Focusrite Scarlett 4th Gen

[DavidR]

Next are the results for Input 0dB, output dial Maximum and the REW Generator at -3dBFS. I chose the latter a bit randomly. However, I later found that around -2dBFS will provide an optimal level to match with the amplifier in putting out 2.83V for driver testing. The distortion characteristics are similar to the -3dB results.

 

[DavidR]

I've continued testing loopback, thinking to post selected full THD graphs, but I found an option in REW to display only a selected curve of all open measurement files in an Overlay window. It helps to highlight differences. These are still all with the input gain at 0db and the output at maximum. Any change in either one of those can make drastic changes in the test results. I may post some selected graphs with changes in those, although I doubt I'll need to do that for testing distortion in drivers.

First up is an overlay of THD noise floor in dBr of the REW generator output from 0dBFS to -30dBFS.

Certainly the results won't be a surprise to most here since the noise floor should be unchanged if the two gain settings are constant, but the relative difference with the signal is interesting to see. What I don't know is how much change is simply due to the D/A output level and is there any influence due to the analog section with signal level.

 

[DavidR]

This overlay was very interesting. There's an option to specify the number of harmonic components in the THD. What is interesting is the change in each when the H2-4 is changed to H2-9. The 0dBFS and -3.0dBFS level change is barely perceptible below about 1kHz when doing it on the fly. The H2-4 has a slight droop above 1kHz, then rises in the typical way. However, changes for -20dBFS to -30dBFS show a significant rise across the spectrum, but the rise is generally more prominent as frequency drops. This is outside of the area above 9kHz that has a typical drop seen for -20dBFS and higher due to the higher harmonics dropping out of the calculation due to the 96kHz sample rate. Those big changes should be ignored.

What else is interesting is that for the measurements below -20dBFS the area from about 4kHz up has practically no change going from H2-4 to H2-9 inclusive. It seems that the lower harmonics increase much more dramatically with increasing absolute output level. I'm not sure what to make of that. It seems that if one is looking for the "sweet spot" one must examine the harmonics of most interest. I am most interested in the H2-4 or maybe H2-6 range.

 

[DavidR]

One more overlay, THD+Noise for generator 0dBFS to -30dBFS.

 

[AnalogSteph]

What else is interesting is that for the measurements below -20dBFS the area from about 4kHz up has practically no change going from H2-4 to H2-9 inclusive. It seems that the lower harmonics increase much more dramatically with increasing absolute output level. I'm not sure what to make of that.

The whole affair seems to be dominant H2 at low freqs (possibly from the circuitry between DAC and output?), changing to dominant H3 with some H2 and H5, and the new H2 seems to partially cancel the previous contribution hence the slight dip. You should find some similarities with the MOTU M2 which uses basically the same THAT6263 input amplifier IC. It's not what one would call the last word in distortion but adequate. I think this comes with the territory of running on +/-5 V supplies. (There's still the Clarett+ series if you don't like that.)

The default behavior of Nth-order harmonics should be an increase of (N-1) dB per dB level. That's why the higher orders tend to fall off rapidly in any system with little negative feedback as level is reduced. (RF tends to make extensive use of these characteristics, see intercept points.) When you're got a system with lots of feedback, you may see something different... with rising level, H2/H3 remains about the same but the dropoff towards higher harmonics becomes flatter until you get a whole zoo.

 

[DavidR]

Thanks for that detail. I still have a lot to learn to understand the intricacies of distortion. It seems like a fairly simple concept until you dig into it.

 

[DavidR]

I've been making more tests, trying to get a better understanding of the limits and what the distortion tests reveal. Some of them seem to be a bit redundant, some not clear to me of the purpose. First is a 31-tone 1/3-octave distortion response at -9.0dBFS generator output. This is a loopback, so the distortion of output and input are involved. I can't test input or output separately.

This is TD+N vs Level, similar to the previous test, but for increasing signal level.

My question on this one is the hump. Is this the ESS Hump I've seen discussed in other reviews? How does it relate to this next one?

 

[ssashton]

My EMU0404 finally quit, so I'm thinking about this 2i2 4th gen. Does anyone know if the noise floor of the analog inputs can be improved by dropping in some top quality opamps? I see the ADC is rated at 120dB DNR yet the line-in is rated at 115dB.

