Two Tone Test with Scope and SA

[mawyatt]

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!

Don't have one so just speculating and why I said "think" it might be. Analog combing would have the advantage of better resolution. The results certainly look good, so either way it's a nice performing AWG.

Best,

[tautech]

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!

Don't have one so just speculating and why I said "think" it might be. Analog combing would have the advantage of better resolution. The results certainly look good, so either way it's a nice performing AWG.

All the SDG X models offer the Wave Combine feature including the 14 bit SDG1000X models.

I checked the SDG2042X I've been using and indeed if channels are ON and Wave Combine is selected a relay click is heard however until someone RE's these we don't yet know where in the signal path this occurs.

Still don't have my SDG6022X back from a mate but will have over the weekend so to push 2 tone tests to much higher frequencies than thus far...... but it might need improving a bit first. 
Will also do 2 tones at different amplitudes like rf-loop suggested and all for shits and giggles although we might just learn a bit more. 

[mawyatt]

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!

Don't have one so just speculating and why I said "think" it might be. Analog combing would have the advantage of better resolution. The results certainly look good, so either way it's a nice performing AWG.

All the SDG X models offer the Wave Combine feature including the 14 bit SDG1000X models.

I checked the SDG2042X I've been using and indeed if channels are ON and Wave Combine is selected a relay click is heard however until someone RE's these we don't yet know where in the signal path this occurs.

Still don't have my SDG6022X back from a mate but will have over the weekend so to push 2 tone tests to much higher frequencies than thus far...... but it might need improving a bit first. 
Will also do 2 tones at different amplitudes like rf-loop suggested and all for shits and giggles although we might just learn a bit more. 

Well if you and JJ are hearing a relay click that's gotta hint at the combining is taking place in the analog section, not digital. Since a relay is used this also hints that this is at a higher analog level where a relay is preferred for linearity over a SS switch, maybe even at the outputs which are resistive summed??

I'll be happy to RE if someone will lend me one

Best,

[mawyatt]

Hey, this is old but think it's the best place to start with a continuation with the new Siglent SDS2000X HD which has true 12 bit ADCs. Nice to compare how well the 12 bit core ADC improves the 2 Tone IMD tests.

Know a couple of you folks out there have these new MSOs and hoping you'll be kind enough to do some additional 2 Tone IMD testing with the basic 12 bit and any additional resolution enhancements available.

Best,

[rf-loop]

Generator SDG1032X. It need need tightly understand that this with splitter is not enough for serious dual tone test. Who knows what it itself generate...
Separate channels externally resistive combined (resistive 50Ω Δ type splitter)
So, SDG outputs are 6dB higher. To SDS2504X HD input ~0dBm signals.



9.9kHz and 10.1kHz  (distance 200Hz)
f1 = 9.9kHz  f2 = 10.1kHz

2*f2 - f1 = 10.3 Peak Marker 4 (-85dBc)
2*f1 - f2 =   9.7 Peak Marker 1 (-83dBc)




99.9 and 100.1kHz  (distance 200Hz)


0.9999MHz and 1.0001MHz  (distance 200Hz)
(note FFT samplerate change)


10.999MHz and 11.001MHz (distance 2kHz diff)
(note FFT samplerate change)

[mawyatt]

Generator SDG1032X. It need need tightly understand that this with splitter is not enough for serious dual tone test. Who knows what it itself generate...

2*f2 - f1 = 10.3 Peak Marker 4 (-85dBc)
2*f1 - f2 =   9.7 Peak Marker 1 (-83dBc)

Well we may not know just how good the signal from the SDG1032X is (could be verified with a quality SA tho), but the result you've shown is quite good indeed!! If we "assume" the IMD result is root sum square of the AWG and the DSO, then the DSO worst case IMD is as shown (assuming AWG IMD is zero)!!!

Siglent evidently got this right just like they did with the 8 bit SDS2000X+

Thanks for taking the time to show this

Best,

[David Hess]

Run the same test with a low distortion summing amplifier and see if the results agree.

