USB isolation

oscroft

Prominent Member
Joined
Sep 24, 2019
Messages
1,660
Reaction score
1,579
Points
523
Location
Liverpool, UK
Here's a topic that generates polarized opinions - USB noise, and its possible effect on audio. It came to mind because of this YouTube...


The guy's an audio engineer, and I tend to trust people like that as hopefully being a bit more objective. Obviously you can't tell by listening to comparisons over the audio horror that is YouTube (and I'm still amazed by the number of people who present their comparisons as YT videos and expect us to be able to hear them). But he captured test files in .wav format, so I downloaded those to try.

First problem - it's obvious from the file names which is which. So I sent them to a friend and asked him to make me 10 copies (some, not necessarily five, of each) with file names I can't guess. He used a random name generator, and sent me 10 files. I loaded them up in Vox, the music player I use on my dedicated Mac Mini, and set it to shuffle mode. I sat and listened to all 10, then I did the same again to get a different shuffle sequence. And I couldn't hear any difference.

While this was going on, I'd ordered a cheap "USB to USB Isolator Board Audio Noise Eliminator Industrial Isolator Protection" from a Chinese eBay seller - I won't pay £120 for the iFi thing to test, but £14.50 (inc postage) is fine. That arrived a few days ago, I inserted it between the Mac and my Audiolab M-DAC... and I swear it sounds better!

So, a real effect or a placebo? I have no realistic way to test blind, so I really can't say. But I thought I'd offer it for discussion/entertainment anyway, partly in case anyone else wants to try those iFi Purifier test files.
 
.... and I swear it sounds better!

USB or any file digitally transmitted, can have problems, but we have to ask what those potential problem are.

Jitter for example, moving digital files for playback, and cause the sound to get slightly out of sync, but you can't hear it. That is, you don't hear a hum or buzz, or echo or anything that would imply that something is wrong. Yet, the music will not be bit perfect.

If we assume CD quality, that means 44,100 samples per second. So, if we have 1,000 consecutive read errors or inaccuracies, how much time is that? I think if I calculated correctly that is about 0.022 seconds. Can you hear that, can your really hear something that last 0.022 second?

So, the question isn't that there are or aren't read or conversion error, the underlying question is, can you hear them? How do you hear them? And what is it that you are really hearing? I suspect the changes to the audio, short of a massive failure, are going to be very subtle, and you might technically be able to find them, but can you really hear them?

Audioquest has had one of these on the market for quite a while now.

Amazon product ASIN B00YTA78FW
Going back to the 1,000 sample example, how big does an error have to be to be able to be heard.

As to these USB Filter Devices, fine if it works for you, fine, not a problem.

But consider that Music is not store, and music is not transferred, and music is not converted. What you have is a list of numbers that represent voltages that are translated back into voltages so they can be played back as music. So, what we are asking is, do the numbers get through or don't they?

Keep in mind that critical data is contained in a series of numbers. If you pass a Database, Formatted Document, or a Spreadsheet though a USB connection, do you expect it to arrive precisely on the other side?

If the Spreadsheet says your company made £1,000,000, but the USB mangles that into £100 is that acceptable? And how often does that actually happen? I really don't think it happens that often. I think if the Spreadsheet says £1,000,000 then it arrives on the other side as £1,000,000.

I think if the song is the Beatles, on the other side of the USB Connection, it arrives as the Beatles. For it to not arrive as the Beatles would take a near catastrophic transmission error beyond anything that any reasonable equipment would tolerate. Remember that 1,000 consecutive errors is just 0.022 seconds.

It is possible that noise on the line could cause a bit here or there to be misread or dropped, but those errors would be beyond microscopic. So, yes, noise filtering could potentially result in a cleaner signal ... In theory. Hard to say what it does in practice though.

Steve/bluewizard
 
Interesting thoughts Steve, thanks. There's just one point I want to pick up on, and that's the mistake of thinking digital music transfer and computer network file transfer are the same. They are not, in a crucial way.

Computer network file transfers use error checking and correcting protocols. That includes packet checksums, and any lost or damaged packets are resent. And it doesn't matter how many retries it needs to do, as it's not real time and there's no rush. That's why your million pound spreadsheets are pretty safe.

Digital music transfer does not use the same error checking and correcting protocols (at least, not to the extent of file transfer protocols), and does not resend lost or damaged data - the receiving equipment just has to do the best it can if it can't recover in real time. It's effectively still 1980s redbook standard (with higher bit depths and sample rates tacked on these days), from the days when the bandwidth and memory simply wasn't there to use a full error-correcting packet-based protocol at real-time music transfer rates.

The other thing I've seen claimed is that noise on a USB connection can leak into the analogue circuitry of a DAC (and computer USB environments can be very noisy indeed). That's certainly possible if the D and A sides of a DAC are not sufficiently screened and filtered (and some high end manufacturers even keep them galvanically isolated), but I've no idea how much of a problem that is in practice.

UPDATE: It seems CD-style error correction can correct up to 4,000 successive bits (as opposed to file transfer protocols having no upper limit). I think that's about 0.005 seconds (or 1/180 second). Errors beyond that use interpolation/concealment (or filling in by estimation). (This is from reading about CD/red book protocol, and doesn't take into account errors caused by USB noise/timing/jitter - it's just the base protocol correction ability. I have no idea what error rates jitter might introduce.)
 
Last edited:
Another bit of speculation while I'm sitting here idling away my time with rambling thoughts...

I have all my CDs ripped to FLAC files. Whenever I rip CDs, different ones seem to rip at different speeds, and there are often obvious retries going on. That suggests that errors from CDs (at least from mine, most of which are quite old) are common, and that FLAC files after multiple retries should be better than playing straight from CD? (Straight from CD doesn't use USB, mind, so I don't know if the comparison is relevant).

One thing I've never seen is a proper treatment of all this by someone who has expertise in both file transfer protocols and digital music transfer protocols. I think that would be interesting.
 
I use a galvanic isolator between my MacBook and usb to optical converter and I used to use one between my M-DAC and MacBook.

I originally bought it for an old W7 laptop I used to use. I thought it maybe made a difference then. Less so with my Mac.

But, as you say, I can’t do a realistic double blind.

I’ve kept it in use because why not?

I’ve got one that can take a separate 5V power supply too. And I’ve used a battery to check if I could hear a difference. And, again, possibly.

But in the end I just use a quiet UPS power supply to the 5V input, bypassing the macs usb supply, so I know it’s as quiet as I can be bothered to mess about with.

It was all cheap, and I don’t like the idea of electrical noise being lobbed into my dac so just because I can I did.
 
Ps, one of these £25 used, from eBay.

 
Ps, one of these £25 used, from eBay.

My USB thingy is similar to that - it has a separate connection for power with an internal/external switch. I did try it with a battery pack, but it kept cutting out presumably because there wasn't enough current to make it think something was connected. I should try it with a mains to USB power supply (though the only ones I have are probably as noisy as the Mac).

It's marketed as Nobsound by some sellers (like Amazon), though it doesn't say Nobsound on it...
USBthing2.jpg
 
Last edited:
USB uses the same checksum protection as TCP, so your spreadsheet has the same level of protection on a network transmission as USB. If a bad packet is received it is resent, at least for data such as spreadsheets. For audio and video, protocols for both often drop bad packets rather than retransmit them, to keep latency constant. This is true for both network and USB In theory, but most network audio probably uses TCP, and retransmits bad packets. However you have to ask how often these error occur. I can tell you fhat for modern Ethernet they are very, very rare. Over the Internet that’s not always true. For USB 3 the allowable error rate is 1 in 10^12 Bit Error Rates for USB 3.2 | To USB or Not to USB I havent found this number for USB 2, but I’d be shocked if it is large enough to cause audible problems.

This does not address the question of jitter, which is timing rather than an error, but tests with DACs in places such as audioscience suggest that these are also inaudible.
 
USB uses the same checksum protection as TCP, so your spreadsheet has the same level of protection on a network transmission as USB. If a bad packet is received it is resent, at least for data such as spreadsheets. For audio and video, protocols for both often drop bad packets rather than retransmit them, to keep latency constant. This is true for both network and USB In theory, but most network audio probably uses TCP, and retransmits bad packets. However you have to ask how often these error occur. I can tell you fhat for modern Ethernet they are very, very rare. Over the Internet that’s not always true. For USB 3 the allowable error rate is 1 in 10^12 Bit Error Rates for USB 3.2 | To USB or Not to USB I havent found this number for USB 2, but I’d be shocked if it is large enough to cause audible problems.

This does not address the question of jitter, which is timing rather than an error, but tests with DACs in places such as audioscience suggest that these are also inaudible.

10^12... :laugh::rotfl: I can’t tell you how much these numbers always make me laugh...

1in a thousand billion. It’s just to HUGE to contemplate.

And people still buy ‘special’ ‘audio’ usb cables.

I’d be happy to believe that in anything but the most appallingly made dacs jitter hasn’t been a significant problem for well more than a decade.

The only thing I can think of that might be an issue from a computers usb output is electrical noise... but even then I’d imagine any but the cheapest dacs have that covered.

I used the galvanic isolator simply because it’s cheap and it might give the dac an easier time.

I must say I’ve only noticed obvious noise (an actual honest to god crackle; sounded like an LP) on a very old laptop years ago, and that wasn’t fixable by the isolator I had at the time.

I fixed it by dumping the ancient thing and getting a MacBook.

A galvanic isolator I simply see as insurance. Probably not worth it, but so cheap, why not?
 
Well, the jitter timing thing can cause bit errors - if it measures a voltage at the wrong time, it can miss the bit (or, at least, that's what jitter explanations I've read say). But yes, I think any problems that might exist over USB are not due to bit errors - or at least, bit errors are so vanishingly small that I'm not going to hear them.

That leaves the possibility of noise getting into the analogue side of the DAC, which does seem plausible. Or, it might be that my USB isolator sounds better because I paid money for it, while I didn't pay anything for the iFi Purifier test files from YouTube 😉
 
I assume you've seen this:

 
I hadn’t. I have now. Seems pretty unarguable to me :smashin:
 
The only thing he doesn’t address is the usb source.

