Electronics > Beginners
TL071 distortion - bad amplifier design?
spec:
UPDATE #1 of 2018_12_19 (remove annotations from schematic)
AngraMelo
As promised, attached is a schematic for a slightly modified Self Simple Blameless Amplifier (SSBA).
I have added some notes to show various circuit functions, which. no doubt will be be discussed ;D
This architecture, and versions of it, are ubiquitous.
I will say no more for the time being (social duties most of the day), except that, in my opinion, this design provides the best bang for the buck in home audio power amplifiers with an excellent balance of cost, complexity, and above, all audio quality: it will blow away the original opamp version.
spec:
Also, as promised, here is some information on polypropylene (PP) capacitors that are suitable for use in amplifiers, pulse generators, and general electronics, as opposed to the PP capacitors widely used for power applications: snubbing, power factor correction, motor start, etc. Other manufacturers, make suitable PP capacitors but I like the specification for the Panasonic ECWFA (250V) series, although they are relatively new to the market and hence more expensive. The ECEFD (450V) and ECWFE(450V) series would be fine as well though.
Panasonic
ECWFA (100nF to 6u8F) 250V, 450V, 630V
ECWFD (47nF to 4u7F) 450V, 630V
ECWFE (100nF to 4u7F) 450V, 630V
To give some idea of cost (£UK, from DigiKey UK):
ECWFA 100nF (£0.51), 1uF (£0.78), 4u7(£2.58) 6u8 (£3.66)
ECWFD 100nF (£0.42), 1uF (£0.64), 4u7 (£0.99)
ECWFE 100nF (£0.38), 1uF (£0.60), 4u7 (£0.99)
https://industrial.panasonic.com/cdbs/www-data/pdf/RDI0000/ABD0000C183.pdf
https://industrial.panasonic.com/cdbs/www-data/pdf/RDI0000/ABD0000C255.pdf
https://industrial.panasonic.com/cdbs/www-data/pdf/RDI0000/ABD0000C202.pdf
Zero999:
spec,
The circuit you've posted is an operational amplifier! It looks perfectly decent to me. I agree will be better than the circuit posted at the start of the thread (at least it'll work!) but don't see why it should give superior performance, to an IC op-amp, with a suitable power amplifier on the output stage?
I noticed the lack of a significant bias on C2 and immediately thought that this could be an issue with distortion. There are two schools of thought on this: one, always ensure electrolytic capacitors have a significant DC bias, which is what I've been brought up on and two, just make sure the reverse voltage never exceeds a volt or so and all will be well. The reason why I've subscribed to #1 is the knowledge that an electrolytic capacitor has a parasitic diode element, in reverse parallel and even if the forward voltage of the diode isn't exceeded, causing significant current flow, it will be non-linear, hence distortion. Unfortunately reverse bias isn't accurately modelled in SPICE or shown on datasheets, presumably because capacitor manufacturers don't recommend it. In light of this decided to do some research to see whether anyone has actually done some tests to quantify this and after some digging through the depths of the Internet I found the paper linked below:
https://linearaudio.nl/sites/linearaudio.net/files/Bateman EW 12 2002 mar2003 1uF electrolytic or film.pdf
The results of the experiment are interesting. Although the distortion, due to the unbiased electrolytic capacitor is much worse, than a near-perfect film capacitor, it's nowhere need as bad as I thought. I was also wrong that the bias voltage across the capacitor should be as near the maximum rating, as possible to produce optimum distortion.
The main point which the experiment highlighted is the levels of distortion, even in the unbiased electrolytic are tiny <0.002% THD. In the circuit you posted, the distortion due to capacitors, whether aluminium electrolytic, polyester or polypropylene will be negligible. Even though C2 may introduce more distortion than any of the other capacitors, it will be tiny, compared to the rest of the circuit. The distortion due to the C2 will be much lower than the figures given in the experiment because the AC voltage across it will be tiny, it's oversized, FC = 1.2Hz and there will be the small positive bias voltage due to the current from Q3's base and R8, which will help matters, even though it's suboptimal.
Conclusion: use whatever capacitors you like in that circuit, it will make no difference to the distortion.
Kleinstein:
The last circuit shown is a well known audio amplifier circuit.
The circuit with an OP at the input could work as well. However it is not the really easy way for high power: One would need a power stage with additional voltage gain. It needs some attention to make the two gain parts to work together. The simple 1 pole compensation is usually no more suitable as this would need only one slow stage. Compared to many of the normal OPs the simple transistor longtail pair is also relatively fast and allows thus for good overall speed. The use of inner loops with local feedback is kind of controversial - not everybody likes it. Well done, the concept with OP for the input and discrete power stage can work. Though the general concept may look simple, I consider the classical amplifier with discrete transistors easier.
However the TL07x are relatively high noise and might thus produce some noticeable hiss.
I also see a slight problem with an electrolytic capacitor with no DC bias and possible slight ( 10s of mV) reverse bias from an offset. In addition to possible distortion, there might be an issue with durability. The oxide layer will slowly get thinner and the distortion might only show up after longer use.
Mark Hennessy:
I referred to Cyril Bateman's work earlier - it really is excellent.
I'll mentioned Douglas Self again. Search out "Small Signal Audio Design" and jump to chapter 2.
The important point is that most types of capacitor are pretty linear providing they are not being used to set a time constant (acting as a filter).
With that in mind, it is usual to over-size electrolytic capacitors in audio paths. Just picking a value based on the usual 1/(2piCR) way is not a good idea.
Best practice is to have just one capacitor that defines the -3dB of the whole amplifier, and make sure this is a film type rather than an electrolytic. Having done that, all other capacitors should be at least an order of magnitude bigger than what might otherwise appear to be needed. The aim being to minimise signal voltages across them so that they don't produce distortion. Of course, at higher frequencies the capacitors are hopefully akin to a short circuit, so no voltage is being dropped across them during the signal excursions.
Someone who is not aware of that distortion mechanism could easily assume that the capacitors are unreasonably large for no good reason.
It's easy to test for capacitor distortion if you can measure THD. You don't need a high-end audio analyser to see the distortion from an electrolytic capacitor. Self describes an experiment (page 58) involving a 47uF cap and 1k resistor forming a high-pass filter with a -3dB point of 3.4Hz, which would appear to be pretty reasonable for audio coupling; at 20Hz, the loss is a mere 0.12dB, so you'd think that would be OK given that there is minimal voltage being lost across the capacitor, but there is enough there to cause the THD to be 0.01%. Whereas a 470uF cap is indistinguishable from the AP test system at 20Hz.
As always, one could argue about the audibility of 0.01%, especially down at 20Hz where no sub-woofer will be anything like as clean as that, but once you're aware of the problem and the way to solve it, then few good engineers would be comfortable to leave it alone. A larger capacitor is not all that expensive in the scheme of things.
Self also shows that polyester caps aren't as good as polystyrene and polypropylene, but they are still much, much better than electrolytics, and certainly better than a TL071. This presents a dilemma, but on balance I'd be happy with polyester most of the time. For a top-flight design - which the circuits in this thread are not - then it probably makes sense to fit a polypropylene. No-one will hear the difference, but it's not expensive and doesn't cause other problems, so why not? Marketing will thank you for it ;)
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version