Focusrite Scarlett 4th Gen

[AnalogSteph]

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

From @Julian Krause:

The dynamic range is there but distortion performance isn't. Even if we assume you only need +14 dBV (4 Vrms), -80something dB is not exactly measurement grade.

For comparison, here's the 4i4:

Front input is obviously identical to the 2i2. Rear is approaching chip limits (CS5381 spec: -110 dB @ -1 dBFS) but is fixed level so you have no way of shifting DR around without an external preamp.

 

[DavidR]

For comparison, here's the 4i4:
View attachment 396546
Front input is obviously identical to the 2i2. Rear is approaching chip limits (CS5381 spec: -110 dB @ -1 dBFS) but is fixed level so you have no way of shifting DR around without an external preamp.

Doesn't this clearly indicate that the issue is in the analog input circuitry? The front input has a gain stage. In my loopback tests I had thought that the issue was with the generator, but this makes it appear that the generator output is good, the variable input not so much. It also raises another question that I had dismissed. That is, could the "better" opamps improve the upper range? Since objective measurements could demonstrate that aspect, it wouldn't be based on the subjective.

 

[AnalogSteph]

Doesn't this clearly indicate that the issue is in the analog input circuitry?

Obviously. Like I said, much the same as for the MOTU M2.

It also raises another question that I had dismissed. That is, could the "better" opamps improve the upper range?

Given that the 4i4 back line-in bottoms out around -105 dB THD+N right now and that's still a bit short of chip specs, I would guess so. Only for that input, of course.

There's not much that can be done for the front inputs, it seems to be a limitation of the preamp IC. This is a relatively compact, very low-power 2-channel part at 6 mA idle specifically designed for use on +/-5 V supplies so it even features some negative gain - by contrast, the classic PGA2500 as used e.g. in the MOTU Ultralite Mk5 and other higher-end interfaces contains only one amplifier, needs a typical 30 mA, costs about the same as the THAT 6263 so cost per channel is roughly double, not to mention two of them take up a fair bit more space.

It is abundantly clear why the designers would have chosen this particular part for a relatively inexpensive, bus-powered interface, and also why quite a number of manufacturers in this price range (e.g. Audient, SSL etc.) have been sticking with classic all-analog input amps with pots being in the actual signal path despite some associated inconveniences.

 

[ssashton]

Right, the.comments about the 2i2 quality make a lot more sense to me now! I saw the spec is claimed to be THD -100dB @ 8dB gain. I figured at 0dB gain it would meet the converter specs, but obviously not, what a shame. Good that I asked about it here rather than just hitting buy (which I nearly did!).

 

[DavidR]

Not that I'm immediately inclined to do it, but I wonder if it's possible (and how difficult) to open up the 2i2 and bypass the variable line input gain stage to then match (almost) the distortion vs level of the 4i4 fixed input. I have no need of the variable input, I'll always be using it at zero gain.

 

[DavidR]

After more experimentation I've found that I won't even be needing the Scarlett 2i2 Gen 4 input. My usage is primarily to be for speaker distortion testing and an amp or two. I tested the Scarlett generator output to an M-Audio Delta 1010LT balanced input. They both functioned well using the Java driver (Windows 10). The results show what I think are exceptional input harmonic distortion components for an audio card design from 2002, although it's got a not-so-good noise level compared to the Scarlett. Since the harmonic distortion can be separated from the noise it looks like the combination of Scarlett output and 1010LT input is better than the Scarlett alone for my purposes. The Scarlett generator output is much better than the 1010LT. The results are below.

I included a loopback test of the 1010LT for perspective. It only has balanced inputs, so that loopback used the single-ended I/O.

Next is a comparison of the Scarlett-to-1010LT balanced connection. One test had the Scarlett Output Level knob set to maximum and at 3 o'clock (as well as I could eyeball it). The overlay of the two shows them nearly identical. Keep in mind that the level is dominated by the 1010LT input. The 1010LT and Scarlett input specs are rather close. The benefit of the 1010LT is that it has no variable input gain that compromises the Scarlett.

