Power Amp output inductance or not?
I include 2 scope traces of my diy power amp output:
1) Load = 8ohms parallel to 2uF at 4vac with LR output filter. R=1 ohm, L = ?
(By the way , I left the RC in series between the output and ground for the 2 waveforms. R = 10, C = 68nf.)
With no filter now
2) Load = 8ohms parallel to 2uF at 4vac with no LR output filter. R=0 ohm, L = ?
Which one should be the best?
where are you looking? - what you really want to see not ring is the output Q/feedback connection - looking at the L-speaker term will show ringing when the L is doing its job
the Thiel network on the output is there to do a number of jobs.
The inductor separates load capacitance from the NFB loop to help keep the stability correct. i.e. phase margin above design limit.
The cap gives the amp a load at very high frequencies, again to aid stability. Can someone explain this mechanism to me?
The combination provides a filter to remove back emf and interference from entering the NFB loop and modulating the inverting input.
As Jcx said the ringing between the inductor and the load is normal and cannot be eliminated but can be damped, as your wave shows.
If you omit it you eliminate the ringing but open up the amp to the stability and back emf problems that must be solved by other design options.
I am monitoring on the Power amp circuit board directly, not at speaker terminals. I want to characterize the amp. By the way, there is no real speaker connected but a dummy load of course.
I notice that with the LR filter, the waveform is damped. I also notice that without the LR filter, the damping is less obvious but the peak is a lot less . My questions is then: what is the best way to get the best more natural sound into the speaker and desensibilizing the amp circuit to load impedance. I use voltage overall negative feedback of about 40 db. I have noticed that most amps today (from what I have seen) do not include the LR circuit. Remember that I use a dummy load, does someone might consider a real speaker load to adjust amp parameters?
In the first picture you can see the "natural" response of the amplifier when driving a purely resistive load above the audio band because the load capacitor is isolated by the inductor (it suggests to use a lower impedance RC network, may be 4.7ohm and 220nF, or stronger frequency compensation in order to damp better that ringing).
In the second picture, the capacitor connected as a load is interacting with the amplifier at higher frequencies (there is no inductor to isolate it) and reducing phase margin but also gain, so the overall result is no overshoot. However, nasty things may happen in the Mhz range.
By the way, some modern amplifiers lack a RL network in series with the output to reduce costs and save space (and you may see people complaining about them oscillating when things like exotic speaker cables are employed). Anyway, adding that RL network is a good design practice because it prevents external stuff to interact with the amplifier above audio frequencies and leaves it only loaded with its internal RC network thus yielding consistent RF behaviour regardless of the load.
try sinewave into your 8R//cap. Use a range of voltage from 25% to 90% of max. and 105% of max.
use a range of caps 100nF to 2uF. these combinations will take some time.
Look for instability on the sinewave. @zero crossing, @halfway up/down the wave, @the top of wave form.
You might need to do this for a range of frequency, no guidance. Be careful you don't apply high voltage and high frequency to the output RC (Zobel).
Then repeat for with and without the inductor.
Hear from you in two or three days with your results.
Use your ears!
The only way to be sure about such choices as you have to make here is to actually listen to the results.
The late JLH (UK audio Guru) told me in private correspondence that provided there were not more than 3 'wiggles' on the square wave, he considered this to be "sonically benign", but this is one listening trial I have not carried out for myself (yet!) and so I cannot verify this. I strive for 'clean' square waves, using my chosen speakers and cables.
I have never had reason to doubt anything JLH has said, and he made no mention of the amplitude of any overshoot, so I presumed that 'time' is more significant here than 'quantity', where any sonic effects are concerned.
Assuming this is so, then the first 'scope pic (with the output network) should sound better, but in any such listening trials and/or measurements like these, it should certainly give the best results if the intended speakers *and* cables are used. They will always present different complex impedances etc. to the amp's output, and are likely to have some effect on the choices you end up with for the best results.
The "mechanism" you query is all due to the defined 'loading' of the amp's output at HF by the series cap and R. With the (most?) frequently used 100n & 10R here as the Zobel network, this will roll-off the output (by grounding it) with a -3dB point at about 160kHz, off the top of my head. This therefore overcomes any potential problems due to 'unusual' speaker/cable effects at HF, where the amp could be looking into a very high impedance, otherwise.
