Focusrite Scarlett 2i2 (4th Gen) Interface Review

[torgeirs]

But I also see cirrus have a new preamp-adc chip that is maybe more difficult to mess up with lowcost analog components
Hopefully it will find its way into som interface toghether with those lowcost superlow distortion dac headphone driver solutions out there.

 

[restorer-john]

I think it is possible, but that's two boxes to buy and manage. I haven't checked the prices, but I bet the two combined approach the cost of a QuantAsylum QA403.

Tom

The Quantasylum QA403 is a single box analyzer. Sample perfect sync A/D and D/A, with precisely calibrated input and outputs across a wide range.

200k/100k bal/unbal consistent across all ranges for the input.
up to ~15V output (bal).
Up to 384k FS on the A/D side, 192k for D/A.
Fully isolated- something the APs are not.

The product, the software and support are truly excellent. I've retired virtually all my legacy analyzers (and I have a few...) as the QA (as I like to call it) has obsoleted them all. And it's the same size as a Focusrite 2i2 for scale.

 

[tomchr]

More like 200$ for ADC and 50% to 150$ for DAC. My dac gives only a a few dB better signal and is overkill

As @restorer-john points out, you lose the clock sync between ADC and DAC when you put them in separate boxes, unless the boxes are fancy enough to take a 10 MHz house reference or something similar, which they aren't. That can limit your measurement capability.

So $350 vs $600. For the additional $250 you get something that works well across a relatively wide range of input voltages, frequencies. With the external sound cards you get one sweet spot for the DAC and one sweet spot for the ADC. Maybe the will line up. Maybe they won't.

Guess many will put those parts inside som old lab wreck and put some protection cirquitry around it

I'm guessing you mean an old lab rack. Sure. You can do that. But now you're talking $350 for the ADC/DAC boxes, plus the cost of the rack, plus the work, additional circuitry, etc. Versus just spending $600 on something that works.

I'm not here to say how you should spend your money or your time, but I don't see the logic in cobbling something together when a known good solution is not much more expensive. In my general life experience I have always spent more money buying cheap tools than it would cost to just buy the right tool in the first place. It usually starts with, "oh I only need to use it a few times on this one task, so I'll buy the cheap tool". Then I'll end up using the tool more than I thought, hating it the whole time. Then I go out and buy the real tool that I should have bought to begin with. I'm trying to learn from this and buy quality tools from the start. But that's me.

When I started to realize that my audio business had potential, I bought an APx525. At the time it cost about $12k. I lost sleep for a few days after spending that amount of money. But my circuits were too good for the '525. I really needed the '555. I tried for years with external filters, precision oscillators, etc. to get close to the performance of the APx555. And I got close. At one voltage/frequency combination I could measure down around -145 dBc THD, but not reliably and not across a range of voltages, frequencies. So I ended up buying the APx555 when I finally had a really strong year. By then AP had come out with the B-series, which is a smidge better than the original.

Fully isolated- something the APs are not.

True. Though I've never had issues with that on the APx555. I would be exceptionally surprised if AP hasn't thought about noise from PC power supplies entering via the USB cable when they designed their analyzers. They've always relied on PCs as the number cruncher so this isn't new to them.

I've retired virtually all my legacy analyzers (and I have a few...) as the QA (as I like to call it) has obsoleted them all.

I still hang on to my HP 3563A dynamic signal analyzer. It's handy for measuring infrasonic stuff like DC servos. The 3562/63 will measure down to 64 µHz if I recall correctly. That said, it's been quite a while since I last fired it up. Same can be said for my HP 3577 network analyzer. That one is handy for loop gain measurements. It's bulky, heavy, and loud so I've been thinking to replace it with one of the many lower cost VNAs available.

Tom

 

[torgeirs]

Thanks for shearing. Luckily the separate solution can be used for other general purpose audio if the thd is to high for the measured circuit.
Then a 600$ box is well earned

The sync issue is interesting
The ADC-iso is sometimes sold as a unboxed card for 149$
Then a I2S header is exposed. So then it could be synced to a i2s dac of choise
The lab wreck is often given away. Or laying there collecting dust.
But the bnc connectors and case are often still good

 

[tomchr]

Then a I2S header is exposed. So then it could be synced to a i2s dac of choise

Maybe. That's an assumption that you make.

