Strange noise problem with Focusrite Scarlett

[Tomás Mulcahy]

Interesting issue here, you can see it on the pics. The one with the harmonics is Scarlett 6i6 1st generation. The other is a Scarlett 2i2 2nd generation. On same computer.




Even with everything in the building switched off, the harmonics appear. Tried 3 different computers. Tried 2 different 12V PSUs for the 6i6. Same USB cable every time. Nothing connected to the inputs or outputs.

I had created a ground loop (using two different plugboards) and this increased the level of the harmonics. I'm not really sure why that is?

What's confusing me is, it never happens with the 2i2?

Have I exhausted all possibilities? Is it most likely a fault in the 6i6?

[Hugh Robjohns]

I would suspect DC-DC converter noise from the internal power supply circuitry. Does the level of harmonics vary with preamp gain?

[Tomás Mulcahy]

Hi Hugh, thanks for giving this some thought

Yes, the level of harmonics changes with the gain. With nothing plugged in. What does that tell us?

I could check the output filter on the DC-DC converter, I have some experience with them... hopefully I'd spot it on the board without a schematic

[Hugh Robjohns]

Yes, the level of harmonics changes with the gain. With nothing plugged in. What does that tell us?

It's not really fair to test an preamp with an unterminated input, so definitely worth testing again with 150 Ohms strapped between pins 2-3 inside an XLR plug. But if the levels of harmonic change with preamp gain we know the nasties are getting into the front of the analogue stage, either via the power rails or the input circuitry somewhere, and so it's not a problem with the converter or its clocking. So that lends some more weight to the possibility of DC-DC converter noise, as does the shape and starting frequency of the harmonics.

H

[Tomás Mulcahy]

The same harmonics are there with a low output, single coil passive bass into the instrument input (it's a Hofner B-Bass Hi Series). That's how I first became aware of it- they're audible . Their level is greatly reduced with an AT-2020, but again obvious and audible (some of them, I'm not a bat !) with a Naiant omni capacitor mic.

So I guess that means you're correct about the DC-DC converter issue?

[James Perrett]

Most modern DC/DC convertors that I've used have worked at 100kHz or more but it could be a beat frequency between two of them.

[Tomás Mulcahy]

I've seen other reports of the same harmonics, some people find that wiggling the USB cable changes it. I've opened it up to inspect the board but there's nothing obvious- except that it's very well made! I've opened a support ticket, we'll see how it goes.

Might be just the excuse I need to upgrade to an RME BabyFace!

[Tomás Mulcahy]

Update: Focusrite have diagnosed the issue and supplied notes and schematics but I've had to sign an NDA. They have seen this issue before for a small number of customers.

Kudos to them for amazing after-market service! With Covid and Brexit returning it for repair was not an option. More efficient to DIY (I am a qualified tech). I've time next week to try the fix they suggested.

[muzines]

Some ultimately it's some kind of hardware problem with a limited number of units, I guess, rather than a more general design thing.

Sounds like great service from them, though.

[Tomás Mulcahy]

Some ultimately it's some kind of hardware problem with a limited number of units, I guess, rather than a more general design thing.

Exactly!

Sounds like great service from them, though.

This will be the fourth time for me that they've gone out of their way to help with an old product. Seems like a nice company to work for!

[Tomás Mulcahy]

Update: so the suggested fix worked fairly well. The 8kHz harmonic is reduced to 4dB above the noise floor, the 16kHz is 12dB above noise floor. Others above 20kHz also attenuated. Quite satisfactory.

[Martin Walker]

Update: so the suggested fix worked fairly well. The 8kHz harmonic is reduced to 4dB above the noise floor, the 16kHz is 12dB above noise floor. Others above 20kHz also attenuated. Quite satisfactory.

Only just caught up with this thread, but what a resounding finish, and all credit to Focusrite for resolving it.

It's totally understandable about the NDA, but it's nevertheless a shame, as I'd love to have known the reason for these stray lines that only affect a tiny proportion of units. To me the most likely reason is a certain batch of a particular component that didn't meet the intended spec, but we will never know

Martin

[Tomás Mulcahy]

Well I am still chatting with them. Did something slightly different and noise is basically eliminated on mic inputs, but still there on line inputs but vanishingly low as you can see. This is a tough test, no input and set to "Hi Gain" with 48kHz bandwidth. If I turn up the monitor level by 30dB I hear digital/ RFI noise.


