Galvanic isolation of your DAC


#1

When the zeros and ones of digital audio are as correct as possible and — where relevant — the timing is as good as you can get it still isn’t all over!

Ted Smith has mentioned galvanic isolation a few times: it clearly matters to him. Electrical interference can get into your DAC via any of the electrical connections. Galvanic isolation is about getting rid of electrically conductive connections to avoid that kind of interference. I’m interested in ways of achieving it for my DirectStream Junior; I dare say that the same solutions will be applicable to the DirectStream ‘Senior’.

Mains power

We can’t provide galvanic isolation for the main power input though we can clean up the power as best we can. I use a dedicated mains spur from the house’s distribution unit (US, breaker box) and clean the power going to my audiophile source and power amp using a PS Audio power plant.

I²S

This uses electrical conductors. I’ve never heard of another way of getting the signal in although those that use it can take heart from the fact that I²S sources are almost always going to be of audiophile quality.

USB

I believe there’s a way of using an optical cable by Corning to get the signal in but I’m not clear whether it achieves full galvanic isolation. There’s a thread about this. From what I’ve read reliability is an issue.

SPDIF — electrical (phono plugs on a co-axial interconnect)

This uses electrical conductors.

SPDIF — optical often known as TosLink

There’s no electrical connection, so whoopee! However there are technical limitations. It would be good for somebody to summarise those limitations here.

Ethernet

Again there are electrical conductors. Also there are devices that generate electrical noise — computers and switches — along the route from the server’s drive to the DirectStream. Also Ethernet cables may be long which is great for putting a perhaps acoustically noisy server in another room but increases the chance of the cable acting as an aerial for interference.

WiFi

The wireless cousin of Ethernet is often eschewed by audiophiles because of concerns about data throughput, especially when the domestic WiFi is likely shared by family members watching streamed video. However I’m interested in setting up a WiFi channel dedicated to getting music files from server to DirectStream. The DirectStream has no WiFi connectivity (unlike, for example, the old but much-loved Logitech Transporter [originally Slim Devices Transporter] which had a couple of aerials on the back). However I’m sure there’s a WiFi expert out there who could advise us how to use WiFi access points or range extenders or some such to overcome that.

Audio out

Finally we all use electrical connections to get the analogue out of our DirectStreams. I suppose there’s an opportunity for electrical interference to get in by that route, along the neutral connection.

I hope I’ve summarised the issue correctly. If electronic engineers need to correct me I welcome it. I now hope to see some discussion about ways of achieving galvanic isolation.

Peter


#2

I have an I²S PCIe sound card with it’s own linear power supply and it sounds great. The PC has no bearing on power supply and even the PCIe data lanes are filtered by the card to keep noise down to a minimum. I’ve thought about purchasing some kind of optical HDMI linking solution to try and galvanically isolate things. The only thing I’m weary about is that the whole point of using I²S is thrown out the window when using such equipment. Or at least I assume that’s the case. I²S is a great solution as you don’t need to encode and decode to various transport formats to get the audio information to your DAC so the signal integrity remains more intact compared to other transport methods. Also I²S is nice in that it keeps the clock information and audio information separate. I just don’t know enough about optical HDMI solutions to know if it will do more harm than good. Will these solutions harm the benefits that I²S has? Does it keep the I²S benefits in tact or do these HDMI optical solutions “encode” the information in a transport format that needs to then be decoded back to I²S once it reaches the receiving end of the optics? My philosophy has been to use a good quality, well shielded, HDMI cable and keep the run short. I don’t know if I’d hear much of an audible improvement (or possibly an audio downgrade) if I went with such a solution. With the other options, especially USB, I can see the argument where an optical solution may prove to be useful.


#3
Peter said

Ethernet

Again there are electrical conductors. Also there are devices that generate electrical noise — computers and switches — along the route from the server’s drive to the DirectStream. Also Ethernet cables may be long which is great for putting a perhaps acoustically noisy server in another room but increases the chance of the cable acting as an aerial for interference.

WiFi

The wireless cousin of Ethernet is often eschewed by audiophiles because of concerns about data throughput, especially when the domestic WiFi is likely shared by family members watching streamed video. However I’m interested in setting up a WiFi channel dedicated to getting music files from server to DirectStream. The DirectStream has no WiFi connectivity (unlike, for example, the old but much-loved Logitech Transporter [originally Slim Devices Transporter] which had a couple of aerials on the back). However I’m sure there’s a WiFi expert out there who could advise us how to use WiFi access points or range extenders or some such to overcome that.

