AD3 thd vs AD2

[dvb]

Hi there,

 I have just a quick and simple question (I hope), before making my mind into buying an AD3 :) - I want to use it for some basic audio measurements

In my current AD2 setup, if I feed 1Vrms from W1 into C1 via BNC adapter I have 0.02-0.03% THD in spectrum analyzer... is AD3 capable of lower THD?

Unfortunately it seems I can't insert a printscreen here, neither an external URL...

Thank you!

 

 

[attila]

Hi @dvb

The performances of AD1,2,3 are similar. Here you can find a feature update list:

 

The signal is generated with a 3rd device:

 

The ADP2230 has deep buffer and could offer better results:

[dvb]

@attila - Thank you! - I've done extensive reading before posting this, AD3 vs AD2, the only new features I don't understand completely are 15bit@50MHz and 16bit@25MHz sampling and the behaviour of the hardware filters...

As prerequisite, with my AD2, I measured W1 source with a Keithley 2015 THD multimeter; for 1Vrms@1kHz I obtain 0.02% THD+N, while activating the CCITT filter in the multimeter gives me about 0.005% THD+N - not bad at all! :). Repeating these for THD only I receive 0.006% and 0.002% respectively, w/o and with CCITT filter, but these are probably too optimistic to be true... here's how internal filters into DMM work, they are obviously centered around 1kHz measuremets :)

 

OK, then I made this test, W1 into C1 in AD2, config. 1 (4k) and config. 2 (8k)  and gives me something like in attach ("Channel options" set to "average" too);

The main question is that a combination in 16bit sampling + filters can do better in AD3?

For these "static" 1kHz measurements (de facto "standard" until 10-15 years ago) it's obvious that a simple soundcard can do better but the instrument that shines in WaveForms in conjunction with ADx is the Network Analyzer! In this respect the range 10Hz-100kHz (or even 1MHz) is brilliantly covered both by hardware and software, what I can only wish is about 6-12dB lower in THD

 

 

 

Edited May 6 by dvb

[dvb]

@attila here's an exemple how I use the Network Analyzer, my own interstage xfmr driven by a DHT triode 801A :)

 

[dvb]

... I'm talking of course about this loopback W1 into C1 and C2, there's no soundcard with such frequency linearity! ... if only the THD was a little lower... :)

 

 

[attila]

Hi @dvb

The average sample mode simply takes the average of N ADC values, like at 10MHz with 100MHz system clock each stored sample is the average of 10 conversions. The AD3, ADP2230, ADP3X50 stores up to 16 bit and the AD1,2 only 14 bit samples. The Network Analyzer enables this by default but it is only effective at lower rates, like bellow 200kHz with 16Ki samples 32 min periods.

 

BNC loopback with 100ms coherent averaging and 1M samples for ADP2230.

[dvb]

@attila - thank you!

Still, looking at the big picture, do you think a 9th configuration, like keeping the Scope settings from 2nd config. but allowing all the remaining memory to AWG1 will help the accuracy of Network and Spectrum Analyzers?

... meaning nothing for AWG2, Logic, Patterns & Others

... or any other new config which will help specifically the 2 instruments mentioned above

Edited May 7 by dvb

[attila]

Hi @dvb

I don't see difference between 1k (green) and 4k (red) AWG buffer.

[dvb]

@attila - Yes, I also did a comparison of the first 3 configurations from AD2 (1Vpk, rigurously same conditions) but including also THD+N... I know I can not expect wonders from a new config but still... :)

 

[dvb]

... of course, maybe I messed up the options settings in the cropped cartouche in the right, but is the same mess up...

[attila]

Hi @dvb

The Network Analyzer uses average sampling mode for the ADC to eliminate/reduce higher frequency aliases. 
The signal generated with deeper DAC buffer looks cleaner with decimated sampling, but with averaging I see minimal difference, 0.2dB

[dvb]

@attila - I have to confess I'm far for proficient in WF, but at least I learnt today how to "hold" a measurement in Spectrum Analyzer ("overlay" :) ) 

So, in my first test below I tried to repeat yours but starting  "small" :), only comparing "Channel options" Decimate (dec) vs Average (avg) for config. 1 in AD2, 2x(8k+4k)

then I switched to config. 2 in AD2, 2x(16k+1k)

So I'm glad we misunderstood each other (because I learnt a couple of things today); while in my 3 configs comparison I was trying to underline the bigger difference in THD+N, you were thinking about THD all along... however, in my tests there is also a notable gap in THD too - my initial (uneducated) logic told me that the increase in THD+N in second case is only because of smaller AWG's memory buffer... I don't know...