Up until now I had been using the Scarlett DAC at 48kHz. The next graph shows little variation in the analog output for the 48kHz setting in the Scarlett. The overlay shows the THD vs Level for the GainDial set at Max and at 3 o'clock. While using the analog output from the Scarlett for the source I wondered what the results would be with the Scarlett DAC at 96kHz. I then added the THD vs Level with the Scarlett when set for 96kHz.

This surprised me. The Scarlett output THD is roughly 5db higher. The only change made was switching the DAC from 48kHz to 96kHz. The analog output THD with DAC at 96kHz is about 5db higher than the DAC at 48kHz. Unless I've made an error somehow in measuring, but I've been careful to try to prevent that.

My conclusion is that for my purposes I'll use the Scarlett 2i2 Gen 4 as source with the Delta 1010LT for input. I'm a bit disappointed in the Scarlet variable input.

 

[AnalogSteph]

The results show what I think are exceptional input harmonic distortion components for an audio card design from 2002, although it's got a not-so-good noise level compared to the Scarlett.

Which would be where exactly?

The card is littered in AK4524s, whose ADC specs are relatively modest: DR 100 dB(A), THD+N -90 dB @-1 dBFS. (As opposed to the DAC side with 110 and -94 dB, respectively.) Now from experience the ADCs in AKM codecs tend to perform a bit better than spec, and that seems to be the case here as well. The NJM5532s in the analog stage are more than adequate, too. Being an internal PCI card, it also had the advantage of having access to +12V and -12V supplies, which seemingly are used for the analog stage as-is.

The 1010LT's Achilles heel seem to be crappy electrolytics. Not unusual.

I suppose you'd be even more impressed with the "normal" 1010 and its AK5383s. Those are also pin-compatible with AK5393s (with relatively few differences to the even better AK5394A) so further upgrades are theoretically possible if you can find one of these now-obsolete parts. Modding these oldies would not be unprecedented, here's an even older model:

Echo Mona Audio Interface RMAA Result

Dear All, May I share the RMAA result of my Echo Mona? I prefer this one to my ASUS Xonar D2 which has 115db dynamic range. (Will post its RMAA result if interest is showed, if it is OK and if I am posting at the right place.) It is still working nicely today, but I cannot understand why its...

www.audiosciencereview.com

 

[DavidR]

Which would be where exactly?

Fair question. I had not made a complete set of tests for the 1010LT, made a few. I had the Scarlett sample rate at 96kHz briefly for one test. When I switched it back, REW barfed and now does not work with it through Java. Works fine with the Delta ASIO driver. This prevents me from using both Scarlett out and 1010LT in (until I find the problem). In the mean time I'll post what I have to date.

This comparison demonstrates why I find the 1010LT input to be a better performer for my testing. I have no concern with THD+N, I'll be testing for harmonic components only. I've never place a lot of importance on the single THD+N. I've been using first the Delta 410 and now the 1010LT in a 3-way dipole using the Bodzio Ultimate Equalizer DSP. I've never noticed any noise from the 1010LT, with my amp it's not an issue, it doesn't have enough gain and the system is sensitivity is about 86dB. But I digress. I'll be testing raw drivers and some finished systems at some point for the harmonic distortion components.

The graphs below show the input of each device using the Scarlett 2i2 as source for the tests. The output gain dial was set to maximum. The Scarlett generator was set 4 volts for both. This was actually slightly above the 1010LT input upper limit of 14.0dBu (+4dBU setting). The Scarlett's limit is 22dBu, so 4V is far below the maximum input. In that regard the Scarlett has more leeway, but I have no need of that range. I'll be using a voltage divider probe.