[G0HZU]

For comparison, a fairly modern lab spectrum analyser should achieve about 95dB SFDR in a two tone test with a 100Hz RBW selected. Greater SFDR than this is possible with a narrower RBW but the measurement time will obviously be longer.

[Performa01]

Since we have seen some LF results so far, I thought I’ll add some measurement at the upper end of the spectrum, specifically at 90% of the specified bandwidth of 500 MHz.

So, this IMD test is about 450 MHz and 200 kHz frequency offset.

It has already been brought up several times before: for comparable results, we need a bunch of conventions and preconditions. I cannot meet all of them, for instance I cannot guarantee that the dual tone test signal has any lower than -74 dBc IMD.

As has been already discovered, the results also depend on the scope settings. Like a “real” spectrum analyzer, we won’t get impressive results if we feed the scope input with full scale signals (in SA terms: right below the compression level). In general, the lower the input signal(s), the lower the distortion. For testing the third order dynamic range of an SA, we usually have signals in the realm of at least 30 dB below the 1 dB compression point. Consequently, I think it is only fair to use signal levels at least about 10 dB below full scale for the scope as well.

Now let’s calculate the settings for 0 dBm test signals (sine waves assumed throughout these calculations).

0 dBm into 50 ohms means 223,61 mVrms or 632,46 mVpp.

With two 0 dBm signals combined, the peak output voltage doubles to 1.265 Vpp.

The SDS2000X HD ADC input range is roughly 9.6 divisions, so we need at least 132 mV/div to have the dual tone signal within full scale. If we want to keep the signal at or below -10 dBFS, this makes for 4 Vpp full scale.

In the end I chose 500 mV/div, which is equivalent to 4.8 Vpp full scale.

We can get ~74 dBc IMD3 in this scenario, this is equivalent to a TOI (third order input intercept point) of +36.7 dBm.

See attached screenshot:

SDS2504X HD_IMD_500mV_C450MHz_O200kHz_0dBm_L

Please note that this is not a bad result at all. There were (and still are) plenty of low(er) end spectrum analyzers that might not even reach 70 dB third order dynamic range, whereas top instruments can provide up to 110 dB.

For example, the ancient HP8590A has a third order dynamic range of 70 dB at -30 dBm input.

 

[mawyatt]

Since we have seen some LF results so far, I thought I’ll add some measurement at the upper end of the spectrum, specifically at 90% of the specified bandwidth of 500 MHz.

So, this IMD test is about 450 MHz and 200 kHz frequency offset.

It has already been brought up several times before: for comparable results, we need a bunch of conventions and preconditions. I cannot meet all of them, for instance I cannot guarantee that the dual tone test signal has any lower than -74 dBc IMD.

However you can guarantee that the DSO (or the AWG) is better than this....which is impressive for the DSO, and not so bad for the AWG either IMO

As has been already discovered, the results also depend on the scope settings. Like a “real” spectrum analyzer, we won’t get impressive results if we feed the scope input with full scale signals (in SA terms: right below the compression level). In general, the lower the input signal(s), the lower the distortion. For testing the third order dynamic range of an SA, we usually have signals in the realm of at least 30 dB below the 1 dB compression point. Consequently, I think it is only fair to use signal levels at least about 10 dB below full scale for the scope as well.

For the DSO case the lower signal level may not be the "Sweet Spot" of IMD, this is because you need to partially fill up the ADC range to capture as many "bits" as possible and this conflicts with the smaller the signal the better the amplifiers IMD performance. So the overall DSO "Sweet Spot" is more likely somewhere in the middle DSO range where you've positioned your input signal.

We can get ~74 dBc IMD3 in this scenario, this is equivalent to a TOI (third order input intercept point) of +36.7 dBm.

See attached screenshot:

SDS2504X HD_IMD_500mV_C450MHz_O200kHz_0dBm_L

Please note that this is not a bad result at all. There were (and still are) plenty of low(er) end spectrum analyzers that might not even reach 70 dB third order dynamic range, whereas top instruments can provide up to 110 dB.

Agree, this is very good indeed and indicative that the Siglent engineers got this front end design right just like the X+ version, of course they had the advantage of that excellent design to start with!!