To quote from Ben Duncan:
" Delivering HF is a problem to amps with high NFB, especially when it is global. The amount of feedback (loop-gain) is reducing in most amps well before 20kHz, and often, in high NFB amps, from 10kHz. The output impedance begins to rise at typically +6dB per Octave at frequencies above this point. Thus the reducing NFB makes the amp's output look inductive, and also develops an output impedance that increasingly interacts with any load capacitance..... to cause a lagging phase shift. If the extra phase shift is very much, and occurs at a low enough frequency, RF oscillation will result."
JLH explained this, similarly, to me before I had read Ben D.'s book "High Performance Audio Power Amplifiers".
Ben D. also points out it is imperative that any 'output network' should be placed outside of the feedback network if it is to be of any benefit, and that amps without output inductors (except non, or very low feedback designs) are renowned for becoming unstable and even blowing up, when low inductance speaker cables are used.
One of JLH's mosfet designs (ETI 1984) initially had no output inductor//resistor arrangement and he specifically pointed out that with this design, these components were unnecessary, although all of his designs used the Zobel series cap and R to ground. When this same design was published subsequently (ETI 1989) with some minor modifications, the inductor had re-appeared, and being surpirised by this, I queried it's later inclusion.
JLH's response to this was that, in the intervening years, he had received a few grumbles from some constructors who had used 'unusual' speaker cables with the earlier amp, and they had experienced some oscillation problems, so this was the easiest way of avoiding this possibility.
I hope this helps.:)
A few night thoughts from me.
Firstly it would help, fab, if you can give the horizontal time scale (I did not notice you giving the square wave frequency). Meaning that those oscillations may be totally above the audio band and irrelevant. And yes, it will help to load with a speaker equivalent load; although the impedance usually rises with frequency and has less effect than a fixed load resistor.
The Zobel network is a way of "bringing in" a load to the amplifier (as Bobken explained) above the audio band, so that it does not see a high impedance because of loudspeaker impedance rise as stated above, but also does not see this load in the band so as to waste power in it. The probelm often overlooked is that it is fine once the C-reactance is small compared to the 10 odd ohm of the resistor, but somewhere on the way there this load will present a 45 degree phase angle [where Z(C) = R] which some circuits do not like. It is not a universal cure-all.
I presume the L-R network mentioned is an L with damping R in parallel, all in serie with the load. This should be effective only above the audio band and thus not influence the sound at all. In my own design (8 ohm load) I use 20 uH with 22 ohm in parallel; this L becoming 8 ohm only at 60 KHz. My amp does not like the Zobel; I connect a 680 ohm 5W resistor permanently over the loudspeaker terminals, wasting about 1,2% of energy all the way. With the L it can tolerate any sort of load, even short-circuit (obviously not with any degree of input signal because the feedback is then off). Without the L the output transistors blow in 1 mS flat because of h.f. instability with output shorted, and will not tolerate an output capacity of over 15 nF - this just to illustrate the effectiveness of that L.
As Ben Duncan stated, such an "undesirable" state can be piously cured by an early enough roll-off (e.g. by Cdom). But if this cuts well inside the audio band you have a high-order harmonic generator, which WILL give an audible effect. There are few circuits with negligible phase shift up to 20 KHz; the price of that is careful stability design and a lower margin than over-compensated circuits. But this is slightly off-thread.
Negative feedback + open loop gain makes the output resistance/impendance of an amp small. This resistance and the load capacitor interact to provide another pole. If the frequency is low, that pole is far away and does not mess with your amp. As frequency increases, the open loop gain drops. The outpuut resistance of the amp rises (beacause of NFB and reduced open loop gain) and moves that pole that did not cause troubles at LowFreq. It is not imperative that it will cause trouble with a given capacitance, but that is the mechanism and is what always happens. Note that for every amplifier there is an arbitrary value that makes the amplifier unstable beyound that margin. Sometimes a similar effect is used to stabilise the amp. The series RC connection (on the output) is used to affect the amplifier's gain and phase and make it stable. If done wrong wil make your amp burn, sorry. :hot:
|All times are GMT. The time now is 10:45 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio