Galvanic isolation

Vincent Kars

WBF Technical Expert: Computer Audio
Jul 1, 2010
860
1
0
Often galvanic isolation is recommended
Shield your computer from the noise generated by the PC
Lots of claims about improvement when using some kind of isolator but very few measurements.

Exasound measured the noise level with and without galvanic isolation of the USB

According to Exasound
exaSound-e12-DAC-Noise-Floor.jpg
FFT spectrum (1024K points averaged 8 times) showing the noise floor measured on the e12 DAC RCA line outputs. The e12 DAC is connected with a standard USB cable to a low-cost desktop PC. The noise floor is extremely clean. There are no visible traces from the computer high-frequency noise usually transmitted via the USB connection. The power line related noise (60Hz and its harmonics) is below -157 dB (0.0000014%).

exaSound-e12-DAC-No-Galvanic-Isolation.jpg
For comparison, the same measurement is taken without Galvanic ground isolation. The noise level is increased by 30dB. This experiment clearly shows the benefits of using Galvanic isolation.

Source: Exasound http://www.exasound.com/e12/e12Measurements.aspx

Obvious the pic above show a substantial increase in the noise floor if you don’t isolate the DAC from the computer.
On the other hand even without galvanic isolation the noise is below -120, a very low value.
Makes me wonder if the difference is audible at all.
 
Last edited:

Speedskater

Well-Known Member
Sep 30, 2010
941
15
368
Cleveland Ohio
It makes me wonder too! (but I won't get to their site until morning)
For a test like this, it's no fair touching anything or any control setting. And the replacement cable has to be the same length and dress (placement). The tests also need to be done back to back.

A person on another forum does all sorts of power line measurements and scope traces. But while doing the tests, I think that he makes other changes and waits till the 'good' cable shows nice results and waits till the 'bad' cable shows poor results.
 

Tam Lin

Well-Known Member
Mar 20, 2011
96
36
923
79
North Texas
Actually, I would like to see the FFT jitter spectrum plots with and without galvanic isolation because, unless it's done right, and it rarely is, using galvanic isoators with async USB adds considerable jitter to the sample clock.
 

amirm

Banned
Apr 2, 2010
15,813
37
0
Seattle, WA
Nice find Vincent. Thanks for posting it. Wish the graph was not such an eye chart :). So I just took them into Photoshop and overlaid the two and enlarged:



Some other bits:

1. The difference is not in "noise." Visually the noise floor looks the same to me. When performing FFT, the noise level can be made higher and lower based on FFT parameters. Unless matched to 1 Hz/bin, the actual value cannot be trusted. You can make comparisons as they have done but can't say, "noise floor is -140 db" etc.

2. What they are talking about are deterministic tones represented by single frequency spikes. These are not subject to FFT parameters so their value can be discussed in the absolute.

3. What bleeds into a device is completely PC and DAC specific. So while -120 is indeed small numbers, change the DAC and PC and those values will completely revamp. That is the beauty of galvanic isolation. It highly reduces the impact of the differences between PCs and DACs in this regard. When you hear a DAC on one PC, it is liable to sound similar on another PC/DAC. Not so when isolation is missing. One machine may have very clean ground while another having a dirty one. You just don't know.
 

amirm

Banned
Apr 2, 2010
15,813
37
0
Seattle, WA
Actually, I would like to see the FFT jitter spectrum plots with and without galvanic isolation because, unless it's done right, and it rarely is, using galvanic isoators with async USB adds considerable jitter to the sample clock.
Did I read that right? You are saying isolation makes clock jitter worse? How so?
 

Tam Lin

Well-Known Member
Mar 20, 2011
96
36
923
79
North Texas
Galvanic isolators are not recommended for use with clocks because they add significant skew and jitter. Look at the specs for typical added jitter:
ADuM7440 ... 2 ns
ISO7240 ...... 450 ps
Si8640 ........ 350 ps
IL715 .......... 100 ps (5v)

Many USB DAC makers brag about their femto-clocks. The oscillator may be femto but the clock signal that gets to the DAC chip is pico or nano. Figure #8 on page 20 of the Trinity DAC User Manual shows a canonical async USB audio implementation. The precious ‘femto-clock’ makes two trips across the galvanic barrier picking up a fresh dose of jitter each time, not to mention the jitter added at other points on its journey to the DAC chips. Isochronous USB is probably the worst digital audio transfer protocol ever devised. BTW, because it is closer, the noise created by the DSP is much worse than the noise created by the PC.
 

amirm

Banned
Apr 2, 2010
15,813
37
0
Seattle, WA
Thanks for the data but I am still not clear. If I have an async USB DAC, then the clock is local and not derived from USB. Galvanic isolation inside such a DAC can only do good in providing ground isolation.

