Two Tone Test with Scope and SA

[mawyatt]

Hello,

I've been looking at various scopes from the usual sources and wanting a scope that could give high Dynamic Range for the standard 2 tone IMD tests using an FFT. The moderately priced scopes mostly seem to have 8 bit ADCs with exception of the R&S which has a 10 bit ADC. Has anyone done any FFT based 2 tone IMD tests at low (~10KHz & 100KHz) and medium (1MHz) and higher frequencies (>10MHz)?

I know this test is more in line with a Spectrum Analyzer, has anyone done these test and frequency range with a modest SA?

I think the new dual channel 16 bit AWGs from Rigol and Siglent might be a good source for the 2 tones, using a resistive combiner at the outputs. 

Anyway, interested if folks have done any of these test and any comments, recommendations are welcome.

Seems this would be an interesting comparison for the various scopes, SA and AWGs now available.

Thanks in advance.

Best,

[xmo]

"... standard 2 tone IMD tests..."
--------------------------------------

Standard?  What standard?

IMD tetsts?  Testing what?  Ferrites?  Antennas?  Coax cables? Receivers?  Amplifiers?  Transmitters?

[mawyatt]

To help evaluate the instruments, especially the scopes front end circuitry and ADCs. Given quality 2 tone AWG outputs (or other quality signal sources), how do the various scopes compare with each other and with a modest SA.

Are the newer 16 bit AWGs good enough sources for these 2 tone IMD tests for typical scopes?

Best,

[tautech]

After some emails chatting about other stuff I got the rundown on what mawyatt considers a Two Tone Test...easy it is too using a single output from both channels of a SDG2042X and its Wave Combine feature.

FYI mawyatt word for word instructions so anyone can join the fun:
The test are very simple, they use two close spaced tones of equal amplitude and show the result of an FFT or direct SA output. Example create a tone of 10KHz in one AWG channel, another tone of 10.1KHz is the other channel. Resistively sum these together and set amplitudes to 1V and look at the FFT result, or SA result. You will see the 2 tones at 1V (or +10dBm(10mw) in 50 ohm system) at 10KHz and 10.1KHz and two smaller tones at 9.9KHz (10KHz -Difference frequency) and 10.2KHz (10.1KHZ + Difference Frequency), these side tones are due to Intermodulation Distortion and tell how good the DUT is, in this case the Scope or SA (assuming tones are really good to start with from AWG). Now do this at 100KHZ and 101KHz, then 1MHZ and 1.0MHz and 1.01MHz and so on. You can use different levels and tone spacing to revel things about the system under test.

The math behind this involves the 3rd Order non-linearity which creates these side tones. Very important test for just about any system, audio, RF or Microwave.

Behind the scenes mawyatt asked me for FFT 2 tone test results using SDS2000X Plus and now I have the 100 MHz model in stock again, here they are and there should be more than enough info in each screenshot to establish the frequencies used that BTW are based on 10 KHz, 100 KHz, 1 MHz and 10 MHz.

Signal source
SDG2042X 50 Ohm output mode 2 active channels combined. 500mV p-p.
Siglent BNC cable
SDS2104X Plus in FFT split screen mode.









And as some sanity check the 10 MHz 2 tone into a SVA1032X.

[mawyatt]

Rob,

Thanks for the tests.

These tests show the difference between the two tones and the 3rd Order products is over 50dB indicating the SDS2104 has a respectable front end and ADC performance. Was this with the 2104 in 10 bit or 8 bit mode?

The SA displays shows the AWG output is good. Since the two tones were combined within the AWG in the digital domain, this shows that the AWG DAC and output amplifier have good linearity. BTW, what model SA did you use?

All in all nice performance of the SDS2104 & SDG2042. Hopefully others can do this simple test with other similar priced and featured scopes and we can see how they compare.

Just a note, the individual tones need to set at 1V peak loaded with 50 ohms (+10dBm into 50 ohms, for 0dBm reference 0.316V peak), however this is fine as shown.

Thanks again for running these 2 tone tests!!

Best,

[mawyatt]

Rob,

Thanks for the tests.

These tests show the difference between the two tones and the 3rd Order products is over 50dB indicating the SDS2104 has a respectable front end and ADC performance. Was this with the 2104 in 10 bit or 8 bit mode?

And how big is the range of values so you can say this is a good frontend or a bad one?
Is this test really meaningful or comparable for an oscilloscope?

Edit: I think I could do the test with an R&S RTB2004, Rigol DG4162 and DSA815.

Simple, if the front end doesn't degrade the ADC too much then it's a respectable front end IMO!! The scope is a system and uses an ADC and has a limit based upon the type and number of bits, so you don't want the front end to degrade the ADC performance too much. You can do this simple test at a number of levels & frequencies to "see" how the scope preforms. If you want to consolidate this into a meaningful number, then the classic IIP3 comes into play.

This is a simple way to compare instruments of similar price and feature points, and for me a decision on what scope to purchase.

