SDS2074X actual scope bandwidth

[vt100]

Curious if anyone out there has ever measured the SDS2074X's actual bandwidth.

Today, I took a SDG6022X and fed a 1Vpp signal starting at 1KHz into a SDS2074X with the supplied BNC cable and a 50 ohm feed through terminator. The scope measured 1.03Vpp (3% error?) @ 1kHz.

3db drop at 1.03Vpp would be .728Vpp target value.

I increased the frequency on the SDG continuously until I got to 200MHz. At 200MHz I dropped below the target value.

Backed off... somewhere in the 198-199MHz range seems to be the upper 3db limit.


Is the scope's rated bandwidth actually limited by the 70MHz probes they include with it, and the scope is actually a 200MHz scope? Or, did I do something obviously wrong (I am still a plebe learning about these things) when doing this measurement?

Both scope and SDG are factory firmware, unmodified by various hacks published in the forums.

See video at:

Thx,
wyse60

[Performa01]

Traditionally, the specified bandwidth of Siglent X-series scopes appears to be much closer to -1dB rather than -3dB amplitude drop – at least for the majority of practical vertical gain settings. See the bandwidth measurement data for the 300MHz model SDS2304X for reference:


SDS2304X_bandwidth

Up to 200mV/div vertical gain, the -1dB amplitude drop is close to 300MHz.
The -3dB drop is at an only slightly higher frequency, simply because we’re approaching the physical bandwidth limit of the frontend with its higher order roll-off here.

In contrast, the artificial bandwidth limit for the lower bandwidth models has just a first order roll-off, so it might well be that the -3dB bandwidth is up to twice as much as specified.

In your case it seems to be even more, but maybe Siglent factory have not bothered fitting different filter capacitors in the frontend than for the 100MHz model.

[David Hess]

Oscilloscope bandwidth is usually specified at the probe tip with the specified probe.

Usually a leveled signal generator is used to make these measurements but the SDG6022X is specified to be +/-0.3dB which is unusually good.  Has anybody verified this performance?

This specification is only given for 0.5 volts peak-to-peak so I would be cautious at other output levels and use an external RF attenuator for lower outputs.

Your results look reasonable to me.

[Performa01]

All modern oscilloscopes that I’m aware of have separate bandwidth specifications for the scope itself and the probes. This makes perfect sense, since the gain flatness over frequency will always be significantly worse with a probe when compared to a direct coax connection. There are quite a few tasks where we don’t need a probe and then we’re certainly interested in the frequency response of the scope alone.

Of course a leveled signal generator has been used for the measurement; the specification for the amplitude flatness is +/- 0.5dB, but that sort of A-brand (R&S) gear usually exceeds its specifications for most applications.

My measurement data for the SDS2304(X) are rather old, from about 3 years ago, some 18 months before the SDG6000X has been introduced.

Since you mentioned the SDG6022X and its amplitude flatness specification, you’ve got me curious. I don’t have my old Boonton HF power meter (with 200kHz to 18GHz sensor) here in this lab, but I thought I’ll just take a chance and use an appropriate DSO with fast FFT processing to verify the frequency response of an SDG6052X.


SDS5104X_SDG6052X_FFT_FRG3b

The difference between minimum and maximum of the frequency response is just 0.7dB, and this includes the frequency response error of the SDG6052X, adaptors, cable, ext. attenuator and DSO.

The detailed test setup is as follows:

1.   SDG6052X sweeping from 1MHz to 500MHz with 500s sweep time and 0dBm output level into 50 ohms. Note that the output level is thus 632.45mVpp.
2.   BNC to SMA adaptor.
3.   60cm SMA-SMA cable manufactured by “C.E. Precision Assi, Inc.”, specified up to 18GHz (but only 2.66mm outer diameter, so we can expect some noticeable loss at 600MHz).
4.   Narda 10dB micro-Pad, 18GHz, 2W in order to ensure perfect input matching and avoid ripple in the frequency response.
5.   SMA to BNC adaptor.
6.   1GHz DSO with input switched to 50 ohms, running a 16kpts FFT up to 2.5GHz in “Max. Hold” mode (measurement only shown up to 500MHz). Reference level is -8dBm and vertical scale is 0.5dB/div. Flattop window.

The DSO bandwidth flatness is specified as follows:

10kHz  to 100MHz:   +/-0.5dB
100MHz to 333MHz: +/-0.8dB
333MHz to 666MHz: +1.0dB, -1.2dB

This clearly shows that the DSO is supposed to contribute the majority of the total inaccuracies in this test.

I think this is not bad at all. The generator is set to 0dBm and because of the inline attenuator (“micro-pad”) we expect a level of -10dBm at the scope input. The DSO measurement is almost spot-on (+0.175dB error, equivalent +2%) at very low frequencies.

With this result, I’m rather confident that the SDG6052X actually meets its specifications, but I shall bring my power meter sometimes and take a few measurements to be absolutely sure.

[David Hess]

All modern oscilloscopes that I’m aware of have separate bandwidth specifications for the scope itself and the probes.

Tektronix specifies bandwidth at the probe tip (1) and *not* oscilloscope bandwidth except in special cases.  If you use one of their 200 MHz oscilloscopes with the specified probe, the bandwidth at the probe tip (in the coaxial feedthrough test fixture) will be the oscilloscope bandwidth specification.

This makes perfect sense, since the gain flatness over frequency will always be significantly worse with a probe when compared to a direct coax connection.

For high impedance inputs with external or internal 50 ohm terminations, this breaks down at higher frequencies because the probe compensation includes corrections to the oscilloscopes input response which is where all of those old A and B oscilloscope probe versions came from.  Now they are more likely to use disparate part numbers.  Most users would never notice the difference but the oscilloscope is unlikely to meet its transient response specifications with the wrong probe unless the probe's high frequency compensation is adjusted.  (2) At least some of the various smart probes handle this through digital calibration.

Since you mentioned the SDG6022X and its amplitude flatness specification, you’ve got me curious. I don’t have my old Boonton HF power meter (with 200kHz to 18GHz sensor) here in this lab, but I thought I’ll just take a chance and use an appropriate DSO with fast FFT processing to verify the frequency response of an SDG6052X.

With this result, I’m rather confident that the SDG6052X actually meets its specifications, but I shall bring my power meter sometimes and take a few measurements to be absolutely sure.

The Tektronix SG503 leveled oscillator is suppose to be within 3% up to 250 MHz or about +/-0.25dB which is why I wondered.  The unleveled sources I have seen up to now were at least twice as bad.  Testing showed that SG503s are usually half as bad as their specification.

A couple years ago on the Tektronix list we had a discussion about how to verify the performance of a leveled signal sources and the easiest solution I recommended was to use a sampling oscilloscope because its frequency response can be calculated from its strobe width which can be directly measured by finding the first null in its frequency response.  I figure they were more common than power meters or sampling RMS voltmeters of sufficient performance and easier to calibrate than a power meter.

(1) PDF page 8.  I would have quoted it but I am too lazy to crack the PDF's DRM at the moment.  I have a collection of older versions of this document going back to 1990 which are more informative as the trend over time has been for technical documentation to be less informative.

(2) PDF page 8 again.

[Performa01]

Tektronix specifies bandwidth at the probe tip (1) and *not* oscilloscope bandwidth except in special cases.

If you use one of their 200 MHz oscilloscopes with the specified probe, the bandwidth at the probe tip (in the coaxial feedthrough test fixture) will be the oscilloscope bandwidth specification.

Sorry, but I read it the other way round. This well known document clearly states the obvious:

In general, the bandwidth and rise time interactions
between probes and oscilloscopes are complex. Because
of this complexity, most oscilloscope manufacturers
specify oscilloscope bandwidth and rise time to the
probe tip for specific probe models designed for use with
specific oscilloscopes.

Specific probe models, designed for specific scopes, will not have an isolated bandwidth specification, but a system specification, i.e. their bandwidth is always specified in conjunction with the scope they are designed for. This is an important difference to your interpretation. The fact that specific probes are not specified without the corresponding scope does not mean that the scope cannot have a bandwidth specification on its own without probe.

Consequently, I looked up the datasheets for TBS2000 and DPO/MSO2000 on www.tek.com and they specify the bandwidth of their scopes just like any other manufacturer does. There is absolutely no hint that this specification includes or requires any probe (if this were actually the case, it has to be stated in a datasheet!); furthermore, the datasheet doesn’t even specify the supplied standard probes.

For high impedance inputs with external or internal 50 ohm terminations, this breaks down at higher frequencies because the probe compensation includes corrections to the oscilloscopes input response which is where all of those old A and B oscilloscope probe versions came from.  Now they are more likely to use disparate part numbers.  Most users would never notice the difference but the oscilloscope is unlikely to meet its transient response specifications with the wrong probe unless the probe's high frequency compensation is adjusted.  (2) At least some of the various smart probes handle this through digital calibration.

Nothing breaks down at higher frequencies. I’ve yet to come across a scope that wouldn’t meet its specifications when using a direct connection – but then again, I have not tested this on any Tek scope. Anyway, It would be really sad if Tek scopes would not meet their transient response specifications with direct coax connection but required some special probe for that.

We are talking about the SDS2000X here, which is a general purpose DSO/MSO where I have tested the top model with 300MHz specified bandwidth (without probe). Of course we expect such a scope to have its 3dB bandwidth some 5-10% higher than specified when measured directly. In fact I was happy to see an actual 3dB bandwidth between 10.67% and 17.33% higher than specified, depending on the vertical gain setting (up to 100mV/div).

With the supplied SP2030A probes, the 3dB bandwidth expands by more than 50%, so there is indeed some massive frequency response correction going on. Many other probes from various manufacturers have been tested and none of them provided a comparable result, not even a fancy Keysight N2843A (500MHz) probe borrowed from an MSO-X3000A. An old Tektronix P3010 (100MHz) probe that belongs to some TPS2000 (one of these isolated channel scopes with no sample memory) performed not too bad overall, slightly lower bandwidth (still >400MHz -3dB), but better signal fidelity than the Keysight on the SDS2304X. Of course I’m convinced that these (and all other tested) probes work beautifully together with their associated scopes, just haven’t bothered to measure them all.

