Frequency response of your DSO

[Wim13]

@ The Electrician,

Thanks for this measurement and very nice pictures... very intresting, and explained a lot.......

[JoeyP]

I have no reason to suspect a problem with the Marconi attenuator - certainly at lower frequencies it appears to be spot on. The graph 0dB line is pretty

It dawned on me that there is an easier way to determine whether reflections are the cause. You could simply make the measurement with a different cable length. If the peak moves, it's due to impedance mismatch, if not, it's part of the scope's response.

[DrBob]

One of the ways you can estimate the frequency response (-3 dB Bandwidth = BW) of a DSO, is to measure the rise time or a very fast rise time pulse and use the formula:

BW ~= 350/Tr

where Tr is the measured rise time of the pulse.

So we would expect the rise time of the DS2 series to be:

Scope   BW MHz  ~Rise Time
DS2072     70        5.0ns
DS2102    100       3.5ns
DS2202    200       1.75ns

The rise time of the test pulse should be less than 20% of the Tr of the scope to create less than about 2% error. The problem for many, is where to find an appropriately fast rise pulse to apply to the oscilloscope.

The relatively slow rise time calibrator pulse, which is only intended to calibrate amplitude, is entirely unsuitable for Tr measurements. Fortunately, the Rigol DS Series has a trigger out pulse available on the rear panel that can serve moderately well if you do not have access to something faster.

Lets look at it.

To minimize the cable reflections, the 42" 50 ohm good commercial coaxial cable transferring the pulse to the measuring scope was terminated with a good quality 50 ohm feedthru, like the Rigol PL-50, if the scope does not have a 50 ohm input option - like the Rigol DS Series.

My measurements of the trigger out pulse were done on a 1990s vintage, but still very functional, Tektronix 7904A using a Tr=25ps (0.025ns) type S-4 head in a 7S12 sampling plugin (photo 1). It's accuracy was affirmed by measuring a Tektronix S-2 25 ps pulse generator head correctly. A BNC to SMA adapter and 10dB attenuator was used at the sampling head to reduce the DS2202's trigger voltage.

Since it is an analog scope, we have to do some things the old fashion way. The input signal is set to be 5 divisions high and the time to transverse the center 4 vertical divisions (10% to 90%) is the rise time. The accompanying photo shows this is 1.0ns (photo 2) through the 42" BNC cable. A 50 ohm feedthru is not necessary, as the input impedance of the sampling head is 50 ohms.

An additional measurement using a military quality 9" BNC cable gave 0.9ns for the pulse rise time - but that cable is too short to reach from the rear output to the DS scope's input.

So there we have it - a readily available moderately fast rise time pulse for Rigol DS Series users. It would be much better if it were, say 0.3ns, but we will see there is a way to compensate for it and calculate the true rise time of a scope with a bandwidth into the 200 MHz region.

Measuring a scope's BW using risetime, in my opinion, is not the best way. But scopes measure pulses as well as sine waves, so knowing your scope's risetime is a very good thing.

[EV]

the Rigol DS Series has a trigger out pulse available on the rear panel that can serve moderately well if you do not have access to something faster.

An additional measurement using a military quality 9" BNC cable gave 0.9ns for the pulse rise time - but that cable is too short to reach from the rear output to the DS scope's input.

I have also measured this Rigol DS2202 trigger out rise time. I used Tektronix TDS3032 scope with FET Probe P6243 and 50 ohm feed thru terminator between trigger out BNC connector and probe tip.

Rise time is 848 ps. Look at the attached picture. This TDS3032 is fastest scope I have.

[DrBob]

I have also measured this Rigol DS2202 trigger out rise time. I used Tektronix TDS3032 scope with FET Probe P6243 and 50 ohm feed thru terminator between trigger out BNC connector and probe tip.

Rise time is 848 ps. Look at the attached picture. This TDS3032 is fastest scope I have.

Good technique and nice work with a 500 MHz DSO. That agrees well with my ~900ps measurement at the end of a 9" cable.

[EV]

Good technique and nice work with a 500 MHz DSO.

It is a 300 MHz scope. However measured BW (-3dB) with RG58 cable is about 410 MHz and with FET probe P6243 BW is over 500 MHz. Look at earlier post:
https://www.eevblog.com/forum/testgear/frequency-response-of-your-dso/msg192342/#msg192342

[tinhead]

One of the ways you can estimate the frequency response (-3 dB Bandwidth = BW) of a DSO, is to measure the rise time or a very fast rise time pulse and use the formula:

BW ~= 350/Tr

where Tr is the measured rise time of the pulse.

but only on scopes with gaussian response, on flat-type response DSOs we need to use 0.4 to 0.5 instead of 0.35.
And even 0.35 is only truth when the response is real gaussian,  today you can get an (not only chinese) DSO with
mixed response, where the first part is coming from e.g. input stage and the second part of the response
(e.g. butterworth fir) is coming from the digital part of DSO.

So yes, rise time can be used to measure bandwidth but only WHEN you exactly know the frequency response of your DSO,
which is what i trying to achieve in this thread.