Hi Peter,

Thanks for starting this interesting thread.

Couple small comments to add. I like many others use Fibre Optic isolation in the Ethernet path to try and isolate noise. But like many of these things it comes with a trade off.

On the one hand no copper at all between the FMC’s - woo hoo!

On the other hand you need two FMC’s which were designed for networking and not audiophile purposes. I have mine powered by battery power supplies to try and get them off the mains network and mitigate how much noise they add back in after the optical cable run. But it’s beyond my knowledge level to know whether this is fool proof or whether the FMC can still introduce any unwanted effects.

I’m also intrigued by the Wireless thing as I’ve seen posts on CA stating a dedicated Wireless channel can beat Fibre Optic for full isolation. But again surely that comes at a trade off as you need kit to run the wireless sender/receiver set-up so what power supplies etc are needed there to stop reintroducing noise.

Be interesting to see what the folks here think works best.

Cheers,

Alan


#4
Seegs108 said

I have an I²S PCIe sound card with it’s own linear power supply and it sounds great. The PC has no bearing on power supply …


Thanks for your comments.

So your I²S PCIe sound card is inside your computer but has a separate (external?) linear power supply. It’ll be picking up some computer noise merely by virtue of being located inside the computer’s case so I guess that’s what you’ve thought of an optical HDMI solution.

I speculate that I²S is susceptible to jitter just like any other digital audio signal. I suppose that conversions from electrical to optical and back can worsen jitter so there might be a problem there. Having said that, the DS/Jnr reject jitter very well so that might not matter.

As for other reasons why the strengths I²S might be degraded by such a solution… like you I don’t know.


#5
BigAlMc said

I like many others use Fibre Optic isolation in the Ethernet path to try and isolate noise. But like many of these things it comes with a trade off.

Hi Alan. Thanks for your post.

Please can you tell us a little more about the model of fibre-optic isolators you use for your Ethernet?

The battery power supply is intriguing. Did you try a mains supply first (and if so would you tell us something about your mains setup) or did you go straight for battery?

Regards, Peter


#6

What kind of trade off?

I’m using a fibre optic connection (two Allied Telesyn media converter) as isolator, between router and streamer. It works flawless.


#7

Hi Peter,

The implementation is as follows:

QNAP NAS >Ethernet>TP-Link SG108 switch>Ethernet>TP-Link MC100CM (FMC)>1 metre Fibre Optic cable>TP-Link MC100CM>Ethernet > “end point” (end point in quotations as playing around between BII, MR & MR/SU-1 combo as per my other post)

The NAS is on stock PSU.

The two FMC’s are powered using TP-Link Vivid 6700mAh PS battery power supplies.

The switch is powered by an iFi Ipower.

I had posted a thread on CA asking whether Optical Isolation meant you did not need to worry about the power supplies upstream.

The floating battery PSU’s were a recommendation from a helpful person there and I went straight to these. So in answer to your question - no I haven’t compared to the stock supplies. I just accepted it made sense that the battery power supplies isolated the FMC from the mains. I do however hear a distinct drop in quality if I listen to music whilst the battery power supplies are charging, compared to not charging.

Also - ironically given I was trying to avoid investing in power supplies upstream I underestimated the old nervosa and I ended up buying an LPS (edit - fairly cheap one - not the Uptone LPS-1) for the NAS anyway embarassed. It should arrive soon as I ordered it off Ebay. Not sure it will make much difference other than one more ‘piece of mind’ item that I tried to care of one more ‘weakest link’ factor…

Hi @il-carletto

By “trade off” I meant the following. The optical isolation has definitely made a noticeable improvement in SQ and it worked perfectly from the start. However it comes at the price of having even more boxes and PSU’s behind my rack. And it means the signal is being converted multiple times. So whilst it’s a beneficial addition I can’t say for certain that the FMC’s for example don’t influence the signal in other ways compared to a more direct path.

Cheers,

Alan


#8
The switch is powered by an iFi Ipower
Alan - it's interesting that you say that. I used to have a pair of TP-Link MC100CS's (and fiber) between my network switch and my DSJ. And because I was paranoid about noise, I bought a pair of the iFi iPower adapters for the MC100's.

