Something I thought I knew, but am left puzzling...
Take a simple class B output pair driven directly by an opamp. Result is na sty crossover distortion. Now, the nfb is taken from the circuit's output t erminal, not from the opamp's output, so surely an ideal opamp should corre ct for the 0.6v Vbe drops. Question is why doesn't it?
The other question is what strategies exist to tackle it.
- bias the output trs to class AB
- helper resistor from opamp output to circuit output, so the opamp powers the output directly from -0.6 to +0.6v
- low Vbe transistors, ie germanium
- a diode drop in the feedback path seems to reduce it to a degree
nasty crossover distortion. Now, the nfb is taken from the circuit's output terminal, not from the opamp's output, so surely an ideal opamp should cor rect for the 0.6v Vbe drops. Question is why doesn't it?
I cannot imagine anyone getting involved in a class B output stage unless it was DIY?
But anyway, whatever you do with the drive input to the output pair you can never overcome the o.6 V threshold.
I cannot recollect how it was done but there were many ways of baising an output pair. The simplest and almost universally adopted system was to use two transformers to couple the output pair and input/ speaker.
It does, within the limitation of the operating frequency of the opamp, and most particularly, the power output transistors - the drive capability of the opamp and capacitance and switching time of the audio power transistors are usually the limiting factor.
That's why bias current was normally added - lower crossover distortion in exchange for some power wasted in the output stage by avoiding ever switching the transistors completely off, so you keep them in their faster operating area.
Hmm. You need to be up at 100Khz switching to get reasonable HF performance, and the RF emitted is phenomenal. Huge problem, in terms of screening and filtering.
Get it wrong and there goes the neighbourhood (broadband) ...:-)
Its the non linear part of the curve though, is it not, hence your ref to class ab. However some crossover isn't, its what happens when the high gain low signal bits are running from the same supply as the high power stuff is, even with decoupling. The worst case of this I encountered in the past was Sinclairs intigrated crcuit amps IC10 and IC12, which were then made by Texas under a number I forgot. I could never get those to be well behaved, too much gain in too small a space. Brian
Didn't Sinclair produce a range of quite effective broadband jammers from LF to UHF masquerading as audio amplifiers in the 70's? I seem to recall helping someone solder one into a biscuit tin in an attempt to silence the RF emissions before his mother found out what was causing her to miss Coronation Street.
s nasty crossover distortion. Now, the nfb is taken from the circuit's outp ut terminal, not from the opamp's output, so surely an ideal opamp should c orrect for the 0.6v Vbe drops. Question is why doesn't it?
Yes, that's what I thought. But seeing the distortion in Spice it looks ver y much like 0.6v has been cut out. It does explaint he asymmetric shape fur ther up, ie above 0.6v, where the output plays catch-up.
ers the output directly from -0.6 to +0.6v
If its just down to speed, presumably a fast opamp and fast trs followed by output filtering should more or less clean it up.
I intended to rely ont he opamp driving the load direct upto 0.6v out via a resistor, but R has to be excessively low to just reduce the distortion so me, so that isnt really working. Simplicity and efficiency are all with the matchbox size amp I'm designing.
That reads to me like you're attempting to emulate the Op-amp based Power Amplifier shown here:
which strikes me as being overly complex and a "A Little Bit Shitty"(tm) imo.
or were you thinking of something like this?
The 12W transitor amplifier cct diagram except for C1 being replaced by a shorting link so that at very low output levels, the op-amp itself drives the load, significantly reducing, if not completely eliminating, low level signal cross-over distortion. Gawd knows why there's a 10nF cap in place of a wire link in that cct.
This second version is a clever 'servo styled' arrangement of complementary darlington power transistor 'assisted' op-amp power.
The trick to using this technique successfully is to determine the quiescent no load supply rail currents of the chosen op-amp and to select R2 and R3 such that the volt drop is just under the total Vbe voltage for the transistors used (probably 0.5v per be junction, in this case 1v total per darlington pair).
Furthermore, this 'trick' also relies upon effective heatsinking and careful control of the rail voltages for the sake of the op-amp's voltage rating limits. It will also help to use higher power rated transistors than you'd normally expect to get away with in more conventional designs along with a fast blow fuse (or a transistor analogue of a polyfuse, even faster and resettable). You could also embellish the output cct with the addition of emitter feedback resistors to improve thermal stability if you can afford the reduction in Pk to Pk voltage swing.
I first used this power transistor servo assisted op-amp design that was published over thirty years ago as a half watt per channel headphone driver (8 ohm voice coils) as an add-on to the built in RIAA pre-amp in my TD125MK1 record deck.
I later used it to provide a 2W per channel heaphone driver in a homebrewed portable mixing desk designed for 'Live' recording where, with earplugs fitted, the closed back headphones would allow me to monitor the mixing desk output without too much pollution from the local sound field of the venue.
The last time I used this cct was in a miniature 50W per channel stereo amp. In this case, I used bridged output to quadruple the maximum output over a single ended design limited by the modest Pk to Pk swing afforded by the op-amps (+/-15v max working supply rails with
+/-18v absolute maximum peak overvolt rating) using a regulated 35v PSU.
This last exercise in 'servo assisted op-amp' power amplifier design did prove a little bit problematical regarding the frequency at which it would blow output transistors but fine tuning resistor values and a much better heatsink seems to have tamed this problem somewhat (though I daresay I can still blow it up if I really tried).
Opamps are normlly quite low output power. Indeed, if the opamp has a class B or AB output, you may be seeing some distortion in the opamp output. A class A opamp output would be better, but have even lower output drive capability.
Didn't power amps move to FET outputs? I kind of moved from analogue to digital before that happened, but I would have guessed power MOSFETs in a class B output stage might be able to produce a significantly better result.
Certainly for switching, I use power MOSFETs and they are much easier to drive and much faster than power bipolar transistors. (Class D amps will be using power MOSFETs for switching at well above audio frequencies, and these have been commonplace for a decade or more.)
nasty crossover distortion. Now, the nfb is taken from the circuit's outpu t terminal, not from the opamp's output, so surely an ideal opamp should co rrect for the 0.6v Vbe drops. Question is why doesn't it?
rs the output directly from -0.6 to +0.6v
No, both far too complex. Its a minimal amp consisting of opamp driving a t r pair, plus 2 nfb resistors. That's it.
The cap means the opamp only drives the output for short spikes, not for th e rest of the time.
Yup, thats the problem with that approach. The design needs to be built by people with no testgear or knowledge of how to use it.
If its not essential its not going in. There won't be any short protection, unless it can be arranged without increasing part count, which seems unlik ely.
Well, that's the other problem. I like the upsides but that's not livable w ith.
I'm tempted to think a faster opamp might get rid of a lot of the xover dis tortion.
In that case, I think they've left out a low value resistor (circa 10 ohms?) that would be in parallel with the cap. Unless the op-amp output is actually connected to the speaker load, how are you going to see the necessary drive current from the supply rail bias resistor connections?
My problem with this design, I now recall, was the (not so) "clever use" of a dual circuit breaker to act both as a mains on / off switch and DC current overload protection on the single ended 35 volt supply.
This is one of the rather neat benefits of a bridge output amp design when it comes to overload protection - it can be placed where it can't hurt the normal performance of the amp and simplified to just a single 'fuse' (electronic or otherwise) per channel.
Other benefits of bridged output are a halving of the required slew rate performance for a given output voltage swing, frequency response from DC up, elimination of the need for a muting relay to suppress switch on thumps and better immunity to strong sources of RFI.
The thermal stability with bi-polar output devices does leave a lot to be desired, hence the need to 'over-engineer' the power ratings and heatsinking of the output devices.
I could have avoided burning out transistors if I hadn't been so lazy as to use an electromechanical protection device on the common 35v rail instead of choosing an electronic 'polyfuse' per channel power rail arrangement. It may not have solved the odd case of thermal runaway mediated shutdown but at least it would have demoted such events from "Disasterous" to merely being inconvenient annoyances.
A faster op-amp _and_ power output transistors would also let you get away with more biasing margin, allowing you to forgo any complicated thermal tracking.
As it stands, my "50W RMS per 8 ohm channel stereo amplifier marvel" does leave a lot to be desired. :-(
HomeOwnersHub website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.