Art: After reading more of the threads on this website, I have noted an overriding preference for simple designs around here.
OTOH, personally I enjoy things that are challenging enough to make me think seriously and sweat to get them to actually work, and that extends to the kinds of schematics and board layouts that I like to cook up.
But although much of my input may prove useless to this website, I really don't mind giving advice or sharing ideas where it may be warranted. So I'll probably continue hanging around, post a few things, and if the response is active, maybe I will post some more. Conversely, if the response is dead, or if there are too many responses that show a lack of basic understanding or thought, I'll likely limit my posts.
In the meantime, here's a circuit sketch for a line amp (20dB). Not too complicated, uses opamps, but not altogether uninteresting. Maybe even useful to someone.
FYI, you probably already now it, but AD846 is a discontinued
product. Recommended replacements according to AD are
AD8001 and AD811, if memory doesn't fail me. I haven't
checked the datasheets, so I have no idea how compatible
Christer: You are right, and thanks for the correction. Oh well, it wouldn't be the first time that I have designed something using out-of-production (OOPs) components ;-).
In order to make the design that I posted work, you need external access to the internal node that drives the output buffer. AD844 is equipped with this (pin 5, same as the AD846), and I think it should work. So swap AD846 with AD844 and try it out (perhaps some fine-tuning of the time constants may be in order).
If you want to make sure that the time constants have all been trial-and-errored and optimized (so you don't have to do much debugging work yourself), I believe that the AD846 may still be available as old stock (price may be somewhat higher than when new).
"But although much of my input may prove useless to this website, I really don't mind giving advice or sharing ideas where it may be warranted. So I'll probably continue hanging around, post a few things, and if the response is active, maybe I will post some more. Conversely, if the response is dead, or if there are too many responses that show a lack of basic understanding or thought, I'll likely limit my posts."
I think it was pretty clear that he meant it to to be exactly that.
A composite op amp with two nested feedback loops, one around compensation node a current feedback op amp. Quite a test I'd say.
"In the meantime, here's a circuit sketch for a line amp (20dB). Not too complicated, uses opamps, but not altogether uninteresting. Maybe even useful to someone."
uses opamps, but not altogether uninteresting
Well one out of four isn't to bad. The gain is actually 20.83 dB.
Man I hate to see what complicated looks like.
Gentleman start your Spice engines and warm up your 350MHz scopes!
Just a bit of fun Mr. Carr. I guess any of the good stuff that goes into a $30.000 preamp is WAY beyond mortals like us.
Hartmut, Art: Although I don't enjoy working with opamps as components so much, many of the circuits I find immensely interesting. Some IC designers are very, very clever and talented.
And often I find it useful to try out an idea in the easiest, most painless way possible. So if I can locate an opamp that provides the necessary access nodes, it will definitely be my guinea pig (grin). Then if the idea works in practice, I can consider transferring it to a discrete implementation.
If you build and measure the posted circuit, I think you will find that the high-order distortion is _very_ much lower than either the 844 or 848 can manage on their own. The circuit can swing quite a bit of voltage, too - about 25V p-p out to 2MHz, tapering off to 10V p-p at 10MHz. The slew rate isn't bad, either, something like 300V/uS going either up or down.
Hartmut, no, I haven't gone overboard in applying this exact type of thinking to the Connoisseurs yet. The air-dielectric construction that we use is bloody hard to make, and this is both our strength and our weakness. Although I love interesting design ideas, in reality I must think very carefully about how much circuitry I introduce, what it is intended to accomplish, and whether the benefit is worth the greater complexity.
In this respect, I am no different than any DIYer. Before I can afford to get too liberal with the parts count, I need to cook up some new circuit board structural concepts that provide an acceptable level of sonic performance but won't cause too much grief in production.
Nonetheless, I always like to work on something new, and so on my next "3.1" design, in addition to a more advanced circuit construction system, I want to try a Cob-cancelling concept, push-pull self-biasing MOSFET vreg, current-sensing self-biasing JFET buffer stage, stacked/split grounding system, add another round of general topological refinements, and...and...
Just had a look at the circuit with the active feedback you posted. I *think* I understand most of it, except the 47k load at pin 5. The way I see it, this decreases the transconductance of the 846 from, what is it, a couple of megs to the 47k, losing 30dB or so of open loop gain. If so, can you tell me the reason you did this?
Jan: Sounds like it is time for a hint. I suggest that you reverse the order in which you are looking at the circuit.
Assume that we first are using the 846/844 alone, per a "regular" circuit application, with wide-band low-distortion as a major design goal. In our quest for lower overall distortion, we add a resistive load (of suitably smaller impedance than the internal phase compensation caps) to the collector-collector node that drives the output buffer. One outcome is reduced open-loop distortion, which is a desirable outcome. But another outcome is reduced open-loop gain, which, as you rightfully pointed out, is not desireable. For when we close the loop, the amount of global feedback will plummet, leaving the closed-loop distortion pretty much back where we started.
But what if there was a way to apply that resistive load but also keep the amount of feedback intact?
Well, as you say, there are several ways to look at the circuit. Initaially I also looked at it as a feedforward circuit through the transcond stage of the 846, but realised that is not correct.
You are making up for the lost open loop gain by the introduction of the 848's open loop gain in the closed loop. What counts is the excess gain in the loop, not (for this discussion) where in the loop it is located. Are you saying that the result of all this is reduced open loop distortion, despite the added open loop distortion of the 848?
BTW, you may want to take a look at the AD830. That has the conceptual topology of your circuit on a single chip.
Jan: Um, I think that it may be faster if you just simulated the circuit yourself. It would be instructive if you can plot the frequency curves of the 846/844 and the 848 individually as well as of the composite output, and you can see the contribution of each part.
With that in mind, you can then study the distortion of the 848 by itself, the 846/844 by itself, and finally, the distortion of the composite circuit. Look at the distortion at 10 Hz, 100Hz, 1kHz, 10kHz, and 20kHz, respectively (of course, you are free to go higher). Out to 1kHz, you probably won't notice much difference between individual and composite operation, but by 20kHz, the differences should be very, very clear.
BTW, thank you for the note on the AD830. '' I like the design concepts; as you point out, there are fundamental conceptual similarities, and I look forward to trying the device out!
>I was hoping for some very clever insight of why things are as they are without having to go to the simulator.<
Ah. But I find that simulations are not only instructive, they are fun to do. At least when I don't bang into convergence problems. ''
>The 830 is also already many years old.<
For whatever reason, I haven't used it for anything yet.
>I think they have a successor now, not sure.<
Sometimes, what the manufacturer recommends as a replacement may not work for a given application. Although Analog Devices suggests the AD8001 as a replacement for the AD846, it doesn't bring the collector-collector node out, and is therefore unsuitable. AD may not recommend the AD844 as an 846 replacement, but it does have the collector-collector node on an externally accessible pin, and therefore IS suitable.
In addition to what the manufacturer recommends, you have to understand your circuit, know what it needs to work, and overlap that knowledge with the internal schematic of the opamp.
Nonetheless, please do let me know if you find something along similar lines as the AD830. I don't think it is so useful for the kind of composite circuit that I was proposing (both front-ends appear to be of identical type, while I am looking for something with different types of front-ends), but it could be a very useful device in its own right.