Affordable silent line out sound card

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
If your amplifier is galvanically isolated from the protection ground, most soundcards are dead silent, incl. internal ones. See my measurements of a thin client PC internal soundcard connected to class-T amp powered by independent adapter without the PE line hifi.slovanet.sk :: Zobrazi� tému - AIO Internetové rádio + CD (the picture names + labels are self-explaining). Nothing special (50 USD overall costs) but certainly not very noisy.
 
If your amplifier is galvanically isolated from the protection ground, most soundcards are dead silent, incl. internal ones. See my measurements of a thin client PC internal soundcard connected to class-T amp powered by independent adapter without the PE line hifi.slovanet.sk :: Zobrazi� tému - AIO Internetové rádio + CD (the picture names + labels are self-explaining). Nothing special (50 USD overall costs) but certainly not very noisy.

Yes, toroidal transformers PSU and ungrounded (dead silent with an ibasso DAC)
I always hear small scratching noises in the background approximately 20 cards tested now... and all of them have fantastic reviews/measurements and people saying that their cards were silent ;) :eek:
 
Have you considered that the problem may be the amplifier(s) that you are connecting to?

I have experienced widely differing noise and hum pickup levels when connecting computer based audio gear to consumer amplifiers that I own. It seems to come down to grounding schemes and whether or not there is any RF filtering on the input. At least that is what I attributed to the differences that I discovered...
 
What is the purpose of the line transformer, if the ungrounded amplifier has no ground loop? I do not understand why the line transformer should make any difference. Unless the amp is not so much ungrounded...

I think it is due to the low impedance output loading and the natural RF rejection of the coil.

Have you considered that the problem may be the amplifier(s) that you are connecting to?

I have experienced widely differing noise and hum pickup levels when connecting computer based audio gear to consumer amplifiers that I own. It seems to come down to grounding schemes and whether or not there is any RF filtering on the input. At least that is what I attributed to the differences that I discovered...

On the lynx cards, the analog section seems to be independent of the rest of the board and totally decoupled (FB111/114/119/101 caps)

http://prosound.ixbt.com/interfaces/lynxstudio/e44/image002.jpg


Wow, what a mystery ! :devilr::p;)
 
Last edited:
It cannot be fully decoupled unless the analog section is powered by another galvanically separated power supply. It would have to be powered from the DB25 connector. Is it the case?

HF damping does not remove the audible LF noise caused by the motherboard activity. I have experienced the noise almost always when the amplifier shared another ground connection with the PC than just the line-out shield. Or if the soundcard output was touching the PC case in the output slot, causing a ground loop within the card itself. What amp are you using?
 
It cannot be fully decoupled unless the analog section is powered by another galvanically separated power supply. It would have to be powered from the DB25 connector. Is it the case?

HF damping does not remove the audible LF noise caused by the motherboard activity. I have experienced the noise almost always when the amplifier shared another ground connection with the PC than just the line-out shield. Or if the soundcard output was touching the PC case in the output slot, causing a ground loop within the card itself. What amp are you using?

PSU regulators have a very high PSRR, they don't need decoupling.
Line out are symetrical and each section on the signal path is decoupled... it is a fully decoupled floating desing.
 
So is this the case?

A) symmetrical output from lynx to your symmetrical amp input - no noise

B) single-ended output from any other soundcard to your single--ended amp input - noise

That would be a typical example of ground loop IMO. Which if broken by the line transformer - no noise

IMO you are having a ground loop. Even if your amp is galvanically separated from PE (did you actually measure it with an ohmmeter?), some capacitance coupling possible?
 
Long winded reply --

Basically, what happens in most cases is that there is noise current traveling on the lineinput ground return wire (usually, the shield in coax wire). There are lots of things that power off the the AC mains in the house, and most of that stuff also connects to the safety ground ("green wire" in the USA, to reduce risk of electrocution) and passes noise currents, such as from switching supplies or capacitive coupling of AC line waveforms, into the safety ground.

