What I really want is for engineers to make sure they know what they are trying to avoid or fix in the first place. If one does not have an intuitive grasp of why amplifiers oscillate, it is difficult to find the optimum solution with or without a coil. Some seem to think that a coil is absolutely necessary, others would omit it just because it looks in the way. If someone wants to lay out the problem with loading an amp, or preamp with a capacitive load in detail, please do so, or find a link that we can all relate to. If I knew a link, I would have mentioned it.
Cap loading the way Bryston is just fine, in my opinion. Why did this come up in the first place?
Cap loading the way Bryston is just fine, in my opinion. Why did this come up in the first place?
John, I think you were being objectively correct and politically incorrect. 😀
I'm struggling with the value of the proposed RFI fix of adding a shunt cap at the output. How important would you say RFI pick-up is on the speaker cable to an amp, and if it is of any significance at all, whether a simple series resistor on the output creates a significantly higher vulnerability compared to using an inductor and/or shunt cap?
I'm struggling with the value of the proposed RFI fix of adding a shunt cap at the output. How important would you say RFI pick-up is on the speaker cable to an amp, and if it is of any significance at all, whether a simple series resistor on the output creates a significantly higher vulnerability compared to using an inductor and/or shunt cap?
traderbam said:
(today)
Sorry, I was off by a factor of two.
Although what Bryston has done is not what I would do, what is your reason for characterizing their approach as dumb?
(as you know, I would have an RC Zobel in front of the coil, would have a smaller coil, and would have a small resistance in series with their 100nF output capacitor.)
Cheers,
Bob
Oh, for two main reasons. Firstly, I spend a lot of my time trying to remove electrical "mass" from my circuits, so adding a lump directly across the signal path goes against this, and secondly because it creates an opportunity for resonances, as Eva has eleaborated upon.Bob Cordell said:
In general, adding mass to a system to improve its stability (if that is what Bryston have done this for - I don't know) is not as good as correcting the stability in other ways.
I'm with you that an RC shunt to gnd, with care about path inductance, should be used before the series R and optional L.
Why do you think Bryston chose this arrangement?
Brian
traderbam said:
Thanks, Brian. I agree that 0.1 uF is a bit much.
I really don't know why Bryston did it this way, but I really do wonder whether they would have needed to be this heavy-handed to achieve stability if they had put a zobel in front of the inductor.
Here's one more thought in regard to concerns about a direct capacitance to ground at the output possibly getting into a resonance with the load. In an amplifier that incorporates a Zobel in front of the coil as well, note that at high frequencies the resistor across the coil in combination with the resistor of the front zobel provide a lossy resistive shunt to ground for any parallel resonance that might occur between the output capacitor and the load. This is a good thing.
In the case of the Bryston, where they have no Zobel in front of the coil, they may not get much if any such damping because the emitter follower output of the amplifier will itself begin to look inductive at high frequencies. So in this regard I do tend to think that maybe they goofed or at least did not pursue the best overall solution. Of course, in the commercial world, we must always recognize that an approach using a Zobel in front of the coil adds two components and cost. And, of course, components like these in output networks must be of high quality and may be called upon to withstand high currents under some conditions.
Cheers,
Bob
phase_accurate said:
Right, Charles! 😉
In my current hybrid design I have 1 Ohm resistor against 8 Ohm load up to 10 W (class A MOSFET); 0.33 Ohm up to 100W (class C BJT), and zero Ohm for the rest (class C MOSFET). However, I had to use a negative feedback to smoothen crossover points of approximation, so real output resistance is lower (0.1 Ohm max, and very stable below 10W across all audio frequency band). Errors are compensated by very fast amp on 2 tubes: one pentode and one triode, with local NFB and about 20 DB of extra gain to compensate output stage errors. Now I am waiting for a PCB soaked on backyard to build a prototype to check it in the real life on festival on campground: I tryed a breadboard with record of a helicopter only because helicopter's sound is normal for neighbors, but I badly want to test it on full loudness with a real live music.
Why so much effort when you can make full-range multi-kilowatt class-D pocket amplifiers with uniform 0.05% THD and just a few pairs of TO-220 MOSFET? I've lost most of my interest in linear amps since I got into class D and loudspeakers (which seem far more challenging subjects).
Eva said:
Because I am not interested in absolute THD numbers. 😀
Why do you think signal levels are measured in Bells instead of voltage percents? 😉
My current amp is very similar to 3-state 3-bit D/A where each bit is analog (0-10W,10-100W,100-1000W). 😀
It uses 2 pairs of MOSFETs and 1 pair of BJTs. Plus some other parts, like one BJT for voltage controlled current source and one small MOSFET in current mirror. Plus one tube designed for latest TV sets containing pentode and triode, for very fast very linear very wide frequency amplification and error correction.
Actually, very few parts, may be even less than typical class D uses.
However, I pay by an electricity consumption for such sonic results, but much less than typical audiophile amp with such power and quality would consume...
Anatoliy
One advantage of class-D is that no one would dare to discuss if it is wise to have a capacitor right across the output ! 😉
Regards
Charles
phase_accurate said:
One advantage of class-D is that no one would dare to discuss if it is wise to have a capacitor right across the output ! 😉
Regards
Charles
Earlier in the discussion I thought about asking opinions of removing the coil in class D amps. 🙂
Guess one could do that if using a resistor instead, but that's hardly very practical.
I guess there would be some dumb..s audiophools that would buy a class-d amp that is using an RC lowpass at its output if you point to all the (non-existing !) advantages in a high-gloss brochure.
And yes - making it expensive would definitely be selling point !
Regards
Charles
Actually you can also disturb UcD-style self-oscillating class D amplifiers with a load that resonates at RF resulting in a dip in impedance... In these amplifiers leakage inductance of output capacitor is very critical, it should self resonate above 1/(twice_the_propagation_delay) which only allows for a few nanohenries. There are plenty of poor implementations.
I have yet to find a way to decouple the load at RF without disturbing the modulator.
I have yet to find a way to decouple the load at RF without disturbing the modulator.
phase_accurate said:
It would probably be hot like a class A amp, so it must be good, mustn't it? 🙂
There's something I hadn't thought of. Perhaps Bryston are trying to stop bad stuff leaking out of their amps. 😱phase_accurate said:
One advantage of class-D is that no one would dare to discuss if it is wise to have a capacitor right across the output ! 😉
Regards
Charles
There would be a simple test for that: Just play the latest Madonna CD. If there is silence then it works - if not it doesn't !
Regards
Charles
traderbam said:
Brian, your attempt to bash Bryston's Zobel network fails miserably, by implying that a Bryston amp must slew at an (almost) impossible 600V/us, in order to comply with their specs.
1. Slew rates are always measured right at the output of the amp proper, NOT after some LPF like a Zobel network.
2. Even if the slew rate was measured after the Zobel network, then your calculation also fails. At max. output (about110Vpp), that network is capable of slewing at a max. rate of 250V/us, so your figure of 600V/us, needed at the input to get 120V/us at the output, is wrong.
BTW, don't confuse the slew rate of an amp with the rise and fall times of a filter.
Also read John Curl's comment on slew rate:
http://www.diyaudio.com/forums/showthread.php?postid=1558681#post1558681
Edmond, you really need to go back and read the preceeding posts more carefully before casting aspersions on my concerns about Bryston's output network and before you question my ability to read John Curl's posts. Also, you need to find the correct specs for the 7B amp.
However, this prompts me to point out that I don't really care about slew rate. A zero distortion amplifier driving 8-ohms at 300W from a CD source should exhibit a maximum slew rate < 10V/us. A designer can achieve a high slew rate simply by using high voltages and high currents. On it's own, a bold output slew rate spec. means nothing at all when it comes to sound quality. I recall reading about a fashion for high slew rate in the late 1970s...a rush of amps with ridiculous specs emerged, even a Japanese one that claimed 500V/us, if memory serves. The sound quality didn't correlate positively.
I have no concerns about output inductors from a slew rate point of view.
Stereophile
However, this prompts me to point out that I don't really care about slew rate. A zero distortion amplifier driving 8-ohms at 300W from a CD source should exhibit a maximum slew rate < 10V/us. A designer can achieve a high slew rate simply by using high voltages and high currents. On it's own, a bold output slew rate spec. means nothing at all when it comes to sound quality. I recall reading about a fashion for high slew rate in the late 1970s...a rush of amps with ridiculous specs emerged, even a Japanese one that claimed 500V/us, if memory serves. The sound quality didn't correlate positively.
I have no concerns about output inductors from a slew rate point of view.
Stereophile