Phantom Power Capacitors

[john12ax7]

You can use bi-polar caps with DC bias, you don't need to use polar. The Nichicon MUSE ES are a common choice. Though I've never tried them with significant bias to see what if any degradation in performance there is. Another issue is they are only 85deg and up to 50V

 

[thor.zmt]

Just so I understand what's being said here, that the phantom voltage blocking caps on the input of a mic pre (where there's 48v on one side of the cap, and very few on the other) don't have to be polarized ones?

No, that is not what is being said.

What is being said that is that using polarised capacitors in this position results in a LOT more distortion than using nonplarised types, despite being DC biased.

And that actually using capacitor coupling is quite a stupid way of dealing with the problem in 1985.

And that still doing it four decades later with all our further advances in electronics and not just having IC's with on board charge pump switchers to make the P48 power and power the high side circuitry and droping it in beggars belief.

To see designs that should have never been designed in the 1980's lauded as Ultimo Ratio is laughable.

Thor

 

[gyraf]

unless you use really crappy electrolytics, the high bias voltage present when p48 is on makes them practically invisible here.

Try a blind test for yourself, and you'd be surprised how much hifi bs has slowly crept over here

/Jakob E.
who also believed this to be a problem once

 

[MicUlli]

unless you use really crappy electrolytics, the high bias voltage present when p48 is on makes them practically invisible here.

Try a blind test for yourself, and you'd be surprised how much hifi bs has slowly crept over here

/Jakob E.
who also believed this to be a problem once

I fully agree. Electrolytics in the signal path produce negligible distortion with small signal currents of only a few mA max. No necessity to avoid them by building a lot of ******** silicon as substitute..

 

[Matt Syson]

Doug Self wrote about the efects of electrolytic and other( capacitors and a handy and repeatable conclusion is that as long as the AC component across a coupling capacitor is LESS than about 80 millivolts at ALL frequencies of interest then distortion is very close to negligible and certainly is in the same territory as other device distortions and circuit layout effects. After all these musicians/recording folk have been demanding LOW distortion for decades but now digital recording has reduced distortion to VErY low levels they are now whining that it sounds 'sterile' and are hankering for the old days of distortion. You can't please some people but apparently writing about it on the internet makes things 'true'.
Like when Elvis leaves the stage, the 'magic' is gone, the same woith audio gear, after it has been switched off a while the electrons have dissipated so no more 'original' magic (which may have been influenced by drugs or alcohol in the first place). Vintage gear can NEVER sound as it did back in the day because time has moved forward and no recordings can reveal exactly what they were like.

 

[JohnRoberts]

I fully agree. Electrolytics in the signal path produce negligible distortion with small signal currents of only a few mA max. No necessity to avoid them by building a lot of ******** silicon as substitute..

This is a mature topic. Electrolytic capacitors in passive loudspeaker crossovers stressed by passing amps of current will reveal more non-ideal behavior. In phantom voltage blocking applications the caps are not stressed that much.

DC servos can be as simple as one op amp and a couple high quality film capacitors. That said flying a mic preamp up to phantom voltage to eliminate the phantom capacitors is a bit more complicated.

Better is better if cost is no object but we are talking about small amounts of nonlinearity.

JR

 

[Cqwet Dbdfte]

What did you read? Like this:

https://linearaudio.net/sites/linearaudio.net/files/Bateman EW 01 2003 mar2003 10 to 100uF caps and 100 Hz measurements_0.pdf

That is pretty much the standard work, originally published in wireless world (UK) two decades ago.

All articles here:

https://linearaudio.net/cyril-batemans-capacitor-sound-articles

Thor

PS, now don't bl00dy get me started on resistors.

Sadly Bateman (RIP) passed away 10 years ago, his nice capacitor tester has some key components that are obsolete.
There are better parts available now, that could be used, including a nice small SMT board.
MELF resistors supposedly has less noise than thick film SMDs, and better at HV.

 

[MicUlli]

I have done some measurements to show that electrolytics in the signal path produce negligible distortion. The test objects are Panasonic 22uF/50V polar caps (rated to 105°C). A signal voltage of 1Veff was produced and fed into the XLR plug of the mic input of my audio interface.
Without caps: direct injection, phantom power switched off.
With caps: caps in the signal path, 48V phantom power switched on (this produced real 48VDC across the caps). Find attached the results. And never discuss about THD or any other signal degradation of electrolytics for phantom power decoupling any more

 

[JohnRoberts]

I am pretty sure I posted earlier that the distortion from this is very small. IIRC Sam Groner had to design his own extremely low distortion sine wave generator to resolve distortion components this small.

