Reactance of incput coupling cap and noise

[Fancy Lime]

Hi there,

doing some experimentation with input coupling caps before a NE5532 gain stage, I realized that low value caps tend to produce a low rumbling noise. The smaller the louder. Trying to figure out why, I realized that this must be so because the BJT inputs of the NE5532 have a non-negligible input current and therefore any series resistance before the input will produce current noise. And a capacitor is (among other things) a frequency dependent resistor to AC signals. A quick lipstick-on-napkin calculation of the reactance tells us, that with a 100nF input cap, the reactance is ~80kΩ @ 20Hz. That is one big-a$$ in-line resistance in the audio band! And it probably means that the low-frequency noise contribution from the coupling cap is quite substantial in a lot of designs. There are of course ways around this problem: JFET input opamps, directly coupled FET buffer stages before the NE5532, and so on, which may or may not be practical solutions in any given case.

But I am wondering why I have never seen this noise contribution discussed. If my testing (1n to 1µ) is any indication, the noise contribution is *very* substantial on a high gain input stage, even with common values like 22n or 47n. If I am not mistaken, the cap should contribute a 1/f noise profile on top of everything else. Admittedly, this is not nearly as annoying as high-frequency hiss, but still... Should we be more careful with that?

Thanks and cheers,
Andy

[amptramp]

One of my favourite subjects!

With any amplifier (not just op amps) you have a voltage noise and a current noise to be concerned with.  JFET op amps have a low current noise but a higher voltage noise than bipolar op amps which can be made with low noise voltage but higher current noise.  Obviously, these numbers can vary but in general, this is the tendency.  The current noise works with the input resistance it sees to create a separate voltage noise component.  If you have a voltage noise of 10 nV/Hz^0.5 and a current noise of 1 pA/Hz^0.5 and if the input resistance is 20K then the voltage as a result of current noise is 20 nV/Hz^0.5 and they add algebraicly to give you a combined noise of 22.36 nV/Hz^0.5 because the noise sources are not correlated and therefore they add algebraically (square root of the sum of the squares).

With a low value capacitor, the input is isolated from the output of the previous stage and there is no possibility of paralleling the output resistance with the input resistance to get a lower value of effective resistance at the input.  The 5532 in your example is good for voltage noise but not so good for current noise and is optimized for source impedances of about 20K.

Suppose you have an input capacitor as you say with an 80K impedance.  If you have a source of 1K and the op amp is biased through an input impedance of 100K, your 5532 definitely has a large current noise component even though the voltage noise remains low.  As you go up in frequency, the input capacitance provides a lower impedance connection to the 1K source and you get the low noise performance you are looking for.  The capacitor itself is not creating noise, it is isolating the resistance of the previous stage that can parallel its resistance with the input resistance of the previous stage to reduce the current noise.

There is a phenomenon called 1/f noise that increases a frequency decreases.  This may be the rumbling you hear.  You can also get popcorn noise where the signal level randomly flickers between two constant voltages and this is visible on a scope.  Good luck with your noise performance - after a while it gets to be fun.

[PRR]

> 100nF input cap, the reactance is ~80kΩ @ 20Hz.

You don't havea 20Hz speaker. It is 8k at 200Hz. And many guitar chains cut this high. It is 20k at 80Hz, few g-amps go much lower, and typically with a steep cut-off.

Yes, I do not think the '5532 is the Bee's Knee for guitar cord work, and I(hiss) is one reason.

Where 1/f rumble really hurts is hi-fi phono. While speakers don't reproduce 20Hz, the amps do and the speakers try. I had an unfortunate combination of a very high-end medium-size system, with a "low price" phono preamp (external; this was the decade when phono was obsolete). Good low V(hiss) bipolar chip, but modest input cap, "DC Coupled" amplifier. When turned up the speaker cone fluttered and then the amp cut-out, not knowing <20Hz from DC. In a panick I swopped the input and output cap, put the bass-cut in the middle instead of the front.

[highwater]

Where 1/f rumble really hurts is hi-fi phono. While speakers don't reproduce 20Hz, the amps do and the speakers try. I had an unfortunate combination of a very high-end medium-size system, with a "low price" phono preamp (external; this was the decade when phono was obsolete).

Well, shoot, it really *was* only about one decade, wasn't it? That's going in my sig.

Also, I really need to get myself a better record player... one that wasn't built just after that decade, with an internal preamp, for less than $100.

[amptramp]

Following what Paul said, I got a magnificent deal on a couple of JBL G-730 speakers and thought it would be a great basis for a stereo system.  After all, why not listen to music on a typical speaker system that musicians use in live concerts?  But these are lead guitar speakers with a bass turnover of 70 Hz.  Stuff made for guitar is designed to reproduce the 82 Hz lowest note of a guitar.  If you look at guitar tube amps, the output transformer is usually a puny piece of iron that looks like it was designed for something with half the power you would get from a high-fidelity amplifier.  But for their intended use, they are just fine.

[Fancy Lime]

Hi guys!

Yes, 20Hz is not really relevant for guitar or even standard tuned bass. But with a 5-string bass, 30Hz is pretty relevant and I generally try to "consider worst case first" and back off to more relevant cases later. makes me feel a lot better than the other way round   Also, it's not like people aren't tuning down their B-string, So I normally would not like to have a low cut above 20-25Hz in anything that is intended for bass use.

I tested the "current noise of BJT opamps" hypothesis by swapping the 5532 for an TL072. The noise is about the same (only tested by ear, so shovel on heaps of salt) and more importantly, the change in noise when changing the input cap value is the same. Sooo... hypothesis busted? Dunno, need to think more about this.

Cheers,
Andy