Also, are these balanced line inputs and outputs? Can't tell on the TRS sockets by looking.

 

[Blumlein 88]

Yeah pretty sure inputs and outputs are balanced. Don't think an op amp swap will help with noise.

 

[ssashton]

Do we know which opamps it uses?

 

[Blumlein 88]

Do we know which opamps it uses?

No, but unless it were an incompetent choice, it won't be a big factor on noise levels. So the idea you pick up 5 db of better noise from swapping op amps is very unlikely. It likely is being lost due to circuit details or possibly has something to do with leaving headroom for various gain levels. Op amp swapping won't make a difference of any substance on that.

 

[ssashton]

No, but unless it were an incompetent choice, it won't be a big factor on noise levels. So the idea you pick up 5 db of better noise from swapping op amps is very unlikely. It likely is being lost due to circuit details or possibly has something to do with leaving headroom for various gain levels. Op amp swapping won't make a difference of any substance on that.

I'm mostly thinking they had to save cost somewhere given the top quality dac and adc (which is marketable) but low price. Something like 4pcs opa1612 might be quite an expense to add on, but anything less could introduce a few dB extra noise floor, especilly if they used resistances in the 10k range on the input side. Just my complete gestimate!

 

[AnalogSteph]

My EMU0404 finally quit, so I'm thinking about this 2i2 4th gen.

IMO the one to be looking at would be the 4i4 which has a well-performing line-in that is more in the spirit of the 0404's basically measurement-grade input, not to mention a better headphone amp. (It also costs about the same inflation-adjusted.) The front inputs are fine but much like the MOTU M2 with essentially the same THAT 6263 chip, distortion gets a bit so-so towards the treble in particular. Should you be using dynamic microphones, you'd likely appreciate the better EIN when compared to the 0404 though.

It doesn't appear like there's a teardown of a 4th gen Scarlett yet, but about the fanciest opamps I'd be expecting would be the OPA1678s that are gracing newer MOTUs. And probably a 1688 in the headphone amp.

Something like 4pcs opa1612 might be quite an expense to add on, but anything less could introduce a few dB extra noise floor, especilly if they used resistances in the 10k range on the input side.

When faced with high(ish) surrounding impedances, about the last thing you want when trying to keep noise down is super-low voltage noise bipolar input opamps. OPA1612s have 1.7 pA/√(Hz) worth of current noise density (which actually is really good for a part that sports 1.1 nV/√(Hz) of voltage noise, but still on the high side absolutely speaking). OPA1642s would be a better bet under these circumstances.

 

[ssashton]

IMO the one to be looking at would be the 4i4 which has a well-performing line-in that is more in the spirit of the 0404's basically measurement-grade input, not to mention a better headphone amp. (It also costs about the same inflation-adjusted.) The front inputs are fine but much like the MOTU M2 with essentially the same THAT 6263 chip, distortion gets a bit so-so towards the treble in particular. Should you be using dynamic microphones, you'd likely appreciate the better EIN when compared to the 0404 though.

It doesn't appear like there's a teardown of a 4th gen Scarlett yet, but about the fanciest opamps I'd be expecting would be the OPA1678s that are gracing newer MOTUs. And probably a 1688 in the headphone amp.

When faced with high(ish) surrounding impedances, about the last thing you want when trying to keep noise down is super-low voltage noise bipolar input opamps. OPA1612s have 1.7 pA/√(Hz) worth of current noise density (which actually is really good for a part that sports 1.1 nV/√(Hz) of voltage noise, but still on the high side absolutely speaking). OPA1642s would be a better bet under these circumstances.

I thought the line Input of the solo, 2i2 and 4i4 are identical? Is that not the case case?

Indeed my main purpose is to measure DACs, ADCs and amps during prototyping, so I do want every bit of distortion and noise performance. I'm not bothered about mic inputs or headphone outputs.

Going a bit off topic there is a neat opamp noise calculator where you enter the current and voltage noise specs for the opamp and the resistors used around it. Then it gives a noise floor for the circuit. I'd probably use that once knowing what resistor values are in the line-in circuit.

Website of Wayne Stegall - Download Op Amp Noise Calculator

 

[DavidR]

I thought the line Input of the solo, 2i2 and 4i4 are identical? Is that not the case case?

Indeed my main purpose is to measure DACs, ADCs and amps during prototyping, so I do want every bit of distortion and noise performance. I'm not bothered about mic inputs or headphone outputs.

I thought that the 2i2 and 4i4 were the same other than number of inputs, but evidently not. The 4i4 does have more I/O, midi and fixed line. So it has fixed and variable line inputs. Due to the descriptions the 4i4 does have slightly better specs for some of the THD+N compared to the 2i2. Whether or not it's worth another $100 to get that improvement if there's no need for the other I/O of the 4i4 is the other factor. What I'd like to know is the 2i2 variable line input spec for 0 gain rather than the specified for 8dB input gain. Is it the same as the 4i4 fixed line input? Even so, it seems that the 4i4 has marginally better line out THD+N as well.

If anyone has info on the 2i2/4i4 variable line input distortion at 0dB gain, please chime in. This was loopback, so the gain of the output noise was a factor. I've tested the 2i2 at 0dB gain for most testing, but made one at 8dB gain (the way the 2i2 spec is published) and the result was definitely worse. So it makes me think that the variable input at 0dB may be equivalent to the fixed gain.

 

[Blumlein 88]

The quoted specs are good not great for this on the ADC side. The 4i4 is quoted as better by a few db than the 2i2 in the 4th gen. The Topping audio interface was tested by Amir and had better results on the ADC side.
https://www.audiosciencereview.com/...ds/topping-e2x2-audio-interface-review.49127/

Now if you don't care about mic inputs something like the RME ADI 2 Pro FS is about as good as you'll get, but cost about 8 times more. Another route would be using the Cosmos ADC of which there are several threads here on ASR. You could put together a measurement rig with near SOTA performance for not a huge amount of money.

@ssashton

 

[ssashton]

Thanks for the info. The line input -115dB DNR is the appealing part for me, as I'm currently playing with my own DAC design. I am using my BehringerMods upgraded DEQ2496 as an ADC with optical to my PC (since the emu0404 died). It gives me -110dB SNR but my DAC board is exceeding that so further optimisation is limited. If the 2i2 could hit the ADC chips spec's 120dB DNR with an opamp swap, that would be really cool.

The 4i4 is still missing digital I/O which really sucks.

The Cosmos is probably the way to go. Maybe I'll get the ADC and add more boxes as I go. I was a bit put-off by needing to use Asio4All rather than dedicated drivers, but I've been using Asio4All just now for the digital input on my laptop and it seems okay.

 

[AnalogSteph]

What level of performance evaluation are you looking for?

If you want to take entire "legit" THD+N / IMD sweeps at super wide bandwidths --> Cosmos ADC + Scaler + associated cabling
If you just need a good static dynamic range measurement @-60 dBFS or the like and janky is fine --> you just need some sort of preamp with non-potato EIN to strap in front of the DEQ, a cheap used Mackie VLZ4 series or Behringer mixer will do (I reckon my 402 VL4 would make a super clean mic preamp if one could just bypass the tone stage)

Unless you also happen to require an audio interface for other purposes, I can't see a Scarlett making too much sense for you. The 2i2 doesn't have measurement-grade distortion performance, and by the time you're getting to the 4i4 you're paying as much as for the (better) E1DA-based solution.

 

[ssashton]

Thanks that's a good idea, I actually have a spare PCB with some lm4562 on it that I could hook up as a gain stage at the front of the Behringer Input.

I do like the looks of the Cosmos gear though.

I'm curious why comments about the input quality of the 2i2 are not particularly favourable here, isn't it actually a good 6dB better than my old EMU0404 that seems to be respected as a hobbyist solution?

 

[Blumlein 88]

Thanks that's a good idea, I actually have a spare PCB with some lm4562 on it that I could hook up as a gain stage at the front of the Behringer Input.

I do like the looks of the Cosmos gear though.

I'm curious why comments about the input quality of the 2i2 are not particularly favourable here, isn't it actually a good 6dB better than my old EMU0404 that seems to be respected as a hobbyist solution?

Are you looking solely at dynamic range? I've not looked but seem to remember the EMU was lower distortion.