator (t_rise = 70 ps) and get figures closer to the claimed numbers in the datasheet t_r(10-90) ca. 413 ps.
 

Offline jpb

  • Super Contributor
  • ***
  • Posts: 1620
  • Country: gb
Re: Frequency response of your DSO
« Reply #77 on: April 17, 2014, 02:35:30 pm »
The Agilent spec sheet is for a 10-90% rise time of 450psecs giving the product as 0.45.

You're right, the datasheet lists 1 GHz BW with a calculated 10%-90% rise time of ~450 ns, so that would not be Gaussian. The user manual manual however, mentions the Gaussian response of 0.35/f_BW for 10%-90% so maybe the manual is wrong.

I did more tests with a Tek 284 Pulse generator (t_rise = 70 ps) and get figures closer to the claimed numbers in the datasheet t_r(10-90) ca. 413 ps.
The user manual is probably designed for the full range from 100MHz to 1GHz scopes and the ones up to 500MHz Bandwidth probably are Gaussian.
 

Offline Neganur

  • Supporter
  • ****
  • Posts: 1114
  • Country: fi
Re: Frequency response of your DSO
« Reply #78 on: April 17, 2014, 02:43:09 pm »
Here is my test with this pulser. According to Vince the its rise time is 190 ps. I used TEK R7103 scope. Its BW is 1 GHz. I get rise time 10-90% to about 400 ps. It gives BW= 0.350 / 400 = 875 MHz. I have attached pictures with time base 1 ns, 500 ps and 200 ps about the pulse. The pulser is connected with 20 dB attenuator to the scope.

I'm not sure if I measured the 067-0587-02 cal plugin or my Vince pulser #43 with my old Tek 7104 here (with 7B10 and 7A29), bit sloppy photo but I think I can read 1.8 div, so 0.35/360 ps  ~972 MHz 1.4 div, so 0.35/280 ps  ~1.25 GHz

EDIT: Needed to clean my glasses, hehe.
« Last Edit: April 17, 2014, 11:48:13 pm by Neganur »
 

Offline EV

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: fi
Re: Frequency response of your DSO
« Reply #79 on: April 17, 2014, 02:50:54 pm »
I did more tests with a Tek 284 Pulse generator (t_rise = 70 ps) and get figures closer to the claimed numbers in the datasheet t_r(10-90) ca. 413 ps.

You have quite big overshoot in this pulse. Have you connected the pulse directly to 50 Ohm input cnnector with BNC cable.

I have done so and the rise time i about 350 ps with TEK 284 generator.
 

Offline EV

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: fi
Re: Frequency response of your DSO
« Reply #80 on: April 17, 2014, 03:05:52 pm »

I'm not sure if I measured the 067-0587-02 cal plugin or my Vince pulser #43 with my old Tek 7104 here (with 7B10 and 7A29), bit sloppy photo but I think I can read 1.4 1.8 div, so 0.35/360 ps  ~972 MHz

It is 067-0587-02 and 1.4* 200 = 280 ps is correct. BW is 350 / 280 = 1.25 GHz. I have also this 067-0587-02.

Edit: It gives faster rise time because it is connected directly to the plugin connector without any cables.
« Last Edit: April 17, 2014, 03:32:17 pm by EV »
 

Offline Neganur

  • Supporter
  • ****
  • Posts: 1114
  • Country: fi
Re: Frequency response of your DSO
« Reply #81 on: April 17, 2014, 08:49:06 pm »
Quote from: EV
It is 067-0587-02 and 1.4* 200 = 280 ps is correct.
Ah yes you're right, for some reason I edited my post and thought "darn, I skipped the 0.4 Div at the bottom", but it's placed to cross the 10% graticule in that intersection on purpose.

The connection from the 284 is through 30 cm Air line directly into the scope's 50 Ohm BNC, I didn't have any good coax handy at the moment. I could try again with an external feed-through load but I have no knowledge of its bandwidth, but it should be good enough.
 

Offline EV

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: fi
Re: Frequency response of your DSO
« Reply #82 on: April 17, 2014, 09:35:25 pm »
The over shoot has been usually under 3 % when tek 284 is connected with BNC cable to 50 ohm input as you can see also in my  Reply #79 on: Today at 12:50:54 AM.

The connection from the 284 is through 30 cm Air line directly into the scope's 50 Ohm BNC, I didn't have any good coax handy at the moment. I could try again with an external feed-through load but I have no knowledge of its bandwidth, but it should be good enough.
 

