Two Tone Test with Scope and SA

[_Wim_]

AFAIK the two-tone test is primarily a way to see how good the mixer is in a spectrum analyser. A DSO doesn't have a mixer so it is not a surprise this effect doesn't occur.

The scope's ADC also has non neglectable IMD distortion, but probably only significant at higher frequencies than the 10kHz typically tested here.

For example the datasheet of TI ADC08D500 (8-bit 1GSPS) states an IMD of -50db at 121Mhz. 

[jjoonathan]

I think the RTO1024 is an 8-bit scope

Yep, it's 8 bit! The "16 bit HD Mode" marketing wank is one of the more egregious examples out there. It was good sport to watch Lecroy and Keysight take shots at R&S for it. Of course, in the scheme of things, it's not *that* big of an issue. Obviously it didn't stop me from buying one (for unrelated reasons).

Still, there actually is a kernel of truth at the center of the marketing turd: R&S did a better than usual job at averaging. Many implementations are stuck at a pre-Signal-Processing-101 level. They use boxcar filters before downsampling and don't calibrate their PGA settings to ensure good dithering across the instrument bandwidth. R&S brought things up to Signal-Processing-101 spec. That's nice. It plays especially well with the fully digital trigger, which still works great when you've got de-embedding and filtering on. The third timebase knob (record length) also shines in these situations.

I was hoping to provoke a response from someone with an Infiniium S. There was a big split in ADC philosophy this generation, with Keysight taking the "farm out to a bunch of slow CMOS cores, individually calibrated" strategy and R&S (Lecroy too, I think) taking the "one core and step on the gas" strategy. I wanted to see what that looked like in the low frequency "average to the moon" regime. There's still time, I suppose

[egonotto]

Hello,

jjoonathan wrote: "I was hoping to provoke a response from someone with an Infiniium S"

perhaps you can ask Shahriar. He can make test with Infiniium S and Infiniium MXR and Tek MSO 5 and MSO 6.

Best regards
egonotto

[mawyatt]

In my case I was interested in the linearity performance of the scope front end and ADC. Also looking at the AWG output with a SA provides a view into how good the AWG signal is. The 2 tone tests was a simple means to show such and why I asked. If you are interested in the limits of the scopes FFT capability, then other tests like you've suggested might be considered.

Honestly I'm impressed by all the scopes behavior, even the low cost "entry' level devices seem to be performing better than I expected.

AFAIK the two-tone test is primarily a way to see how good the mixer is in a spectrum analyser. A DSO doesn't have a mixer so it is not a surprise this effect doesn't occur.

Don't know about others, but we often used the 2 Tone Test as a verification of the custom chips we developed (for example to compare against simulations) and the linear systems involved (LNA, active filters, attenuators, VGAs and so on) well before any intended non-linear functions (mixers) were encountered, we used the 2 Tone Test on these (mixers) as well. Also used the 2 Tone Test to characterize the high resolution DACs and ADCs we were involved with. In fact this 2 Tone Test was one of the main tests involved in a new type ADC developed (non-uniform sampling ADC) that utilized simultaneous time and amplitude quantization. This ADC has the unique property on not requiring a pre anti-aliasing filter as the input signal itself imposes the time & amplitude quantization that allows post conversion anti-aliasing filtering. Someday these may end up in our test equipment 

In the area I recently retired from (chip design), the 2 tone test was utilized for just about any "analogish" circuit or system, linear or non-linear, to help characterize performance. This test is also a very valuable tool for calculating the classic IIP3 and OIP3 points without having to sweep the input power levels assuming a well behaved system.

Anyway, don't think the 2 Tone Test should only be relegated to systems involving mixers (like SA), but can be highly beneficial in all sorts of "linear" and "non-linear" systems, including the data acquisition systems we call Digital OscilloScopes

Best,

[jjoonathan]

> perhaps you can ask Shahriar. He can make test with Infiniium S and Infiniium MXR and Tek MSO 5 and MSO 6.

