I got that recently when going from dead bug to tag layout . I lost a lot and will just tidy the dead bug design as I don't plan to make a PCB . I had a very good PCB which would have adapted well . Alas I didn't have enough made when I did . It was pre Gerber in Windows 3.1 for making masks .
I am puzzled by the "secret" of the QUAD 303 sound. I got hands on one, one channel cut off the power obviously defunct, the other apparently original. That beast sounds better than a class A but ...the bias of the power BJTs is just 12 mA! And in fact Quad recommends 10mA to 15mA so it is correct. But how? I could not detect any crossover distortion at frequencies up to 1 kHz. There, odd harmonics begin to rise. Not a miracle
as the power BJTs have fT way below 1 mhz. I don't have any idea why there is no crossover distortion at that low bias.
as the power BJTs have fT way below 1 mhz. I don't have any idea why there is no crossover distortion at that low bias.
100 % right . Bryston know it also . Even 5 mA bias starts to work . I remember this being well covered in D Self ( ? Ben Duncan , ESP also ) and is a result of output triples being used . Distortion spectrum is in exponential decay and starts form a low point . The single input transistor doing that is my best guess ( 99% sure ) ? . The Naim 250 I always suspected was a rejected Quad design . The 405 pushed it away ? The intention of the strange Naim LTP was to have a similar balance ? I spoke of valves before . The harmonic distortion decay is a big factor as I see it . Jean Hiraga was the first I remember saying so . The D Self book helps one choose that as it is said how to avoid second harmonic distortion , just reverse the logic . Use 1K + 12 K as I think Naim did and you get almost a 303 ( LTP balnce ) . Bob Stewart is said to have given the idea of loss of information as a distortion mechanism in amplifiers ? Tone controls were seen as extra circuitry and would loose information . As Bob is from Cambridge information is a serious word . I would use relativity in a similar way . Exponential decay keeps harmonics in a musical realtionship which is related in a way that makes things seem relatively correct to the ear . The reason Quad 303 can sound better than class A is the power supply is having a better time of it . 303 has a slight MP3 quality against the better class A which isn't Krell to my ears . One needs Magneplanars to find out . Even the 303 current limiting ( 2 diodes ) works !!! It sounds like a valve amp clipping . Doubtless the series regulator 3055 also helps that ? Run your tweeter on it's own output capacitor if using 303 . It might be < 10 uF .
Exponential decay and loss of information make much sense to me. Because hardly anyone has recently analyzed the harmonics spectrum of acoustical musical instruments, the attack and decay of harmonics. To my knowledge the last such analysis dates back to 1955 or so. The harmonics spectrum of a piano is not at all static, it varies to a huge extent , while it is static for an amp. There may be so far uncovered effects of auditory psychology of sound reception that are responsible for good or bad sounding amps which cannot be explained with necessarily static measurements.
The 303 has a quasi complementary output stage.
Low output bias is normal for a quasi to minimise crossover distortion.
Self and others confirm this.
Similarly, CFP output stage requires low bias for low crossover.
It's the EF output stage that needs ~25mVre for minimum crossover distortion.
Again Self and others confirm all this.
I am rather surprised to hear that, as everything was carefully checked by several people. Please let me know which ones you think are wrong.
Nice to know it is confirmed...but only the 25mVre is obvious at least to me
Hiraga was when I read it a conjecture . He studied MC pick ups and concluded that the harmonics strongly related to preference . He further concluded that the higher distortion of many MC's needed relating to perception . From this I bought a Coral 777 EX which was cheaper than most alternatives . Hiraga said as far as I can remember pick ups with a curve that looked as if it might be linear or more likely exponential ( the conjecture ) would be perceived as low distortion as long as a threshold of good performance is not exceeded . 1 % THD was said by others to be that threshold . Soon the conjecture was said to relate to other devices . If I am correct the early hi fi studies after the Willlaimsson amp concluded 1% THD as hi fi to most listeners with similar observations about harmonics . 0.1% THD was selected as 20 dB better than perception . As the Quad 303 fits 2 descriptions of hi fi it is therefore 40 dB better than perception as it has " correct " harmonics . That means the Quad could be degraded by 20 dB if it retained the harmonic structure and qualify easily ( damping factor aside ) . An amp with lets say only -80 dB 5 th harmonic by similar conjecture is said to be noticeably more distorted and unpleasant . This seems fair to believe . Harder to know as to create exclusive 5 th harmonic might be difficult .
