On my opinion, it is not so much the length of the connections that matters, but the AC leakages between grounds of various equipment connected together.
Sometimes distortion specs are better with unbalanced connections, for equipment that has both balanced and unbalanced. Maybe preferable go unbalanced where possible for that reason.
all of the above. distortion and noise slightly higher, cost and complexity slightly higher and no real ground issues or field coupled EMI issues in home systems.
-RM
Last edited:
This is where my 'quest' concerning 'input current distortion' started, the subject was introduced (at the least to me) by Bob Cordell.
In reaction to the, above referred post, I did some simulations and posted the results, I even tried to build a 'input-current-distortion' measurement device (posted earlier), this failed while my environment seems to noisy and my lab equipment seems not fitting the job. After that I offered to send the measurement device to anyone that has an interest.
After this post and the work done by me no (real) discussion about the subject emerged, so (by now, and after the post referred above) I give up. Thanks for the flowers
Frans.
Last edited:
Hi Jan, in my opinion everyone talking about amplifiers should have this book on his bookshelf
as it sits (annotated and all) on mine.
Last edited:
I think whether one believes the ground issues are significant is where it gets into personal belief on where the 'next veil' is. You believe in the efficacy of mains filters and have gone on record on how important they are to you. But you also use power line ethernet which I would never put in my house due to the HF grot it add that you them have to filter. But UK homes use a ring main system so you are inside the grot antenna unless you rewire the listening room.
I've copped out and am looking to battery power all source equipment (thanks to Jan) and if I can't use optical (TV and bluray). But I know its warm fuzzy rather than hard measurement
In my system, I can hear a difference simply by changing the phase of any AC power cord.
(Easy to measure the best one looking at Voltages between the two grounds before connecting them together.)
I have a little headphone amplifier that I used professionally witch can be powered both by AC or by batteries. The difference is amazing.
I'd like to add one point. The input stage, being operated as transconductance element, transfers a voltage into a current, which then is feeding the transimpedance stage ("VAS") to yield the output voltage. BJT have an exponential basic input law, when driven by a voltage. So, the input stage output signal, considered in each of the both branches on its own, has to be distorted. It is now common belief that summing both signals at the input to the transimpedance stage will cancel distortions of certain orders.
One should ask whether this is exactly true. Even if it is, the control loop actions will produce further components of non-linear input current in order to correct remaining distortions.
Matthias
Last edited:
I do not agree at all. Kolumni's book was a great disappointment for me as, in his approach, I did not find anything original in it compared to Hephaïstos's writings between 1980 and 1995 in L'Audiophile (in french) :Hi Jan, in my opinion everyone talking about amplifiers should have this book on his bookshelf
index
index
Thanks for the references, I am not able to read the French language directly but google translate is of great help.
Still I would like to re-express my view about Kolumni's book, even if there was nothing new for you, getting confirmations for one's knowledge, or [some kind of] a second opinion (even a different view) is also worthwhile.
P.s. Searching for "site:asrr.org Hephaïstos" does not yield any results.
Last edited:
It is certainly true that the output currents of the differential input stage collectors must be distorted. It follows that the net input current to the VAS (mainly flowing into the Miller compensation capacitor in many cases) must also be distorted to compensate for later distortions in the open loop. Indeed, this is the case because these currents are, for the most part, error currents.
The summation of the two currents from the collectors is usually done by a current mirror. That is indeed where some distortions and common-mode effects are canceled. One does need to be mindful that simple current mirrors are not perfect, and therefore that cancellation may not be perfect. For example, in a simple 2T mirror, even if the transistors and emitter degeneration resistors are perfectly matched, the signal output of the current mirror is less than the signal input to the current mirror by the amount of the two base currents of the transistors in the mirror.
It is also the case that the input transistor in the 2T current mirror is diode-connected, meaning that Vcb is zero. This can put some transistors near the region of quasi-saturation, with current gain lower, and less than that of the output transistor of the current mirror. Of course, the output node of the current mirror is often at a higher voltage than the input node, especially when a 2T VAS is used. These are just examples of some things that can detract from the perfection of the current mirror.
If a 3T current mirror is used, with an EF "helper" transistor, some of these effects are mitigated, including the base current error, the differences in Vcb of the two mirror transitors, and the closeness to quasi-saturation of the mirror input transistor that was otherwise diode-connected (it now has an additional Vbe of Vcb headroom).
This approach works very well with a 2T VAS, putting both mirror collectors 2 Vbe + degeneration drop above the negative rail. If the voltage drop across the VAS degeneration resistor is the same as the degeneration drop at the mirror emitters, the 2 mirror collector voltages will be essentially the same (within a tweak to account for some small differences in Vbe between the mirror transistors and VAS transistor).
Cheers,
Bob
We are in accord here. I didn't want to refer to the imperfect real-world implementation of the mirror and summing point. Rather, I was wondering whether exactly all distortions would cancel in a thought experiment. Since the transistors of the input diff pair are always driven into different directions of the exponential curve with different transconductance slopes, I really would be surprised.
The main point, related to some simulation results presented on the previous pages, was: two nonlinear elements, surrounded by a finite-gain ideal loop, probably won't lead to exact linear behaviour.
BTW, do you see a reason to worry about stability of a 3T mirror "with EF helper"? Simulating it, one can easily get phase margins of only 30 degrees, and some people even talk of peaking in such a mirror.
The fitting of 2T VAS and 3T mirror is indeed nice. With a 2T mirror, one has to increase emitter resistors. But balancing also becomes more critical, since headrooms are smaller.
Cheers,
Matthias
Last edited:
The links I gave are valid.
I do not think Google translate will work well because all the pages are scanned images. However, even if you do not understand the french text, the schematics are very instructive and will show how Hephaïstos was in advance in the search of the highest linearity. For example, look at this :
L'étage d'entrée de l'ampli -3- différentiels insolites (Héphaïstos)
There is a Caprio's input stage. Who knew it at that time ? By he way, Caprio is a DiyAudio member ! You also find other topologies which are sometimes discussed by now.
If I gave that impression then I apologize, my comment was just that the site is poorly indexed and that the search engine can not provide any help in finding useful information.
Yes I tried, google translate is of no help, and, although looking at the pictures may be nice, having the insights of the auteur is needed to get full appreciation of the data presented.
I am very disappointed that you have no appetite for further discussion.
We disagree, certainly.