How to identify a quasi-complementary amplifier?

This is a good question.

All output devices are the same type and number.
Either all npn, or all pnp.

Really? Then I must even call the CSPP (Circlotron) and the topologies from Ollson (Better Audio from non complements? from Bengt Olsson, Electronics World + Wireless World, December 1994 page 988) quasi complementary?

I don't think so. I call only such output stages as "Quasi complementary", if it looks like PP true complementary buffer except the lower last power transistor. This means for me, that the upper halve of PP buffer looks like a darlington and the lower looks like a sziklai (CFP) darlington

An complete overview of topologies, where are only NPN-BjT's and N-CH MOSFET's in the last part of power buffer you will find here:
http://www.diyaudio.com/forums/soli...better-audio-non-complements-audio-power.html

also of interest in this case are some articles from the diyaudio member "X-PRO", e. g.
http://www.diyaudio.com/forums/solid-state/94030-quasi-not-quasi-question.html
 
Last edited:
How to identify a quasi-complementary amplifier? is it it contains 4 transistors 2 in 1 pair. And each pair is in complement? Does anyone can share some info about quasi-complementary amplifier?

thank you.

Quasi means acts as if.
So quasi complimentary means the NPN's act as if they were complimentary.

In reality that isnt quite true as one side has more gain than the other due to one side requiring an inverting driver which has gain, where as the other side is a simple driver buffer.

I found the inverting side required a capacitor from collector to base on teh driver transistor to stop oscilation.
 
In reality that isnt quite true as one side has more gain than the other due to one side requiring an inverting driver which has gain, where as the other side is a simple driver buffer.

Yes, and therefore some manufacturers like Quad (model 303, not to interch. to the current dumping version 405) creates three stage super ß devices to avoid (or reduce at least) this disadvantage (see attachement).
But the disadvantage there is the risc of oscillation of the buffer stages itself (not so by use of two stage, darlington for upper and sziklai darlington for lower buffer part.
Only truly experienced (and unfortunately very few) developers (like Quad) achieve a design without any oscillation under any load condition. By many others there are oscillation and hum/buzzing effects through the sensitivity of parasitic influences. If I get such amp devices I must remove one stage in any cases, i. e. the reduce to a two stage PP buffer

By the way, after right service and maintenance the Quad 303 is a very fine sounded amplifier, (much more better than the current dumping version Quad 405 and most true complementary versions). Only power stage versions with much more idle current trough the output can outperform the Quad 303

http://www.net-audio.co.uk/quad303upgrade.html
Quad 33 and Quad 303
http://www.diyaudio.com/forums/soli...upgrade-output-stage-fully-complementary.html
http://www.diyaudio.com/forums/solid-state/113943-quad-303-triple-cascade.html
 

Attachments

  • quad303 power PP super ß buffer.jpg
    quad303 power PP super ß buffer.jpg
    28.9 KB · Views: 2,047
  • quad303m.jpg
    quad303m.jpg
    98.3 KB · Views: 2,016
  • Quad 303 inside.jpg
    Quad 303 inside.jpg
    212.9 KB · Views: 2,081
Last edited:
  • Like
Reactions: 1 user
Does this claim about quasi-complementary correct?

Designers also sometimes use a "quasi-complementary" configuration, which uses a Darlington push pair (i.e., two NPN transistors) and a Sziklai pull pair (i.e., one PNP and one NPN transistor). This configuration, which uses three NPN transistors and one PNP transistor, is advantageous because:

  • Silicon PNP transistors have historically been more expensive than their NPN counterparts.[citation needed]
  • The performance of the lower pull pair, which uses a single NPN transistor, more closely matches the performance of the upper push pair, which consists of two NPN transistors.[citation needed]
 
And how to build a good, simple one

Hi,

quasi-complementary output stages are fine, if done right.

Quasi means acts as if.
So quasi complimentary means the NPN's act as if they were complimentary.
Quasi is Italian for nearly. For me, both swings matched well at all levels and frequencies like this:

* PNP driver transistor uses emitter and collector resistors of equal and same value as NPN driver shunt resistor, hence inverting stage has same small-signal gain as non-inverting stage, for you know unity gain. In a 25 W design this was 220 Ohms.

* A diode is layed in parallel and conducting polarity to PNP emitter resistor. I guess this was 1N4001. This matches great-signal behaviour.

* A capacitor is also layed in parallel at this place. I vaguely remember, this was a few Nanofarads. Value was found by looking at square wave responses.

This turned out non-resonating yet fast enuf, so even using old transistor types, performance turned out fine after setting of a few dozend Milliampères bias, output HF decoupling and non-ringing global feedback.

Unfortunately for a small-multi-channel-system builder like me, discrete Darlington output stages need at least three mounting holes per channel, two for the power transistors and one bracketing both drivers. Is there an integrated chip for this with seven pins, needing only the compensation capacitor and two power shunt resistors for a complete output stage, or even with built-in thermal feedback? Or a truely complementary Darlington pair chip? Or a chip amplifier of this design with most basic voltage amplification stage (single transistor input stage, single transistor second amplification stage with external, bootstrapped collector resistors) for a single supply up to 60V? With small-signal radio, ICs of this "integrate only what profits from integration" philosophy were common, but rare with audio.

Uli
 
I got caught out with a quasi design.
Because the driver levels are different for upper and lower output transistors this results in a negative dc voltage being fed back into the LTP.
I had forgotten to turn the electrolytic capacitor around in the feedback network.
The amp was messing around when I tried to adjust the dc offset.
The negative volts acted like a short through the electrolytic and upset things.
 