[DrBob]

but only on scopes with gaussian response, on flat-type response DSOs we need to use 0.4 to 0.5 instead of 0.35.
And even 0.35 is only truth when the response is real gaussian
So yes, rise time can be used to measure bandwidth but only WHEN you exactly know the frequency response of your DSO,
which is what i trying to achieve in this thread.

Yes - that formula only applies to a gaussian response front end. I will soon show how well it applies to the DS2202.

I have just one more follow-up post on making rise time measurements, then I will get to real meat on the bone and the main thrust of the tread - measuring BW:

1. The old fashion way,
2. A newer way,
3. The newest automated way

[dougg]

the Rigol DS Series has a trigger out pulse available on the rear panel that can serve moderately well if you do not have access to something faster.

An additional measurement using a military quality 9" BNC cable gave 0.9ns for the pulse rise time - but that cable is too short to reach from the rear output to the DS scope's input.

I have also measured this Rigol DS2202 trigger out rise time. I used Tektronix TDS3032 scope with FET Probe P6243 and 50 ohm feed thru terminator between trigger out BNC connector and probe tip.

Rise time is 848 ps. Look at the attached picture. This TDS3032 is fastest scope I have.

When I measured the same thing (DS2202's trig out with another Rigol (DSO1024A)) the waiting to be triggered voltage level was around 0 volts and the triggered voltage level was around 3.3 volts. The DS2000 specs don't say what the trig out can drive so I did not try a 50 ohm termination.

[eurofox]

Hi,

Here is the graph from my TDS784D, basicaly 1GHz scope but trigger til 1,9 Ghz 

eurofox

[Carrington]

What does DS2072_1st and DS2072_2nd mean?


By the way, the Tekway DTS1202B's input stage is very similar to that used in the DS2072. [J1=U6, Q1=R2W (HW1.0) and Q3=Q2=1B].



Unfortunately still undetermined some values and connections for the DS2072's input stage.
So, if someone already dismantled their DS2072A and want to collaborate these are some of the connections to be determined (Red circles).



And also would be interesting to know C and R values (without marking code), but perhaps this, is asking too much, because must be unsoldered.

So yes, rise time can be used to measure bandwidth but only WHEN you exactly know the frequency response of your DSO,
which is what i trying to achieve in this thread.

Yes - that formula only applies to a gaussian response front end. I will soon show how well it applies to the DS2202.

How we classify the response for the DS2000? The theory say this:
https://www.eevblog.com/forum/testgear/sniffing-the-rigol%27s-internal-i2c-bus/?action=dlattach;attach=73448

[tinhead]

Here is the graph from my TDS784D, basicaly 1GHz scope but trigger til 1,9 Ghz 

this is crazy, i always loved TDS7xx series (even if the f* loud)

[tinhead]

What does DS2072_1st and DS2072_2nd mean?

that measurment results from two persons.

By the way, the Tekway DTS1202B's input stage is very similar to that used in the DS2072. [J1=U6, Q1=R2W (HW1.0) and Q3=Q2=1B].

it is actually that way:

DS2000 does have similar frontend to Tekway
Tekway does have similar frontend to DS1000E
DS1000E does have similar frontend to DS1000CA
DS1000CA does have similar frontend to DS1000B (i think)
DS1000B and Instek (don't remember the name) they have similar frontend ^^

... and between all that point you can add the other manufacturers, as they copied it as well,
where Siglent added e.g. 50R termination without clamp diodes but Rigol added some, etc.

How we classify the response for the DS2000? The theory say this:

kind of mixed response, where the first part is gaussian and then fallsing rapidly down as known from flat response (especially when 300MHz enabled). For the bandwidth is however the first part responsible (as it goes far below -3dB), the rest is let's call it a better low pass

[Carrington]

that measurment results from two persons.

OK.

it is actually that way:

DS2000 does have similar frontend to Tekway
Tekway does have similar frontend to DS1000E
DS1000E does have similar frontend to DS1000CA
DS1000CA does have similar frontend to DS1000B (i think)
DS1000B and Instek (don't remember the name) they have similar frontend ^^

... and between all that point you can add the other manufacturers, as they copied it as well,
where Siglent added e.g. 50R termination without clamp diodes but Rigol added some, etc.

LOL...

kind of mixed response, where the first part is gaussian and then fallsing rapidly down as known from flat response (especially when 300MHz enabled). For the bandwidth is however the first part responsible (as it goes far below -3dB), the rest is let's call it a better low pass

So, better measure the BW using the traditional method (-3dB).


Thanks. 

[eurofox]

Here is the graph from my TDS784D, basicaly 1GHz scope but trigger til 1,9 Ghz 

this is crazy, i always loved TDS7xx series (even if the f* loud)

Yes it is quite impressive and I would like to add that BNC cable and connectors have  normaly a bandwith limited to 1Ghz and I'm sure that many DB's are lost this way in the test above 1Ghz.