So I have been the one who has repeatedly warned people to not swap their supplied probes for “better” ones, because the probes need to be matched with the input characteristic of the specific scope. It is not only bandwidth and rise time, but also signal fidelity which is usually close to an optimum with the original probes and might degrade significantly with a random one, even if it has a much higher bandwidth specification.

The probes with separate HF-adjust should be more universal in theory, but my tests with one of these were a rather mixed bag and not really convincing overall.

Finally it should be mentioned that the practical value of a passive high impedance probe at 450MHz (this is the bandwidth I get with my SDS2304X and the supplied SP2030A probes) is rather doubtful. It works well when probing a low impedance (25 ohms) source, but if we have such a low source impedance we could as well use an L-attenuator to match a coax cable or build a Z0-probe. With an input capacitance around 10pF or even higher, there are not many nodes carrying signals at several hundred MHz that could be connected to a high impedance probe without throwing the circuit completely off track.

For the direct connection bandwidth test, I’ve mentioned the vertical gain settings up to 100mV/div so far (actually it applies up to 148mV/div). Above that, the first attenuator kicks in; this is also why the frequency response changes significantly at 200mV/div and 500mV/div. This attenuator is also designed to correct the frequency response of the scope (of course, it has to), and we see even higher 3dB bandwidth limits.

The Tektronix SG503 leveled oscillator is suppose to be within 3% up to 250 MHz or about +/-0.25dB which is why I wondered.  The unleveled sources I have seen up to now were at least twice as bad.  Testing showed that SG503s are usually half as bad as their specification.

3% up to 250MHz sounds really good, but 250MHz just doesn’t get you very far nowadays, where even the cheapest entry level scopes might start at 200MHz (Siglent SDS1202X-E) with their 6dB bandwidth close to 300MHz.

[tautech]

Tektronix specifies bandwidth at the probe tip (1) and *not* oscilloscope bandwidth except in special cases.

If you use one of their 200 MHz oscilloscopes with the specified probe, the bandwidth at the probe tip (in the coaxial feedthrough test fixture) will be the oscilloscope bandwidth specification.

Sorry, but I read it the other way round. This well known document clearly states the obvious:

In general, the bandwidth and rise time interactions
between probes and oscilloscopes are complex. Because
of this complexity, most oscilloscope manufacturers
specify oscilloscope bandwidth and rise time to the
probe tip for specific probe models designed for use with
specific oscilloscopes.

Specific probe models, designed for specific scopes, will not have an isolated bandwidth specification, but a system specification, i.e. their bandwidth is always specified in conjunction with the scope they are designed for. This is an important difference to your interpretation. The fact that specific probes are not specified without the corresponding scope does not mean that the scope cannot have a bandwidth specification on its own without probe.

This ^.
Furthermore, the section; Why So Many Probes? on P14/60 paints a totally different picture to what David was trying to portray.

ABCs of Probes
The wide selection of oscilloscope models and capabilities is
one of the fundamental reasons for the number of available
probes. Different oscilloscopes require different probes. A
400 MHz oscilloscope requires probes that will support that
400 MHz bandwidth.
However, those same probes would be overkill, both in
capability and cost, for a 100 MHz oscilloscope. Thus, a
different set of probes designed to support a 100 MHz
bandwidth is needed.
As a general rule, probes should be selected to match the
oscilloscope’s bandwidth whenever possible. Failing that, the
selection should be in favor of exceeding the oscilloscope’s
bandwidth.[/b] ..............

[Performa01]

Hehe, there must have been different authors contributing to this document, because on page 29 we can learn quite the contrary:

Bandwidth and Rise Time
It’s important to realize that the oscilloscope and its probes act
together as a measurement system. Thus, the oscilloscope
used should have bandwidth and rise time specifications
that equal or exceed those of the probe used and that are
adequate for the signals to be examined.

Anyway, in my experience the bandwidth specification of a probe is not comparable to anything else. On a passive network or any active device like amplifiers or mixers one can measure the frequency respone and will get a fairly repeatable -3dB bandwidth as a result. For probes, one can never tell in advance whether its a good match for any specific scope input or not (except when it's specified of course) and the actual bandwidth is also absolutely unpredictable. In my test back then, the lowly Tektronix P3010 probe rated to 100MHz performed quite similar to the rest of the bunch, including the 500MHz Keysight N2843A. The real important difference is not the 3dB bandwidth, but the gain flatness and signal fidelity that goes with it (and there the 100MHz Tek was an even better match). A 1dB peak or dip somewhere within the scope bandwidth already causes a remarkable visual pulse distortion on the scope screen.

Of course also the impedance at the tip may vary, i.e. the tip capacitance of a 500MHz probe might be slightly lower than one rated for only 100MHz.