But another path for that noise current is via the other stuff through the audio input ground return wire of your soundcard since the computer connects that internally to safety ground at some point, too. The noise current could be sourced from inside the computer (its power supply, noise conducted out to other stuff to get to the stuff's safety ground) or outside (other power supplies conducting noise into the soundcard en route to its safety ground). While the major part of the noise current flows via the green wires in the power cords or adaptors, a finite part of it necessarily flows through your line input ground return (by being a leg of a current divider), and since that return line has finite impedance (milliohms) a noise voltage is generated across the return line, which then appears directly across the soundcard input. Heavier input shield line could help, but it's hard to drop a milliohm-level impedance by much, and even then, you'd have to cut it in half to even drop the noise by only 6dB. Milliohms are small and the noise current is small, but it doesn't take all that much to kill the effective SNR of a soundcard that is supposed to be >96dB. (edit: another entry for noise current is through inductive pickup on the shield lines conducting between both audio and safety ground lines, but the same discussion still applies).

It is very unsafe to remove safety ground from equipment (and your insurance company will just laugh at you if you do and then something bad happens!). Even isolation transformers on the AC power feeding your computer may not fix the issue since a lot of the noise has higher frequency or RF components and isolation transformers could still capacitively couple HF noise.

The best fix for this kind of noise is to use balanced line driven inputs since they don't include the impedance of the line input return in the effective input circuit. This can also be obtained by use of an external unbalanced to balanced audio line transformer at the sound card (with some loss of bandwidth and some impedance complications).

Another approach is to have a quasi-balanced input system built into the gear. This is like the "Unbalancer" circuit John Broskie talks about at tubcad.com, or like the circuit I showed in my "ground loop noise" blog here. The idea there is to suppress the noise current from traveling down the input return line by including a series resistance in the gear that appears in the connection between input return and safety ground, but that doesn't appear in the audio input return circuit. I've also seen such series-impedance managed input return circuits in other audio gear.

That kind of input circuit might be what Lynx uses. I had a Lynx 22 for several years and it is not completely immune to ground loop noise if connected via single ended circuits! Better than some cards, if I remember correctly, but not total. So don't assume a given brand will just 'work' for you because it did in someone else's setup...
 
Last edited:
By no means I am suggesting to disconnect the PE wire!

My audio systems fall into two categories:

1. Low power thin client PC (Fujitsu Siemens Futros, HP T5740) powered by original AC adapters without the PE wire, feeding full-blown amplifiers with 3-wire power cords - no ground loop possible

2. Regular desktop PCs (3-wire power cords) feeding efficient class-T amps powered by external properly isolated PSUs without the PE wire - no ground loop possible either.

Every time I tried to power the amp from the PC or hook a PC to an amp powered from a different power socket, I experienced the ground loop to some extent (PC activity audible). On the other hand, when either the sources or the amp uses no PE wire (by design!), there was no PC activity audible even with on-board sound cards.

With the amp all the way up the sound cards did differ in white noise level (just like any source/amp does), but not the typical PC-activity scratches, cracks, snaps.
 
Last edited:
Moderator
Joined 2002
Paid Member
I am unsure what you mean. My Asus Xonar DX - by no means an expensive card has about -110dB DR and a pretty flat noise floor that maxes at about -112dB. Some artifacts in LF because of measurement loops do show up and the card does not meet manufacturer spec, but is still quite low noise.

My e-Mu 1212m has a noise floor of about -130dB. There are measurements made by many forum members. The eMu 0404 which I use for measurement duties temporarily, is 'only' about -125dB, but I consider that pretty decent for a such a low cost card.

Can you give some idea about what number you are shooting for when you say 'silent'? Does it mean a spec, or something you can hear through the speakers? If so, what kind of noise? Is it a hum, or low-level white noise? Or something else? Or do you mean an actual verified measurement?

The only time I have heard PC activity noise through a sound device is when the power regulators on the motherboard are so poor they couple noise from one load to the other. This is by no means the cards' fault, but the motherboard or the computer's power supply.
 
