Hello.
Since I could not find information, I will ask here:
I am wondering if I can use 3.5mm stereo sockets for balanced signal?
I mean one channel per connection of course, since stereo 3.5mm connection has 3 contacts, just like XLR. Yes i know that XLR has gruond pin, which is not connected to socket itself, and 3.5mm ground is by default conected to socket. But in wiring diagrams I always se XLR ground conected to enclosure.
Do I miss something?
Since I could not find information, I will ask here:
I am wondering if I can use 3.5mm stereo sockets for balanced signal?
I mean one channel per connection of course, since stereo 3.5mm connection has 3 contacts, just like XLR. Yes i know that XLR has gruond pin, which is not connected to socket itself, and 3.5mm ground is by default conected to socket. But in wiring diagrams I always se XLR ground conected to enclosure.
Do I miss something?
¼" TRS is often used for balanced impedance connections.
3.5mm TRS would be just as good.
Remember that screen/shield goes to enclosure at both ends.
Worth checking that your balanced impedance gear complies with Pin1 requirements !
3.5mm TRS would be just as good.
Remember that screen/shield goes to enclosure at both ends.
The sleeve is not "ground". It is part of the enclosure.
Worth checking that your balanced impedance gear complies with Pin1 requirements !
Last edited:
I have another question:
How to add RF protection to INA134? I can't figure it out. All methods i found have negative effects just because of non identical caps. I am thinking of not adding it, because preamp unbalanced input will have it, and amp will always be connected to that preamp. But i don't know if this good way. Feels wrong to have unprotected inputs.
http://www.ti.com/lit/ds/symlink/ina2134.pdf
How to add RF protection to INA134? I can't figure it out. All methods i found have negative effects just because of non identical caps. I am thinking of not adding it, because preamp unbalanced input will have it, and amp will always be connected to that preamp. But i don't know if this good way. Feels wrong to have unprotected inputs.
http://www.ti.com/lit/ds/symlink/ina2134.pdf
What kind of levels and impedances are we talking about here, typically? A consumer or even pro-level interconnection with low source impedance does not generally require super-duper-high CMRR. 50-60 dB tends to be just fine. Microphones on long cables may be a different story!
Anyway, you can try a T connection, similar to what's being done with resistors. Like 470 pF + 470 pF between inverting and noninverting input, with 100 pF or so branching off to shield in between.
This gives a differential-mode input capacitance of 235 pF (so we could actually up the 470pFs to 1 nF and it would still be acceptable), while a common-mode signal would see 940 pF in series with 100 pF, or a bit under 100 pF. This not only keeps high-frequency common-mode input impedance higher, any mismatch between the 470pFs also is of less importance since only a fraction of the CM signal is dropping over them to begin with.
I would complement this with two 100pFs from inverting and noninverting directly to pin 1 at the input - you know, inductance and stuff. And maybe two series resistors of like 100-220 ohms <1% after those, and then your T-connected caps.
Are you already using buffers in front of the INA134? While opamp types would have to be picked carefully (maybe OPA2132 or OPA1642/1652 or something?), this can be used to up common-mode input impedance a fair bit. Adding resistors to your T, you could have 2x 10k||1n and 1Meg||100p branching off. Low-frequency differential input impedance then comes out as 20k, common-mode as ~1 meg.
You could get even fancier with a common-mode bootstrap, but I don't think that would be needed. Plus I'm not sure whether it's necessarily stable with open input...
Anyway, you can try a T connection, similar to what's being done with resistors. Like 470 pF + 470 pF between inverting and noninverting input, with 100 pF or so branching off to shield in between.
This gives a differential-mode input capacitance of 235 pF (so we could actually up the 470pFs to 1 nF and it would still be acceptable), while a common-mode signal would see 940 pF in series with 100 pF, or a bit under 100 pF. This not only keeps high-frequency common-mode input impedance higher, any mismatch between the 470pFs also is of less importance since only a fraction of the CM signal is dropping over them to begin with.
I would complement this with two 100pFs from inverting and noninverting directly to pin 1 at the input - you know, inductance and stuff. And maybe two series resistors of like 100-220 ohms <1% after those, and then your T-connected caps.
