We want the bridge to be balanced throughout the audio band. C is in one leg, L in the other. The condition is: X(C)/Ra=Rb/X(L) with X being the impedance. If freq goes up, X(C) gets smaller impedance, X(L) gets higher impedance, and if the values are right the equation stays true and the bridge remains balanced through the audio band.
You really need to read that article.
And that C also does double duty as comp cap.
The only thing is where are these 4 parts in this new circuit? L2 is easy because there only is one. ;-)
Jan
You really need to read that article.
And that C also does double duty as comp cap.
The only thing is where are these 4 parts in this new circuit? L2 is easy because there only is one. ;-)
Jan
I understand that Impedance has a broader meaning than "resistance", where resistance is a vector purely on the real axis, as opposed to impedance is a superposition of two vectors, one being on the real axis, and the other on the orthogonal imaginary axis.
Further, I do understand that the Bridge consisting of two resistors, a capacitor and an inductor can be balanced, but I am under the impression that such balance will be possible only at one, very specific frequency, being 1,3 MHz.
Meaning that the bridge will be unbalance at all other frequencies.
Unless, of course, that there are "other" parasitic capacitance, or inductance, that "modify" the elements of the bridge, hence allowing it to be balanced throughout a broader range of frequencies. But I do not "see it".
Is it safe to assume that this design allows for switching noise from the current dumpers to penetrate into the speaker / load, intentionally?
Eureka!
OK! Now I get it !!!
It is the division of two values that actually needs to be equal to a division of the two other values. The way the formula looks, the nominators and the denominators, it should be possible to maintain the equality across a broader range.
Many thanks. Now I see it.
R1, L2, C6 and R17 form the bridge.
C6 will have a trimmer capacitor in parallel for fine adjust.
R22 will be adjustable to set the Class-A and Class-B bias.
I tried every TL071 model you posted and the circuit doesn't work correctly.
As an aside is it good to use a topology so dependent on Op-amp supply current? One cannot drop in a substitute and expect the circuit to work.
Try the AD795 which is a default model in LTXVII
I have exactly the same thoughts about the circuit simulation being extremely component (model) dependent.
I'm doing this project just to use the components I already have. It is an upgrade to an old amplifier.
But you are free to simulate any variations and improvements you want.
Last edited:
Read the thread, I posted a variation without the opamp. But it needs improvement. Try it if you like!
Be aware in this thread there are multiple schematics named "Current Dumping Sch.asc". They are not the same....
The "latest" version of Current Dumping Sch.asc only simulates properly if the spice model supply current draw is the same as what was originally used in resistor calculations. In this Thread there were multiple "TL071" spice models presented. Some are macro models and some are discrete device models. I tried all the discrete models presented and none work with the latest Current Dumping Sch.asc. The original Current Dumping Sch.asc only worked with TL071_Qlevel.asy.
The "new" Current Dumping Sch.asc seems to work with the LTSpice built-in AD795 macro model.
The other current dumping amps in the thread that do not rely on op-amp current draw work with macro and discrete models.
The AD795 works. It needs some minor resistor adjustment to obtain the currents on the schematic.
As I said, I'm doing this project to use components I already have, just to upgrade an old amplifier.
My main amplifier is a Leach 4.5. Two of them.
R6?
I think R6 needs to be 100 Ohms? Something less than R8 anyway.
I have been studying this and I'm still not clear about what's going on:
The bridge balances the feedback but is that the objective, not he output? I think that pretty much all the audio frequencies are delivered via the choke.
I also note that the zero from the 120pF cap is required for stability with this much feedback, which I would definitely have a problem with except that it comes from the driver and not the output so maybe it's safe? How do you check the phase margin with two feedback paths?
I any case, thank you for some interesting ideas!
I think R6 needs to be 100 Ohms? Something less than R8 anyway.
I have been studying this and I'm still not clear about what's going on:
The bridge balances the feedback but is that the objective, not he output? I think that pretty much all the audio frequencies are delivered via the choke.
I also note that the zero from the 120pF cap is required for stability with this much feedback, which I would definitely have a problem with except that it comes from the driver and not the output so maybe it's safe? How do you check the phase margin with two feedback paths?
I any case, thank you for some interesting ideas!
Signal-wise it should not make a difference. The input signal and the opamp output load determine the signal current in the power supply pins, no matter what opamp you use.
The difference lies in the opamp DC current, because that will vary with type. Maybe try to make the circuit such that it can work with a range of DC currents in the supply pins
.Jan
Steve, with all due respect, this is a design from last century and fully explained in several articles that can be downloaded for free.
In a sense the object IS to balance the feedback in the two situations: dumpers off, dumpers on. If the dumpers change state, so does the loop gain, and you normally get either xover distortion or gain doubling or a combination. By precisely adjusting the feedback to keep the loop gain the same in both situations, these distortions are, to a first effect, cancelled. The balanced bridge makes sure that the feedback ratios are precisely adjusted to counteract the changes in loop gain.
Jan
Last edited:
With a few exceptions, amps with op-amp front ends are unstable because the op-amp is designed for stability using feedback around themselves only, so when you add more gain and more poles in the loop, it becomes unstable. You can waste the unwanted gain but you can't undo the phase lag of the extra stages, except to some degree with a lead cap such as used in the current dump bridge.
I thought that the grounded op-amp output might cancel the op-amp dominant pole but if the rails move then not. It seems to me that stability in this amp depends on the zero in the feedback that comes from the 120pF feedback cap. Which leads to the question could the same thing be better applied at the op-amp output and not the final output? An idea for us to play with!
I think the current dumping bridge concept is flawed because maintaining the loop gain is irrelevant. What matters is the output gain. That and potential stability problems that I suspect are behind the frequent failure of these amps is why this concept failed in the marketplace. This is a good example of an idea that goes astray by focusing on a complex theory and loosing sight of the real objective. Academics do this in spades all the time. I think it's required to get your Phd (LOL). We used to call it "BSBB" (BS baffles brains).
I much prefer LV's auto bias idea that moves the "current dumping" ahead to faster stages so that the switching is invisible at audio frequencies.