Focusrite Scarlett 2i2 (4th Gen) Interface Review

[NTTY]

Wow. Thank you so much for taking the time and patience to gather this data. This confirms that the volume pot is behaving weirdly by design.

Forgive me if I’m asking too much, but would you be able to take a look at these test signals on the new 2i2?

https://reference-audio-analyzer.pro/en/demo-cs131.php#gsc.tab=0

The test file is supposed to reveal the presence of clicks in a spectrum analyzer, such as the one found in Audacity.

Sure, here you go with the flac played by the Focus and recorded with the Motu (in Audacity) @24bits/96kHz. No dither applied.
I can't hear any clicks. And I believe I inverted left and right, sorry, but you can change in Audacity.

Alternatively (AFAIU), it’s possible to detect the clicks by running a multitone test and slowly decreasing the input level. If the harmonics start "bouncing around" at some point, it means clicks are present.

In the video below, the right graph illustrates the "bouncing" starting from the 0:25 mark.

I’m only asking because the gen4 Scarletts utilize the CS43198 as a DAC chip, and I’ve personally verified that clicks are audible on its "sibling" (CS43131—see post #21). I think having this issue in a professional product would be very concerning.

That being said, I'll stop asking now since people don't seem too concerned about this issue.

I ran the same test and I could repeat the issue as shown in the video when reaching -18dBFS. This is not permanent though. It will do it for few seconds only, but that's back regularly. Weird!

No video for me but three recordings overlaid (200Hz to 5000Hz for better viewing):

These three are taken at -18dBFS. So the noise bounces between the two limits (green and orange) as shown, that's roughly 5dB. The red is one more in between that I captured.

Now, at -19dBFS things get suddenly worse:

What is that sudden increase of the noise floor? It continues to be elevated like that even when decreasing the digital output (I stopped the test @-30dBFS).

 

[NTTY]

I’ve found for dynamic range measurements it is best to use the UL Mk5 front panel XLRs. I’ve been able to measure ~120 dB DR at 4 V output. This seems to work better because the Mic input has a much lower input sensitivity than the TRS line inputs which is perfect for -60 dBFS DR test.

Michael

Thanks, I did as you said and measured the same 110dB (and indeed with -60dBFS test tone undithered), which probably means the Motu was not the limit.

 

[tccalvin]

Sure, here you go with the flac played by the Focus and recorded with the Motu (in Audacity) @24bits/96kHz. No dither applied.
I can't hear any clicks. And I believe I inverted left and right, sorry, but you can change in Audacity.


I ran the same test and I could repeat the issue as shown in the video when reaching -18dBFS. This is not permanent though. It will do it for few seconds only, but that's back regularly. Weird!

No video for me but three recordings overlaid (200Hz to 5000Hz for better viewing):

View attachment 409951

These three are taken at -18dBFS. So the noise bounces between the two limits (green and orange) as shown, that's roughly 5dB. The red is one more in between that I captured.

Now, at -19dBFS things get suddenly worse:

View attachment 409953

What is that sudden increase of low level noise?

Many thanks.

This is the file you provided:

The clicks are clearly there—they’re the vertical lines extending to 48kHz, even beyond the Scarlett’s reconstruction filter. I set the spectrum analysis to only show content down to -96dBFS, so they fall within the audible range. I visually boosted the gain to make them easier to see; the output isn’t actually that noisy in reality. Unfortunately, I couldn’t listen to the output to confirm audibility because all the DACs I own exhibit audible clicking.

The reason you see the noise "bouncing" is that impulse-type signals (our clicks) behave just like noise: they contain all frequencies, even though they only manifest for a fraction of a second. Because they’re so brief, it’s hard to judge their amplitude using real-time FFT since it’s limited by the program’s frame rate.

IMHO, this might be the biggest flaw of these Cirrus Logic chips. They perform exceptionally well in standard tests, but when subjected to more "esoteric" ones, they fail in this way. Chips from ESS and AKM, for whatever reason, don’t seem to exhibit the same problem.

 

[MaxwellsEq]

Sure, here you go with the flac played by the Focus and recorded with the Motu (in Audacity) @24bits/96kHz. No dither applied.
I can't hear any clicks. And I believe I inverted left and right, sorry, but you can change in Audacity.


I ran the same test and I could repeat the issue as shown in the video when reaching -18dBFS. This is not permanent though. It will do it for few seconds only, but that's back regularly. Weird!

