Question from a too-long digital type: Given a run-of-the-mill voltage feedback OpAmp (e.g. LM833) is there a benefit of certain ranges of resistor values, other than the obvious (bias/offset currents)? ...like stability? I was trying to save supply current, but that requirement was been lifted. The stability requirement hasn't.
On a sunny day (Mon, 19 Jan 2009 09:04:15 -0600) it happened krw wrote in :
At lower values the circuit is less sensitive to for example conduction path because of humidity, some sort of dirt, also capacitance in the circuit has less influence then. Sometimes you want lower values to be less sensitive to alien signal pickup, although that depends on the type of circuit too.
100 Ohm is power, 1k is usually OK, 10k is nice, 100k is OK, 1M is ehh, 10M is trouble etc. So, in my resistor box, 10k is almost finished again.Sorry I am but a neural net, I did the PID calc for that ripple filter by looking at the ceiling, and got it right first time. then I tried to improve it in spice and could not. So for what this neural net suggestion is worth.
Others will insist on computer math and buy a book.. like the Art of Electronics. I will never buy the book (with all respect for Winfield), because do not want to mess with the neural net.... Never fix something that works.
In article , snipped-for-privacy@yahoo.com says...>
I would have thought the opposite. Higher resistance makes for lower bandwidth. No?
In this case, 100K isn't so good. 10K seems to work, as does even a 4.7pF cap across the 100K. 1) Why (does the lower resistor help)? 2) how can I be sure (the lower resistor) will stay fixed? I really don't want to swamp the bandwidth with a cap for various reasons, but will if I need to.
Understood. I wanted to save the milliamp, or so. Not so important anymore though. If I can be sure (understanding helps) that the resistor change will solve the problem, the change is easy. Another dozen or so 10K resistors won't even be seen on the BOM. ;-)
looking
spice
If I had reliable models... Unfortunately they suck for this sort of thing.
I'll likely buy the book, though not until the new edition is out. Hint, hint. ;-)
If the resistors are too big, it can add a pole in the feedback loop (at w=1/RC, where C is the total capacitance at the inverting node) which can make the loop unstable. A bit of cap across the feedback resistor can fix that.
Also, opamp input current noise can become lots of voltage noise when dumped into a big feedback resistor.
In extreme cases, like 1.8 GHz current-feedback amps, we use 0603 resistors and keep the inverting node connections as tight as possible, and cut away ground plane in that region to reduce pcb capacitance. The quick-turn board houses tend to have terrible stackups on 4-layer boards.
John
On a sunny day (Mon, 19 Jan 2009 09:58:56 -0600) it happened krw wrote in :
Just a wild guess, you are compensating the input cap with the 4.7p. I would really need to see the circuit before making more wild guesses :-)\\
2) how can I be sure (the lower resistor) will stay fixed?This question is not clear to me.
On a sunny day (Mon, 19 Jan 2009 16:51:29 GMT) it happened Jan Panteltje wrote in :
PS, in this example circuit, the feedback becomes frequency independent for C1 x R1 = C2 x R2, given driven from a zero impedance source. ----- C2 ---- | | |----- R2 ----| | | -------- R1 ------------- - | | | out --------- Uout ~ Uin C1 -- + | Z = 0 | | opamp /// /// ///
C1 is made up of capacitance of the components to ground plus the input cap of the opamp. Without C2, for high values of R1 and R2, there may be enough phase shift to cause instability. If you make C2 too big, then you will get an integrator, lose high frequency response.
In article , snipped-for-privacy@yahoo.com says...>
Nope. The attempt was to roll off the bandwidth so it didn't oscillate. I generally do this but in this case didn't because I didn't want to introduce any phase shift. in the signal (though 5pf doesn't introduce enough to worry about).
Just a unity gain inverting amplifier ('+' to ground through Rf//Ri), driven by the output of another LM833. The layout looks pretty tight to me.
input cap of the opamp.
response.
I don't need the high frequency response but any phase shift is undesirable. With moderate C2 it's not a significant problem but my board has no C2 pad and while stacking 0603s is certainly possible, it's ugly.
BTW, the National FAE's response was that 100K feedback resistor is "iffy" for an LM833. He offered the 100K number before we told him the values we were attempting.
Seems so. As I said in another post, even a 4.7pF cap killed it. I was going to use a 39pF cap (smallest already in the BOM) when I found a smaller resistor also did the trick.
That was the FAE's point.
Our standard resistor is 0603 but I didn't do anything special to the planes. Wiring should be pretty tight though. I'll look closer. Terrible stackups? Don't you define them?
The fast-turn houses we use tend to make the outer dielectrics about
12 mils. Since we tend to make layer 2 our ground, that gives us a lot of capacitance, and low impedance/lossier traces, on the top side. I'd prefer L2 and L3 (usually power pours) to be closer, and the top/bottom dielectrics to be thicker. They do the opposite. If we give them our preferred stack, the price about triples.It's probably not a big deal on an audio-range opamp, which will probably have several pF of input c anyhow. But keeping the traces short is a good idea.
John
If you want to use the cheap services like Advanced Circuits $66/ea 4-layer boards, the board house defines the stack-up because they're running your board with many other peoples' as well. (Their typical stackups are shown here:
Personally I don't think Advanced Circuits' 4-layer stackup is all that awful... the outer layers are ~10mils from the inner plane layer, and while having it be somewhat tighter (perhaps 6mils) is often preferable, I can live with it. I'm curious to hear what John doesn't like about it...
---Joel
If the resistors are too big it becomes sensitive to noise. Too small and the circuit uses a lot of power. And other effects mentioned elsewhere.
-- Dirk http://www.transcendence.me.uk/ - Transcendence UK http://www.theconsensus.org/ - A UK political party http://www.onetribe.me.uk/wordpress/?cat=5 - Our podcasts on weird stuff
A 10 mil 1-2 and 3-4 spacing makes outer layer capacitances to the planes (usually 2 and 3) high and makes matched-impedance outer-layer traces skinnier and lossier at a given impedance. 6 mils would be worse! And the best power bypassing would be if 2-3 were close together to increase plane-plane capacitance there. But it's usually not a big deal if you plan for it.
John
looking
Electronics.
Fixing it? You are right, bad move; how about amping it up though (gently at first to be sure)?
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.