This speaks very well for the new HD version, and shows that Siglent took the time and energy to not just replace the 8 bit ADC with a 12 bit version, but to reengineer the front end to support the new 12 bits ADC and not degrade it's performance!!

Makes one wonder just how good this DSO is, we know it's better than -74dBc IMD at 450MHz from your tests. Maybe someone with a really good SA and a couple quality signal sources with much better than -74dBc IMD when resistive attenuator summed can duplicate your tests. Also especially interested if the lower frequency Two Tone Test can be done with the ERES function (if possible within FFT) which I understand is hardware based in the HD version.

Anyway, thanks so much for the testing and reports!!

Edit: Maybe this front end design is from the 6000 series that they did for LeCroy with the 12 Bit ADC??

Best,

[G0HZU]

However you can guarantee that the DSO (or the AWG) is better than this....which is impressive for the DSO, and not so bad for the AWG either IMO

Note that the displayed result on the scope or analyser screen can appear artificially higher or lower than the real result due to the linearity limitations of the scope/analyser.

Normally, the distortion performance of the analyser will be fairly well known and understood beforehand and the usual aim is to ensure that the IMD level in the analyser is >20dB lower than the IMD caused by the device under test. This keeps the overall uncertainty of the displayed result below about +/- 1dB.

So to measure -74dBc within +/- 1dB  the scope or analyser needs to be capable of -94dBc IMD. To measure -74dBc within about +/- 3dB  the scope or analyser needs to be capable of -85dBc IMD.

[gf]

However you can guarantee that the DSO (or the AWG) is better than this....which is impressive for the DSO, and not so bad for the AWG either IMO

Note that the displayed result on the scope or analyser screen can appear artificially higher or lower than the real result due to the linearity limitations of the scope/analyser.

Normally, the distortion performance of the analyser will be fairly well known and understood beforehand and the usual aim is to ensure that the IMD level in the analyser is >20dB lower than the IMD caused by the device under test. This keeps the overall uncertainty of the displayed result below about +/- 1dB.

So to measure -74dBc within +/- 1dB  the scope or analyser needs to be capable of -94dBc IMD. To measure -74dBc within about +/- 3dB  the scope or analyser needs to be capable of -85dBc IMD.

But if A+B = -74dBc, can't we still conclude that A <= -74dBc and B <= -74dBc?
[ Btw, this leads to the question: How do IMD products from two sources actually add up? ]

[mawyatt]

However you can guarantee that the DSO (or the AWG) is better than this....which is impressive for the DSO, and not so bad for the AWG either IMO

Note that the displayed result on the scope or analyser screen can appear artificially higher or lower than the real result due to the linearity limitations of the scope/analyser.

The scope or any instrument will always show a result for the Two Tone IMD that is the Root Sum Square (RSS) of the signal source and measuring instrument. Thus, the shown result is the worst case of such, either source or measuring instrument. If the instrument were an older analog display types then there could be a display uncertainty in reading, however these DSOs are actually digital data acquisition systems which utilize ADCs and digital techniques so this isn't an issue.

Can't think of a case were the instrument linearity issues would cause an "artificially lower" IMD result (read better), if you have such a case please provide details.

Normally, the distortion performance of the analyser will be fairly well known and understood beforehand and the usual aim is to ensure that the IMD level in the analyser is >20dB lower than the IMD caused by the device under test. This keeps the overall uncertainty of the displayed result below about +/- 1dB.

Here the device under test IS the analyzer, so the source needs to be better than the analyzer. Again the result is always the RSS of both the source and the analyzer.

So to measure -74dBc within +/- 1dB  the scope or analyser needs to be capable of -94dBc IMD. To measure -74dBc within about +/- 3dB  the scope or analyser needs to be capable of -85dBc IMD.

See above, the scope IS the DUT.

Normally when measuring a device the actual measured IMD result is square root of {(Source IMD)^2 + (DUT IMD)^2 + (Measurement Instrument IMD)^2}, so if the Source and Instrument IMD is << than DUT IMD, then the result becomes the DUT IMD.