Offline Neganur

  • Supporter
  • ****
  • Posts: 1114
  • Country: fi
Re: Frequency response of your DSO
« Reply #83 on: April 18, 2014, 12:30:59 am »
Took a little longer than anticipated but here is the frequency response of My MSO-X 3104A.

Signal generator used is an HP8648B, supposedly not calibrated in magnitude and unknown flatness/ripple (the frequency is verified to be accurate with a good counter) 
Each magnitude value is a mean value over at least 40 samples, from 1 MHz to 1500 MHz in steps of 1 MHz.

Also, another shot with the Tek284, with just a GR874 to BNC adapter. I can't seem to lose the significant overshoot so maybe the unit itself is faulty. I have another Tek284 I could try at some point but it got damaged during transport and I don't have the time to fix it right now.
« Last Edit: April 19, 2014, 03:08:41 am by Neganur »
 

Offline EV

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: fi
Re: Frequency response of your DSO
« Reply #84 on: April 18, 2014, 07:54:14 am »
You can try your Tek 284 with your Tek 7104 scope using 7B10 and 7A29 plugins. You should get rising pulse which look same as in my picture. If so your Tek 284 is not faulty.

Also, another shot with the Tek284, with just a GR874 to BNC adapter. I can't seem to lose the significant overshoot so maybe the unit itself is faulty. I have another Tek284 I could try at some point but it got damaged during transport and I don't have the time to fix it right now.
 

Offline eurofox

  • Supporter
  • ****
  • Posts: 835
  • Country: be
    • Music
Re: Frequency response of your DSO
« Reply #85 on: April 18, 2014, 08:28:35 am »
Took a little longer than anticipated but here is the frequency response of My MSO-X 3104A.

Signal generator used is an HP8648B, supposedly not calibrated in magnitude and unknown flatness/ripple (the frequency is verified to be accurate with a good counter) 
Each magnitude value is a mean value over at least 40 samples, from 1 MHz to 1500 MHz in steps of 1 MHz.

Also, another shot with the Tek284, with just a GR874 to BNC adapter. I can't seem to lose the significant overshoot so maybe the unit itself is faulty. I have another Tek284 I could try at some point but it got damaged during transport and I don't have the time to fix it right now.

I did the test with my MSO-X 3104A and HP8648C and got the same results.

For the rise time testing I'm building now a 35ps pulse generator.

eurofox
eurofox
 

Offline Neganur

  • Supporter
  • ****
  • Posts: 1114
  • Country: fi
Re: Frequency response of your DSO
« Reply #86 on: April 19, 2014, 02:37:53 am »
I needed some time to think and also re-read some application notes regarding the results I was getting with the Agilent scope.

My Tek 284 might actually not have an issue after all, but my perception of flat roll-off oscilloscopes does.
In fact, I'm a bit irritated that I wasn't quite aware of the fact that the 1 GHz 3000X isn't Gaussian. This changes a few things for me:

  • The Oscilloscope Vendor needs to specify system bandwidth for me (specific probe+scope), since the inverse RMS formula* for Gaussian scopes does not apply to Flat systems!
    *: BW(system) = SQRT((1/BW_probe)^2 + (1/BW_scope)^2)
  • I am able to measure rise time down to about 500 ps more precise than with a Gaussian scope of same bandwidth.
    Any faster than 500 ps and the Gaussian scope will be more accurate, although both types will have greater than 15% error at that point.
  • I will have rather massive overshoot and ripple when it comes to transient behaviour such as a step response.
  • If my signal rise time is significantly faster than my system rise time (scope+probe), I can estimate the signal rise time by solving for t_r(signal) in
    t_r(measured)=SQRT(t_r(signal)^2+r_t(system)^2)
    But is this valid for the Flat response scope too?
  • The Flat scope has less sampling alias errors, and also requires a less minimum sampling rate than a Gaussian response scope to reconstruct a signal for the same bandwidth
  • I will be able to measure and capture complex waveforms that consist of frequencies with maximum practical components of 714 MHz or less (i.e. up to 700 ps rise times) accurately, i.e. 3% accuracy.
    For a 1 GHz Gaussian scope it is 526 MHz (950 ps rise times) at 3% accuracy.

This is my first flat response scope, so I'm wondering how much of those gotchas I will run into still. It also makes me wish I had not sold the Tek7104 after all, it would be good to have something to compare against.

I'm really worried by the fact that I will see overshoot where a Gaussian scope would not.
Mainly because I am not sure I have fully understood what that will mean for me.

Like, is this signal really looking like this? How do I know it is my circuit and not the scope behaving differently from what I would expect.
What if I am specifically interested in measuring overshoot, I guess I better make sure my signals aren't too fast?
Other than that, I think this means that for in-band signals, this scope will serve me extremely well.

