circuit design cap/impedance question.

[Fp-www.Tonepad.com]

I have been doing some simple opamp design and I have a question regarding impedance and input cap value:

If the impedance of the opamp stage can be defined by the designer, a smaller input cap could be used to get the 'same' frequency response by rising the impedance. I can see advantages on using a higher impedance: for example not using aluminum electros as input caps, or even using smaller sized film (as a 0.1 is about twice as large as a 0.05), but as everything, I'm sure there's a catch...

Where are the limits for setting a high impedance gain stage to be able to use smaller input caps and have the same frequency response?

... confused...
Fp

[R.G.]

The catches are not too bad. If the opamp itself has a very high input impedance, you can run the impedance right up in to 1M territory just fine, even a couple of megs.  What does catch up with you is the rising sensitivity to parasitic capacitance and surface leakage.

Take a 10M resistor, and figure the capacitor that causes a 20kHz rolloff with it. Lessee... C = 1/(2*pi*20E3*10E6) = 7.9pF

So at 20kHz, any combination of shunting capacitance over 8pF around a 10M resistor starts cutting treble. The resistor *body*  has maybe 4pF distributed between the coiled parts of the spiraled film trace to make the resistance. The leads add a couple more pF. Any traces it connects to have another couple... I can readily convince myself that I dont want to use any resistor more than ... 2M? 4.7M? 10M? because of the uncertainty of what parasitic capacitance does.  That doesn't even count the interference and feedback oscillation potential of parasitics coming in.

Over time dust, dirt and what my amp-tech friend calls "bar funk" build up on circuit boards. The combined exhalation of sweaty patrons, heavy tobacco smoke and the various ketones and aldehydes in perfume (let along other ... um organic scents) get in to electronics that gets used in bars.  It's easy to tell the life a mixer has lived by opening it up. This stuff is conductive at maybe a few megohms per square. It can (and has!) play havoc with circuits that expect resistances over a few meg.

My opinion, you don't use over 2.2M in a signal path resistor without clear and well conceived ideas about layout, shielding, and surface passivation (like urethane or lacquer coatings).  You'll get away with it most times, but there's going to be the one time that just ... subtly... goes wrong and you'll never find it.

You can do better than that with special guarded layouts, passivation, and other stuff. It's better to skip the fancy stuff unless you really, truly need it.

R.G.

[Fp-www.Tonepad.com]

Ok, that makes a lot of things more clear.

So, with higher impedance, layout becomes more important in terms of keeping capacitance between traces small (short parallel traces and right angle routing?) and also keeping 'bar funk' out...

Your example with the 10M impedance is very enlightening. The circuit I started off from uses impedances in the ~10k range, which leads to using 1uF-4.7uF caps to keep low freq response. I intend to rise the impedances to around 200k, so I can have smaller caps and not have to use electros in the signal path.

I already made a couple of breadboard experiments and it seems to work out. Which makes me think of another question: for example, on our beloved Tube Screamer circuit, why would the designer choose to have 1uF NP capacitors in the signal path, when the same frequency response could be achieved by rising the impedance of the gain stage and using smaller capacitors?

Fp

ps: I'm working with the PT2395 delay IC.

[Johan]

also, most IC need not only a biasvoltage but also a tiny, tiny biascurrent. for your typical 4558 its only about 50nA, so it really does not come into play most of the time. but if you increase the bias resistor too much, the output will start to stray away from 1/2 voltage point intended with the biasing....in the Tubescreamer, I believe they use a 10k resistor to bias the op, and in the BOSS super overdrive they use 100k and all is fine but if you try a 1M or larger on a 4558 you might start to see an unstable output.
I've used this effect in an overdrive I built for a friend a few years back and it sounded great..sort of a dynamicly changing clippingresponce. but it wasnt predictable or fully repeatable even with 4558's from the same maker, so its probably not a good thing when designing for production...
...so most of the time it doesnt matter but when designing things, you should be aware of it...

Johan

[Fp-www.Tonepad.com]

I should've checked the Boss Super Overdrive schematic out!

But actually my current design issue involves only inverting gain stages, so impedance is set different on those, right? (series input resistor to inv input) instead of resistor to vbias.

I guess I'll keep the impedances around 240k or so where possible, that should alow small enough caps...

Fp

[R.G.]

Yep, on inverting stages, the inverting input is driven to a virtual ground by the feedback, so the input impedance is just the input resistor as long as you go with simple inverting stages. There are designs that give you large input impedance for inverting input stages without huge resistors.

The problem with inverting stages is that if the stage has any gain, the feedback resistor is what gets out of hand even faster than the input resistor. If you want a gain of 100 and you also want an input resistance of 100K, the feedback resistor just hit 10M.

There are some ways around this, but you have to design for them.

[Fp-www.Tonepad.com]

Of course I hadn't thought of the feedback resistor... but that's because in the design I'm working on the gains are around 2x.

This is great info to keep in mind, as there always is a tradeoff in circuit design.

Thanks
Fp