Firewire Ground

[pcy]

Hi,

I have a firewire cable that is completing a ground loop, and causing a lot of nasty noise.


The Firewire cable is connecting a PC to an external sound card with its own power supply and its own connection to the mains, so I tried cutting the ground in the cable.

But it seems theat Firewire needs the ground - without it the PC does not detect any external F'Wire device. Correction - it detects that soemthing is there, but cannot recognize it or communicate with it properly until you reconnect the ground.


So two questions arise:

1. What is the ground used for, other than 0V in a powerd F'Wire Cable and shield?
2. Can you add an impdance to the ground (resistance, capacitance, indictance) to cut out the noise but retain whatever function the ground performs.




Thanks




Peter

Moved to General Hardware - Moderator Rubycon

 

[Rubycon]

Are you hearing the noise with just the interface connected to the PC and listening to one of the outputs with headphones? Most loops are caused by peripherals connected to interfaces. Best bet is to used balanced inputs only. If your inputs are unbalanced a converter can be used that will prevent ground loops and improve SNR considerably.

The ground is required and should NEVER have resistance added.

 

[mpilchfamily]

If your getting noise its proabably from lack of sheilding not a ground issue. As you may know firewire supplies some voltage. The ground is the voltage ground or negative side of the voltage. Its kind of a must have.

You may want to try a feritte ring or bead to help sheild the cable.
http://en.wikipedia.org/wiki/Ferrite_bead

 

[pcy]

Hi,

The noise occurrs whenver any significant change is made to the screen.

This is a professional Audio setup. All the analogue cables are already balanced.

A 6 pin F'wire cable does indeed supply power, but this is not being used. The system works fine if you cut the (white) power cable and all the shields in the F'wire cable, provided you leave the black ground cable connected.

This is a well known problem. Only the solution is in question.

I'll try a ferrite core, though the F'Wire cable is already well shielded.

The ground in the F'Wire cable is not not providing a ground to the F'wire device, which is powerd from the mains, and grounded via the mains plug.

Originally posted by: RubyconThe ground is required and should NEVER have resistance added.

Yes, but what is it required for? The ieee1394 documentation does refer to the ground being connected via a 1Mohm resistor to limit the current flowing in some F'wire circuits.

The question is:

1. What, exactly is the ground being used for - the ieee 1394 specs do not mention it's use
2. What conmination of resistance, capacitance or induction, if any, can be placed in the ground connection to stop it from picking up whaterver EMF is coming from the Graphics Card or Monitor that cases this problem.

 

[Rubycon]

What interface is this?

We have dozens of pro audio stations using 1394 and never have had this problem. The problem may be within your interface itself. You should NEVER hear buzzing when things change on the screen.

Do you have another port to connect your interface to on the PC and interface as well? Most interfaces will have a pair of six pin connectors.

 

[pcy]

Hi,

The interface is a RME Fireface 800.

The Motherhboard is an Intel DP35DP

The Graphics Card (which is the most likely culprit to be pushing out the EMF in teh first place) is a GigaByte 8600GT.


Yes - I have tried several different cables (and cut one up to check it was constucted correctly) connected to all available F'Wire ports.

Originally posted by: Rubycon
We have dozens of pro audio stations using 1394 and never have had this problem. The problem may be within your interface itself. You should NEVER hear buzzing when things change on the screen.

I completely agree that you should never hear a buzzing of this sort, but in fact it's a well known (though rare) problem.

For instance:
www.soundonsound.com/sos/dec08/articles/qa1208_2.htm


I'd say this problem has occured twice in machines I've built over the last 10 years, but my sample size is too small to know if that's statistically significant. You need to be shipping 100s units per annum to know how often this problem actually occurrs.


Thanks



Peter

 

[Rubycon]

Does the interface have this issue on another PC? Have a firewire equipped notebook handy? There will be absolutely no loops if the computer can run off a battery.

Some cards are definitely noisier than others. I've used extremely noisy 8800 ultras (the card makes physical noise - oscillating, inductor sing/coil whine, etc.) yet nothing is perceptible in the recorded output(s). The board itself may just allow some of this noise to creep in through the interface or the (firewire) grounds themselves may be partially floating, etc.

 

[Schmide]

Go in your bios and disable the CPU, PCI-E, LDT, SATA Spread Spectrum. You can play around with various combination to see if one affects another. If it solves your problem, it's doubtful that you would get a major hit from disabling it.

 

[Googer]

Try a dedicated PCI-e based 1394b card with a Texas Instruments Chipset. It may just solve your problem.

 

[pcy]

Originally posted by: Rubycon
Does the interface have this issue on another PC?

No

Have a firewire equipped notebook handy? There will be absolutely no loops if the computer can run off a battery.