A while later I found myself chasing down some bad interference I was getting from my turntable. It’s got an unshielded Rega cart, so it’s pretty sensitive to RFI and such. And a very loud part of the noise was a vibration like sound. Kind of an electric rattle sound, as it were.

To my shock, I traced it back to the iPower units. So while they may supply clean power, be aware of the fact that they appear to kick up some pretty nasty RFI. At least mine did.


#9
Peter said

Thanks for your comments.

So your I²S PCIe sound card is inside your computer but has a separate (external?) linear power supply. It’ll be picking up some computer noise merely by virtue of being located inside the computer’s case so I guess that’s what you’ve thought of an optical HDMI solution.

I speculate that I²S is susceptible to jitter just like any other digital audio signal. I suppose that conversions from electrical to optical and back can worsen jitter so there might be a problem there. Having said that, the DS/Jnr reject jitter very well so that might not matter.

As for other reasons why the strengths I²S might be degraded by such a solution… like you I don’t know.

That's correct. The card was built for audiophiles in mind. It has a 4 pin molex connector from which it receives all it's power. If you simply have it plugged into the PCIe port without something connected to the molex connector the motherboard doesn't even know it's there. It only uses the 12 volt power portion of the molex connection and any 1 amp linear power supply will suffice to power it. The one I have does 1.25A at 12 volts so it's more than enough. You can choose to power it from your computers power supply as it uses a standard molex connector or use an external power supply like I've done if you want to take things to another level. All you need is a DC Female to Molex female cable adapter so you can use a regular 5.5x2.1 DC power plug that would come with any linear power supply. Such a cable only costs $5. It doesn't draw any power for the cards functionality from the PCIe lanes and the PCIe data lanes are heavily filtered by the card to get noise down to an extremely low level (if such noise is even present). Tie this in with the extremely low noise linear power supply I'm using and I don't see how it could get much better? But I suppose that's the point of this thread? Will an optical HDMI cable solution help?

With that said, I don’t know how much “noise” extra from my PC that it would add into the data. Considering I’m using a Seasonic power supply for all other components in my system, which are well regarded PSUs that have low measured ripple noise and top performance in other areas related to non-linear power supplies, that there are no mechanical devices in my PC (I’m using solid state harddrives and it’s passively cooled so no fans), and that there are no other components being shared by the PCIe lanes, I don’t see where extra noise would be coming from?


#10
scolley said
The switch is powered by an iFi Ipower

Alan - it’s interesting that you say that. I used to have a pair of TP-Link MC100CS’s (and fiber) between my network switch and my DSJ. And because I was paranoid about noise, I bought a pair of the iFi iPower adapters for the MC100’s.

A while later I found myself chasing down some bad interference I was getting from my turntable. It’s got an unshielded Rega cart, so it’s pretty sensitive to RFI and such. And a very loud part of the noise was a vibration like sound. Kind of an electric rattle sound, as it were.

To my shock, I traced it back to the iPower units. So while they may supply clean power, be aware of the fact that they appear to kick up some pretty nasty RFI. At least mine did.


Thanks scolley,

Yeah I’ve read mixed reports about the Ipower but figured the cost to cost ratio of what I was trying to improve was worth a shot. I have two. One powering my switch. The other powering my router. [My thinking being I didn’t want to buy a 400 buck LPS to improve a 50 buck bit of kit].

I’m not entirely sure that both/either really needs an LPS. But I’ve read stuff along the lines of whilst it’s great to try and remove noise (isolation, regenerator, reclocking etc) but it’s better to not introduce that noise in the first place.

So far so good with my Ipowers. But good to know that RFI is something to watch for.

Cheers,

Alan


#11
BigAlMc said

Hi @il-carletto

By “trade off” I meant the following. The optical isolation has definitely made a noticeable improvement in SQ and it worked perfectly from the start. However it comes at the price of having even more boxes and PSU’s behind my rack. And it means the signal is being converted multiple times. So whilst it’s a beneficial addition I can’t say for certain that the FMC’s for example don’t influence the signal in other ways compared to a more direct path.

Cheers,

Alan

I agree with you that we have more boxes and PSU’s. I’m using the stock switching Psu with the FMC on the router side and a linear HD-Plex 100W on the streamer side.