Member
Joined 2010
Paid Member
That seems a reasonable summary to me but some readers of English would understand the meaning of Quasi-complementary better if we first defined what a complementary or mirrored output device pair is. Simply put, Q-C push-pull output stages don't have complementary P/N output devices - they fake or approximate the complementary device operation and can do so very successfully, assuming we aren't too fussed about higher distortion, bias anomalies and thermal stability worries.
I think the justification for QC techniques became history as soon as Japanese manufacturers ditched it back at the end of the 1960s. They went on to dominate the audio market globally so I think that manufacturers have actually had only lame excuses for using QC designs for a very long time now. Sure, some have continued to scrape the bottom of the barrel on price to remain competitive but how many remain successful or increase their market this way?

I suspect the only good performance reason for keeping QC alive now, as highlighted in a few recent threads that focus on it, is to utilise the higher harmonic distortion intentionally for sound enhancement. That's at least interesting and has been fun for DIY and audiophiles for as long as solid-state has been popular. QC design in chip amps is testimony to what can be achieved but of course, there's no chance of tweaking chip amps, is there ;)
 
I think the justification for QC techniques became history as soon as Japanese manufacturers ditched it back at the end of the 1960s. They went on to dominate the audio market globally so I think that manufacturers have actually had only lame excuses for using QC designs for a very long time now. Sure, some have continued to scrape the bottom of the barrel on price to remain competitive but how many remain successful or increase their market this way?

I suspect the only good performance reason for keeping QC alive now, as highlighted in a few recent threads that focus on it, is to utilise the higher harmonic distortion intentionally for sound enhancement. That's at least interesting and has been fun for DIY and audiophiles for as long as solid-state has been popular.
Actually, full complementary designs strike me as a conspiracy to burn the speaker up too as soon as there is any problem with the field wiring or amp. Designs that successfully employ a speaker disconnect relay seem to sell in the kilobuck range. Below that price, speaker relays seem to be planned obsolescence devices - doomed to cause failure in the 5-8 year time frame. The second bipolar solution, the IC with integrated protection circuits, seems to be limited to 7 amp peak speaker currents. The third solution, the RGKeen FET rail disconect circuit, seems to be a cult instead of a mainstream design element. My rail disconnect circuit is exceeding 30 parts, more than an entire quasi comp driver board. The triac crowbar the rail disconnect is intended to replace seems mainly competent at melting protection circuits off the PC board, and was a complete failure at blowing the breaker.
I'm still building quasi-comp because the electrolytic speaker capacitor is a simple, elegant, and 30 year life solution to speaker preservation. The higher distortion strikes me as totally buried in the speaker distortions, and negligable. Also my supply of non-import tranformers univerally lacks center taps. What really puzzles me is, is a single supply amp without center tap, inevitably quasi comp? It seems to be so far. Putting electrolytics in series with the output of a +-supply amp makes it sound stupid, quite audibly.
 
Ian, i suspect that it is the opposite, well-done QC had lower distortion than FC. Because opposing something is easier with small than with large signals. I dream of a simple amplifier with off-the-shelf, few transistors (i also suspect that PNP ones are better than NPN, at least they seem to have shorter switching times) and high reliability. As soon as output stage is right, the rest is a piece of cake, is it not?
 
Founder of XSA-Labs
Joined 2012
Paid Member
Here is an interesting point though - same outputs in quasi complementary will cancel even order harmonic distortions more so than odd orders. So it may not sound as sweet or euphonic. As we saw in the quasi thread where asymmetry is introduced by using a FET and BJT combo. Alternatively we can use same sex FETs or BJTs from different manufacturers or series it introduce asymmetry back into the outputs to enhance the harmonic profile of the distortion back to euphonic type.

So maybe IRFP240 and IRFP250's or Vishay on one side and AlFET on the other side? Or Motorola BJT on one side and Toshiba on other side.

You get the idea.
 
Member
Joined 2010
Paid Member
well-done QC had lower distortion than FC
Not really. Self's experimental evidence, at least with discrete designs (and even my own measurements), is that lowest THD is achieved with the best complementary components because the crossover transition approaches a symmetric, smooth "hump" with least disturbance in the transfer of conduction at switchover between devices. Quasi complementary output stages show a marked discontinuity at crossover, giving rise to a bigger spray of harmonics. Mosfets have an altogether larger hump again than BJTs and take special feedback treatment like error correction and TMC to bring THD down in the crossover region for acceptable low level performance. Of course, we would need some expensive instruments and model designs that work with varying amounts of feedback to see these effects for ourselves but simulation can show us quite a bit provided the models are sufficiently accurate.

Definitely though, NPN BJTs are superior where it matters, to their PNP complements of all sizes. Lets stay with recognized facts because suspicions about good science or wildly different opinions to accepted evidence needs to be qualified. If you haven't studied the mechanisms of crossover distortion or been able to read any edition of Doug Self's Power Amplifier Design book yet, I suggest you make an opportunity to. There is a short version of the topic of output stage distortion on his website if you want to get an impression of the detail, beginning at Figure 14 here: Distortion In Power Amplifiers

Indianajo, I sympathise with your bad experiences with protection circuits. I have had my share too and I often don't use anything in my personal gear now - just like my first kit builds and attempts with home etched PCBs back in the 1970s. The good news, if it happens to work for you, is that whether the output stage is quasi-comp. or not has no bearing on the need for a capacitor output. It's predicated on using a single rail supply and so was common practice with early solid state audio design. I guess the association of the two features with old designs gives us the impression they need to be implemented together. The generic Naim amplifier design is an example of a Q-C design with a DC output but there are plenty more.