Indeed

Some cards are definitely noisier than others. I've used extremely noisy 8800 ultras (the card makes physical noise - oscillating, inductor sing/coil whine, etc.) yet nothing is perceptible in the recorded output(s). The board itself may just allow some of this noise to creep in through the interface or the (firewire) grounds themselves may be partially floating, etc.

I would not expect it to. There may be some statistical relationship between noisy cards in the sense you use it, and silent EMF production that teh F'Wire cable is reacting to, but it is stistical, no necessary. This is a rare problem - you'd need 100s of noisy 8800s all using an external F'Wire interface to establish the correleation.

Originally posted by: RubyconThe ground is required and should NEVER have resistance added.

Please, if you know what the ground actually does as your initial response implies, can you please answer my question and tell me what it's being used for. WE know that it is not required as:
1. The 0v of teh F'Wire Power (not bing used in this case)
2. A shield (cutting those has no effect)
3. To provide a Ground to the FFF 800 (independantly grounded via the mains)



Thanks


Peter

 

[Rubycon]

The ground is required for bus powered devices. If your interface is self powered the ground should not be required. Your device should be detected without bus power BUT you need to connect it with a four to six pin cable plugged into a four pin connector on the PC. Notebooks with integrated 1394 will have these type of connectors as do some PC's and breakout brackets used on some PC's. Try using this type of cable.

Getting back to the root of the problem - the ground loop itself - should not be ocurring in the first place. Your PC ground and interface ground are connected together with sufficient differential to allow EMI/RFI to creep into you audio. Always make sure the PC power supply as well as the power supplies on everything in your chain are connected to the same mains ground.

 

[pcy]

Hi Rubycon,

Originally posted by: RubyconThe ground is required for bus powered devices.

Indeed, obvioulsy.

If your interface is self powered the ground should not be required.

That is what I would have thought. But it seems that in fact the ground is always required. Once I had identified the problem, I found a number of people on the net reporting the same finding. Hence my question:

What is the ground being used for?

Your device should be detected without bus power BUT you need to connect it with a four to six pin cable plugged into a four pin connector on the PC.

Indeed - that is in effect the cable I have constructed.

However, a four pin F'Wire cable does in fact have five connections according to the 1394 spec - there is a ground as well as the two twisted pairs; and as we now know this is in fact required.

Notebooks with integrated 1394 will have these type of connectors as do some PC's and breakout brackets used on some PC's. Try using this type of cable.

In effect, I have.

When the cable with just the four twisted pair cables connected failed to work I built a second cable which had all the connections pass through a cct board, so that I could make and break the connections easily. Using this cct I was able to establish, for a whole range of F'Wire devices, that the ground plus the four twisted pair connections are needed, but nothing else is necessarily essential.

I must emphasize that this is a design issue in electrical/electronic engineering. not a computer problem per se.

The question remains:
What is the ground being used for?

Getting back to the root of the problem - the ground loop itself - should not be ocurring in the first place. Your PC ground and interface ground are connected together with sufficient differential to allow EMI/RFI to creep into you audio. Always make sure the PC power supply as well as the power supplies on everything in your chain are connected to the same mains ground.

Yes, of course...

Yes - we agree this should not be occurring: but the fact is that it does, and to a significant number of people, though still a tiny proportion...
Yes - trying to use the same mains ground may help. Of course I have tried that. However - since the PSU enclosure alone in a PC is actually grounded, and Motherboard ground is in fact simply 0v off the PSU low voltage side, there is ample opportunity for the grounds in different (independantly powered) devices to be at different levels.



Peter

 

[pcy]

Hi,

Originally posted by: Schmide
Go in your bios and disable the CPU, PCI-E, LDT, SATA Spread Spectrum. You can play around with various combination to see if one affects another. If it solves your problem, it's doubtful that you would get a major hit from disabling it.

The Motehbaord uses a TI chipset for F'Wire, but I'll give it a shot, thanks.


Peter

 

[pcy]

Originally posted by: mpilchfamily
If your getting noise its proabably from lack of sheilding not a ground issue. ...

You may want to try a feritte ring or bead to help sheild the cable.
http://en.wikipedia.org/wiki/Ferrite_bead

The F'Wire cable seems pretty well shielded, but I've ordered a ferrit core to see it it helps.


Thanks.



Peter

 

[Rubycon]

Have you tried floating the ground on your pc? I would only recommend doing this for testing purposes ONLY as fault can light up the box and cause a hazardous condition.

The ground on the six pin is required for powered devices hence it not showing on specs. It has nothing to do with 1394 other than provide power for devices attached to the bus which need it and are not self powered. Onboard 1394 has limited power whereas an expansion card can provide more power if equipped with an aux power connector (floppy type power connector for PSU).

 

[pcy]

Hi,

Originally posted by: Rubycon
Have you tried floating the ground on your pc? I would only recommend doing this for testing purposes ONLY as fault can light up the box and cause a hazardous condition.

