I beg to differ here a bit. If you have a solid state amplifier with gain, you need feedback to make it usable. Otherwise the result is horrible due to intrinsic transistor transfer function. And you have feedback even if you do not see it at 1st sight, like in the emitter follower with almost 100% voltage feedback. For some strange reason, audiophiles consider feedback only as Global FB from output to input. This misconcept is strategically misused by some marketing oriented engineers. Local feedback has same consequences as global feedback, but it was never demonized. So it goes, lack of knowledge is easily misused.
I know, that's a part of what Peter Garde wrote back in 1977 (reprinted in 1978), see http://linearaudio.net/sites/linearaudio.net/files/volume1ltemvdg.pdf for exact literature references. In fact the input stage needs to handle differential voltages up to twice the peak input voltage of the amplifier if you want to keep it out of slewing when subjected to a big zero-rise-time square wave without any input filtering, no matter what the amplifier's bandwidth may be. You can relax it a lot with some input filtering, depending on the ratio between the filter bandwidth and the bandwidth the amplifier has without the filter - as was also explained by Garde.
It's a pity Peter Garde's articles appear to be less well-known than those of Matti Otala's research group. Garde's articles were spot on, while Otala's group only reiterated the mistake made by Daugherty and Greiner in 1966.
Designing for square waves is actually overdesign for an audio amplifier, as an amplifier that can handle a 10 kHz sine wave with low distortion can handle the rates of change of almost all normal audio signals, bat recordings excepted. In fact, American FM radio has a power bandwidth of only 2122 Hz and European FM radio 3183 Hz.
Still, even if you want to handle such signals, I think you could put a suitable inductor across the emitter degeneration resistors. The inductor will essentially be an open branch for short transients, like the edges of the square wave.
My main amplifier is designed to handle square waves without slewing, has low distortion at 10 kHz and has no input stage emitter degeneration. Its input filter limits the bandwidth to 135 kHz and that was enough to limit the peak error signal to levels a simple bipolar differential pair could handle. Mind you, I didn't use any 3 MHz fT transistors.
Bart did not understand that wide open loop bandwidth is not needed and can be even a bad idea. Bruno convincingly debunked that.
I'll have a column on that next week!
Edit: Christmas Market in Koblenz, Germany today. They say the city is one large xmas show. I'll put in a good word for you guys ;-)
Jan
Last edited:
Same here it was. Hopefully, back in 2009, I had Agilent GHz HF spectrum analyzer in my lab and it was the only type of instrument that was able to show it, to show the real origin of the issue.
Where do you get that from?
A couple of posts back I even said and showed that in the 50s and 60s there were plenty of tube amplifiers with feedback.
In fact, most better designs all had feedback in them.
Does leave the question where that all of a sudden stopped?
Although, I am with Jan on this.
He replied that the general knowledge on designers was generally much higher.
Addition to post #182, which was a reply to @Bonsai :
There was not much attention paid to slew rate limiting and slewing induced distortion/transient intermodulation distortion/input stage distortion in the amplifier course based on Ernst Nordholt's work that I got at university, though. Just a few sentences in Nordholt's book and links to Peter Garde's articles.
There was not much attention paid to slew rate limiting and slewing induced distortion/transient intermodulation distortion/input stage distortion in the amplifier course based on Ernst Nordholt's work that I got at university, though. Just a few sentences in Nordholt's book and links to Peter Garde's articles.
Enough loopgain at frequency of interest is needed. I am not sure that the term “wide open loop bandwidth” has the same meaning to all of us. It might be a semantic game again. BTW, Bruno is a great engineer AND great marketeer as well. So it goes when in business.
I have a 150 MHz analogue scope now. It is wideband enough to show the 200 MHz...300 MHz oscillations you typically get from BC546...BC560-like transistors when the wiring parasitics turn them into oscillators. Of course the amplitude and waveform are incorrect, but at least it shows something...
I usually use a passive 1:1 probe with the ground lead connected to the tip, and use that as a loop antenna to find the spot where the high-frequency magnetic field is the strongest.
My post does not say there were no other feedback amplifiers. There were feedback amplifiers in the 1940’s already - maybe earlier. The point I made was there wasn’t the general knowledge in the industry about how to apply it successfully the way it is now - only a few people had the experience and I pointed to the very high feedback amplifiers of the 1970’s that suffered from problems as an example.
Last edited:
I was not aware Locanthi felt that was specifically necessary. The road to the general understanding in amongst the audio fraternity took yrs after that - I’m sure we can forgive Locanthi for that misunderstanding. Leading lights on this site just a few yrs ago were claiming exactly the same thing and large swathes of the industry still think feedback goes ‘round and round’. 🙂
Jan,
Almost 200 posts in four days.
I genuinely hope that your thread is not becoming another Blowtorch thread of John's.
Patrick
Almost 200 posts in four days.
I genuinely hope that your thread is not becoming another Blowtorch thread of John's.
Patrick
You might be right. 🙂
It came across that there weren't any amplifiers with feedback.
It's pretty hard to find any decent history and evidence of this.
So it stays a bit on either anecdotal or personal interpretation.
Fact is that in WW2 there were plenty of systems with feedback in them, so the knowledge must have been there.
I don't know what oscilloscopes were capable of in the 50s and 60s.
It's always easy to talk back with the knowledge we have these days.
I think a big portion of people here, remember the days without internet.
Finding proper examples and literature can be done within minutes nowadays.
Back than people could have been totally unaware of stuff that was being done somewhere else.
We don't know the entire context around certain inventions.
Things like deadlines, investments etc, can play crucial roles.
I should have clearer - apologies. I am specifically talking about solid state amplifiers in the 1960’ and 1970’s (and even the ‘80’s in some cases). Tube amplifier loop gains were much lower, but with the arrival of solid state amplifiers following the Lin architecture (gm front end > VAS/TIS > OPS) the loop gains and bandwidths this went up and this posed challenges. Tube amps were transformer coupled, so the bandwidths we’re quite limited by contrast.
🙂
Triode amplifying stage with unbypassed cathode resistor HAS feedback. I doubt that any usable audio amplifier had only bypassed cathode resistors. Cathode follower has strong negative feedback. Emitter follower has strong negative feedback. I cannot see any usable amplifier with no kind of feedback. GNFB is just only one example of feedback, from many.