I looked at page 20 of that DAC and it says nothing different:



The DSP is running asynchronous and slave to the master clock on its back-end. The clock is not going across the isolator and coming back to drive the DAC.

But yes, the accuracy of the local oscillator means nothing. What matters is what the DAC sees as the total product which manifests itself in the analog output. Plenty of things can bleed onto that clean clock.
 

amirm

Banned
Apr 2, 2010
15,813
37
0
Seattle, WA
Hmmm. That diagram does show some potential issues. I am not sure why it does not show a direct connection between the clock oscillator and outbound line driver.
 

Tam Lin

Well-Known Member
Mar 20, 2011
96
36
923
79
North Texas
The clock is not going across the isolator and coming back to drive the DAC.

YES IT IS!

Async USB is the same as Adaptive except the DAC provides PLL-like feed back to adjust the incoming sample rate. The PC is in total control and sends just enough data for the DAC to use in the next polling period. And, as with all Isochronous methods, there is no retry for lost or corrupted data.

In figure 8, the Clock Management block is a mux that selects the oscillator based on the sample rate. As displayed, the circuit is round about. The clock selection starts in the DSP, goes through the ISO, to the line driver and to the central gray box, which represents the user controls and display. Then to the line receivers and the Clock Management. The selected clock, 22MHz or 24MHz goes through the ISO to the DSP where it is divided down to create the bit and word clocks, which are synchronized with the sample data and sent back through the ISO and the line drivers to the LIANOTEC shift registers and the DAC chips.

It is a really crappy design but that's the way (almost) everybody does it. Few DAC designers have the desire to write the DSP code and OS drivers to do it right. And why should they? Their audiophile customers think async USB is the greatest thing since stereo; especially if it includes a femto-clock! It's easier for the builder to licence the firmware from XMOS, or another supplier, who has already done the heavy lifting.
 

Speedskater

Well-Known Member
Sep 30, 2010
941
15
368
Cleveland Ohio
One wonder's why a $2000 unit with more than excellent specs even has RCA connectors? If I were building a component like this, it would only have XLR outputs. Differential outputs and/or inputs are the best way to deal with unbalanced interconnects anyway. Just include a nice pair of XLR>Shielded Twisted Pair>RCA interconnects.
Although a balanced output may place great demands on a little power supply.
 
Last edited:

Brucemck2

Member Sponsor
May 10, 2010
426
102
1,598
Houston area
YES IT IS!

Async USB is the same as Adaptive except the DAC provides PLL-like feed back to adjust the incoming sample rate. The PC is in total control and sends just enough data for the DAC to use in the next polling period. And, as with all Isochronous methods, there is no retry for lost or corrupted data.

It is a really crappy design but that's the way (almost) everybody does it. Few DAC designers have the desire to write the DSP code and OS drivers to do it right. And why should they? Their audiophile customers think async USB is the greatest thing since stereo; especially if it includes a femto-clock! It's easier for the builder to licence the firmware from XMOS, or another supplier, who has already done the heavy lifting.

Which vendors/products do it "right" (non-crappy design)?
 

jkeny

Industry Expert, Member Sponsor
Feb 9, 2012
3,374
42
383
Ireland
Which vendors/products do it "right" (non-crappy design)?
If the I2S data is reclocked by the local clock on the isolated side of the barrier then it is a step closer to doing it right. This then becomes synchronous clocking where the local clock is fed back through the isolation barrier to the DSP.
When using asynchronous clocking the PLL needed to ensure buffer never over/underflows, introduces it's own audible disturbances.


Even with synchronous clocking there still appear to be some issues remaining - see here, although I don't agree with all that's stated, I do think it's a considered view from John Swenson
 
Last edited:

Brucemck2

Member Sponsor
May 10, 2010
426
102
1,598
Houston area
If the I2S data is reclocked by the local clock on the isolated side of the barrier then it is a step closer to doing it right. This then becomes synchronous clocking where the local clock is fed back through the isolation barrier to the DSP.
When using asynchronous clocking the PLL needed to ensure buffer never over/underflows, introduces it's own audible disturbances.


Even with synchronous clocking there still appear to be some issues remaining - see here, although I don't agree with all that's stated, I do think it's a considered view from John Swenson

Question remains, which vendors/products do it "right" (non-crappy design)?
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing