Aargh! We were talking about open loop gain, thus not the loop gain at the FB resistor. Read again: http://www.diyaudio.com/forums/soli...lls-power-amplifier-book-103.html#post2402111 where Wahab explicitly stated: 'not loop gain'. Furthermore, I've already explained the difference between these two gains here: http://www.diyaudio.com/forums/soli...erview-negative-feedback-322.html#post2374660For a typical audio power amp the difference is about 30dB!
Next, I was talking about a typical blameless amp. As such an amp, compared to Bob's fig 3.10, has much lower degen. resistors in the IPS and is supplied with a much higher tail current, the open loop gain is (obviously) much higher, hence a UOLGF of 30MHz or so and hence about 130dB DC gain.
Many thanks for making that page accessible; I have not had time to dig my old HFNs out of the cellar.
The circuit does not look like TPC as we know it, Jim; the R1-C1 network looks like a lead-lag network of the sort used in the notorious Otala amplifier. The previous website reference to three gain stages could be more evidence that the circuit might be an Otala, errm, tribute.
I suppose no-one has the schematic? I've not been able to find it.
Hi Edmond, you are probably right (although the schematic with the gain probe would always do a much better job in clarifying the author's intention, compared to stray/isolated graphs, language semantics and words phrasing). However, assuming it's about open loop gain and open loop unity gain frequency, I fail in understanding the relevance of those, in the context of a Miller-TPC-TMC comparison.
Last time I've checked, both the linearity and Bode stability criteria (phase and gain margins) are relative to the loop gain and unity loop gain frequency. The open loop parameters (gain, unity gain frequency), as used by wahab and yourself are, at best, only indirectly relevant. A higher open loop gain for the TMC, compared to TPC, (if that's possible, I haven't checked) doesn't mean that TMC creates more (useful for linearizing) loop gain compared to TPC.
Last edited:
I thought that 'the usefull loop gain for linearization' is simply the difference between closed loop gain and open loop gain. So, for the same closed loop gain, if you get more open loop gain (at a certain freq) you have more gain you can sacrifice for linearization.
jan didden
Mr Self,
question is if the DPA-50S model is All that different, you indicated that Deltec Precision merely "claimed" they used 2-pole compensation.
The DPA-50S is a development from/out of the 100S, the step from double-pole to two-pole is in the margin of semantics.
Seems odd that DPA does not mention it anywhere currently, while sticking to their Sziklai output stage arrangement.
YouTube - Dillinger - Cocaine In My Brain
Last edited:
Hi Jan, yes and no.
What you call "the sum" (in the time domain) is in fact (in the frequency domain) the product between the transfer functions (Laplace transforms) of the feedback network and the amplifier open loop gain. That's not necessary as simple as it may look, because the feedback network may add to, or modify (by (un)loading the amplifier), the pole-zero distribution. If the feedback network is strictly passive, then only loading effects (if any) are to be considered. If the loading effects can be neglected, and the feedback network is strictly passive, only then the loop gain is a scaled down (by a constant) version of the open loop gain.
An analysis on a specific schematic is the only way to find if any of the above approximations are valid - that's why I was asking wahab for the schematic, and stated that an open loop gain graph in isolation doesn't tell much about the loop gain itself.
As an example, if the feedback network has a phase lead cap in parallel, then the loop gain is totally different to the open loop gain.
Hi Andrew,
You raise a couple of good points here. One of the nice things about the Locanthi Triple is that the predriver and driver always run in class A. Indeed, one of the disadvantages of the CFP output stage is that the driver cannot run in class A.
However, when a CFP is used as pre-driver and driver, the CFP can be run in class A, just like the Darlinton emitter follower pre-driver and driver in the Locanthi Triple if the outputs of the top and bottom CFP are connected together with a single emitter resistor.
Cheers,
Bob
Hi Doug,
Well, it is true to me. Minority carriers cause trouble leading to severe crossover distortion; ordinary switching distortion spiced with cross-conduction, a very annoying combination even for a hardened ear. Bipolars are poor transconductors, have less favorable input impedance, large base-to-collector capacitance, exhibit inferior switching characteristics, and numerous disturbances originated from the bipolar structure, there`s no good reason to use them as slave drivers anyway.
loop gain = nfb ratio = open loop gain/closed loop gain
unity gain frequency : frequency at wich the open loop gain is at 0db
unity loop gain frequency : frequency at wich OLG = CLG
Btw, your sentence is not very logic.
Why should the loop gain be equal to the open loop gain ?....
That would be right only if the amp has a CLG of 0db,
not exactly the case with the amps we are discussing in this
thread, be it a poorly designed blameless...
Last edited:
Hi Jan,
Basically, you are right, even with lead compensation in the FB network. However, in case of TMC, things are a bit more complicated. If you simply disable the global NFB you will get (almost) the same OLG as with conventional Miller compensation (CMC). IOW, you don't see the additional gain that linearizes the OPS. As I already pointed out elsewhere, the crux of TMC (compared to TPC) is that only the error created by the OPS is 'TPCed', so to speak, while TPC acts on the whole signal.
If the aforementioned error is only 1%, you hardly see any differences in the Bode plots of TMC vs CMC. In order to make the additional gain 'visible', you have to maximize the error signal. i.e. to make it equal to the output signal itself. How? Simply, by connecting the TMC resistor to ground instead of the output.
Now you probably would say: hey, isn't that TPC? Indeed that's also TPC and that's why TMC and TPC, at least distortion wise, are closely related to each other.
OTOH, as TMC acts only on the error signal, which is relative small !, TMC is not a seccond order system, hence (almost) no phase dip and no overshoot.
BTW, there's one caveat: by tying the TMC resistor to gnd, you also alters the VAS loading. So this method of assessing the OLG is not 100% exact.
Cheers,
E.
Maybe the reason DP audio rather prefers to name it Current Coupled Class A ?
(for those interested : the 2nd page of the Hi-Fi N&R article i linked to describes the Yves Bernard André YBA1 model, first power amp i've seen in the late 80s that employs a BJT output stage without emitter resistors. Current YBA models do, btw)
Attachments
Jan, 'at your service' as the late Pim F. would say.
It didn't bring good fortune to him.
JPV
</p>
Hi Edmond, that is incorrect. Set aside not all second order systems are overshooting and exhibit a phase dip (but only the overdamped systems, or otherwise said in the frequency domain, with close pole/zero pairs), the Bode integrals theory shows clearly that for a constant unity loop gain frequency, and while keeping the closed loop system stable, you can't get more loop gain than what Miller compensation (as a first order system) can provide. Or otherwise said again, for a first order system, the gain times bandwidth product is a constant.
Obviously TMC provides more loop gain than Miller. That extra loop gain has no other way to appear other than from a pole-pole-zero distribution. That is, roll the gain steep and late (in frequency) and bring the roll off to 20dB/decade around the unity loop gain frequency, to keep the closed system stable.
What you can say is that TMC, as a second order system, can provide the same loop gain as TPC (also a second order system) while not requiring the system to be overdamped. This property is indeed an advantage of TMC over TPC, and is one reason why I was looking after an analysis on if and how TMC implements pole splitting.
I would love to further discuss these things, they are indeed fascinating, but it's holydays and I need to get a life