References: Agilent Application Note 1420
« Last Edit: April 19, 2014, 04:01:04 am by Neganur »
 

Offline EV

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: fi
Re: Frequency response of your DSO
« Reply #87 on: April 19, 2014, 06:57:39 am »
Yes, picture 2 in this application note explains the over shoot of your scope. Probably I have seen this note but did not remember it any more.

References: Agilent Application Note 1420
 

Offline rf-design

  • Regular Contributor
  • *
  • Posts: 57
  • Country: de
Re: Frequency response of your DSO
« Reply #88 on: April 19, 2014, 07:46:50 am »
From the users view the choice is very clear. If you test or adjust your circuit or DUT for overshoot or optimal damping near the bandwidth of the scope a gaussian is optimum. Flat scopes overlap a little bit the capabilities for flat response of other instruments. So I think the trend to flat scopes is driven by the small plus which is given to marketing. The inherent speed of the frontend will not change but you get a better number.

For today mid and high-end scopes with digital FIR  filters it would be a software option to switch from gauss to flat. But as it seems in scope market there are a very small number of vendors driving the engineering only to incremental, baby steps. That will change if there are new entries in the market for new vendors which are made open by the baby steps.
 

Offline Carrington

  • Super Contributor
  • ***
  • Posts: 1201
  • Country: es
Re: Frequency response of your DSO
« Reply #89 on: April 19, 2014, 02:04:49 pm »
...
The Flat scope has less sampling alias errors, and also requires a less minimum sampling rate than a Gaussian response scope to reconstruct a signal for the same bandwidth.
...
So I think the trend to flat scopes is driven by the small plus which is given to marketing.
...
But as it seems in scope market there are a very small number of vendors driving the engineering only to incremental, baby steps. That will change if there are new entries in the market for new vendors which are made open by the baby steps.
...
I'm really worried by the fact that I will see overshoot where a Gaussian scope would not.
...
I better make sure my signals aren't too fast.
...
I think exactly the same.
More of the same (Selecting-the-right-bandwidth-scope.wmv):
http://www.element14.com/community/servlet/JiveServlet/downloadBody/35773-102-1-217101/T%26M-Learning%20Center-Oscilloscopes-Video-Agilent.Training_Videos_10.wmv

Inevitably, you will need more and more BW.
My English can be pretty bad, so suggestions are welcome. ;)
Space Weather.
Lightning & Thunderstorms in Real Time.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9902
  • Country: us
  • DavidH
Re: Frequency response of your DSO
« Reply #90 on: April 19, 2014, 05:00:12 pm »
From the users view the choice is very clear. If you test or adjust your circuit or DUT for overshoot or optimal damping near the bandwidth of the scope a gaussian is optimum. Flat scopes overlap a little bit the capabilities for flat response of other instruments. So I think the trend to flat scopes is driven by the small plus which is given to marketing. The inherent speed of the frontend will not change but you get a better number.

For today mid and high-end scopes with digital FIR  filters it would be a software option to switch from gauss to flat. But as it seems in scope market there are a very small number of vendors driving the engineering only to incremental, baby steps. That will change if there are new entries in the market for new vendors which are made open by the baby steps.

I wondered if this would come up in the discussion.  I had an opportunity to evaluate a Tektronix MSO5204 (2 GHz and 5 GS/s) a couple of years ago which *sometimes* displayed this problem.  When the DSP based bandwidth filters were used, fast transition edges displayed overshoot.  When hardware based bandwidth limiting with the same cutoff frequency was used, there was no overshoot.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9902
  • Country: us
  • DavidH
Re: Frequency response of your DSO
« Reply #91 on: April 19, 2014, 05:35:07 pm »
For you, as owner of Owon, the kind of measurment is still important. You was looking for active probe, which would be then  anyway terminated external with 50R.
I kinda assumed an active probe designed for high impedance scopes either put a buffer on the scope side connection or put a 10x compensated attenuator at the scope side which should get you close to a much more bearable 1 pF in parallel with the terminator, that's how I would do it.

All of the active probes I am familiar with have a 50 ohm output and rely on a 50 ohm termination when used with a high impedance oscilloscope input.

There were at one time low input impedance probes intended for use with high impedance inputs which used a transmission line higher than 50 ohms with a similarly high termination resistance included as part of the compensation box but I have not seen one in a long time.  Their advantage over standard passive probes was lower input capacitance.  With a 100 ohm transmission line and termination, a x10 probe would have a 1k input resistance instead of the 500 ohm input resistance of a 50 ohm probe.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9902
  • Country: us
  • DavidH
Re: Frequency response of your DSO
« Reply #92 on: April 19, 2014, 06:52:22 pm »
Quote
The SG503 50kHz to 250 MHz calibration generator did not have a levelling head.

