Focusrite Scarlett 2i2 Gen3 - Measurements in loopback mode and with Topping D10s

[pma]

Focusrite Scarlett 2i2 Gen3

The main review by @amirm is here:
https://www.audiosciencereview.com/...rlett-2i2-audio-interface-gen-3-review.10187/

I have decided to add several measurements in loopback mode with this soundcard

1. THD and THD+N at 1kHz, XLR input

This is not bad, considering the price. THD is quite low, THD+N (SINAD) is higher but still below 0.001%. The main contributor to higher THD+N is DAC output noise. ADC is quieter.

2. CCIF IMD 19+20kHz

I am not very happy with this result. Though difference tone 1kHz is low, we can see skirts of H3 and higher harmonics near the test tones that indicate to worsen high frequency linearity. This is my favourite test and I miss it in most reviews here.

3. THD vs. frequency at -3dBFS

Good up to 1kHz and then we can see rise of distortion with frequency. Mostly the 3rd harmonic. We can see the effect in CCIF measurement here above.

4. THD and THD+N vs. amplitude, at 1kHz

THD result is quite good, THD+N result is not so good and is affected mostly by the DAC noise. Pity.

Conclusion

This is not a bad soundcard, taking into account its price about $140. I appreciate the absence of mains spuriae, balanced inputs and outputs and reasonable THD. DAC noise could be lower.

 

[pma]

Two ways to go to standard XLR cable from the 2i2 TRS outputs

Left channel goes through Neutrik TRS/XLRmale adapter, Right channel through Neutrik TRS/XLR 1m cable


Immunity to signal ground loops (CMR) is very good. Below is a measurement of a preamplifier with balanced I/O that is placed in another room and connected via 2 XLR-XLR cables of 10m length with the Focusrite. No sign of mains hum frequencies. The PC notebook is supplied from its 2 prong plug adapter, that means that there is no additional loop created by mains PE wire between the PC and the preamp. The preamp is class I device.

 

[restorer-john]

Two ways to go to standard XLR cable from the 2i2 TRS outputs

View attachment 142591

Left channel goes through Neutrik TRS/XLRmale adapter, Right channel through Neutrik TRS/XLR 1m cable


Immunity to signal ground loops (CMR) is very good. Below is a measurement of a preamplifier with balanced I/O that is placed in another room and connected via 2 XLR-XLR cables of 10m length with the Focusrite. No sign of mains hum frequencies. The PC notebook is supplied from its 2 prong plug adapter, that means that there is no additional loop created by mains PE wire between the PC and the preamp. The preamp is class I device.

View attachment 142592

@pma your last plot (arta) above is 96kHz sampling but only showing 20kHz upper most freq. The noise rise is chopped off the display, unlike the IMD where you display a 40kHz upper most freq.

I have both the 2i2v2 and the 2i2v3 and they perform best at 48kHz. Their sweet spots are also much lower output voltages, around 381mV (2i2v2) and no gain on the input pots (put them at zero)

2i2v2 (unweighted):

I'll run the same on the 2i2v3 tomorrow and find the sweet spot for THD and noise and levels.

 

[pma]

I have both the 2i2v2 and the 2i2v3 and they perform best at 48kHz. Their sweet spots are also much lower output voltages, around 381mV (2i2v2) and no gain on the input pots.

I tend to have a different opinion. In your "sweet spot" you get THD=0.00034%, but THD+N is high = 0.0038%. This is the result of low level + high intrinsic noise. On contrary, if you see my post #1, I get THD = 0.00035% (same as you) at much lower THD+N = 0.00089%, in a loop, close to full XLR input level. This is a significant difference of 13dB less noise. Try to think about it thoroughly .

Regarding the plot in post #2 that you have mentioned, the DAC noise is amplified by the preamp measured so the THD+N is higher. It is not a loopback measurement of the soundcard itself, but the preamp with gain is inserted, in a loop of 20m total length. This was to demonstrate very good CMR of the 2i2 input amplifier.

 

[restorer-john]

Don't forget, yours is A weighted, mine (2i2v2) is not. A-weighting gives an almost 10dB response reduction by 20KHz.

Anyway, we'll do an apples to apples comparison with my 2i2v3 and see how close they are. Will be interesting. They really are great little interfaces aren't they?

 

[pma]

Don't forget, yours is A weighted, mine (2i2v2) is not. A-weighting gives an almost 10dB response reduction by 20KHz.

Again wrong, John. The THD+N in post#1 is measured without weight with BW = 20Hz - 20kHz. You are referring to measurement with PRE inserted which is irrelevant to card noise. My humble suggestion for you is to read thoroughly before you write an opinion, because your opinion is valid for many members. However, you have to suppose that I will be strict in engineering terms and interpretations of results.

BTW, my measurements strictly correspond to Amir's results in his review. I have just added CCIF and THD(freq) to show the worse linearity at higher frequencies.

 

[restorer-john]

Again wrong, John. The THD+N in post#1

Ah. I just looked your post above which is A-WTD:

Apologies. In the interests of like comparisons, make them either all UWTD or A-WTD, not a mixture of the two.

 

[pma]

Ah. I just looked your post above which is A-WTD:

Apologies. But in the interests of like comparisons, make them either all UWTD or WTD, not a mixture of the two.

Well, accepted. But the key is that you picked up the measurement of the preamp and not the card loopback. And it all was explained in the posts. The problem of web discussions is their shallowness, especially if many different topics are to be replied.