He's already gearing up for a shootout. That's what the preselector was for. Can't wait 

[David Hess]

AFAIK the two-tone test is primarily a way to see how good the mixer is in a spectrum analyser. A DSO doesn't have a mixer so it is not a surprise this effect doesn't occur.

The sampling process before conversion *is* mixing, and the same types of circuits are used.  (1) The difference is that an impulse replaces the local oscillator signal so while a mixer has fine resolution in frequency, a sampler has fine resolution in time.

Aliasing in a sampling analog-to-digital converter is identical to downconversion, and subsampling receivers take advantage of this.  The impulse being used in place of the continuous local oscillator signal contains harmonics which mix with higher frequencies to allow this.

(1) Microwave mixers are a good source for microwave samplers when the transformers are removed from the RF and IF ports.

[SilverSolder]

Just for fun, here is what the tone test looked like in 1975!

I fired up an ancient HP 3571A Spectrum Analyzer -  a 100% analog spectrum analyzer full of gold, crystal filters, and engineering love. 

The two tones are generated by HP3325A x2 summed into a 50 ohm load.

The analog output from the 3571A was tracked on the scope in roll mode for display.

It seems to me the performance is not embarrassing at all, considering this thing is 45 years old.  The noise floor is a bit high, but the 3325A signal generators are not exactly low noise...   The 3571A is capable of -130dB, perhaps a little better, performance on a clean signal.  The ancients knew how to build stuff!  (It probably didn't cost $450)

[mawyatt]

Just for fun, here is what the tone test looked like in 1975!

I fired up an ancient HP 3571A Spectrum Analyzer -  a 100% analog spectrum analyzer full of gold, crystal filters, and engineering love. 

The two tones are generated by HP3325A x2 summed into a 50 ohm load.

The analog output from the 3571A was tracked on the scope in roll mode for display.

It seems to me the performance is not embarrassing at all, considering this thing is 45 years old.  The noise floor is a bit high, but the 3325A signal generators are not exactly low noise...   The 3571A is capable of -130dB, perhaps a little better, performance on a clean signal.  The ancients knew how to build stuff!  (It probably didn't cost $450)


(Attachment Link)

Those old instruments can still hold their own

Brings back fond memories, along with the superb Tek analog scopes of that era

Best,

[rf-loop]

In my case I was interested in the linearity performance of the scope front end and ADC. Also looking at the AWG output with a SA provides a view into how good the AWG signal is. The 2 tone tests was a simple means to show such and why I asked. If you are interested in the limits of the scopes FFT capability, then other tests like you've suggested might be considered.

Honestly I'm impressed by all the scopes behavior, even the low cost "entry' level devices seem to be performing better than I expected.

AFAIK the two-tone test is primarily a way to see how good the mixer is in a spectrum analyser. A DSO doesn't have a mixer so it is not a surprise this effect doesn't occur.

They are not named as Mixer but other way, how many "mixer" you find in one oscilloscope... think about it. Do you know what all circuits can produce mix. Sorry... I feel it was necessary..  but not so seriously.

Is it good to know even basic fundamentals...

Basic: When two frequency signals,  f1 and  f2,  are input to any nonlinear device.... etc...  do you know any other nonlinear devices than these what are commercially named as Mixer. Ok lets start from bottom, do you know diode... how about transistor etc etc.

[tggzzz]

Is it good to know even basic fundamentals...

Basic: When two frequency signals,  f1 and  f2,  are input to any nonlinear device.... etc...  do you know any other nonlinear devices than these what are commercially named as Mixer. Ok lets start from bottom, do you know diode... how about transistor etc etc.

Precisely.

Don't forget to add "capacitor" to that list.

[jjoonathan]

I've seen a rusty cable mix surprisingly well even at low power.

[2N3055]

I've seen a rusty cable mix surprisingly well even at low power.

Many chemical compounds are semiconductors..

[SilverSolder]

10 kHz and 10.1 kHz 0 dBm from SDG6052X with digital WaveCombine.

[...]   Picoscope 4262 (16bit)  [...]

That looks like pretty impressive performance!

[2N3055]

10 kHz and 10.1 kHz 0 dBm from SDG6052X with digital WaveCombine.

