Electrolytic vs film caps for smaller values.

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
I typically use as few coupling capacitors as possible and, when I do, film is my first choice.

On my current boards the only audio path electrolytics are THAT1646 sense capacitors which are typically 10/35V bipolar.

But for someone wanting to use electrolytics for coupling I try to provide this footprint or one similar.

I like the socket pin idea.

Electrolytic_Or_Film.JPG
 
In a mic preamp or moving coil phono preamp the input coupling capacitors' reactance allows low frequency input noise voltage to develop across the capacitor(s). The reactance(s) appear in series with the source impedance.
That is true, but how significant is that?
The only noise increase results from the circulation of EIN current through the reactance.
Assuming two 100uF caps, there is a 3dB increase at 21Hz, compared to a 150r source. At 32Hz, the increase is marginal, and negligible above. And that's only a tiny fraction of the EIN voltage.
Larger values of course minimize this but phantom fault stored charge limits practical input values to 100 µF or less for mic preamps.
This is one particular case where reality puts physical limitations to what should be. The addition of a coupling cap should not alter significantly the impedance of an output stage or a source.
In the case of a 150r mic, the capacitor impedance should be significantly lower than the real component at the lowest frequency of interest.
In order to fulfill this condition, the coupling caps in a mic pre should be no less than 330uF, resulting in 50 ohms reactance at 20Hz.
Moving-coil preamps are a different animal. Dealing with sub-ohm source impedance introduces many compromises, and yes, DC coupling is certainly a way that makes the least compromises.
 
But If you want to trade notes on mathmatical formulas used in coupling circuits and have a true scientific discussion about this you are wlcome to start another thread on your site particularly for this.
There are not so many mathematical formulas related to this. Zc=1/j.C.omega is the only one I can think of, and may be F=1/2.pi.R.C
You need to unlearn that. Transformers can be engineered however an EE wants them designed.
Certainly not. At least not within the constraints of reality. There are limits to acceptable bulk and cost, particularly when alternative solutions provide similar performance at a fraction of the cost.
I remember someone asking me once my advice on transformer coupling designs at one of my lectures that I gave over RF transformer core constructions
RF is another subject. RF xfmrs are used because they combine optimum power transfer and selectivity, which no cap-coupled or DC coupled circuit can provide.
, later, I found out that man was one of the engineers at Jensen and about a decade or so later, he made a series of high end interconnect transformers using the standard industry 8-12MHz core construction techniques I lectured on.
We are lucky enough here to have Mr CMRR (Bill Whitlock) as a member. He could chime in.
Maybe your view of transformers was limited to the splattering of the mid-low end UTC transformers audio transformers of the past?
No. My appreciation of xfmrs come from deep knowledge of how they work and their limitations. I have seen enough xfmrs in my life to know what they're good for, and try as much as I can to avoid using them for signal transmission.
Transformers are still used today and in even different designs. From power supplies to network switches, and routers, to electricity power meters.
Rightly so. Because no direct or capacitor coupling can replace a mains transformer, and the galvanic isolation provided by xfmrs is essential in systems where connections can be subject to all sorts of mishandling. The subject here is inter-stage coupling of audio signals, though.
I never said anything about improving noise. Resistors can be selected for the best low noise profile while offering low insertion loss when DC coupling.
Then can you explain "nor add distortion, besides self noise in resistors, that can be selected for better noise profile."
The trick in all of this is to do this outside the operating range so the design does not alter the signal in an undesirable way.
"do this outside the operating range"... ?
 
A quick glance at their site revealed one that DIN-PB that he designed.

What does that mean ? Mis-typed ?

Transformers are still used today and in even different designs. From power supplies to network switches, and routers, to electricity power meters. It is not my fault low end audio only use them as 'audio gimmicks' with inexpensive transformers.

Yes. I think we all know that transformers are used in a huge range of applications from power distribution to small signal digital comms. I work with them daily in power supply/conversion systems. But that's of little relevance to audio interstage coupling.
 
That is true, but how significant is that?
The only noise increase results from the circulation of EIN current through the reactance.
Assuming two 100uF caps, there is a 3dB increase at 21Hz, compared to a 150r source. At 32Hz, the increase is marginal, and negligible above. And that's only a tiny fraction of the EIN voltage.

The 100 µF assumption is for designers that know what they're doing.

I guess you have been lucky enough to not have seen designs where 10-22 µF input coupling caps were used or attempted. In my former applications role I have.

Designers lower the input capacitor values, raise the value of the bias resistors, and do so without thinking about looking at the resulting source impedance the preamp "sees."
 
Last edited:
The 100 µF assumption is for designers that know what they're doing.
For sure.

I guess you have been lucky enough to not have seen designs where 10-22 µF input coupling caps were used or attempted. In my former applications role I have.
Actually, I have seen this - a lot- even from reputed mfgrs such as SSL Neve...
Designers lower the input capacitor values, raise the value of the bias resistors, and do so without thinking about looking at the resulting source impedance the preamp "sees."
As one member often says, how many records did not sell because of the wrong capacitor resistor transistor opamp value? ;)
Fortunately, with technological improvements in electrolytics, using 330uF/100V caps in this role does not seem weird.
 
A nylon P clip with a pin lock push rivet makes a very solid component mount to PCB or chassis ,
with the pin socket it would make caps user serviceable without the need for any special tools ,

Lead outs and solder joints usually bear the weight of the cap .Any temperature related movement inside the cap also physically stresses the solder joints over time when the rubber seal is set against the PCB .