I am unsure what you mean. My Asus Xonar DX - by no means an expensive card has about -110dB DR and a pretty flat noise floor that maxes at about -112dB. Some artifacts in LF because of measurement loops do show up and the card does not meet manufacturer spec, but is still quite low noise.

My e-Mu 1212m has a noise floor of about -130dB. There are measurements made by many forum members. The eMu 0404 which I use for measurement duties temporarily, is 'only' about -125dB, but I consider that pretty decent for a such a low cost card.

Can you give some idea about what number you are shooting for when you say 'silent'? Does it mean a spec, or something you can hear through the speakers? If so, what kind of noise? Is it a hum, or low-level white noise? Or something else? Or do you mean an actual verified measurement?

The only time I have heard PC activity noise through a sound device is when the power regulators on the motherboard are so poor they couple noise from one load to the other. This is by no means the cards' fault, but the motherboard or the computer's power supply.

Did u get those numbers via self loopback?

Try having the two cards measure each other. I won't be surprised if they measure like onboard sound.

Add: single ended connections. Going balanced will avoid the problem as already mentioned.

Add2: To quote some guy, "If you can hear the noise from the speakers, something is really wrong". Onboard sound typically handle at least -80dB and up to -100dB, being limited by common mode noise, but that is at full range output. The gain structure some people use throughout their entire setup can make the noise end up perhaps 20dB at the speaker.
 
Last edited:
I've read all your answers THX (too much to quote).

You shoud perhaps stay aware that we all have very disparate AC quality, and a simple bad decoupled motor in your house commodities can ruin your (ideal situation) measurements.

My AC plugs are ungrounded (i must deal with), so i've designed all my audio stuff with double or triple insulation.
I've tested a balanced output of my mackie sound card and it is perfectly silent. It seems to be the right solution but balanced output professional cards are painfully expensive:mad:
 
Last edited:
Moderator
Joined 2002
Paid Member
Did u get those numbers via self loopback?

Try having the two cards measure each other. I won't be surprised if they measure like onboard sound.

Can't measure two e-mu cards with each other, the drivers do not allow two cards in the same PC.

However I am attaching an FFT of a preamplifier being measured for performance at 4V out. The 0404 is both the source and A/D for this test.

The 1W/8ohm test is a real-world connection example using a F5-type amplifier and unbalanced connections. For this test the 0404 is the only unit in service apart from the amplifier.

Also attaching a loopback baseline for the 0404 using RMAA. For this the digital output is fed to a Benchmark DAC2 and then the output is fed back to the 0404's A/D.

I did measure some onboard solutions at one point, but I rarely got a noise floor better than about -90dB. Even that would not create noise (that I know of or could reliably hear).
 

Attachments

  • FFT 1k 4VRMS.png
    FFT 1k 4VRMS.png
    36.7 KB · Views: 128
  • 0404 baseline.JPG
    0404 baseline.JPG
    118.9 KB · Views: 135
  • THD vs FR 1W 8 ohm - log freq.png
    THD vs FR 1W 8 ohm - log freq.png
    91.8 KB · Views: 128
However I am attaching an FFT of a preamplifier being measured for performance at 4V out. The 0404 is both the source and A/D for this test.

The 1W/8ohm test is a real-world connection example using a F5-type amplifier and unbalanced connections. For this test the 0404 is the only unit in service apart from the amplifier.

Not a real-world connection example. Your final receiver and your initial source are sharing the same ground point, hence avoiding the common mode noise. But I agree that in a real-world connection, even with single-ended connections, if done correctly, it is possible to avoid the common mode noise, even with a computer source.

But we don't always get to decide whether the receiver + source combination will cause problem or not. That's why isolating the connection with SPDIF is useful as you have demonstrated.

Below is a CS4398 DAC measured with Xonar DX. SPDIF is used with both, but the better one is a proper isolated USB-SPDIF converter, while the top is the cheap unisolated connection from onboard sound.

ljm-cs4398-spdif-noise2.png
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.