No video for me but three recordings overlaid (200Hz to 5000Hz for better viewing):

View attachment 409951

These three are taken at -18dBFS. So the noise bounces between the two limits (green and orange) as shown, that's roughly 5dB. The red is one more in between that I captured.

Now, at -19dBFS things get suddenly worse:

View attachment 409953

What is that sudden increase of low noise floor? It continues to be elevated like that the when decreasing the digital output (I stopped the test @-30dBFS).

There's some speculation about the DAC having a Class H setting or a Class AB setting. The latter doesn't exhibit this behaviour, I believe.

 

[tccalvin]

There's some speculation about the DAC having a Class H setting or a Class AB setting. The latter doesn't exhibit this behaviour, I believe.

That's what's stated in the RAA page. We still haven't been able to verify that claim.

 

[NTTY]

Testing the inputs is much longer, sorry I won't have the time today. Basically it's better to keep 3dB headroom at the input, as it improves things, I'll try to report so that it's not too boring to read.

In the meantime, just for info, this is a loopback (analog out, line in, TRS) on my Motu Mk5 with the output set at roughly 2Vrms (8.24dBu) and input with gain increase (+15dB) so that to reach near 0dBFS at the input:

And now if I send the 2Vrms output of the Motu to the Focusrite (Line input) adjusting the gain to reach near 0dBFS too:

Noise floor is lower but distortion is much higher.

Things get better with input gain lowered by 3dBFS which sounds reasonable per planned utilization. But it’s nowhere near the Gen3. Something might escape me (user manual says to aim for -12dBFS at the input…). More tests to come anyways.

 

[CedarX]

There's some speculation about the DAC having a Class H setting or a Class AB setting. The latter doesn't exhibit this behaviour, I believe.

The Class AB /H selection is documented in Cirrus’ datasheet. I believe Class H is the default used by many OEMs, the reason being the lower power usage and the claim by Cirrus that these clicks are too short to be audible. I have a dongle where you can choose Claas AB or H.

 

[tccalvin]

The Class AB /H selection is documented in Cirrus’ datasheet. I believe Class H is the default used by many OEMs, the reason being the lower power usage and the claim by Cirrus that these clicks are too short to be audible. I have a dongle where you can choose Claas AB or H.

Based on my experience, I agree that you won’t hear the clicks most of the time. The presence of this issue doesn’t make Cirrus Logic devices unusable—not by a long shot.

That said, while the audibility of these clicks is highly situational, I’d argue that it’s still more noticeable than subpar SINAD numbers. On specific signals, you might hear a “crackling” noise, similar to a guitar amp starting to clip or the sound of an old turntable.

For most other signals, you likely won’t notice the clicks audibly, but they’ll still be embedded in the DAC’s output (they're very frequent). This is worth keeping in mind if, for any reason, you need to record a clean analog signal from that output.

 

[CedarX]

Based on my experience, I agree that you won’t hear the clicks most of the time. The presence of this issue doesn’t make Cirrus Logic devices unusable—not by a long shot.

That said, while the audibility of these clicks is highly situational, I’d argue that it’s still more noticeable than subpar SINAD numbers. On specific signals, you might hear a “crackling” noise, similar to a guitar amp starting to clip or the sound of an old turntable.

For most other signals, you likely won’t notice the clicks audibly, but they’ll still be embedded in the DAC’s output (they're very frequent). This is worth keeping in mind if, for any reason, you need to record a clean analog signal from that output.

Definitely a concern in the case of an interface like the Scarlett, even if it has no “pro” aspirations.

 

[Waxx]

That is about the same specs as my 12 years old Steinberg UR22 (made by Yamaha). That should be better by now, especially with the modern chips around. Steinberg still sells theirs (now the MKII, but almost the same design) and it still sells a lot.

But today you got better, a Motu M2 has a similar price but is a lot better. That is what i would get now.

 

[kencreten]

I have a second gen Focusrite interface... I run everything through it. Sounds fantastic. I'm also lucky in that the music I create... at the outside might have 10db of dynamic range.

 

[AnalogSteph]

That is about the same specs as my 12 years old Steinberg UR22 (made by Yamaha).

Yeah nah, as folks down under would say. The 2i2 gen4 can hit a higher SINAD than the old Steinberg manages in terms of dynamic range on the output side. It'll also handle higher mic input levels than the first-gen UR22 can even dream of (that still sported a lot of rough edges, including being relatively easy to damage when hotplugging mics with phantom power on), and sports lower input noise. The Steinberg preamps are good in terms of GBW and distortion at high gain, I'll give 'em that.