I'm not suggesting that you can't do any of the things you say. I'm not here to judge how you spend your time or your money. I just don't see the logic in buying $350 worth of DAC/ADCs, having to find some lab rack to put them in, cobble things together, hack into the I2S, yada-yada when I can buy a calibrated solution that works well for $600. Note also that the QA403 has a built-in attenuator that's controllable from the software. That's handy for power amp measurements. You'd have to add that to your solution too.

For a cheap one-box solution I think the Scarlett 2i2 is a good choice. I'm able to get pretty decent performance out of it with balanced connections. It doesn't work so well with unbalanced connections, but oddly the 3rd generation Solo works for that. Go figure. I'll share some data in a while.

Tom

 

[torgeirs]

I both agree and disagree.
The QA403 seems like a nobrainer if you spend large amount of time analysing analog cirquits and have 600$.
But access to the state of the art ADC audio chip with MELF resistors on input and design that dont degrade the ADC spec for 179$ is not a bad deal either.
It will take a lot of effort to make a pro grade measuring system around it. But for a hobbyist or on the move solution it is great.
And it doubles as a digitalizing marvel

 

[torgeirs]

One function on REW that is underrated is the stepped measurements. By looking at the harmonics and not the noise the sweetspot if the DAC measured can be found. Se a lot of dacs that has lowest distortion dBr long before 0dBFS
The measurements takes hours though.
0.2 db increments and a lot of averaging and wide FFT

 

[torgeirs]

And by adjusting the DAC output a few DBs a low for each harmonic can be found and used for single freq harmonic measurement

 

[torgeirs]

But for single tone measurements the THD issue is solved for those putting in the effort

Digital Distortion Compensation for Measurement Setup

Software Distortion Compensation for Measurement Setup Hi, I am thinking about options for some sort of digital compensation of harmonic distortion within the measurement loop (DAC -> ADC). E.g. ESS DAC chips offer second and third harmonic distortion compensation. I have played with virtual...

www.diyaudio.com

Guess a scarlett would be good enough for AD/DA part of setup

 

[Xyrium]

They're not designed as a piece of HiFi and they shouldn't be tested as that. It has to accommodate wildly varying input levels on the analog side, balance them and then drive (usually) a pair active monitor speakers, an amp, or into headphones. As such you never would max out the monitor output. I have a few of the previous models here and lowest distortion and noise is with the inputs significantly retarded for line level inputs and the output set around 2 o'clock to give rated and lowest THD+N even right up to 0dBFS from the digital source (USB/PC)

But this 4th gen has a totally different front end AFAIK. I wouldn't expect the operation to be greatly different- it is after all the most popular ongoing interface in the world...

Tough call IMO. Since they are used to record the stuff a consumer would be playing back, why wouldn't we want to ensure the highest quality input ADC to ensure the best consumer or pro output?

I have the Clarett 4 Pre, the headphone output sucks, the overall headroom isn't amazing, but the sound is....fine. I have two guitar rigs and an e-drum kit hooked up to it. Focusrite makes decent stuff, just not in line with RME, or perhaps even MOTU, IMHO.

 

[torgeirs]

I have the RME babyface first generation.
This is aparently made to measure the bestregarding THD between -25 dBFS and -5sBFS for the DAC
Right in the sweetspot where i have used it for years. I have used the nice big wheel as digital volume control all the years.
Measured with ADC-iso

 

[torgeirs]

As of the non synced issue with seperate USB DAC and USB ADC for measuring.
Hvere does the problem manifest itself? Do not REW compensate good enough?
This is a measurement from another measuring site. (ADC so need to be isolated. That problem should go away with ADC-iso)

EARLY LOOK (Part I): E1DA #9039S USB Balanced DAC/headphone amp - Super Hi-Res, Tiny Package! Also, let's tweak...