I’m aware that modern dacs are largely vice free.

And also that a decently made usb output should not have any noise on it. Some might. Cheap windows machines perhaps.
 
The only thing he doesn’t address is the usb source.

I’m aware that modern dacs are largely vice free.

And also that a decently made usb output should not have any noise on it. Some might. Cheap windows machines perhaps.
Ah, interesting point, which I'd overlooked. I've no idea how noisy the USB on my Mac Mini is, but it is a pretty old one.
 
Ah, interesting point, which I'd overlooked. I've no idea how noisy the USB on my Mac Mini is, but it is a pretty old one.

I’ve no real idea if it has any merit as a notion.

But I’ve also no reason to think all USB ports are splendidly engineered.

I imagine a good dac or usb converter has noise rejection too, or at least certainly should have, so I don’t know if a galvanic isolator is required on many usb outputs.

However, as injecting noise into an audio device can at the very best be neutral sounding, my isolator remains.
 
I’ve no real idea if it has any merit as a notion.

But I’ve also no reason to think all USB ports are splendidly engineered.

I imagine a good dac or usb converter has noise rejection too, or at least certainly should have, so I don’t know if a galvanic isolator is required on many usb outputs.

However, as injecting noise into an audio device can at the very best be neutral sounding, my isolator remains.
I keep thinking I should build a Carlson Super Probe (YouTube, Mr Carlson's Lab), just to check for noise from various sources - maybe I'll get round to it. And yes, I'd have expected my Audiolab M-DAC to have some decent noise filtering.
 
I keep thinking I should build a Carlson Super Probe (YouTube, Mr Carlson's Lab), just to check for noise from various sources - maybe I'll get round to it. And yes, I'd have expected my Audiolab M-DAC to have some decent noise filtering.

I had a chat with John Westlake when I bought my original M-DAC and he was rather pleased with his noise isolation for the dac inputs, but he still recommended a galvanic isolator for the usb input.

He wanted to build one into it, but space considerations on the board prohibited him.

He even sent me one he’d been testing with.

His position was that expensive boutique filters and de-jitter claims for usb ‘accessories’ were bollocks.

Except for galvanic isolation.

He made the simple point of why put any noise into sensitive audio equipment when it’s so cheap to not do so.

That was good enough for me.

So I don’t. :smashin:
 
Last edited:
I keep thinking I should build a Carlson Super Probe (YouTube, Mr Carlson's Lab), just to check for noise from various sources - maybe I'll get round to it. And yes, I'd have expected my Audiolab M-DAC to have some decent noise filtering.

If you ever do build one I’d be intrigued to hear what you find.
 
I think the assumption in the test is that the USB signal is VERY noisy.

It doesn't matter how noisy it is coming in - part of the job of the DAC is to clean up that noise. Every manufacturer should clean the signal.

Even the poorest DAC he'd ever tested wasn't improved by the USB filters.

Fine by me - plenty of other things to spend money on. :) Get a more expensive something else where it will make a difference.
 
I had a chat with a John Westlake when I bought my original M-DAC and he was rather pleased with his noise isolation for the dac inputs, but he still recommended a galvanic isolator for the usb input.

He wanted to build one into it, but space considerations on the board prohibited him.

He even sent me one he’d been testing with.

His position was that expensive boutique filters and de-jitter claims for usb ‘accessories’ were bollocks.

Except for galvanic isolation.

He made the simple point of why put any noise into sensitive audio equipment when it’s so cheap to not do so.

That was good enough for me.

So I don’t. :smashin:
Now that is very interesting - he's certainly the one who'll know. And yes, if it's cheap to keep noise out, why not just do it and don't worry about whether it's strictly needed?

If you ever do build one I’d be intrigued to hear what you find.
I'll let you know, for sure.
 
I think the assumption in the test is that the USB signal is VERY noisy.

It doesn't matter how noisy it is coming in - part of the job of the DAC is to clean up that noise. Every manufacturer should clean the signal.

Even the poorest DAC he'd ever tested wasn't improved by the USB filters.

Fine by me - plenty of other things to spend money on. :) Get a more expensive something else where it will make a difference.

Can’t say I disagree.

But as I already have one...

As it happens I use an optical cable from the usb converter to the 1120 so there’s no noise getting to the Lyngdorf anyway.

The isolator is only used from the macbook to an SMSL usb to optical/phono converter.

I’ve no idea if such a thing is susceptible to noise from a usb or not, and if it was I’d expect the manufacturers to take care of that, as you say.

:laugh: but I still use the isolator...
 
I’ve just remembered why I first bought an isolator with an extra power input.

I once had an older NOS dac which wouldn’t work with the current from 0.5v from the usb output from the old laptop, it just want happy.

So, to kill the annoyance, I bought an isolator (about £15) that could also take a power supply from a phone charger and a 3.5mm pin.

That sorted it out straight away with a max of 2.2 amps.

Stopped the dropouts and they never returned. Nothing to do with the isolation of course, it was all about current supply.
 

The latest video from AVForums

Is 4K Blu-ray Worth It?
Subscribe to our YouTube channel
Back
Top Bottom