A P clip of 6mm diameter closed still holds a 10mm dia cap very securely , the excess can be trimmed easily ,using only the lower part (as shown) of the clip to secure it down to the deck provides more than enough mechanical strenght for smaller caps .

A small loop might be wound in the lead outs so as to prevent any thermal/physical stresses being transferred from inside the cap to the sockets/pcb/solder or maybe even to help reduce vibrations in the PCB reaching the cap mechanically if the need arose .

Ever tap on the frame of a console with 10,000 electrolytics in it and listen to what comes out ?
it sounds like a bucket of bolts :)




1671111404094.png
1671111477906.png
 
Last edited:
As one member often says, how many records did not sell because of the wrong capacitor resistor transistor opamp value? ;)
Fortunately, with technological improvements in electrolytics, using 330uF/100V caps in this role does not seem weird.
So true.

I had a hard time protecting inputs from destructive differential phantom faults with values >>100 µF due to stored charge.
In addition the phantom turn-on time for quiet operation due to charging currents gets really long as the value is increased.
 
I had a hard time protecting inputs from destructive differential phantom faults with values >>100 µF due to stored charge.
In addition the phantom turn-on time for quiet operation due to charging currents gets really long as the value is increased.
I guess it's the price to pay for the improvement. There is probably a way to provide an active variable time-constant protection scheme, but is it worth it?
Some are willing to wait 15-20 minutes for their Sony C800G to reach adequate temperature equilibrium...
 
In a mic preamp or moving coil phono preamp the input coupling capacitors' reactance allows low frequency input noise voltage to develop across the capacitor(s). The reactance(s) appear in series with the source impedance.
I have long speculated about DC coupled phantom powered mics. You have done far more than speculate.

Talking about noise, have you ever quantified the noise difference between your DC coupled mic preamp and a conventional capacitor coupled preamp. In theory this could net out capacitor related noise. Of course we cannot forget any capacitor in series with the gain pot also.
Larger values of course minimize this but phantom fault stored charge limits practical input values to 100 µF or less for mic preamps.
yup, patch bay preamp killers.. :rolleyes:

JR
 
Meanwhile I have a tray of transformers for a Royer converter circuit on my desk. So I'll just pop those into an audio circuit because that is definitely a good idea....
Ribbon mics are one of the transducers that need a xfmr. I know some have tried to design amplifiers specifically tuned for the very low Z they present.
A basic calculation shows that in order to design a bipolar input stage that has the same noise contribution as a 1 ohm source, it would need to operate at a quiescent current of about 50mA. Since the aim is to have an input stage significantly quieter than the source resistance, and that most ribbons have a resistance that is lower than 1 ohm, the actual current would need to be increased about 10 fold, with about 100+ transistors in parallels.
An FET-based solution would require about the same number of elements in parallels.
This is a rough estimation, in practice it's more demanding.
Apparently, the few attempts i was aware of have not resulted in viable products.
 
The only moving coil preamp I ever designed (decades ago) used a DC coupled input stage and ran some very low noise PNP input devices (ROHM 2sd737s) now out of production. IIRC I used around 3mA of device current density. I don't recall getting any complaints about noise, and favorable comparisons to transformers of the day, but I was not a player in the high end because my products were too low cost to be taken seriously by that crowd. :rolleyes:

Step up transformers offer attractive potential to "transform" low single digit source impedance up to something better managed by modern silicon (source impedance increases with the turns ratio squared). Good transformers are not as inexpensive and relatively easy as solid state designs. The low noise devices I referred to went EOL (end of life) last century due to lack of commercial demand. I suspect they were replaced by integrated circuit mic preamps that didn't suck and are now widely used in professional products.

An FET-based solution would require about the same number of elements in parallels.
This is a rough estimation, in practice it's more demanding.
Apparently, the few attempts i was aware of have not resulted in viable products.
I think one of the last designs Brad (RIP) was working on was a low noise FET input MC preamp. I don't recall details but as I recall he had to design a trick low noise power supply to deal with marginal PSRR of his novel topology (mostly speculation on my part about the why) I don't think I ever saw an actual schematic. I just recall him talking about it around the internets.

JR
 
I think a transformer is about perfect for an MC preamp. Free noiseles gain and better noise rejection than than the best CMRR from a differential input. Low signal levels so distortion is negligible. If I had unlimited time I'd try Wayne's transformerless MC preamp but I'm very happy with a Lundahl in front of his MM preamp.
 
Ribbon mics are one of the transducers that need a xfmr. I know some have tried to design amplifiers specifically tuned for the very low Z they present.
A basic calculation shows that in order to design a bipolar input stage that has the same noise contribution as a 1 ohm source, it would need to operate at a quiescent current of about 50mA. Since the aim is to have an input stage significantly quieter than the source resistance, and that most ribbons have a resistance that is lower than 1 ohm, the actual current would need to be increased about 10 fold, with about 100+ transistors in parallels.
An FET-based solution would require about the same number of elements in parallels.
This is a rough estimation, in practice it's more demanding.
Apparently, the few attempts i was aware of have not resulted in viable products.

Ah! I think I've seen what may have happened here. When I referred to "Royer" I was referring to a Royer Oscillator circuit rather than Royer Microphones.
In fact the transformers are for an arguably mis-named "Resonant Royer" power converter circuit. It's also known as a "Baxandall Converter" which gets us another thing we might assume is audio related on here :) And to add to the fun a variant of the 'proper' Royer Oscillator circuit is also known as a "Jensen" Converter/Oscillator ! Seems you just can't get away from these "low end audio" references !
Apologies for any confusion. I'll refer to it as a "Resonant Tank" converter if I have need to refer to it again here :)
 

Latest posts

Back
Top