In your example of a -74dBc IMD measurement, if the source has no IMD then the Scope IMD is obviously -74dBc. If the source has -77dBc IMD (3dB difference), then the Scope has -77dBc IMD, if the source has -80dBc (6dB difference) then the Scope IMD is -75.3dBc, if the source has -84dBc (10dB difference) then the Scope IMD is -74.5dBc, and if the source has -94dBc (20dB difference), then the Scope has 74.04dBc. Of course you can do this analysis with a true DUT IMD and include the Source and Scope/Analyzer IMD and find the 10dB separation is sufficient for the 1dB uncertainty margin. This tends to confirm the usual 10dB minimum separation in measurements for a reliable resultant, however because the measurement instrument IS the DUT the 10dB margin actually confirms a 1/2dB uncertainty.

Best,

[mawyatt]

But if A+B = -74dBc, can't we still conclude that A <= -74dBc and B <= -74dBc?
[ Btw, this leads to the question: How do IMD products from two sources actually add up? ]

Yes, the result is the Worst Case of -74dBc IMD for all things involved. See directly above for more details.

Best,

[G0HZU]

If the source has -77dBc IMD (3dB difference), then the Scope has -77dBc IMD

It isn't that simple, you have to think in terms of voltage level and phase rather than power.

[mawyatt]

If the source has -77dBc IMD (3dB difference), then the Scope has -77dBc IMD

It isn't that simple, you have to think in terms of voltage level and phase rather than power.

These were all actually computed based upon the RSS method as outlined above. Here one must convert to actual levels (voltage in the case of the DSO) in linear terms not dB, do the proper calculations, then revert back to dB scales. Since you can't add or subtract linear terms in dB (actually a exponential/logarithmic scale, so adding in dB is multiplying, and subtracting in dB is dividing), you can't do the RSS computations without converting to linear terms, do the squares, then summation, then square root, then convert back to dB scales ....elementary signal processing!!

BTW phase has nothing to do with 3rd order IMD as outlined in this thread about Two Tone IMD, it's all about the mathematical relationship between odd order system non-linearity and a method to represent the DUT in a simple way to convey such. This 2 Tone IMD is utilized in all sorts of various fields and why we wanted to see how well these new DSOs behave. It's also useful in extrapolating the well known RF metric of 3rd Order Intercept as was shown earlier (without sweeping the input), however this extrapolation does have a baseline assumption that the DUT has a well behaved non-linearity and exhibits gain compression as signal levels increase. There is a little know amplifier where this metric is not applicable, where the gain actually expands over a range before going into compression, but this is out of the scope for discussion here (will discuss if folks want, but better in another separate thread).

Anyway, hope this helps clear up any confusion regarding the Two Tone IMD test and how they infer the performance of the instruments and DUT involved.

Best

[G0HZU]

BTW phase has nothing to do with 3rd order IMD as outlined in this thread about Two Tone IMD, it's all about the mathematical relationship between odd order system non-linearity and a method to represent the DUT in a simple way to convey such.

When you combine two (same amplitude) IMD tones in phase the voltages sum in phase so you get twice the voltage. This means the level seen on a spectrum analyser goes up 6dB relative to each tone. If the tones are out of phase they cancel so the IMD level will tend to cancel. So I think phase can be important.

Normally with a narrowband test setup the IMD terms tend to be in phase. However, in theory at least, the phase shift can be sufficient to cause significant cancellation of the IMD terms when two non linear stages are connected in series. So you have to consider phase in any analysis.

[G0HZU]

In your example of a -74dBc IMD measurement, if the source has no IMD then the Scope IMD is obviously -74dBc. If the source has -77dBc IMD (3dB difference), then the Scope has -77dBc IMD, if the source has -80dBc (6dB difference) then the Scope IMD is -75.3dBc, if the source has -84dBc (10dB difference) then the Scope IMD is -74.5dBc, and if the source has -94dBc (20dB difference), then the Scope has 74.04dBc. Of course you can do this analysis with a true DUT IMD and include the Source and Scope/Analyzer IMD and find the 10dB separation is sufficient for the 1dB uncertainty margin. This tends to confirm the usual 10dB minimum separation in measurements for a reliable resultant, however because the measurement instrument IS the DUT the 10dB margin actually confirms a 1/2dB uncertainty.