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

The normal audiomeasurements seems quite ok, like my own.
Example of 2 days test of RME babyface stepped response

A few points got noisy, but it only impair the visual.

I understand the problem if old school low bin count FFT is used.

 

[tomchr]

So getting back to the 2i2. I got my paws on a 2i2 Gen 4 and put it through its paces on the APx555B. The sweet spot seems to be with +10 dB input gain and 1.6 V RMS input amplitude on the balanced input. That's similar to what @NTTY found.

TLDR; I would not recommend the 2i2 Gen 4 for audio measurements. The performance with balanced inputs is fine, but the performance (or lack thereof) with single-ended input isn't up to snuff.

First some data from measurements on the ADC input.

That's not really awful. It's a bit disappointing with 25 dB degradation in THD from 1 kHz to 20 kHz, but the 1 kHz THD is quite good.

The limiting factor is really the ultrasonic noise. That makes 192 kHz sampling rate useless, which limits the maximum test frequency to about 15 kHz if you want at least two harmonics (H2 and H3) captured.

Here's the DAC:

The DAC shows the -107 dB THD that can be expected from a half-decent (but not yet high-end) DAC.

Loopback isn't awful either.

It doesn't make me go, "ooh! aaah! Must have one!!" but I think you get your $180-200 worth out of it ... at least with balanced input.

It does fall apart with single-ended input, however. Here's a measurement of the 2i2 Gen 4 with a single-ended connection to the APx555B source.

That's just awful. The mains hum (and other garbage) stems from the PC power supply. It gets in via the USB ground. Adding a USB isolator gets you this:

That's still not great. Basically -80 dB THD+N at the sweet spot (160 mV RMS, +30 dB input gain) with a single-ended input. This was confirmed by @NTTY as well. Most audio sources, especially in the consumer segment, have unbalanced (single-ended) inputs and outputs. The 2i2 Gen 4 would be useless for measurements on such circuits.

Interestingly the third generation of the Scarlett series performs much better with a single-ended input. You can see my measurements on the 3rd Gen Solo here: https://neurochrome.com/pages/measuring-distortion-on-the-cheap. Or you can watch here:

Tom

 

[NTTY]

Thanks Tom,

Your results are indeed consistent with mine, so that means the Gen4 improves a little compared to Gen3 as long as we use balanced inputs.

To mimic the usual view of Amir, reusing data he collected for other interfaces, and as per our measurements (balanced in/out), we have for the DAC:

And for the ADC (from 1.4Vrms to 4Vrms input):

Cheers

 

[AnalogSteph]

Adding a USB isolator gets you this:

That's still not great. Basically -80 dB THD+N at the sweet spot (160 mV RMS, +30 dB input gain) with a single-ended input. This was confirmed by @NTTY as well. Most audio sources, especially in the consumer segment, have unbalanced (single-ended) inputs and outputs. The 2i2 Gen 4 would be useless for measurements on such circuits.

'fraid so - clearly the THAT626x was not designed with such an application in mind. You can get rid of the ground loop noise by impedance-balancing an unbalanced output, but that still doesn't make the signal voltages symmetrical (BTL). I suppose it's still fine for microphones with slightly compromised circuitry. Do you reckon you could run a THD(f) like that? If the H2 remains fairly constant, it would be audibly inconspicuous at least.

The 2i2 may be the least attractive 4th-gen Scarlett for device measurements. There still is the Solo if you want the "classic" preamp, or the 4i4 with its additional fixed line-level input.

 

[tomchr]

'fraid so - clearly the THAT626x was not designed with such an application in mind.

Do you know for a fact that the 2i2 Gen 4 uses the THAT626x? I tried to get some information on the circuitry in the shielded can inside the 2i2, but couldn't tease that out from a quick online search. It's possible that I didn't dig deep enough.

You can get rid of the ground loop noise by impedance-balancing an unbalanced output,

Hmmm... I'm not so sure. With a single-ended connection the ground connection becomes part of the signal path, whereas for a differential connection it isn't. With the relatively low levels of THD+N we're measuring here it doesn't take much error current to add to the +N.