The main post is post #1 and the REW measurement gives all the info. Measuring bandwidth, noise, distortion, noise + distortion. Only you have to open the image and read the description. THD+N = 0.00089% (20 - 20000Hz unweighted) makes 101dB SINAD, at THD = 0.00035% (valid just for the measurement, may differ with warm-up). I think the loop SINAD 101dB cannot be beaten by finding the "sweet spot". Anyway, There are also plots of THD and THD+N vs. level in post #1. THD/level shows some interesting inflection. Most probably if it was measured after longer period of warm-up the numbers below 0.001% might be a bit different.

 

[restorer-john]

It's all good Pavel. I looked at your last ARTA plot and assumed they were all A-WTD after a quick glance.

We'll see how different my setup is with the 2i2v3 tomorrow (same settings- 1M TRS-XLR cables are my standard) My PC is sitting on the same ground/earth as all the test gear and the USB outer/0V, so I get some issues with (what I think is) USB packet noise sometimes when going single ended/unbal. 8/16kHz spikes.

Do you have any experience with high speed USB isolators?

 

[pma]

Loopback TRS ==> XLR, 44.1kHz, unweighted (BW 22kHz). OUT = 0dBFS, level adjusted by monitor volume pot to get input close to 0dBFS. THD+N slightly below 0.001%, SINAD slightly above 100dB. ARTA.

In a long term run, THD slightly varies around 0.0005%.

 

[pma]

Some more measurements .....

Headphone output
is the weakest point of this soundcard.

However, the distortion is acceptable up to some 200mV with 50 ohm load. The good point is low output impedance about 1 ohm. I think they use the ancient cheap JRC4556 opamp in the headphone amp which would explain this distortion behaviour.

Maximum output voltage
from TRS balanced output is 4.8Vrms

Noise measurements

Wideband output ultrasonic noise is about 27mVrms

It is a result of sigma-delta DAC noise shaping which starts above audible frequencies

It has however "noise shape" with no high individual spectral lines even at higher measuring bandwidth

 

[AnalogSteph]

I think they use the ancient cheap JRC4556 opamp in the headphone amp which would explain this distortion behaviour.

Close but no cigar. NJM8065 I think, i.e. the NJM4565 "reboot" - a much better load driver than the 4558 but still nothing special by modern standards, certainly for a headphone driver. Bit of a dubious choice for this position, I don't think the 0.5-0.75 mA more idle current for a 4580/8080 would have killed them (it wouldn't go much louder but distortion would have been a good bit better). Last time I checked it should also be possible to accommodate both SOP-8 and DMP-8 with a single footprint, so even the trusty NJM4556A of O2 fame may potentially have been an option even if its power hunger would have been more significant (almost 4 mA more in idle).

 

[pma]

Close but no cigar. NJM8065 I think, i.e. the NJM4565 "reboot"

Thank you! Yes the 4556A should be better than what I have measured, according to the datasheet curves. I will not bother with re-soldering, as I do not need the headphone output for anything but functional monitoring.

 

[pma]

Some more observations after 2 months of working with this card. Most (but not all) can be deduced from the THD 1kHz spectrum in post #1. The soundcard has acceptable 1kHz THD regarding its price. However, the presence of high order harmonics indicates to worse or poor high frequency linearity. This was, unfortunately, confirmed during the tests.

Acceptable for measurement for THD (and THD+N, though with reservations) 1kHz of medium- parameters amplifiers
THD and THD+N in a loopback looks like this

with my analog measuring interface MSYS (balanced I/O, stepped gain 0 - 12dB, input divider 0 or -21dB) it looks almost same

Unacceptable for THD measurements at higher frequencies. There is a strange multiple THD hump starting on frequencies of several kHz. Below is the 6kHz THD and THD+N. The measurement is not random, it is perfectly repeatable.

Acceptable for SMPTE and DIN IMD measurements of medium-parameters amplifiers.

Unacceptable for any kind of CCIF IMD 19+20kHz measurements.

The card is so-so acceptable for single sine measurements on frequencies up to 1kHz. It is so-so acceptable for SMPTE and DIN IMD measurements of average or worse than average amplifiers. It is unusable for measuring of SOTA amplifiers of any kind. The only chance, though not verified yet, might be with a notch filter to eliminate fundamental frequency.
High frequency non-linearity is very disappointing. Same applies to high intrinsic output noise which degrades THD+N measurements.

Regarding sound, again it is so-so.

 

[pma]

Focusrite 2i2 driven from Topping D10s. We can see big improvement in HF linearity, this shows that ADC part of Focusrite 2i2 is considerably better than the DAC part. Please also note that noise is also lower.

 

[pma]

One more graph with Topping D10s as a signal source - THD vs. input voltage and DIN IMD (250Hz+8kHz) vs. input voltage. Please note no IMD hump. Only rising 2i2 ADC distortion with level above 0.5V (0.7V) input voltage.
Symmetrization transformer was used behind D10s output, otherwise 2i2 ADC distortion would be considerably higher if the balanced input signal was not used. The link transformer also effectively breaks the loop created by 2 USB devices connected simultaneously and interconnected by the analog link cable.

 

[pma]

I am posting measurements of Focusrite 2i2 DAC part only, balanced output , measured with E1DA Cosmos ADC in mono mode with balanced input, 10V range. Fs=44.1kHz.

THD and THD+N vs. output voltage, at 1kHz

THD and THD+N vs. frequency, BW=22kHz, 2.76V output voltage

 

[pma]

Some more measurements of the 2i2 DAC section

CCIF IMD 19+20kHz vs. level

31 tone multitone at 1.083V level

Thank you @JohnPM for addition of IMD sweeps and multitone sweep into REW.