Weird distortion in headphones when speakers are off

[Rantenti]

Hi all,

I have an ifi iDSD Signature connected to Neuman KH80 active speakers via RCA-XLR cables.

There is a weird phenomenon where, when the speakers are OFF, and I plug in my headphones (any headphones), I would hear severe clipping distortion (not ground loop hum noise) when playing loud tracks with minimum headroom. However, the distortion in the headphones immediately and completely disappears if I do any of the following:

1. turn the speakers ON, or
2. disconnect the speakers from the DAC/amp
3. reduce the source (any source, eg. phone, computer etc) output to <50% (the magic number is always 50)
4. a preamp is inserted between the DAC and the speakers

I substituted the iDSD Signature with my older ifi micro iDSD and the same thing occurred.

I had the iDSD Signature checked and it was normal. I had written to ifi and they kindly responded that the problem was likely the Neumanns being sensitive to the voltage output from the iDSD.

A simple solution is to use a preamp but this added unnecessary circuitry to the chain and produces audible degradation of clarity and punch.

Another solution is to use two DACs (well, I do have two), one for the speakers and one for headphones.

However, being curious, I would appreciate if anyone can enlighten me about the technical / scientific cause of the problem and to see if anyone had encountered a similar problem too.

Thanks!

 

[AnalogSteph]

I would guess that the KH80 input may be using (normally reverse-biased) protection diodes going to the analog supplies. When powered down, those obviously are at 0 V, effectively turning the whole affair into a limiter (about 0.6 V peak assuming plain silicon diodes, specifics depending on source impedance).

Seems like the signal path in the iFi DACs is just being split between the pre-out and the headphone amplifier without adding an extra buffer (plus isolation resistor) for the former. This may have saved some power consumption, board space and BOM cost but it also makes the circuit susceptible to rogue loads. There are more units with this same exact problem out there, I have run across it before but can't remember which device it was (possibly a FiiO E10 or E10K?).

Who do you blame for a problem like this? iFi for neglecting to adequately buffer the pre-out? Neumann for potentially failing to include a few hundred ohms in series? Both to some degree, possibly? Good question.

I do see the the KH80s seem to be an IEC Class II device, so you should have few issues with ground loops at least.

 

[Rantenti]

Thank you for your explanation. I wouldn't start blaming any side for not including such buffering/ protective features as they are not obliged to do so, but I see this as an area of caution on the consumer side when buying equipment, and potential improvement on the manufacturer side. However, I wonder if these rogue loads damage the DAC? If so, manufacturers should consider adding a buffer.

As I am buying a new DAC and amp stack, I do consciously screen for output switches that allow me to choose to output to XLR only, RCA only or both, so that I can cut output to any speakers/ headphone amp which I am not using, then there won't be such problems any more (please correct me if I am wrong!).

As for the Neumann KH80s, I can hear an extremely soft hiss only if I put my ears within a few centimeters of the speakers. This is super quiet comparing to my old monitors. I do plug everything into one wall socket, no whines or hums and I'm happy.

By the way, since I was using RCA-XLR cables, I have read that "The easiest way is to solder a link between pins 1 and 3 (shield and negative) of the XLR, rather than trying to solder the shield and negative wire to the sleeve contact of the RCA. "

I wonder if there are clearly audible differences between these two ways of wiring, and does it matter for the cable to be used in either direction, RCA out to XLR in, or XLR out to RCA in? My Chord Clearway RCA to XLR cables were ordered online from a reputable shop, but I have no way of knowing how they were wired inside.

Have a nice day!

 

[EdW]

I would guess that the KH80 input may be using (normally reverse-biased) protection diodes going to the analog supplies. When powered down, those obviously are at 0 V, effectively turning the whole affair into a limiter (about 0.6 V peak assuming plain silicon diodes, specifics depending on source impedance).

Seems like the signal path in the iFi DACs is just being split between the pre-out and the headphone amplifier without adding an extra buffer (plus isolation resistor) for the former. This may have saved some power consumption, board space and BOM cost but it also makes the circuit susceptible to rogue loads. There are more units with this same exact problem out there, I have run across it before but can't remember which device it was (possibly a FiiO E10 or E10K?).