Of course feel free to argue about how many angels can fit on a pin head if you have too much time on your hands. Better is always better but engineering always comes down to cost vs benefit. I was cheap before I went to work at Peavey for 15 years, they just sharpened my cost/benefit pencil.

JR

 

[jp8]

I know capacitor distortions have been discussed to death, yet, I decided to also test some capacitors for distortion. So I fired up my Audio Analyzer this afternoon and measured three different elcaps and a MKP for reference. My goal was not only to measure the amount of distortion, but also to verify whether THD would rise or not when the caps were biased. Cyril Bateman's measurements suggest that THD would rise, but according to this paper from TI, it would decline. Especially at low frequencies, when the voltage drop across the cap is the highest. I tested in a similar way as Miculli did, but only feeding a single ended signal into the Analyzer's preamp. If driven symmetrically and through the same capacitors, the even order harmonics from the caps would partly cancel out.

The preamp input caps are Nichicon EP series bipolar caps. I could not change those for this test. The other caps are listed in the picture below. The zip file contains all the REW distortion graphs, measured at 1000Hz and 20Hz.



Assuming the MKP has negligible voltage dependent THD, we can conclude from the first measurements that the preamp's bipolar Nichicon EP caps show some voltage dependency, especially at 20 Hz, which could be expected. All the measurements at 20 Hz clearly show that at low frequencies, THD rises when bias is applied. So this suggests Cyril was right and the TI paper was wrong. I deliberately say "suggests" here, because perhaps there are other elcaps that react exactly the opposite way. So I don't see my measurements as watertight proof.

But how bad is this THD rise after all? There are probably not many people who can hear THD below 0.1%. The lowest threshold I know of is 0.06% by Arienne (Soliloqueen). The THD at 20Hz remains more than an order of magnitude below that level. So even when taking the THD rise, caused by the phantom power into account, THD remains inaudible by any known standard. As far as I'm concerned, electrolytic capacitor distortion is just an academic discussion and not something to really worry about in audio equipment. But having said this, I have to admit that I still use over-specified elcaps in audio circuits that the rational half of my brain knows make no sense... Nichicon will love this irrational behaviour.

Take this for what you think it's worth. YMMV

Jan

 

[Cqwet Dbdfte]

I would think power supply noise and a circuits power supply rejection is fair game, being phantom or not.
While %THD of electrolytic effects in a PSU, wouldn't be directly translated to signal interference any amplifier, and sort of useless in this context.
Knowing a circuits PSU rejection is valuable, and how many microvolts (hopefully) of noise is generated by the PSU.

 

[john12ax7]

All the measurements at 20 Hz clearly show that at low frequencies, THD rises when bias is applied.

Doesn't the 47uF polar show the opposite? Going down from .0005% to .0004% with bias?

 

[Cqwet Dbdfte]

Would not at low frequency, if any program content, any harmonics if measurable, would be inaudible?

 

[abbey road d enfer]

And that actually using capacitor coupling is quite a stupid way of dealing with the problem in 1985.

And that still doing it four decades later with all our further advances in electronics and not just having IC's with on board charge pump switchers to make the P48 power and power the high side circuitry and droping it in beggars belief.

Then what are you suggesting?
Do you have a receipe for a mic pre to handle 40+V common-mode voltage with standard +/-17V rails?

 

[Cqwet Dbdfte]

I know capacitor distortions have been discussed to death, yet, I decided to also test some capacitors for distortion. So I fired up my Audio Analyzer this afternoon and measured three different elcaps and a MKP for reference. My goal was not only to measure the amount of distortion, but also to verify whether THD would rise or not when the caps were biased. Cyril Bateman's measurements suggest that THD would rise, but according to this paper from TI, it would decline. Especially at low frequencies, when the voltage drop across the cap is the highest. I tested in a similar way as Miculli did, but only feeding a single ended signal into the Analyzer's preamp. If driven symmetrically and through the same capacitors, the even order harmonics from the caps would partly cancel out.

The preamp input caps are Nichicon EP series bipolar caps. I could not change those for this test. The other caps are listed in the picture below. The zip file contains all the REW distortion graphs, measured at 1000Hz and 20Hz.

View attachment 130288

Assuming the MKP has negligible voltage dependent THD, we can conclude from the first measurements that the preamp's bipolar Nichicon EP caps show some voltage dependency, especially at 20 Hz, which could be expected. All the measurements at 20 Hz clearly show that at low frequencies, THD rises when bias is applied. So this suggests Cyril was right and the TI paper was wrong. I deliberately say "suggests" here, because perhaps there are other elcaps that react exactly the opposite way. So I don't see my measurements as watertight proof.

But how bad is this THD rise after all? There are probably not many people who can hear THD below 0.1%. The lowest threshold I know of is 0.06% by Arienne (Soliloqueen). The THD at 20Hz remains more than an order of magnitude below that level. So even when taking the THD rise, caused by the phantom power into account, THD remains inaudible by any known standard. As far as I'm concerned, electrolytic capacitor distortion is just an academic discussion and not something to really worry about in audio equipment. But having said this, I have to admit that I still use over-specified elcaps in audio circuits that the rational half of my brain knows make no sense... Nichicon will love this irrational behaviour.

Take this for what you think it's worth. YMMV

Jan

It looks like the noise floor of the bias supply boost low level noise, more so in the low frequencies, which could account for differences in the -115dB region.
Maybe ganging some 9V batteries together for low noise supply would provide a lower bias noise.
A baseline of harmonics measured without a cap could be useful.
The levels look small enough to be ignored.

 

[abbey road d enfer]

When measuring noise (shorting the input), the source impedance is a capacitor in parallels with a resistance. Such a circuit has an intrinsic noise (check KTC noise) with a spectrum that goes with a density in 1/f. So yes, noise tends to be shifted towards low frequencies.
Making sure that C is large enough compared to R, most of the noise goes to infrasound.
That's why, in a well designed xfmrless preamp, the base resistors are quite higher than the nominal mic impedance (typically two 10-22k resistors).
Basing the capacitor value only on the -3dB LF response is significantly overlooking this aspect.
For example, condenser mics use typically a 1Gohm resistor with a 50-70pF capsule, which results in a -3db LF point of about 2-3 Hz.

 

[abbey road d enfer]

All the measurements at 20 Hz clearly show that at low frequencies, THD rises when bias is applied.

I concur with john12ax7 that there are some exceptions.
In order to pinpoint more accurately, it would be interesting to make the same tests with separate capacitors.
It is quite possible that some effects in one type compensate those in the other type, so it makes it very difficult to analyze.

 

[merlin]

Even a regular electro under the worst conditions will produce distortion well below 0.001% at mic levels.

 

[thor.zmt]

Then what are you suggesting?

Making something that makes sense.

Do you have a receipe for a mic pre to handle 40+V common-mode voltage with standard +/-17V rails?

I actually have one that does it on a singe 5V Rail, actually.

But there I'm cheating and design in a 2MHz micro switcher to give +54V/+18/-18V regulated low noise local power supplies. It also (obviously) gives P48 after a linear 48V Regulator and has a balanced out capable of over +26dBu out into 600 Ohm.

But you define the problem very poorly and you are deliberately creating an unsolvable task by unnecessarily denying available resources to the solution.

You are making Phantom Power 48V available to the Microphone but you are denying it to the Microphone Preamp in the same case. Why do that. So let's formulate the problem correctly:

Given is the following power supply:

+/-18V and +48V regulated and > 54V unregulated under all operating conditions

You are given the following parts:

OPA1679 Quad OPA
MMBT4403 BJT (as many as you like)
STN1NK60Z (4pcs)
any number of BC84X series dual transistors and small passive parts (try not to overdue it, pithiness is a virtue. Any intelligent fool can make things overly complex)

Task:

Make a Microphone Preamplifier that provides P48 or no Phantom Power to the Microphones and requires no coupling capacitors in to out, including any such coupling capacitors inside feedback loops, that offers an Ein of better than -130dBu at maximum gain and extremely low distortion.

Is there any problem? This 2024, not 1980. It should be trivial. Nu?

Thor

 

[thor.zmt]

Even a regular electro under the worst conditions will produce distortion well below 0.001% at mic levels.

Mind you, it shows nicely that BIAS INCREASES HD and does not reduce it, which was the point I made and which kicked off this whole debate.

Thor