[Altemir]

Some mesuarements from our lab... Sorry for my bad english and quality of Hi-freq mesuarements. I have used Hoden HB-B100 BNC-BNC cable, R&S SMB100A (2.2GHz), Tektronix 011-0049-02 Feed-through Terminator (only for 1M inputs scopes and "50Ext", for other - internal 50 Ohm load). Level of signal was 1Vrms at all experiments. Memory depth for MSO4024 was AUTO.

[tinhead]

Some mesuarements from our lab...

thank you very much!

[DL8RI]

Here some values of an erm... "elderly" DSO, the Philips PM3320A, Specified 200MHz BW and 250MS/s.
My source is a Rohde & Schwarz SMS, the VSWR was done with an R&S ESVP with an HP 8502A Opt H26 Reflection Test-Set.

I've done the measurement until the scope showed no reasonable waveform anymore, so it's far beyond the 200MHz. The Input was set to 50 Ohm, no additional Attenuator was used, input-Power 0dBm.

Next week (or somewhat later) my Hantek hand-held Scope should arrive, I will post the results. They may be more interesting.

[babysitter]

Although not yet the requested response measurement, I used a antenna analyzer FA-VA3 to check for the matching of diverse scopes with either a picoscope feedtrhough terminator (my own) or a pomona 4119 (workplace).

This ZIP contains the CSV tables with the real/imaginary part of the impedance seen by the analyzer using the pomona terminator. Similar data will follow for my home scope(s).

Unfortunately, I set the cable compensation to UT-141 semirigid cable by accident but hey, I think you can live with it !

The filenames tell about the scope settings, as the matching usually changes with the V/DIV Setting, not every setting but changes in ranges over several settings. The settings are selected to represent different cases.

Expect me to post data for the picoscope terminator for the above scopes and my own VC 630-2, TEK 453 and DST1062B/200 MHz hacked later, and RF response as soon as I fixed my Marconi 2018A or decided to run it unrepaired (lost FM and PM ability, maybe some latch/multiplexer issue...), otherwise the FA-VA3 will be used as source.

EDIT: Baby 2 has shown up a bit early, further work postponed !

[Co6aka]

...BNC cable and connectors have  normaly a bandwith limited to 1Ghz...

That depends upon materials and construction. I rescued some rather strange and wonderful Huber-Suhner cables from a scrap bin; the cable is marked "6GHz" and the connectors are silver plated with gold plated pins. I tested them all at 3GHz when I got them, but don't recall the loss per foot. Regardless, for accurate measurements you must sweep your cables and make up an attenuation profile, and then offset your device-test measurements accordingly.

When I measured the frequency response of my DS4014-with-hacked-FW, I first measured the generator-plus-cables into the RF power meter, and then adjusted the generator to produce as close to 0dbm as I could get. (The smallest amplitude step of the Agilent E4438C is 0.2dbm so that was the limiting factor.) The E4438C had been calibrated in August, and the MiniCircuits power meter had just been calibrated a couple of weeks before, so I "trusted" it most, but both were in close agreement. Anyway, the cable and connector losses were quite low, even at 1GHz, which was the upper limit of my DS4014 tests.

If you have access to some RF test EQ then you can make up loss profiles for all of your cables and adaptors -- label 'em first, otherwise... 

[Altemir]

New corrected results,  that I have put in my topic.

[rfbroadband]

R&S RTO 1044

[Neganur]

I'll see if I can do a frequency response measurement on the MSO-X3104 later. Unfortunately, I know that the output of my HP8648B signal generator isn't constant over the whole frequency range, and I don't know its flatness either. The measurement report claims that the scope is at - 1.00 dB at 1 GHz with 0.29 dB uncertainty.

I was testing the Vince/Williams pulser on it and was a bit surprised that I get nowhere near the bandwidth I should be getting (0.23/330 ps = 700 MHz). According to the measurement Vince did, the pulser has approximately 200 ps rise time.

[jpb]

I was testing the Vince/Williams pulser on it and was a bit surprised that I get nowhere near the bandwidth I should be getting (0.23/330 ps = 700 MHz). According to the measurement Vince did, the pulser has approximately 200 ps rise time.

The 0.23 constant is based on a single pole filter and 20-80% rise time (more common is the 0.35 relating to single pole and 10-90% rise time). According to various app notes from Tektronix and Agilent for higher frequency scopes with maximally flat pass bands the constant can rise from 0.35 to 0.4 or even 0.5 so this could account for some of it. If you multiply 0.23 by (0.5/0.35) you get around 0.32 and the bandwidth rises to much nearer 1GHz.
(I think that the 0.23 should actually be 0.22 for a single pole - ln(0.8/0.2)/2.pi = ln(4)/2.pi=0.22.

The Agilent spec sheet is for a 10-90% rise time of 450psecs giving the product as 0.45.

It might make more sense to work with 10-90% rise times and compare that with Agilent's spec. directly.

[EV]

I was testing the Vince/Williams pulser on it and was a bit surprised that I get nowhere near the bandwidth I should be getting (0.23/330 ps = 700 MHz). According to the measurement Vince did, the pulser has approximately 200 ps rise time.

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.