If the source and the scope both truly had -77dBc IMD, the resultant IMD would typically appear at -71dBc on the analyser (a 6dB difference and not the 3dB you imply). This assumes the IMD terms are in phase. In reality there might be a small phase shift. If the test was done with wide tone spacings the phase cancellation effects could be more pronounced (in theory at least).

Your other calculations above are wrong as well.

[mawyatt]

BTW phase has nothing to do with 3rd order IMD as outlined in this thread about Two Tone IMD, it's all about the mathematical relationship between odd order system non-linearity and a method to represent the DUT in a simple way to convey such.

When you combine two (same amplitude) IMD tones in phase the voltages sum in phase so you get twice the voltage. This means the level seen on a spectrum analyser goes up 6dB relative to each tone. If the tones are out of phase they cancel so the IMD level will tend to cancel. So I think phase can be important.

Normally with a narrowband test setup the IMD terms tend to be in phase. However, in theory at least, the phase shift can be sufficient to cause significant cancellation of the IMD terms when two non linear stages are connected in series. So you have to consider phase in any analysis.

Don't think you understand the Two Tone IMD, what you are missing here is that the Two Tones are not the same frequency, but separated by a small fraction of the absolute frequency of either. In the last IMD test graph shown the Two Tones were at ~450MHz, with a 200KHz frequency difference, and both Tones at equal amplitude. So the Two Tones wander in and out of phase and the amplitude varies between twice the individual Tone amplitude (in phase) to zero (out of phase) at the difference frequency rate. Note how the composite amplitude varies between the maximum of twice the single tone level to identically zero, this exercises the total amplitude range of the DUT between these extremes and why this simple 2 tone test is so reveling of the DUT non-linearities.

Best,

[G0HZU]

What matters here is the relative phase of a small IMD term generated in stage 1 (and this IMD tone is then amplified by stage 2) vs the phase of the IMD term actually generated in stage 2 from the two main test tones.

If the phase of these is not the same then there will be some cancellation in stage 2. Do you understand it now?

[G0HZU]

In your example of a -74dBc IMD measurement, if the source has no IMD then the Scope IMD is obviously -74dBc. If the source has -77dBc IMD (3dB difference), then the Scope has -77dBc IMD, if the source has -80dBc (6dB difference) then the Scope IMD is -75.3dBc, if the source has -84dBc (10dB difference) then the Scope IMD is -74.5dBc, and if the source has -94dBc (20dB difference), then the Scope has 74.04dBc. Of course you can do this analysis with a true DUT IMD and include the Source and Scope/Analyzer IMD and find the 10dB separation is sufficient for the 1dB uncertainty margin. This tends to confirm the usual 10dB minimum separation in measurements for a reliable resultant, however because the measurement instrument IS the DUT the 10dB margin actually confirms a 1/2dB uncertainty.

Like I said, your calculations above are totally incorrect. I think you should have used this equation for the case where the terms are in phase.

Amplitude Error = 20*log( 1+ 10^(d/20))   where d is a negative number that represents the difference between the two IMD levels.

In your analysis a 10dB separation delivers <1dB uncertainty. This is obviously wrong and the correct answer is just under 2.4dB.

[mawyatt]

In your example of a -74dBc IMD measurement, if the source has no IMD then the Scope IMD is obviously -74dBc. If the source has -77dBc IMD (3dB difference), then the Scope has -77dBc IMD, if the source has -80dBc (6dB difference) then the Scope IMD is -75.3dBc, if the source has -84dBc (10dB difference) then the Scope IMD is -74.5dBc, and if the source has -94dBc (20dB difference), then the Scope has 74.04dBc. Of course you can do this analysis with a true DUT IMD and include the Source and Scope/Analyzer IMD and find the 10dB separation is sufficient for the 1dB uncertainty margin. This tends to confirm the usual 10dB minimum separation in measurements for a reliable resultant, however because the measurement instrument IS the DUT the 10dB margin actually confirms a 1/2dB uncertainty.