True but cable loss is less of a problem for the SG503 as it only goes up to 250MHz where 1m of RG58 has a loss of about 0.25dB

The SG503 does not have an external leveling head but does include one before its two x10 switchable attenuator stages.

It also specifically states on the front panel that calibration occurs with a specific Tektronix part numbered 50 ohm coaxial patch cable so presumably the cable characteristics are calibrated out.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9902
  • Country: us
  • DavidH
Re: Frequency response of your DSO
« Reply #93 on: April 19, 2014, 07:00:55 pm »
I'm not pessimistic, I'm just cognizant of the fact that they are designed for very low impedance sources at high frequencies and a compensated probe to get the correct frequency response. A cheap scope simply can't be tested as a separate system from it's probe unless you are willing to build specialized circuitry, inline termination is not the same as having an input designed to have 50 Ohm impedance over the device's usable frequency range ... the probe and it's compensation are integral parts of what makes up the frequency response.
  • The Oscilloscope Vendor needs to specify system bandwidth for me (specific probe+scope), since the inverse RMS formula* for Gaussian scopes does not apply to Flat systems!
  • If my signal rise time is significantly faster than my system rise time (scope+probe), I can estimate the signal rise time by solving for t_r(signal) in
    t_r(measured)=SQRT(t_r(signal)^2+r_t(system)^2)
    But is this valid for the Flat response scope too?

It is worth noting that lower frequency oscilloscopes, at least Tektronix ones with low frequency extending into the 100s of MHz, were specified for bandwidth *at the probe tip* with specific exceptions.  This is from an older (paper!) version of "ABC's of Probes" by Tektronix which is available online as part of Linear Technology application note 47:

   Most manufacturers of general-purpose oscilloscopes that include standard accessory probes in the package, promise and deliver the advertised scope bandwidth at the probe tip.
   For example, the Tektronix 2465B 400 MHz Portable Oscilloscope and its standard accessory P6137 Passive Probes deliver 400 MHz (-3db) at the probe tip.
   However, not all high performance scopes can offer this feature, even when used with their recommended passive probes.  For example, the Tektronix 11A32 400 MHz plug-in has a system bandwidth of 300 MHz when used with its recommend P6134 passive probe.  This is simply because even the highest impedance passive probes are limited to about 300 to 350 MHz, while still meeting their other specifications.


This makes calculating the system bandwidth from the root sum of the squares of the specified probe and oscilloscope bandwidth marginal at best and impossible at worst and cheap oscilloscope manufacturers and marketing droids are hardly going to make it easier.

It is worth mentioning that the standard probe tip measurement uses a coaxial probe tip connection, 50 ohm signal source, and 50 ohm termination so the probe tip sees 25 ohms in a coaxial environment which is hardly representative of actual use.

Quote
  • The Flat scope has less sampling alias errors, and also requires a less minimum sampling rate than a Gaussian response scope to reconstruct a signal for the same bandwidth

The only aliasing errors I run across are those caused by nonlinearities in the digitizer itself which mix the incoming signal with the sampling frequency producing products above the Nyquist frequency.  Interleaved ADCs are especially prone to this problem as Agilent discusses in one of their application notes.  Equivalent time sampling if available largely negates this issue simply by supporting a sampling rate so high that aliasing becomes a non-problem even with a Gaussian response.

Quote
This is my first flat response scope, so I'm wondering how much of those gotchas I will run into still. It also makes me wish I had not sold the Tek7104 after all, it would be good to have something to compare against.

I am inclined to think that an old style sampling oscilloscope is an even better choice in this case although they lack a Gaussian response as well.  They have the advantage of predictable frequency response and are my go-to tool for calibrating the pulse generators used to calibrate the transient response on analog oscilloscopes.

My fastest analog non-sampling oscilloscope is a 500 MHz 7904 and my fastest vertical amplifiers for it are only 400 MHz but wow, it sure works well and especially so with lower bandwidth vertical amplifiers.  It is difficult to appreciate how visually clear a high acceleration potential oscilloscope CRT is when used at low bandwidths until you see one.  Just that by itself is enough reason to use a 500 MHz CRT oscilloscope in a low bandwidth application and may explain much of the nostalgia for the 2465 series of oscilloscopes.

Quote
I'm really worried by the fact that I will see overshoot where a Gaussian scope would not.