[...]   Picoscope 4262 (16bit)  [...]

That looks like pretty impressive performance!

That is a real 16bit scope. Fantastic little beast.
Even 8 bit one (3406D) is excellent.

[jemangedeslolos]

I made some quick tests on my MSO7014. source is a Siglent SDG2082X  :

[SilverSolder]

Just for fun, here is what the tone test looked like in 1975!

I fired up an ancient HP 3571A Spectrum Analyzer -  a 100% analog spectrum analyzer full of gold, crystal filters, and engineering love. 

The two tones are generated by HP3325A x2 summed into a 50 ohm load.

The analog output from the 3571A was tracked on the scope in roll mode for display.

It seems to me the performance is not embarrassing at all, considering this thing is 45 years old.  The noise floor is a bit high, but the 3325A signal generators are not exactly low noise...   The 3571A is capable of -130dB, perhaps a little better, performance on a clean signal.  The ancients knew how to build stuff!  (It probably didn't cost $450)


(Attachment Link)

Those old instruments can still hold their own

Brings back fond memories, along with the superb Tek analog scopes of that era

Best,

Sometimes I just buy weird stuff on eBay because it looks interesting and doesn't cost much.  It gives the same feeling as starting an old muscle car and taking it for a drive - giving a hard time to the occasional new car! 

I have managed to avoid going down the Tek scope rabbit hole, even though I can see the admirable engineering that is going on with them.   There is only so much space in the man cave!   

[tautech]

I made some quick tests on my MSO7014. source is a Siglent SDG2082X  :

In Dot mode ? 

[jemangedeslolos]

I made some quick tests on my MSO7014. source is a Siglent SDG2082X  :

In Dot mode ? 

No, vector mode on channel 1 but color grading option enabled in FFT settings.
I left the memory depth management in auto.
I hid the channel on most on my screenshot but I can redo the measurements without hiding the channels and without activating color grading option.
In the end, the results are the same.
We miss max hold on FFT settings in Rigol scope so I can't replicate your screenshots.

[tautech]

I made some quick tests on my MSO7014. source is a Siglent SDG2082X  :

In Dot mode ? 

No, vector mode on channel 1 but color grading option enabled in FFT settings.

Then why is there just dots below the 2 highest markers ? 

We miss max hold on FFT settings in Rigol scope so I can't replicate your screenshots.

Can you engage any averaging ?
Enough and it cleans up FFT much like Peak hold albeit at a lower value.

[rf-loop]

This whole two tone test can be just fun playing or it can be serious tests for analyze equipment.
Separate peoples do separate tests with different settings and ... all different. So results are some times just nonsense and or not at all comparable.
First need do some definitions and this also depends what are we testing. if there is under testing oscilloscope front end it is really not 10kHz thing.  After then also signal level is more than important. After then signal source is very important. Example two tones need really be well isolated from each others ... and list continue... endless untiol peoples do some deal and define test parameters what all use. Other way... like old sentence. Garbage in - garbage out.

Different oscilloscopes have different analog front end. Input levels are divided to some bands and inside bands some small steps.
One perhaps important thing is try first find where is front end 1dB compression point. Why. Because more linearity less mixing. So different scopes may be tested with very different front end working point related to its best linearity. Result may be totally different if test same scope with different level. 
So, it need define what is right level for two tone tests so that it is any kind of useful and comparaple or is it better to keep just for fun playing. And now one question is, is it same power.... ups voltage level for different scopes. Or is it same proportional level from one V/div setting full scale.
I am thousands of km far from my test equipments so I can not demonstrate how important this is specify so that results are comparable and give some real useful information.

Used test level.
Two tones sources isolation. Is it isolated when one function generator digitally combine two tones - really.
Oscilloscope BW is mostly least 100MHz. Often 200... 300.... 500 etc etc. Testing with 10kHz and then compare scopes is like fun joke.