I will admit that the 4i4 gen4 is more like it if you want to experience the full capabilities of the converters used, it's got a number of upgrades like an additional fixed line-level input and a lower-impedance headphone output. I don't know why they stuck with analog volume for the 2i2, this makes that circuitry a bottleneck for the DAC.

 

[Waxx]

I mean it's in the same league, which is a shame for a device in a sector where tech moved on and better (altough it already existed) became a lot cheaper and easier to build. This 2I2 is probally good enough for 95% of the users and it's cheap, but so is the Motu M4 and many others that do a lot better for the same price. For today's standards i think this is bad.

btw, i still use my UR22, it does the job. But for my main systems i got better dac's.

 

[DanTheMan]

In truth, looking at Julian Krause's tables, this little guy actually performs really well barring its headphone amp. It's in the upper end of performance in pretty much every other way--pretty close to the Topping, SSL, Motu, Audient, etc... None of them are absolutely the best. Then you look at its haptics: top of the pile IMO barring maybe the SSL (which has a better headphone amp for certain). The reported longevity and support from Focusrite on this line of product actually makes this a pretty good buy if you intend to use it for a long time, and you should. You could combine it with power-sipping closed back headphones like the Meze with Brainwavs pads and really have something nice for the long haul. The elevated bass on the Meze is actually quite useful for recording purposes as it will alert you to unnecessary power wasting from thunking keys, foot stomps, mic pops/plosives, etc. For a musician/home recording-type, I'd put this on my short list of possibles. No question about it.

 

[ampguy]

Nice Amir, I hope you get a chance to test a UA 276 or 176 soon

 

[Vigovsky]

I think Steinberg makes good sound cards.
Sometimes, I used the youngest UR12 model with one input channel in 2016-2018. I made a transformer balanced output
for it in a Direct Box Art Audio Pro case. I completely changed this Box inside and installed Swiss Neutrik NTE-1 transformers.
I measured through the linear/balanced XLR output (NTE-1) and on the Steinberg XLR microphone input.
THD = 0.0007% ( - 6 dBFS ) 24/96kHz. It is attached. The sound card was connected via USB to a PC.

Later I used Neutrik NTL1 in different projects, NTL1 transformers are much better.

 

[NTTY]

Hi everyone,

I’ve been too busy this week to review the inputs, but I have a good set of tests now. I can say for instance that XLR (Mic) input at the back will saturate at 16dBu (4.88Vrms), as per the specs. I think Amir saw it saturate at 1.7Vrms (7dBu) because the input gain was at 9dB, which I kind of recall was the way they were adjusted after installation of Focusrite Control 2 software.

In the meantime, I re-wrote my review of the line outputs. I included tests with unbalanced cables (TS) on top of balanced (TRS). You might want to have a look because performances are lower with TS and it’s informative to know.

I’m preparing a review of the three types of inputs we have here (even if in the end, it’s really two simultaneous inputs only). The 2i2 Gen4 can also be used in stand alone mode, as a preamplifier (not connected to a PC), I’ll test that too.

 

[AnalogSteph]

In the meantime, I re-wrote my review of the line outputs. I included tests with unbalanced cables (TS) on top of balanced (TRS). You might want to have a look because performances are lower with TS and it’s informative to know.

Using TS plugs effectively means that one half of the output stage has to drive 50 ohms. This could be quite detrimental to performance when two halves of a dual opamp are being used.

If you want to get "proper" unbalanced output, the cabling required will be a bit more complex, e.g. TRS to 2x RCA -> 1x RCA-RCA coupler -> RCA to TS or XLR, whatever you need.

 

[NTTY]

Using TS plugs effectively means that one half of the output stage has to drive 50 ohms. This could be quite detrimental to performance when two halves of a dual opamp are being used.

If you want to get "proper" unbalanced output, the cabling required will be a bit more complex, e.g. TRS to 2x RCA -> 1x RCA-RCA coupler -> RCA to TS or XLR, whatever you need.

Yep, but the user manual is not that specific:

Speaker Outputs R and L - 6.35mm (1/4") jack (TS or TRS) sockets to connect your Scarlett to speakers or an amplifier. We recommend you use 6.35mm (1/4") TRS jack cables for balanced connections.

I can tangibly recommend to stick to TRS too

That said, I could try with front input of my Motu that has a way higher impedance load than line input at the back. It could make a difference.

EDIT:

2Vrsm out via TS from the 2i2 to Motu Line 3 input (10kohms) vs Line 1 (1Mohm):

 

[bachatero]

@amirm the ADC section is lacking the comparison chart, just wanted to make sure it's included