Like, is this signal really looking like this? How do I know it is my circuit and not the scope behaving differently from what I would expect.
What if I am specifically interested in measuring overshoot, I guess I better make sure my signals aren't too fast?
Other than that, I think this means that for in-band signals, this scope will serve me extremely well.

I started having the same concern after evaluating a couple of low end and high end DSOs which all displayed this problem in one form or another and I suspect oscilloscopes which have "upgradable" bandwidths will suffer from this to an even greater extent.  I would much rather have good transient response which I cannot fix than better anti-aliasing since that has been a solved problem, at least for those who understood it, until recently.

On one hand, their anti-aliasing prevents aliasing which was not a problem before.  On the other hand, the same anti-aliasing messes up the transient response.  I am failing to see an advantage here.

If I get around to it this weekend, I will post the measured frequency response of my 2232 in DSO mode using an SG503 leveled sine wave oscillator.  I have faster DSOs I could test but lack a faster leveled signal source.
 

Offline EV

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: fi
Re: Frequency response of your DSO
« Reply #94 on: April 20, 2014, 07:29:58 am »
There is overshoot also in the pulse from your Vince's MK2 #43 pulse generator.

I'm really worried by the fact that I will see overshoot where a Gaussian scope would not.
Mainly because I am not sure I have fully understood what that will mean for me.
 

Offline Carrington

  • Super Contributor
  • ***
  • Posts: 1201
  • Country: es
Re: Frequency response of your DSO
« Reply #95 on: April 20, 2014, 05:21:30 pm »
I don't know the JW-32 BW, so suppose that is 1GHz with gaussian response (at least it seems), now the JW-43 have a BW of 1GHz with flat response. So under these conditions, it is clear that for a pulse with a rise time of 200ps both must have an error greater than 20%, in fact, for 500ps already is 20% for both (gaussian and flat). Unfortunately seems that every measure has been taken by a different person, and therefore with different types of generators and cables, so the following comparison is not very accurate, since each system will introduce different artifacts...

Anyway, this is the result:



JW-43 (red) appears to be less accurate that JW-32 (green), right? However, if I'm not mistaken, for a tr of 700ps JW-43 (flat) should be more accurate than JW-32 (gaussian).

Please note that the time axis was modified to match it for all, but the amplitude and delay are incorrect.
« Last Edit: April 20, 2014, 05:30:47 pm by Carrington »
My English can be pretty bad, so suggestions are welcome. ;)
Space Weather.
Lightning & Thunderstorms in Real Time.
 

Offline EV

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: fi
Re: Frequency response of your DSO
« Reply #96 on: April 20, 2014, 06:05:54 pm »
JW-32 BW is 1 GHz and it is gaussian. Its risetime 20%-80% is 190 ps as it is 200 ps for JW-43. It is a little faster than for JW-43. I attach Vince's test also for JW-32.

I don't know the JW-32 BW, so suppose that is 1GHz with gaussian response (at least it seems), now the JW-43 have a BW of 1GHz with flat response. So under these conditions, it is clear that for a pulse with a rise time of 200ps both must have an error greater than 20%, in fact, for 500ps already is 20% for both (gaussian and flat). Unfortunately seems that every measure has been taken by a different person, and therefore with different types of generators and cables, so the following comparison is not very accurate, since each system will introduce different artifacts...
« Last Edit: April 20, 2014, 06:24:26 pm by EV »
 

Offline EV

  • Frequent Contributor
  • **
  • Posts: 479
  • Country: fi
Re: Frequency response of your DSO
« Reply #97 on: April 20, 2014, 06:14:02 pm »
All JW tests can be seen here:

https://www.eevblog.com/forum/blog/eevblog-306-jim-williams-pulse-generator/360/

in « Reply #373 on: May 20, 2013, 01:52:27 AM »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 9902
  • Country: us
  • DavidH
Re: Frequency response of your DSO
« Reply #98 on: April 27, 2014, 09:36:47 pm »
Here is the plotted frequency response of my Tektronix 2232 in DSO mode using an SG503 leveled sine wave generator up to 260 MHz with 2 nanoseconds of RG-400 cable and a 50 ohm feedthrough termination that is suppose to be good to at least 1 GHz.

One curve is with the 2232 alone and the other includes the cheap 100 MHz x10 probe that I normally use with a coaxial probe tip to BNC adapter.  The measured bandwidth is 104 MHz for the oscilloscope alone and 106 MHz with the oscilloscope and probe.  The probe measurement was done using a different input level and attenuator setting which may have affected the results slightly.

Up to about 20 MHz, the oscilloscope absolute level measurement agreed with the output indicator on the signal generator to within 1%.

« Last Edit: April 27, 2014, 09:42:55 pm by David Hess »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf