On that basis, you can argue that all push-pull designs are tainted due to the
nonlinear character of the devices, just as you can argue that miniscule
minimum current makes them Class A.
As a practical matter you might have to decide what minimal percentage of
bias current needs to remain in the lesser conducting device, or perhaps
define your terms as a percentage of distortion present in the current
through a single device.
Personally, I use a 5% or 10% minimum criterion for the purpose of rating
the output power of a square-law push-pull circuit. Less than that but more
than zero is closer to "non-switching".
We can argue endlessly (it seems to be the purpose of this thread anyway), but a criterion is a criterion: for me, we stay in a "pure" class A domain until one of devices has begun to take more than its share of the current (ie when there is a deviation from the constant sum law).
This can only happen in class AB or sliding class circuits: with the Allison, where the crossing is the intersection of two straight lines, the "pure" class A regime is lost as soon as one device goes into cutoff, which means the "general" class A is also lost.
This is why I tend to class topologies according to a hierarchy: on top is true class A where the OP stage as a whole remains in stasis, just below is the shady area of the sliding bias/non switching class B/etc (there are also subcategories there), and at the bottom, conventional class AB (which can be very good in practice, but suffers at least theoretically from variations of dynamic resistance).
Pure class A tends to have a current-invariant linearity (not 100% true, but more so than other configurations).
OK, for a 1kW amp, this is not really an option, but for a line buffer or an headphone amplifier, why not if an uncompromised top quality is required.
In practice, reasonable safety margins are required because of engineering tolerances and because higher order effects begin to show if you stretch your (good) design too close to the limits, I don't argue against this
Elvee, I notice you used a quasi complementary output stage. Is that a must, or a matter of choice?
Also, you used 2N3055. What would ahppen if, instead of him, you used say MJ21196, which can take a lot more voltage and can dissipate more than double the power? Or the very popular Toshiba 2SC5200?
Also, you used 2N3055. What would ahppen if, instead of him, you used say MJ21196, which can take a lot more voltage and can dissipate more than double the power? Or the very popular Toshiba 2SC5200?
In the circlophone? It is a must, and it isn't actually quasi (arguing about terminology once again). Quasi, (for me at least) is a mix of a true NPN darlington with a PNP emulated by a CFP composed of a power NPN and a PNP driver.
In the circlophone, the two OP have the same sex, and can be both CFP's, or darlingtons or even MOSFETs: all of these options have been explored.
To my eyes, complementary symmetry in amplifiers is only superficial, since N and P devices tend to have naturally different characteristics (for silicon at least), and complementary pairs are produced by stretching the processes in opposite directions, with a result that is not really symmetrical (not as much as one would like anyway).
Synthetic symmetry by other means seems therefore preferable. I didn't invent the concept: Bengt Olsonn among others is the originator, but I find it sound.
To summarize, better symmetry can be obtained from same sex devices than from "complementary" ones.
All of these have been tested and with great success, but I always design my BJT amplifiers to be compatible with the 2N3055 or the MJE2955: it is a good reality check. If they achieve good performances by modern standards with these transistors, this means they are really robust and anything-proof: something like the AK47 of amplifiers.
A design using the best available components of the day and lavishing on ridiculous design rules has no merit of its own, it simply relies on the work of others and basically adds nothing into the equation: it simply subtracts from engineering margins created by other, more competent people, essentially the guys from the semi companies.
A good design has added value, and I can recognize such a value instantly when it is present. That is what I strive for
Not quite, Dejan
. It's certainly measurable - and all that it in fact is, is that the spectrum of distortion components is different from the usual - but the "tragedy" of audio, as it currently stands, is that there is almost no interest in better understanding "what's going on". Hence the ongoing thrashing, going around and around and around the same old boring circuits, , etc ...How come? When using modern RET devices, chances to run in output stage stability issues (in particular for triples) are increasing exponentially. Usually, substituting original 2N3055/2955 with modern high Ft devices is a safe receipt for disaster.
600 Ohm loads (input impedances) in pro audio? Today? It's a myth!!!
Pro audio switched to high-ish impedance loads many decades ago. Exceptions are 'vintage' units still in use, their re-issues and perhaps some recent DIY 'pro' gear.
Are-you sure you don't mix cheap home studios with real professional facilities, like big recording studios, video and movie's post production facilities etc.?
Even the portable eight tracks digital recorders some use in movie live miking , on batteries, are still 600 Ohms and XLR. Like this one:
http://www.aeta-audio.com/fileadmin/downloads/press_info/4Minx_Fact%20Sheet.pdf
Have-you ever seen any pro big PA system with something else than XLR and 600 Ohms for lines.
Asymmetrical high impedance and CINCH or Jack plugs ?
Can-you imagine the hum, and how many seconds before some jack unplug itself or cable pulls his plug ?
Last edited:
Yes.
Where do you see 600 Ohms?
Huh? Don't mix 600 Ohm drive capability on output(s) with 600/600 Ohms transmission system. Learn proper terminology
Voltage Audio Distribution
And stop TROLLING!
Over and out.
Attachments
Substituting faster OP stage is not a problem from a global stability perspective: this simply means that the amplifier will be overcompensated but stable. Doing the opposite would be problematic.
With faster devices, there is a real risk of local instabilities, but these can normally be tackled with the usual remedies, stoppers, etc, and I tend to avoid triples anyway.
Yes
Elvee, thank you for the explanations.
So, if I were to use selected MJ21196, which I happen to have a lot of, I will be replacing a nominally 70W (although this depends uch on its origins) trannie for a nominally 250W trannies.
Can I raise the supply voltage from your initial 25V to say 40V? Should I expect anything as a side effect?
How will the measurements do, improve, stay the same, degrade?
So, if I were to use selected MJ21196, which I happen to have a lot of, I will be replacing a nominally 70W (although this depends uch on its origins) trannie for a nominally 250W trannies.
Can I raise the supply voltage from your initial 25V to say 40V? Should I expect anything as a side effect?
How will the measurements do, improve, stay the same, degrade?
Just drawn in an unconventional way. Q1-2 are fed directly by the supply, and Q3-4 via a RC filter.
Don't we were talking about "drive capability on output(s)" and drivers ?
Who said a word about inputs impedances witch are allowed to be between 600 Ohms and infinite...
Who is trolling ?
Last edited:
Not quite, Dejan
Frank, in case of my headphone amps, both the BJT and the FET/MOSFET version, moving on the bias from 44 to 86 mA produced ZERO difference in measurements, so long as I kept the output volume to the limit it takes a headphone of lesser effciency to reach 120 dB SPL. ZERO difference.
On the other hand, I have not tested many headphones, just some Sennheisers, Grados, Sonys and Koss cans, the ones I own or could borrow. It's statistically possible, even probable, that there are some odd units out there with which a difference might have been noticed.
Also, my test gear is reliable only up to 0.001% up to 100 kHz. If I had better measuring gear, I might have seen some differences beyond that. The bottleneck is the oscillator. I have one going up to 30 MHz, but it's truly useful only for bandwidth measurement, as its own THD is nothing to write home about, and another, which is a very low THD unit, less than 0.0001%, but only up to 100 kHz.
The one thing that did produce a measured difference was raising its PSU voltage from +/- 16.5V to say +/- 24V. That produced some nice side effects, specifically lower measured THD.
These days, Ethernet in various flavors. KAESOP, DANTE, that kind of thing.
The MJW21196 (same transistor, plastic) has been tried: http://www.diyaudio.com/forums/solid-state/189599-my-little-cheap-circlophone-10.html#post2639264
Some useful info starting here:
http://www.diyaudio.com/forums/solid-state/189599-my-little-cheap-circlophone-21.html#post2690089
And of course, Daniel has created a nice builder's thread with lots of information, including dimensioning for rail voltage:
http://www.diyaudio.com/forums/soli...tion-parts-accessories-beginner-friendly.html
Measurements stay basically the same, and if you use transistors having a good gain flatness like the 21196, performances will improve somewhat
So it is both distance and angle, thanks to your answer.
I really love the idea... and your prototype
While i'm asking myself questions about the way to schield the coils against external magnetic fields (Mu-metal ?) without adding distortion to the signal.
It seems very impratical, specially if we consider cross talk stereo problems and mechanical vibrations ones to solve, but all the fun of crazy DIY, "on the wild side".
I really love the idea... and your prototype
While i'm asking myself questions about the way to schield the coils against external magnetic fields (Mu-metal ?) without adding distortion to the signal.
It seems very impratical, specially if we consider cross talk stereo problems and mechanical vibrations ones to solve, but all the fun of crazy DIY, "on the wild side".
Last edited:
- Status
- Not open for further replies.