Or are you suggesting adding the balancing impedance from the RCA shell to the TRS ring (IN-) with the RCA shell going to TRS sleeve (GND) as well? RCA centre would go to the tip of the TRS jack (IN+).

Do you reckon you could run a THD(f) like that? If the H2 remains fairly constant, it would be audibly inconspicuous at least.

The 2i2 is on its way back to Bezos. Sorry.

I did try inverting the input, i.e., connecting the signal to IN- and GND with IN+ grounded. Same depressing results.

The other thing that's odd about the 2i2 inputs when used single-ended is that the sweet spot for THD+N moves to 160 mV from 1.6 V. As I mention in the video, I wonder if the 2i2 has an 'intelligent' input that switches over to instrument mode when it senses a single-ended connection. It doesn't light up the INST indicator on the front panel, but it sure behaves as if it was in instrument mode.

The 2i2 may be the least attractive 4th-gen Scarlett for device measurements. There still is the Solo if you want the "classic" preamp, or the 4i4 with its additional fixed line-level input.

It looks like the Solo Gen 4 has the same preamp as the 2i2 Gen 4. At least that would be my guess. Same for Solo Gen 3 vs 2i2 Gen 3. I'll know soon. A friend of mine just picked up a 2i2 Gen 3 and he'll run the loopback test on it.

I don't think there's any doubt that you get more, better circuitry with the devices that have higher input count. The 4i4 is $350ish. That gets into the "why not just get the QA403?" territory. Unless you plan to use the 4i4 as an audio interface instead of or in addition to using it as an audio analyzer.

Tom

 

[bennybbbx]

This is a review and detailed measurements of the Focusrite Scarlett 2i2 Fourth Generation audio interface (DAC, ADC, Mic Pre and Headphone out). It is on kind loan from a member and is on sale for $168.
View attachment 409562
I have a soft spot for the red color of the enclosure. It gives it brand identity which is a value to the company as well. I also like the standard Line input rather than some combo with microphone in:
View attachment 409563
Only a USB-C cable is provided so that is what I used for measurements (no 5 volt adapter).

Focusrite Scarlett 2i2 4th Generation DAC Measurements
Let's start with our usual dashboard after setting the output volume level to 0dBFS (it goes up beyond it):
View attachment 409565
Well, this is a headscratcher. We are miles away from company spec of 109 dB for SINAD. It is distortion bound so not much that can be done about it (or so I thought). This causes the ranking to be at the bottom of all interfaces tested:
View attachment 409566

By accident, I realized that if I set the volume to max (+6 dB?) but then attenuate the level in the source (analyzer output) to still get 4 volts, I get much better results:
View attachment 409567
How could this be? This is inverted as usually distortion is proportional with output level. Here, maxing out the output lowers distortion! And how would the customer know to drop their max level to -2 dBFS? It makes no sense to me.

Anyway, going with max volume, we can see the ideal output is indeed 4 volts although the penalty for max volume of 5 volts is not much at all:
View attachment 409568

IMD vs level shows the difference in the two volume settings:
View attachment 409569

Same for THD+N vs frequency:
View attachment 409570
Above, we also see the classical issue with some DAC ICs using noise shaping as reducing bandwidth helps a lot.

EDIT: here is the multitone:

View attachment 409795
Dynamic range is not impacted by volume setting:
View attachment 409576

For these other tests, I didn't want to redo them:
View attachment 409571
View attachment 409572
View attachment 409573
View attachment 409574

This one would likely improve with max volume:
View attachment 409575

ADC Measurements
Here is our ADC dashboard:
View attachment 409577

Input saturates at 1.7 volt. From my review of first generation Focusrite Scarlett 2i2, lowering the input gain below 0 dB helps. But I expect it to calibrated to 0 dB since we have real line inputs. Company's spec of 100 dB is not great either as it barely clears the bar for 16 bit content.

Dynamic range thankfully is respectable:
View attachment 409578

Frequency response is excellent as it should be:
View attachment 409580

I can't figure out how the 2i2 is beating the others in noise but it does:
View attachment 409581

But per above, it saturates very early. Maybe that is the difference as we are looking at generator level rather than actual output level.