Are you already using buffers in front of the INA134? While opamp types would have to be picked carefully (maybe OPA2132 or OPA1642/1652 or something?), this can be used to up common-mode input impedance a fair bit. Adding resistors to your T, you could have 2x 10k||1n and 1Meg||100p branching off. Low-frequency differential input impedance then comes out as 20k, common-mode as ~1 meg.
You could get even fancier with a common-mode bootstrap, but I don't think that would be needed. Plus I'm not sure whether it's necessarily stable with open input...
Thank you for response.
Levels to drive LM3886 with about 26dB of gain to full power. So quite high.
It will go from DRV134 in preamp to INA134 in amplifier, then From INA134 to LM3886 on same pcb. No additional buffers.
Can I add T caps to input socket itself? If I can, do I still need 100pf caps from hot and cold to pin1?
That's above mic level but certainly not super high. Most listening takes place up to and around 1 W audio output, requiring only about 140 mVrms of input. The classic 50 mW / 4 Ohm DIN reference even requires only 22 mV.
Are you aware that the DRV has a nominal gain of +6 dB, while the INA is a unity gain job, effectively adding 6 dB of gain? And that the INA has about 3 times as much noise as an LM3886 input and still almost twice as much as a good unbalanced preamp? Like this, the maximum SNR re: 50 mW / 4 ohm would be about 70 dB or a bit less (7 µV of INA output noise translates to at least ~140 µV of LM3886 output noise after all), which is a fairly good but by no means exceptional value.
So it would be nice to have 6 dB of attenuation after the INA, but that doesn't make for a particularly good combination of resistor divider input and output impedance. Since, however, you will not even require +6 dBu of input to the LM3886 while the DRV/INA combo will happily support more than +20 dBu on the interface, I suggest you increase preamp gain by 6 dB and include a 12 dB passive resistor divider between INA and LM3886 (like 2.2k/750R or 3.6k/1.2k, depending on how much series R you still have after that - ideally the LM3886 should be seeing the same impedances at its two input pins). Ordinary metal films should do.
The resulting level plan could look as follows:
volume pot: reference @ 0 dBr
unbalanced preamp out: +20 dBr
DRV out: +26 dBr (bal)
INA out: +26 dBr
LM3886 in: +14 dBr
So the end result looks like an unbalanced preamp with 14 dB of gain in this case. Your interface would be run 12 dB hotter than LM3886 input levels. At an estimated +3.5 dBu there or +15.5 dB on the interface, you should be able to get by with +/-12 V for the INA easily. Given the output loading provided by the resistor divider, I would recommend some lowish-ESR ~47 µFs for rail bypassing nearby.
Of course you could also add the 6 dB of extra gain in front of the volume pot, which should give no major problems with level handling yet, just be sure to put no undue load on the opamp output and watch out for impedance balance at the same time.
BTW, some pro amp designs also go "balanced on the last mile", which has its charm. An LM3886 is a big opamp, so why not have it do bal/unbal. Read Bruno Putzeys' "G-Word" whitepaper for details. The balanced receiver would be replaced by a dual opamp buffer then (and I'm pretty sure it's possible to make a balanced buffer with inherent CMRR with a balanced opamp like OPA1632, not so sure how though), and the attenuator would become a balanced 3-resistor job (e.g. 1.8k/1.2k/1.8k). Yes, then resistor matching is up to you, but your LM3886 input circuitry should be a fair bit less susceptible to things like magnetic coupling from the power supply leads and such.
If you manage to squeeze them in, why not. I imagine it may be a pretty tight fit though. Don't skimp on cap quality here, they should be NP0 ceramics or comparable (styro, film). Definitely no crummy dielectric ceramics here. Note that NP0 may only be easily available up to about 820p, though it depends on size (that's for 0805 if memory serves).
Last edited:
Thank you for such elaborate reply. It's really helpfull.
I think I should paint the whole picture what I am trying to do. So it is like this: I am putting 6 channels of LM3886 and I wanted to eliminate ground loops, so I thought it would be good idea to go balanced. I still want to go balanced, and now I am really interested in "last mile" aproach. I will try to put ciruit together when I get time, and when I completely read that nice paper.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.