Yes - I have thought of cutting the ground on either the PC of the FF800.

It's a bit of a performance, as the leads are fully moulded - I'd probably use an extension lead with an ordinary plug so I could take it apart and disconnect the earth.

I will try that, but I don't expect it to work, as the Motherboard of a PC is not in fact earthed. Only the PSU enclosure is actually earthed, and the "ground" of the PC itself is the 0v line of the PSU - so it's bound to the neutral line of the mains as much as anything.

The ground on the six pin is required for powered devices hence it not showing on specs.

Sorry... I cannot figure out what you mean by that sentence. Did you mean "required only for powered devices".

It has nothing to do with 1394 other than provide power for devices attached to the bus which need it and are not self powered.

You keep saying this, but I have compelling (though not, I admit, conclusive) evidence that the F'Wire ground is in fact required for F'Wire to work.

Indeed that was my original question:

What is the the Ground of the F'Wire cable being used for, that makes the Ground necessary for independantly powered F'Wire devices to work?

You seem adamant that the answer to my questiuon is "nothing". I understand why you might think that - it was my view also, until I tried it. Have you tested this - i.e built a F'Wire Cable with just the four twisted pair connections (no power, ground or shield). If you have, and you can tell me the name of any Motherboard + F'Wire device pair that works with such a cable, that would be extremely useful information.

Onboard 1394 has limited power whereas an expansion card can provide more power if equipped with an aux power connector (floppy type power connector for PSU).

Indeed, but:

So?... My question has nothing to do with devices powered via a F'Wire cable: maybe I'm missing something here, but I fail to see the significance.





Peter

 

[Rubycon]

I always have premade cables and never have issues. We also have fully balanced power. Once you determine that you have a power issue you should really correct that issue.

The ground is required for a six pin connector. 1394 only needs four wires to communicate however a bus powered device adds two more (+5V and ground).

Your device may need the presence of 5 volts in order to turn on its interface - without this the PC cannot see it.

If you have a notebook or PC with a four pin interface you can see for yourself that it works without the extra lines.

 

[Modelworks]

A ferrite core will not remove noise from a ground loop, only rf noise.
A really easy way to remove the ground loop is to plug the audio gear into a isolation transformer.
http://www.amazon.com/exec/obidos/ASIN/B00006HPFH/ref=
http://www.pacificgeek.com/pro...16&s=1068&ID=21514&P=F

There are cheaper transformers, just need one that is 120VAC in and out with a high enough current rating.

I looked at that interface and I hope the designers were smart enough not to make the ground on the jacks and connectors the same as the ground for the AC line. They should be completely isolated from AC ground . If not I would definitely buy a isolation transformer.

 

[pcy]

Hi Modelworks,

Thank you. You post moves the discussion forward.

Originally posted by: Modelworks
A ferrite core will not remove noise from a ground loop, only rf noise.

Is it possible, though, that the F'Wire cable is in fact picking up RF interferance being put out by the Graphics Card?

A really easy way to remove the ground loop is to plug the audio gear into a isolation transformer.
http://www.amazon.com/exec/obidos/ASIN/B00006HPFH/ref=
http://www.pacificgeek.com/pro...16&s=1068&ID=21514&P=F

There are cheaper transformers, just need one that is 120VAC in and out with a high enough current rating.

As it happens I'm in the UK, so I need a 240VAC unit, but now I know the sort of thing to look for...

I looked at that interface and I hope the designers were smart enough not to make the ground on the jacks and connectors the same as the ground for the AC line. They should be completely isolated from AC ground . If not I would definitely buy a isolation transformer.

It's a quality piece of equipment - we can only hope though.

Thanks



Peter

 

[pcy]

Hi Rubicon,


I'm afraid I don't understand:

Originally posted by: Rubycon
I always have premade cables and never have issues.

How many systems are we talking obout here. 100s, 1000s. This is a rare problem: you need a big sample to have any confidence that an absense of any issues is statistically significant.

We also have fully balanced power.

What, please is "fully balanced power". I know what balanced cables are: but balanced power is not a term recognized by audio or electrical engineers here in the UK.

Once you determine that you have a power issue you should really correct that issue.

I don't have a power issue, as I have been at pains to point out.

The ground is required for a six pin connector. 1394 only needs four wires to communicate however a bus powered device adds two more (+5V and ground).

How do you know this? If you only use pre-made cables, all your F'Wire cables have a ground. As I have poionted out several times, the four pin F'Wire Cable actually has a ground as well.

Your device may need the presence of 5 volts in order to turn on its interface - without this the PC cannot see it.

No. It works fine if the twitsted pair and ground only are connected. Disconnecting the +ve power has no effect.

This behaviour is repeatable with every combination of F'Wire device and Motherboard I have so far been able to test. This leads me to the conclusion that the ground is being used, as standard, for some purpose. Maybe it is a reference voltage against which the voltages in the twisted pairs are measured. Maybe...