If the source and the scope both truly had -77dBc IMD, the resultant IMD would typically appear at -71dBc on the analyser (a 6dB difference and not the 3dB you imply). This assumes the IMD terms are in phase. In reality there might be a small phase shift. If the test was done with wide tone spacings the phase cancellation effects could be more pronounced (in theory at least).

Your other calculations above are wrong as well.

Well let's just go over those very calculations that you infer are wrong.

Scope IMD is -77dBc and Source IMD is -77dBc,

then -77dBc = 10^ (-77/20) or 10^-3.85 or 141.25 E-6 or 19.953 E-9 squared

So 19.953 E-9 + 19.953 E-9 = 39.905 E-9

Square root of 39.905 E-9 = 199.76 E-6

10 base LOG of 199.76 E-6 = -3.6995

20*(-3.6995) = -73.99 or ~ -74dBc

BTW you can perform the above using dBc as power ratio, i.e. 10*Log rather than voltage ratio which is 20*Log and the result is the same as it should be, since dBc is a ratio of like terms!!

Care to point out where this is wrong!!! How about you post your calculations here and shown how you come to the incorrect conclusion that this is all wrong!!

Best,

[G0HZU]

What you are overlooking is that the IMD terms are coherent. If you combine two coherent terms in phase the level on the spectrum analyser goes up 6dB with respect to the power of each tone and not 3dB.

I'm not sure I need equations to prove this as this should be common knowledge.

You have made the common mistake of just doubling the power. The power goes up by 6dB in the in-phase/coherent case which is a multiplication of four not two.

Put the numbers in to my earlier equation for the case where the IMD terms are at the same level. This is when d = 0.  You should get 6.02dB as the answer (not 3.01dB).

[G0HZU]

A good practical example would be to take two identical 12dB gain amplifiers and put them in series with an attenuator in between them that has 12dB attenuation. So you end up with the same 12dB gain of a single amplifier. Then compare the narrowband IMD level of this arrangement to the IMD seen on a single 12dB gain amplifier with no attenuator. Note that this analysis assumes a perfect measuring tool. Or you could consider the second amplifier stage to be the limiting factor of the measurement tool if you like.

For a narrowband system the IMD of the first and and the second amplifier usually sums in phase so the IMD levels for the two amplifiers in series will usually be about 6dB worse (higher in level) compared to the case where you just measure a single amplifier on its own with no attenuation.

This is because the two IMD terms sum together in phase so you get twice the voltage (four times the power) hence a 6dB increase in IMD level seen with the dual amplifier setup.

[mawyatt]

In your example of a -74dBc IMD measurement, if the source has no IMD then the Scope IMD is obviously -74dBc. If the source has -77dBc IMD (3dB difference), then the Scope has -77dBc IMD, if the source has -80dBc (6dB difference) then the Scope IMD is -75.3dBc, if the source has -84dBc (10dB difference) then the Scope IMD is -74.5dBc, and if the source has -94dBc (20dB difference), then the Scope has 74.04dBc. Of course you can do this analysis with a true DUT IMD and include the Source and Scope/Analyzer IMD and find the 10dB separation is sufficient for the 1dB uncertainty margin. This tends to confirm the usual 10dB minimum separation in measurements for a reliable resultant, however because the measurement instrument IS the DUT the 10dB margin actually confirms a 1/2dB uncertainty.

Like I said, your calculations above are totally incorrect. I think you should have used this equation for the case where the terms are in phase.

Amplitude Error = 20*log( 1+ 10^(d/20))   where d is a negative number that represents the difference between the two IMD levels.

In your analysis a 10dB separation delivers <1dB uncertainty. This is obviously wrong and the correct answer is just under 2.4dB.

Think you are talking about your very own calculations being "totally incorrect"!! And your thinking is totally incorrect!!

The simple straightforward analysis we have shown above shows that the result we arrived at is correct as stated, best review how and where to use the Amplitude Error equation you proposed.

Best,