EQ of speakers is the only pre-distortion that worth discussing, the rest is from the field of fantasy.
I saw a simple MOSFET amp with single input transistor take a distortion analyzer to it's sensible limits . It had a simple moderate current SE FET amp stage drive the class B dead band in feed-forward via a capacitor . This is something I have tried many times without any positive result . The guy if I remember dropped it as he felt he could hear no difference . I was doubly impressed as a piece of Douglas Self dogma was proved wrong . There was no significant distortion by using a single transistor input . I must say the supposed single input extra stability people talk of has no reality in the world I live in . When it works it is due to high input capacitance transistors .
I was told if great care is taken to make the Quad 405 work to the best possible null then magic happens . In the opinion of the man who did it Quad were making a cheap amplifier and ruined the concept . Sansui did it better and my Kenwood KR 750 also ( not sure , no circuit dia ) .
Pre- distrotion is something to know of . I must build a new oscillator as the - 80 dB I get from mine is not good enough .
Last edited:
My memory of the circuit I mentioned seems slightly wrong ( feed-forward ) . However the reading is good . What I like here is all of the no-no's of modern thinking are rejected ( MOSFET's , single input , capacitor coupled ) .
MJR-8 Mosfet Audio Power Amplifier
MJR-8 Mosfet Audio Power Amplifier
I have posted working sims showing Quad's and a "parallel Black's Feedforward version from the Vanderkooy and Lipshitz paper
this web site has publication references http://digilander.libero.it/paeng/feedforward_concepts.htm
this web site has publication references http://digilander.libero.it/paeng/feedforward_concepts.htm
The essence of current dumping is feedback. When 2 transfer functions start overlapping feedback linearizes the result. That "bridge" is the result of usage of slow opamp and slow output transistors. The author did not expect them to work nice, so made frequency response worse, but predictable, to balance it better. In 1970'th I independently made similar amp in class A+C, without any "Bridge": I used faster opamp and faster transistors. Nested feedbacks did the magic charm.
The positive thing I read about feed-forward is it doesn't require massive bandwidths to be able to get a reliable 50 kHz of usable bandwidth ( even in germanium ) . The example I show is a very simple amplifier . Please go to the home page and read on . The complexity is low compared with class D . I feel that the Quad was bad news for a good idea . If the class A was up to 10 watt and used MOSFET's so as to have gradual turn on all would be better . I completely go with no absolute need for the bridge . All my versions were that way .
JCX I read your description before and like it very much . I seem to remember Mr Black was on a Mississippi river boat when negative feedback occurred to him ( 1929 ? ) . Mostly to stabilize gain . Apparently after 5 repeater telephone amplifiers the overall gain was highly unpredictable ( voltage of PSU etc ) . Negative feedback meant telephone lines needed virtually no maintenance after the invention of NF , just the occasional valve change . I as you can understand never met the man so can not vouch for this account .
JCX I read your description before and like it very much . I seem to remember Mr Black was on a Mississippi river boat when negative feedback occurred to him ( 1929 ? ) . Mostly to stabilize gain . Apparently after 5 repeater telephone amplifiers the overall gain was highly unpredictable ( voltage of PSU etc ) . Negative feedback meant telephone lines needed virtually no maintenance after the invention of NF , just the occasional valve change . I as you can understand never met the man so can not vouch for this account .
I like this description . To say feedback is pre distortion is the simplest explanation . If very fast it must be very similar ? I guess complimentary input topologist look to eliminate the work of the feedback circuit ? My guess is very fast feedback might be preferable ? The idea that NPN and PNP might be the same and do cancellations looks a bit dodgy . Good old negative feedabck will sort it out in the end . Doubtless the harmonics die down quicker if good cancellation before feedback ?
http://www.tubecad.com/articles_2001/Inv_Dist_Cancellation/CompInvDisAmp.pdf
http://www.tubecad.com/articles_2001/Inv_Dist_Cancellation/CompInvDisAmp.pdf
Yes. If the amp, each stage, had been linearized before applying global feedback, and if it's bandwidth is good for good feedback on the top end of the frequency band, it is the right direction. However, local feedbacks linearizing each stage decrease open loop gain decreasing "amount of global feedback", but it is the matter of optimization.