Linearity is good enough:
View attachment 409584

Wideband distortion vs frequency is not professional level:
View attachment 409585

Headphone Output Measurements
Company highlights the design of headphone amplifier but what I see is well below my standard for power output:
View attachment 409586
View attachment 409587

Conclusions
I don't understand how the pipeline in 2i2 works. As is, it seems backward in the way distortion decreases at higher internal volume levels. Company specifications needs to indicate this rather than just giving a single number. The distinction takes the internal DAC from lousy to competent so big difference. The ADC is "good enough" for a consumer level product, not professional Headphone output as expected, is just a checklist item, being beat by some low cost dongles!

Overall, I would have wanted to see significant improvements in a 4th generation audio interface. Sadly the Scarlett 2i2 does not deliver on that front so I am not going to recommend it.

------------
As always, questions, comments, recommendations, etc. are welcome.

Any donations are much appreciated using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/

I think there should also test if the left and right channels bring same db output at volume 9 o clock, 12 o clock and 3 o clock. because often hear with level of 8-9 oclock. I never guess that there is at 9 o clock volume level a volume diffrence of 0,7 db between left and right in the scarlet 8i6o . this is lot hearable for me . in the past i guess it is because the t.racks get too few input and so this less clarity sound happen. 0,2 db between left and right i think ok.
i connect the output of 1 and later 2 to input 1 (and use no t.racks for this test) and measure with diffrent position of the volume knob. this are the results i get. really worse i think. maybe because nobody test that, other interfaces are even more worse. pot unlinearity can be a big reason that think a expensive DAC sound better. see the db levels in the 2 lines with the checkboxes between left and right. for the 3 oclock volume i reduce the measure level so be sure no clipping happen. what make things even more worse , there measure speakers with volume set to 9-10 o clock and so have 0.7 db diffrences between left and right measure which is not there

 

[AnalogSteph]

Do you know for a fact that the 2i2 Gen 4 uses the THAT626x?

I am reasonably certain, yes, a THAT6263 to be specific. So pretty much like the MOTU M2. It's not like there are too many options for digitally controlled preamps in a $200ish (US) 2x2 interface with +/-5V supplies to begin with. Using PGA2500s would take up a lot more space and make things substantially more expensive, you don't generally see those below $500ish interfaces.

Or are you suggesting adding the balancing impedance from the RCA shell to the TRS ring (IN-) with the RCA shell going to TRS sleeve (GND) as well? RCA centre would go to the tip of the TRS jack (IN+).

This, although even having no balancing impedance should provide some CMRR over an all-unbalanced connection with a TS plug.

The other thing that's odd about the 2i2 inputs when used single-ended is that the sweet spot for THD+N moves to 160 mV from 1.6 V.

I would chalk that up to missing cancellation of even-order harmonics. Given that SE results are dominant H2 by a long shot and this harmonic scales the slowest with signal level, this would have a major impact on distortion levels.

It looks like the Solo Gen 4 has the same preamp as the 2i2 Gen 4.

The Solo has classic analog gain control and is literally advertised as using "The classic Scarlett mic pre design.", so this seems unlikely.

The "classic" design seems a bit short on GBW and I won't be surprised if (treble) distortion performance towards the high end of the gain range isn't much to write home about, but feel free to prove me wrong (I don't think we have any hard data on that).

 

[skyfly]

In the graph below "Here is our ADC dashboard:"
I see "Gain 0dBFS."
As I understand it, we don't use "FS" (full scale).

 

[tomchr]

The Solo has classic analog gain control and is literally advertised as using "The classic Scarlett mic pre design.", so this seems unlikely.

If the Solo Gen 4 has a potentiometer for the volume control and uses the same circuitry as Gen 3 then that would be good news for those looking to measure as it would likely have the same preamp as Gen 3 but possibly a better ADC and/or DAC.

Anyway. Focusrite posted a comment on my review video saying that they're aware of this thread an of the issues with single-ended performance of the 2i2 Gen 4. They'll go and investigate. So maybe they'll have a fix. Who knows?

Tom