TI has an nice article on ADCs for calculating NF and 3rd Order Intercepts. I'm sure there's lots more information available regarding this on line.

https://www.ti.com/lit/an/slyt090/slyt090.pdf

Best,

[tautech]

Was this with the 2104 in 10 bit or 8 bit mode?

8 bit.

The SA displays shows the AWG output is good. Since the two tones were combined within the AWG in the digital domain, this shows that the AWG DAC and output amplifier have good linearity.

This is the graphic of how it is done:

BTW, what model SA did you use?

SVA1032X in SA mode.

Just a note, the individual tones need to set at 1V peak loaded with 50 ohms (+10dBm into 50 ohms, for 0dBm reference 0.316V peak), however this is fine as shown.

Yeah sorry I stuffed up as when I put the SDG in 50 ohm output mode I forgot the selected amplitude was halved and got half way through the tests before it dawned on me why the FFT peaks were lower than they should be. 
Anyways I could've selected a dB output for things to line up even better. 

The little SDS1104X-E is next.....when I get a chance.....shame member hendorog has my SDG6022X as with that I can push 2 tone tests to 200 MHz or 500 MHz when 'improved'. 

[David Hess]

The results are a couple dB better than I would expect but consistent with the dynamic range of an 8-bit digitizer at about 51 dB.  The SFDR is usually lower but modern integrated ADCs are self calibrating so their linearity should be pretty good.

[mawyatt]

Was this with the 2104 in 10 bit or 8 bit mode?

8 bit.

The SA displays shows the AWG output is good. Since the two tones were combined within the AWG in the digital domain, this shows that the AWG DAC and output amplifier have good linearity.

This is the graphic of how it is done:

BTW, what model SA did you use?

SVA1032X in SA mode.

Just a note, the individual tones need to set at 1V peak loaded with 50 ohms (+10dBm into 50 ohms, for 0dBm reference 0.316V peak), however this is fine as shown.

Yeah sorry I stuffed up as when I put the SDG in 50 ohm output mode I forgot the selected amplitude was halved and got half way through the tests before it dawned on me why the FFT peaks were lower than they should be. 
Anyways I could've selected a dB output for things to line up even better. 

The little SDS1104X-E is next.....when I get a chance.....shame member hendorog has my SDG6022X as with that I can push 2 tone tests to 200 MHz or 500 MHz when 'improved'. 

Rob,

Could I impose on you to do a couple of these tests with the 10 bit ADC mode? Like David mentioned below, these are a little better than one would expect in an 8 bit Data Acquisition System, which is after all what these scopes are!! Maybe the 10 bit performance will be even better!!

Best,

[mawyatt]

The results are a couple dB better than I would expect but consistent with the dynamic range of an 8-bit digitizer at about 51 dB.  The SFDR is usually lower but modern integrated ADCs are self calibrating so their linearity should be pretty good.

Agree these results are a little better than I expected with the 8 bit ADC, speaks well for the input circuit linearity. And since we don't know how much the ADC input range is filled, maybe even a little more possible performance! Wonder what ADC Siglent is using, but the popular 8 bit 1GSPS TI ADC81000 specs a ENOB of 7.5 bits at Nyquist and IMD of -51dB with SFDR of 58.5dB.

The 10 bit results may be even better!!

Best,

[David Hess]

The results are actually so good that I wondered if non-linearity in the source was cancelling non-linearity in the DSO but the SA results rule that out.

8 bit linearity is not particularly demanding of an oscilloscope front end (1 part in 200 or 0.5%, see below) and other errors tend to limit performance to below 8 bits anyway.  (1) Older DSOs tend to include over-ranging so the display only encompasses a range of say 200 counts or 50 points per division for 8 vertical divisions.  This leaves some leeway to align the waveform with the position control and do math on a displayed waveform which is slightly larger than the display without truncation.

(1) Linearity gets worse at higher frequencies, but this particular case is almost ideal.

[egonotto]

Hallo,

several measures:

RTA4004 with 1MOhm (I forget switch to 50Ohm)
Picoscope 4262 with 50 Ohm
Picoscope 5243A with 50 Ohm

Best regards
egonotto

[tautech]

Could I impose on you to do a couple of these tests with the 10 bit ADC mode?

10 bit mode with 10 KHz and 10 MHz 2 tone tests. AWG output now 0 dB.
Envelopes are always fun to get solid triggering on so holdoff added as/if required and done previously. 

Again, SDS2104X Plus and SDG2042X.

[jjoonathan]

Here's a SDG6022X into a RTO1024 oscilloscope and FSP spectrum analyzer.

[mawyatt]

I'm not sure I did everything right. I combined the two signals from the generator with a T-piece and terminated them with 50 Ohm on the oscilloscope. For now only a test with 100 kHz.

Is MaxHold correct for FFT?

Generator Rigol DG4162
Oscilloscope R&S RTB2004
Spectrum Analyzer Rigol DSA815 (with extern 20 dB Attenuator)

Think you want to resistively combine the two sources for the tones unless you can combine them internally like the Siglent. A resistive "T" should work, think this has 16.7 series ohms on each leg to maintain 50 ohms on all three ports.

Don't think you need to use the FFT max hold feature.

Best,

[tautech]

Don't think you need to use the FFT max hold feature.

You certainly don't if you don't mind looking at all the noise at noise floor levels.
I don't as I'm only interested in the fundamental and its associated peaks therefore I mostly use some form of peak/max hold.
All screenshots done thus far here are with max/peak hold.

[mawyatt]

Here's a SDG6022X into a RTO1024 oscilloscope and FSP spectrum analyzer.
(Attachment Link)
(Attachment Link)

This is superb performance, what you would expect from a 16 bit ADC based scope!! However it's well out of the price range of these other scopes, and I see it's been replaced with the RTO2000 starting at only $14,500!!

Best,

[mawyatt]

Could I impose on you to do a couple of these tests with the 10 bit ADC mode?

10 bit mode with 10 KHz and 10 MHz 2 tone tests. AWG output now 0 dB.
Envelopes are always fun to get solid triggering on so holdoff added as/if required and done previously. 

Again, SDS2104X Plus and SDG2042X.

Thanks for running these tests again. Appears the 10 bit doesn't improve over the 8 bit performance which makes me wonder if the 10 bit is achieved with some sort of signal averaging rather than a true direct 10bit digitized result?

Best,

[mawyatt]

Don't think you need to use the FFT max hold feature.

You certainly don't if you don't mind looking at all the noise at noise floor levels.
I don't as I'm only interested in the fundamental and its associated peaks therefore I mostly use some form of peak/max hold.
All screenshots done thus far here are with max/peak hold.

I prefer to see the noise floor, as sometimes there might be some information buried within. Everyone has a preference!!

BTW the 10KHz Time Domain plot doesn't look like the envelope is extending down to zero, is this just an artifact of the display & time base or is the waveform envelope not getting to zero?

Best,

[tautech]

Thanks for running these tests again. Appears the 10 bit doesn't improve over the 8 bit performance which makes me wonder if the 10 bit is achieved with some sort of signal averaging rather than a true direct 10bit digitized result?

10 bit explanation here:
https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg2783928/#msg2783928

I prefer to see the noise floor, as sometimes there might be some information buried within. Everyone has a preference!!

Each to their own.   

BTW the 10KHz Time Domain plot doesn't look like the envelope is extending down to zero, is this just an artifact of the display & time base or is the waveform envelope not getting to zero?

Yeah I spotted that too so will investigate..........
Did it also happen in 8 bit mode ? Haven't looked yet.

[mawyatt]

Thanks for running these tests again. Appears the 10 bit doesn't improve over the 8 bit performance which makes me wonder if the 10 bit is achieved with some sort of signal averaging rather than a true direct 10bit digitized result?

10 bit explanation here:
https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg2783928/#msg2783928

That explains it, not a true 10bit but oversampled. Wonder if Siglent will extended the oversampling to 12 bits like Keysight has done?

Best,

[tautech]

Thanks for running these tests again. Appears the 10 bit doesn't improve over the 8 bit performance which makes me wonder if the 10 bit is achieved with some sort of signal averaging rather than a true direct 10bit digitized result?

10 bit explanation here:
https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/msg2783928/#msg2783928

That explains it, not a true 10bit but oversampled. Wonder if Siglent will extended the oversampling to 12 bits like Keysight has done?

SDS2000X Plus also has ERES of normally up to an additional 3 bits.
Check out those posts more for further info on this. 

[tggzzz]

If you can still get one, the Digilent Analog Discovery 2 has 14bit ADCs, AD9648
https://reference.digilentinc.com/reference/instrumentation/analog-discovery-2/reference-manual#scope_adc

[joeqsmith]

doubt this will be of any help but here is my HP33120A & 2024  set to 10.0 & 10.1MHz with minicircuits splitter.

Shown on my SignalHound  in swept mode and my vintage HP3589A.     HP supports a narrow band sweep but I think this is outside of it's usable range.     

For fun also shown with the HP3589A using 10MHz and 10.0001MHz.   

[David Hess]

I'm not sure I did everything right. I combined the two signals from the generator with a T-piece and terminated them with 50 Ohm on the oscilloscope.

Think you want to resistively combine the two sources for the tones unless you can combine them internally like the Siglent. A resistive "T" should work, think this has 16.7 series ohms on each leg to maintain 50 ohms on all three ports.

You have to be very careful combining the signals because the output amplifiers of the signal generators will produce the same distortion products that you are testing for if they are driven with the other signal.

Ideally a combiner which provides isolation between the input ports is required but it is usually sufficient to pad each output with an attenuator and use minimum loss resistive divider.

Update: I meant to type "minimum loss resistive combiner".