I’m not worried about the signal double conversion, instead. It’s not an analog signal, in an ethernet environment the Frame Check Sequence (FSC) checks the errors in the data packets


#12
Seegs108 said
With that said, I don't know how much "noise" extra from my PC that it would add into the data. Considering I'm using a Seasonic power supply for all other components in my system, which are well regarded PSUs that have low measured ripple noise and top performance in other areas related to non-linear power supplies, that there are no mechanical devices in my PC (I'm using solid state harddrives and it's passively cooled so no fans), and that there are no other components being shared by the PCIe lanes, I don't see where extra noise would be coming from?
Thanks for an interesting and informative description of your computer. (I'm sure it'll set a few forum users off to modify their PCs!)

As I understand it when two pieces of electronics are close to each other, e.g., inside the same case, the possibility exists for inductive, capacitative or electromagnetic (‘radio’) coupling between parts, even when connecting conductors are minimised, filtered or even absent. I think this can be a source of electrical noise which can then pass out of the components along signal or ‘ground’ wires. This is why obsessives like me seek galvanic isolation—no wires! I have no idea whether your carefully put together computer will produce it and if it does whether it could degrade sound sent through the I²S input of a DirectStream but in the general case it’s possible.


#13
scolley said
I used to have a pair of TP-Link MC100CS's (and fiber) between my network switch and my DSJ.
I've been looking up these units (available on Amazon). Although the price differences aren't large the ones that connect to Gigabit Ethernet (on the electrical side of the interface) are more expensive than those that connect only to 10/100 Ethernet. So I looked up the DirectStream Junior spec's but can find no indication of the Ethernet port's speed. Does anybody know?

#14

The Jr and Sr Bridge I and II use 10/100, plenty of bandwidth for a maximum 11MHz bit rate signal - faster just generates more RFI…


#15

I use one of these between BridgeII and a network switch:

http://www.artistic-fidelity.de/index.php/en/giso-isolator

Works great and makes a small but positive improvement to SQ.


#16

I wonder whether there are any interference-blocking differences between a transformer-based isolator such as the artisitic fidelity GISO referred to by markus46 and an optical isolator such as the TP-Link devices referred to by scolley

The other thing to think about is that a transformer-based isolator (presumably) needs no power supply whereas an optical isolator does and power supplies can be trouble!


#17

Hi @markus46

Do you know if this GISO isolater differs from the medical grade LAN isolators such as the Baaske Network Isolator MED MI 1005?

Looks like a similar principle. Am just wondering if the GISO brings more to the table by being developed with picky audiophiles in mind? wink

Cheers,

Alan


#18

More cost?


#19

Hi Alan,

Sorry, I don’t know how the GISO isolator compares to other passive or active solutions.

I originally heard about it in the Meridian Sooloos forums. Most folk that tried one only had positive things to say about it, so I got my own and am glad I did

Cheers

Mark

BigAlMc said

Hi @markus46

Do you know if this GISO isolater differs from the medical grade LAN isolators such as the Baaske Network Isolator MED MI 1005?

Looks like a similar principle. Am just wondering if the GISO brings more to the table by being developed with picky audiophiles in mind? wink

Cheers,

Alan


#20

Thanks Mark,

Yeah I’ve seen similar posts on CA about the Baaske one and other similar ones.

Interestingly I saw a few people recommend placing it between the Modem & Router in order to shield from noise coming into the router from the telephone/cable wires coming into the building.

I bought the Baaske one (hence my question). In my setup I have a Router with inbuilt modem so the CA suggestion wasn’t an option. I therefore added mine between the Router and Switch. Not sure if that’s the best place or not.

tbh I didn’t hear any difference but I can’t say I put much effort into a/b testing it. It cost $50 on ebay so I wasn’t too worried.

Perhaps your implementation of having your filter before the BII makes more sense. I figured since I have optical isolation before the BII I’d try to isolate the LAN signal coming into the switch (for Tidal and whatnot as I have Roon installed on my NAS).

I should probably play around with it in different positions to see if I hear a difference. But meantime it’s one more expense on the ‘not sure if it really adds anything in SQ but gives me peace of mind’ list embarassed

Thanks for the feedback.

Cheers,

Alan