This is the Scarlett:

This is the Delta 1010LT:

The noise floor of the Scarlett is nominally better, but the key for me is the distortion components. In the 1010LT they're all buried in/near the noise floor for the same input voltage. More interesting to me is that no matter the signal level, all the harmonic components remained relatively the same, very little variance. The Scarlett, on the other hand, had dramatically increasing and different changes in the harmonics. From about -2dBFS the change is extreme. The above graph is for ~0dBFS for the 1010LT.

I was actually rather surprised at the 1010LT performance. It may be that the balanced inputs are far better than the single ended that may be what the manufacturer rating references. I hope to find the issue with the card so that I can do additional tests (and actually use it).

 

[DavidR]

The card is littered in AK4524s, whose ADC specs are relatively modest: DR 100 dB(A), THD+N -90 dB @-1 dBFS. (As opposed to the DAC side with 110 and -94 dB, respectively.) Now from experience the ADCs in AKM codecs tend to perform a bit better than spec, and that seems to be the case here as well. The NJM5532s in the analog stage are more than adequate, too. Being an internal PCI card, it also had the advantage of having access to +12V and -12V supplies, which seemingly are used for the analog stage as-is.

According to the M-Audio 1010LT spec sheet, both the line input and output are rated for THD (at 0dBFS) as less than 0.002% (-94dB). It doesn't differentiate between single ended and balanced. Now that I think about it, I'm not sure about it's input voltage limit since it lists a single value, not for each type of input.

The Scarlett is rated as -100dB (-1dBFS @8dB gain), no way to know what it is spec'd for at 0dB gain. All my tests were with input gain set to 0dB.

The 1010LT's Achilles heel seem to be crappy electrolytics. Not unusual.

I considered changing the single large electrolytic. It is very old. I have a second 1010LT, I may do that and test for changes. I'm not (yet) inclined to change them all.

 

[AnalogSteph]

I considered changing the single large electrolytic. It is very old. I have a second 1010LT, I may do that and test for changes. I'm not (yet) inclined to change them all.

There's actually two big ones, in RC filters for the incoming +/-12V rails, and those are arguably the ones I'd be worried about the least, relatively speaking.

This post on AK may be a glimpse of what awaits you:

I use a Delta 1010LT 8 channel audio card in my home studio. I recently upgraded all my equipment including the PC. I discovered the Delta channel 1 was acting up and had bulging power rail caps. Some of the inputs distorted easily and the outputs were weak. The board has about 20 - 22 ufd and 20- 10 ufd caps. There are also a couple 470 and 47 ufd. I replaced them all with premium Nichicon of equal values. I measured all of them and most were very bad. In the nfd range. The card sounds great and all the distortion is gone. I'm surprised it worked at all.

That was from 2015. I would imagine that results may vary dramatically depending on how much use (a) the card itself and (b) the computer with it have had and how warm things got (PC ventilation wasn't always super great in the olden days), with a pinch of sheer pot luck thrown in.

Back in the mid-2000s to early 2010s I used two Terratec Aureon 5.1 Sky cards for radio recording fairly extensively (an Envy24HT job with WM8770, one of them reflashed to Audiotrak ProDigy 7.1), and one of them actually developed some electrolytic-related issues. Using fancy electrolytics became en vogue ca. 2005 (no doubt as the effects of the capacitor plague hit in full force) but wasn't much of a thing yet in 2003.

 

[DavidR]

You're right, two big ones. A lot of small ones. I thought that the big ones being power supply would be the most useful for replacing given the age. I've had to replace many caps in power supply/regulation sections in various electronics. Never had others fail. I'll consider replacing the others since it's a spare at this point. All my M-Audio cards were purchased used, so no idea the conditions of the PCs they were in.

 

[ZestClub]

Do we know which opamps it uses?

Did anyone find out which Opamps it uses?
The Scarlet 3 gen used the very basic Japan Radio JRC2122 and the NJM8068 with very low slew rates 2.4v/uS

 

[JonoJono]

Anyone upgraded from the 2nd gen to the 4th gen?

Have you noticed any large differences on the output?