It occurs to me that pre-disortion should be very slightly nicer than feedback as there is no time delay . Various valves seem very similar so the fit of the antiphase curves can be reasonably good . However output impedance is higher without the loop feedback .
Distortion can almost meet DIN 45500 ( 8 W 1% ) without loop feedback ( SE amp KT 88 ) doing this . Certainly so at 2 watts . Without pre-distortion 10 % is more typical .
Distortion can almost meet DIN 45500 ( 8 W 1% ) without loop feedback ( SE amp KT 88 ) doing this . Certainly so at 2 watts . Without pre-distortion 10 % is more typical .
I don't want this to digress into a discussion on EQ and DRC but could you tell me what benefit you would get doing this in the analog domain instead of the digital domain? It seems to me that the con's far outweight the pros when it comes to EQ.
It would be great to equalize speakers for high-end sound reinforcement, where I don't want latency. At home I am satisfied by digital Audissey.
Ah then you need to roll your own implementations. Digital can be effectively minimum phase, just that most of the popular chips (ADC, DACs) implement linear phase and have high latencies. Its not inherent in digital as a technology though as far as I'm aware.
What latencies do you need to achieve?
What latencies do you need to achieve?
One thing that is possible in digital is to EQ with different phase outcomes . I would imagine sometimes analogue is better and other times digital . A mix might be very good .
I would not say this but for the fact we already have information in digital format . Seems to me that is where to play with it if so .
I was trying to dream up a digital format based approximately on Nicam . The idea would be to have a digitally compressed dynamic range with absolutely defined parameters . The signal at any point can be reconstituted back into the original dynamic range if required or points between . It would use the best of all ideas ( part could be analogue under digital control perhaps ) . DVD as storage medium . This format would be ready to use for radio stations or the car without decompression . The compression arithmetic based on perception and taste ( motor car biased perhaps ) . The chips that process it might have user digital processing included . For example buy two DACs and configure them as active crossovers . The ultimate quality would be rather like buying a posh turntable in the past . It would be incredible . For ordinary people very acceptable in the cheapest format ( meaning me in my motorcar ) .
That's what is wrong today . No big leap from ordinary to hi fi today . Even top end . Same miserable sources of sound . Ones which mostly we bettered in 1957 ( Louis Armstrong , Belefonte etc Decca Peter Grimes 1959 )
It depends on what kind of PA applications. But even for stadium for example floor monitors have to be nicely equalized. My PA systems are famous for "Sensitive microphones". However you understand that that means extremely flat frequency response. Put floor monitors with bumpy frequency response there and everything is screwed.
I think pre-distiortion is only practical in valves from my own tests . It is almost alchemy as it corrects that which hasn't happened yet . Please tell me I am wrong . I know the simplistic geometric reason . What I find hard to accept is how good the fit can be . I am very new to valves so I hardly know anything . All I know is that building near to zero feedback with valves is possible ( loop feedback ) . I can not think how to do it with transistors as easily . 2 common type valves , 0.5 V in 8 V out 8R . THD , 0.03% at 0.6 W 0.1% 1.6 W 0.43% 5 W , 8 W clipping . No loop feedabck ( no Schade either ) . Schade is interesting . It is simple shunt where driver is no great hindrance . It is a slightly better triode ( 30% more gain for identical distortion ) . It measure much like a triode . However the overall gain is poor . I haven't tried all the RH series , RH88 seems less than ideal . If wanting to know about RH 88 measure the cathode of the ECC 81 with various waveforms , very interesting and perhaps V to I converter is correct ? I also found ECC 82 and 12BH7A usable , they were better than expected . A cascoded ECC 82 looked very like a single ECC81 in all respects ( 90 V g1 ECC82 No 2 ) . These are valves I inherited . The interest also . I am notoriously mean so always use what I don' t have to buy . Most of my valves came from the old Chipping Norton Studio .