 

 

Edited May 7 by dvb

[dvb]

Honest mistake in my previous post - but not entirely useless test - same BINs in both cases, so I remade for config 1 & 2, last value in trace label is BINs used, so configuration 2 in AD2 2x(16+1) has 2 cases at 4k and 8k sampling, in all 3 cases Channel option set to average like in Network Analyzer

I also changed and spaced the cables a little, there are 2x1m cables entering BNC rig. I also disconnected a third cable from Keithley DMM.

 

As far as I know only the first 10 harmonics are considered so I reduced the frequency range from above...

 

 

.... reducing even further to the audio domain which I'm interested in, changing from 10k to 1kHz and changing the X scale to logarithmic looks like this (see increase in THD for config 1...)

 

 

 

 

 

 

 

Edited May 8 by dvb

[dvb]

The first "3 configurations comparison" from AD2 compressed in one screen... now that I know how to "hold", "overlay", "reference" :)

only THD and THD+N shown

up to 200kHz

and up to 2MHz, slightly different THD distribution...

 

 

Edited May 8 by dvb

[attila]

Hi @dvb

I tried to make configuration 9 similar to the 2nd but with more AWG buffer but it does not fit.
Later I will try to add cordic sine generator for 2nd config, which should help in AWG THD.

[dvb]

Hi @attila ! ... and thank you for trying!

I had 2 intensive days with WF and things get clearer for me, finaly! :)

From my last 2 graphs and for my practical needs in audio domain Config.2 seems appropiate up to 200kHz in Network Analyzer. Given the resolution ADC/DAC THD+N doesn't help me too much but THD does... below -70-80dB THD in audio the discussion is mostly academical anyway :) ... or dedicated to high-end soundcards below -110dB THD (which will end up in speakers having 2-5-10% distortion :) )

Edited May 8 by dvb

[attila]

Hi @dvb

I'm not sure if you are aware of but WaveForms can be used with sound card too, last device in the manager.

[dvb]

@attila - I bought AD2 to get rid of soundcards :) ... which are totally unreliable except maybe "static" 1kHz Spectrum Analyzer. Try to get a "sweep" 10Hz-50kHz with WF Network Analyzer including THD and THD+N... Soundcards, even if you waste the time to "calibrate" them,  have a very narrow in/out level, narrow and nonlinear bandwidth and flimsy drivers... maybe I'll open a topic with soundcard related issues in WF, compared with REW for example, but this is not a priority right now...

However, check these taken with a Focusrite 2i2;

 

[dvb]

@attila - can't stress enough what important will be any workaround related to "config. 2" up to 0.2MHz, even with marginal results!

Of course, I don't know much about hardware restrains or cordic sine generator, but if helps only C1, C2, W1 and 0.2MHz "limit" should be considered in this new/improved config.

[dvb]

To contribute to the cause I thought some additional measurements will help...

As preparation I measured the 1kHz output of Focusrite 2i2 with Keithley 2015 THD - both mentioned in this topic. I didn't use any filter embedded in instrument, like those mentioned at the beginning.

The following picture came identical with what I measured with WF and SpectraPlus (using solely 2i2 as hardware), around 0.0007%, but this instrument has a refresh rate of few times per second and I didn't capture exactly that precise figure :) ... almost 

 

From this I moved to AD2 and I connected W1 to Keithley's input and here are the results (min and max);

 

 

... and finally I connected Focusrite output to AD2's C1 and I almost meltdown :) ... this is config.2 ADC from AD2

 

I hope it helps.

 

Edited May 9 by dvb

[dvb]

Small update to the previous graphs, I reduced the upper frequency to 20kHz (knowing that Focusrite doesn't solve too good the upper part of the spectrum and Keithley was set to measure the first 10 harmonics) 

... and of course changing from "Linear average" to "Min hold" gives "better" figures :) 

 

Edited May 11 by dvb