For an affordable audio device this is superb performance, and for a seven year old product this is great support. And there's a new free plugin from Acustica in my Focusrite user account. Zero complaints from me!

Will keep you posted

[muzines]

Note - you can't serve images from Dropbox since they stopped this a few years back, so your images don't show.

Let me know what you think of Cream2...

[Tomás Mulcahy]

Note - you can't serve images from Dropbox since they stopped this a few years back, so your images don't show.

Oh that's annoying. i can see them here but that means nothing obvs Imgur?

Let me know what you think of Cream2...

Will do!

[CS70]

Note - you can't serve images from Dropbox since they stopped this a few years back, so your images don't show.

Uh? What do you mean? I can see the images just fine.

[muzines]

Hmm, not showing for me, any of them, and I do not have any ad blockers running. And if I click on them in Safari, I get "the image failed to load."

Let me check Chrome...

Ok, in Chrome, the images in the first post do show, it's because you're using the dl=1 "hack" which tricks the browser into downloading the served image content into an image tag (which is what people did after Dropbox stopped rendering HTML requests precisely to stop people linking to Dropbox hosted images on web pages).

This appears not to work in Safari, at least here, perhaps a security thing, I'm not sure.

I can't check the second post of images of posts, because Chrome, for some reason, currently thinks there is only one page of posts and doesn't let me access the more recent posts - they end with the post "Will keep you posted " (probably a cache thing which will fix itself...)

[CS70]

Ah, a Safari thing. I didn't think of "dl=1" an "hack", as it's simply telling Dropbox to allow direct DownLoad of the file instead of a preview page. However, that has to be the actual "dropbox link".

An even better way is to add raw=1 which simply instruct the HTTP server to return the octet stream with the content.

I think I was the first on this forum to mention the "dl=1" a few years ago, I've long since moved on to raw=1 but didn't occur to me of mentioning here.

[Tomás Mulcahy]

You mean replace dl=1 with raw=1? Can you give an example please?

[CS70]

Sure

[muzines]

Ah, a Safari thing. I didn't think of "dl=1" an "hack", as it's simply telling Dropbox to allow direct DownLoad of the file instead of a preview page.

The reason I quoted "hack" was because it was to get around Dropbox's restrictions about serving images on web pages. The DL method forces a direct browser download of a file to the user, but in the context of an image link, browsers generally accept the "download" as returning image data and displaying it on the page as such. It's not really a "hack", but it is a workaround, and you don't generally serve images on the web this way for a bunch of reasons.

Edit: I see the "raw" method image fine.

[Hugh Robjohns]

I've changed all of Tomas' links to raw=1 so they should now be visible in the thread for everyone.

[Martin Walker]

OK, now to reveal the reason for my special interest in this topic:

This image shows the spectrum analyser display for my Fireface UC mic inputs, set to 20dB gain with a 2-metre open circuit TRS cable plugged in (hence the 50/250/350Hz mains hum spikes).

The two offending spikes (13.5kHz and 16kHz) disappear if you short the far end of the cable to ground, and when you unplug the cable. The spikes go up and down with the mic preamp gain setting, which suggests that something is being picked up by the cable and amplified via the preamp, but at 13.5kHz and 16kHz?????

Not really a problem, and the two lines are really low in level, but it would be good if I could eradicate this pickup - any thoughts?

Martin

[CS70]

Ah, a Safari thing. I didn't think of "dl=1" an "hack", as it's simply telling Dropbox to allow direct DownLoad of the file instead of a preview page.

The reason I quoted "hack" was because it was to get around Dropbox's restrictions about serving images on web pages.

Understood I just never even knew that Dropbox had restrictions.

[Hugh Robjohns]

Not really a problem, and the two lines are really low in level, but it would be good if I could eradicate this pickup - any thoughts?

They look like aliases to me, and the maths would suggest something getting in up around 60kHz. So I'd be looking at switch mode power supplies or maybe backlight inverters in LCD monitor screens. Try turning things off while recording, and then analyse afterwards (with the screens back on)