If you have a notebook or PC with a four pin interface you can see for yourself that it works without the extra lines.

The F'Wire specs I see quite clearly show the 4 pin interface as having 4 pins plus a ground. Maybe those web sites are all wrong; but if you are so certain can you please cut the ground and shields in your 4 pin F'Wire cable and let me know if it still works.



Thanks



Peter

 

[Modelworks]

Originally posted by: pcy

Is it possible, though, that the F'Wire cable is in fact picking up RF interferance being put out by the Graphics Card?

It is definitely possible for RF to get into the wiring. Is the port on the computer connected directly to the chipset handling the firewire ? Or is it connected to it through another cable ? I ask because some motherboards have firewire on the board as a set of pins that connect to another cable the provides the firewire port connector. That is a likely source of noise since those internal cables are usually poorly shielded.



Also about the grounding, it is very similar to usb.
USB has DATA+ , DATA- , 5V, GND. The GND is not needed for usb to work and the DATA - wire is not connected to system GND.

Ground is not required on a firewire cable for communication either. It uses 4 wires to form 2 communication channels , where USB only has one channel. One is TPA+ and TPA- and the other TPB+ and TPB- they do not connect to the main ground at all.

 

[Rubycon]

Balanced wiring basics here.

It should not matter if there is RF noise coming down the line - this should never get to your interface outputs! Something is definitely wrong with your setup or the host PC. Check the interface on another system. The purpose of using 1394 for recording interfaces is to get the box out of the box so to speak and keep noise floors to an absolute minimum PERIOD!

As Modelworks pointed out in the last post ground is NOT required for communications. Cables are shielded and the shield is tied to the chassis ground at both ends. These chassis grounds must be at the same potential or there will be risk of noise from ground loops.

Isolation transformers will work but should not be necessary unless you have a specific wiring problem. If all devices are wired to the same feed and wiring is good then you have an internal issue.

 

[pcy]

Originally posted by: Modelworks

Originally posted by: pcy

Is it possible, though, that the F'Wire cable is in fact picking up RF interferance being put out by the Graphics Card?

It is definitely possible for RF to get into the wiring. Is the port on the computer connected directly to the chipset handling the firewire ? Or is it connected to it through another cable ? I ask because some motherboards have firewire on the board as a set of pins that connect to another cable the provides the firewire port connector. That is a likely source of noise since those internal cables are usually poorly shielded.

Good question. In fact we tried both the F'Wire port on the back IO panel and the front port connected to a header via an internal cable.

It made no difference at all.

Incidentally, the internal cable is in fact well shielded - each twisted pair has it s own shield, plus a third shield for the complete cable. By contast, many pre-made F'Wire Cables have just the overall shield.

Also about the grounding, it is very similar to usb.
USB has DATA+ , DATA- , 5V, GND. The GND is not needed for usb to work and the DATA - wire is not connected to system GND.

Ground is not required on a firewire cable for communication either. It uses 4 wires to form 2 communication channels , where USB only has one channel. One is TPA+ and TPA- and the other TPB+ and TPB- they do not connect to the main ground at all.

That's what I thought until I tried it. I built a F'Wire cable that that passed through a cct board. Withe just TPA+/- and TPB+/- connected it does not work. Connect ground, and it works fine. Disconnect the ground and the device vanishes, re-connect it and it works again...

I only have a few F'Wire devices here, so I don't know how universal this phenomenon is. But I get this behaviour on every combination of Motherbaord and F'Wire device I have available to test.

I have been told that the ground provides a referance voltage to enable the recieving F'Wire chip to measure the voltage in the twisted pairs to resolve a 0 from a 1.

Is this possible?



Thanks




Peter

 

[pcy]

Hi Rubycon,

Originally posted by: Rubycon
Balanced wiring basics here.

Thanks. Makes sense. Sadly it's not available in domestic premises in the UK, and maybe not at all.


Peter

 

[Modelworks]

Originally posted by: pcy


That's what I thought until I tried it. I built a F'Wire cable that that passed through a cct board. Withe just TPA+/- and TPB+/- connected it does not work. Connect ground, and it works fine. Disconnect the ground and the device vanishes, re-connect it and it works again...

I only have a few F'Wire devices here, so I don't know how universal this phenomenon is. But I get this behaviour on every combination of Motherbaord and F'Wire device I have available to test.

I have been told that the ground provides a referance voltage to enable the recieving F'Wire chip to measure the voltage in the twisted pairs to resolve a 0 from a 1.

I haven't worked directly with firewire circuits only USB. It is possible that the device on the other end is using it for reference . I tried to look at the data sheets for several of the firewire chipsets and they are not available without contacting TI. That is one of the reasons firewire is dying off, it is a proprietary interface and getting engineering info on it is like pulling teeth.