Last time I characterize one noname SA using two tone test I have two fully separate identical high quality RF generators and well isolating combine with test lab grade components. Even if I need test only up to 100MHz I have not even think to use function generator what can internally "somehow" combine two carriers if I do not want just keeping some fun.  Yes they can combine and this is suitable and ok for many kind of purposes but imho fully questionable for serious IP3 tests.

https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_application/application_notes/1ma71/1MA71_2e_amplifier_nonlin_meas.pdf

Specially chapter 3 beginning and then there bit later 3.4.

Of course now SA used is now inside DUT, oscilloscope.

[SilverSolder]

[...] Two tones sources isolation. Is it isolated when one function generator digitally combine two tones - really.  [...]

Are you saying, we might really be measuring the distortion of the function generator itself if we do that?

[gf]

[...] Two tones sources isolation. Is it isolated when one function generator digitally combine two tones - really.  [...]

Are you saying, we might really be measuring the distortion of the function generator itself if we do that?

That's why you need to take care, cross-checking the signal on a SA, in order to verify that the a priori distortion of the signal is significantly lower than the distortion of the scope under test. Note that the SA is not supposed to be the DUT for the test which is subject of this thread. Testing a high dynamic range SA is much more demanding.

[2N3055]

[...] Two tones sources isolation. Is it isolated when one function generator digitally combine two tones - really.  [...]

Are you saying, we might really be measuring the distortion of the function generator itself if we do that?

Of course you will measure distortion of the function generator if it is worse than scope or SA noise floor..
I agree with Rf-loop that this is not serious testing... but...
We did get some interesting  and real data from it despite not being done by rigorous method.
Not accurate data, but more of a glimpse of order of magnitude of some behaviour.

First thing I discovered is that digital WaveCombine on my SDG6000X does respectable job on low frequencies at 0dBm . Also most of the scopes shown here, inexpensive and expensive, has shown that pretty much none of them have any appreciable intermod distortion above noise level and resolution of scope A/D, at frequencies and levels tested..

This makes me curious to actually sit down and do more comprehensive test, first to figure out IMD of SDG6000X across frequencies and output levels using digital WaveCombine, to figure out how well it works when D/A isn't full scale, and when switching attenuators, at lower levels when S/N ratio gets low or distortion when I get to higher levels and amplifiers start distorting...

Also here we could see difference in FFT implementation on different scopes, and where are the limitations...

So not a waste of time.

But it is all, so far a discovery journey to learn something about your equipment (Thanks to  Mawyatt who  brought it up as a topic). This type of testing is very common in industry for much more than just mixers and RF, like he said, but his bringing it up here made it visible to many who didn't use it before for this purpose. And it sparked some very good posts, so all very good and well...

If we would like to make standardized test, we would have to define it. Which would need some sort of goal or target result expectations (what do we want to accomplish), some sort of technical measurement equipment spec (what equipment and what spec we would need to do it), and measurement method (what needs to be measured and how exactly).

Problem is, that could be specified, but it would vary for different purposes (IMD distortion of audio front end, or 100MHz scope or 1Ghz scope) and would soon evolve into IEC type of standard, that not many would be able to follow, if not for any other reason than lack of equipment necessary.

Regards,

[mawyatt]

As the OP this request was to help show how good or bad the scope front ends are and and how well they preserve the ADC performance. Having AWGs available at reasonable cost, and modest SAs allows the Two Tone IMD Test to be implemented by just about anyone that's seriously interested, and many folks certainly did contribute...so thanks

There's no need for a complex expensive setup as long as the input signal source IMD is at least 10~20dB better than what you are measuring. The AWGs have shown very good performance even when the two tones are combined digitally (I assume this for the Siglent AWG, very nice feature).

My goal was to help decide on which Digital Scope to purchase for my developing home lab now that I'm retired and this cost is out of my own pocket

Hopefully it may help others interested in new instruments, or just evaluating the instruments they already have on hand.

So many thanks to all that have contributed and helped me spend my retirement savings 

BTW just purchased the Siglent SDS2102X Plus, wanted the 2104X Plus but they are not in stock here is US. Soon will be the AWG and SA, and maybe a programmable PS.

Best,

[jjoonathan]

Is SDG6000X WaveCombine digital? Mine clicks a relay when I engage it. I suppose it's probably range related. In any case, the resulting signals are quite nice!