Who do you blame for a problem like this? iFi for neglecting to adequately buffer the pre-out? Neumann for potentially failing to include a few hundred ohms in series? Both to some degree, possibly? Good question.

I do see the the KH80s seem to be an IEC Class II device, so you should have few issues with ground loops at least.

If Neumann did include a resistor in series with their inputs there would still be a potential problem. The load for signals under +/-600mV would be high but outside this range the impedance would be lower and equal to the value of the resistor. So unless the output impedance of the DAC is very low then this non linear load would distort the DAC output but appreciably less. Increasing the series resistor value would slowly reduce the effect but would of course add noise . . . So there is no easy solution to the non linear impedance of ESD diodes connected to power rails when the rails are off.

 

[restorer-john]

I would guess that the KH80 input may be using (normally reverse-biased) protection diodes going to the analog supplies. When powered down, those obviously are at 0 V, effectively turning the whole affair into a limiter (about 0.6 V peak assuming plain silicon diodes, specifics depending on source impedance).

They may not even need the diodes. The unbiased front ends can do exactly the same thing.

As a young teenager in the 1980s, I first came across the exact same problem with unbuffered tape outs (rec out) distorting the main signal when my cassette decks were turned off. In fact, it was the first time I used my oscilloscope on an audio system. Nobody heard the distortion except me, and I remember showing my father the distortion caused by the (inactive) cassette deck, clipping the peaks. And that was only line level (up to about 1V max)

 

[solderdude]

A simple solution is to use a preamp but this added unnecessary circuitry to the chain and produces audible degradation of clarity and punch.

Another solution is to use two DACs (well, I do have two), one for the speakers and one for headphones.

And the third solution is to add a resistance of say 1k in series with the RCA out from the DAC.
Can be a short extension cable for instance.

 

[Rantenti]

And the third solution is to add a resistance of say 1k in series with the RCA out from the DAC.
Can be a short extension cable for instance.

I tried to connect the RCA output to my integrated amp using RCA inputs, which is further connected to passive speakers. No distortion in the headphones occured.

One thing I didn't understand is that, active speakers are basically a preamp/amp attached to the speakers. If connecting to the preamp of the active speakers (turned off) would cause distortion, why would the distortion not occur when I connect the DAC to an integrated amp when the amp is turned off?

Furthermore, imagine if my active speakers are designed to use RCA input instead of XLR, will the distortion problem in the headphones output remain the same or less so?

Thanks!

 

[solderdude]

Because when the load is switched off all equipment presents a different and often non linear load.
In some cases the outputs are in parallel. In other cases they are not.
Not all devices provide the same 'load' when switched off.

 

[AnalogSteph]

By the way, since I was using RCA-XLR cables, I have read that "The easiest way is to solder a link between pins 1 and 3 (shield and negative) of the XLR, rather than trying to solder the shield and negative wire to the sleeve contact of the RCA. "

I wonder if there are clearly audible differences between these two ways of wiring, and does it matter for the cable to be used in either direction, RCA out to XLR in, or XLR out to RCA in?

Yes and yes.

The "shortcut" of linking pin 1 and 3 together effectively negates any advantage of the balanced input, it turns the cable into a 100% unbalanced affair that's no more than a mechanical adapter. Most cables you could buy even a few years ago were built like this, actually. If you want the input to provide at least some common-mode rejection, the split between shield and audio cold must occur at the RCA end, and the closer you can replicate the source's output impedance on cold the better. Driving just the "hot" side and merely impedance-balancing the cold side actually is quite common in home studio gear - instant balanced output with no more than a few cheap passives, an almost complete no-brainer.

There is not even one single correct pinout for XLR to RCA cables - it depends on the output stage of the device in question. Most often you are likely to need 2 and 1 to signal and ground, but for a transformer-coupled output it's 2 and 3, and some old gear may need 3 and 1.