A single input transistor has a subtle advantage . It has lower noise than LTP . JLH shows how to control DC offset if wanting DC coupling . I found it remarkable stable when trying it .
From what I know triples do have slight advantages when setting bias points . It is so long ago I studied this I will have to build a jig .
My 303 is a sad story . It was repaired by my brother for a Mr Haynes ( Professor ? ) in Oxford . He died shortly afterwards . His wife couldn't bare to keep it . I kept it as a spare amplifier . I hit hard times in my life and it became amplifier No 1 . As I learned more I realized it was technically superior to most . It has plenty of areas of doubt . The 33 pre amp needs carefully understanding to get it to please . All in all 33/303 is true hi fi . My brother died aged 48 ( Flu ) so doubly it is a memory to treasure now . The repair included sensible capacitor changes . The bias pot was replaced by fixed resistors as was DC adjustment ( mid point , 33.5 V ) .
Someone told people like me off for speculating and that rules should be applied . Without speculating nothing becomes science . I build things , they displease me . I want to know why . 90% of what I build is said to be OK . Was it me , was it the device . Why did it sound bad or good . I find every X things that prove something a large quantity Y disproves it . Each case on it's merit is best . Sometimes I reject a thing that might have worked best . For example a single transistor volume pot buffer should be OK in theory , if a valve I wouldn't hesitate except to avoid it totally . One conjecture of mine is a 741 should be capable of hi fi performance if used with a little imagination . That might be 2N4403 inputs and class A CCS trick to output . Equally 2N3055 is a good enough device . People are horrified if I say it ., They even make the highly intellectual statements like there are better devices . That is as simple as saying day follows night . No , I am saying if your passion to succeed is great they can work . Are you able to do it ? That is like making school dinners that taste like restaurant quality . It can be done . An analogy , it is the taste of food that counts . Looks are selling points in both disciplines . How many amps have stacks of components as they are going to a hi fi show as if showing off a V12 of 3960 litres ( Citroen conjecture circa 1928 , 330 cc per cylinder if square bore stroke I believe , Him from Voisin ) ?
The graph shown above came for this .
Quad Spot: Measuring distortion with software
A quote about 303 from above .
The next myth: high bias current
In the Quad service manual it is stated that the DC bias current of the power transistors must be between 10mA and 5mA. A popular “mod” is increasing the current to 35mA and in some cases even higher. Well, in real life the distortion is only 0.05% at a bias current of 0.5mA. At 5mA it is 0.01% and will not get lower at higher bias currents. Also the spectrum analyzer will not give a better (cleaner) picture at high levels of bias current. So Quad was (is) right! For practical reasons, stability of the circuit, 10mA is a good value for the bias current.
Joost Plugge
That seems to suggest a better bias circuit could be worth looking at with 303 . Maybe one extra resistor to aid tracking at the Vbe collector perhaps ? Not bad for 1967 , my guess would be started circa 1962 after Tobey and Dinsdale and better than it or Leak / Bryan . I prefer it to the valve Quad . Like two sisters it becomes apparent the other one who gets less admirers is the better one . Quad always claimed in blind tests they sounded identical . Perhaps so in a narrow group of tests . Ask both for current and 303 wins ( hum modulation of PSU in class A , not completely cured by op amp type circuit of 2 x EF 86 and 2 x KT66 argued not to be true LTP by most , like saying Goose is not a Swan ..... ) .
Some interesting stuff here ( HC Lin link )
http://ochtedete.free.fr/!-amplis/quasi/QuasiComp/QuasiComp.PDF
Quad Spot: Measuring distortion with software
A quote about 303 from above .
The next myth: high bias current
In the Quad service manual it is stated that the DC bias current of the power transistors must be between 10mA and 5mA. A popular “mod” is increasing the current to 35mA and in some cases even higher. Well, in real life the distortion is only 0.05% at a bias current of 0.5mA. At 5mA it is 0.01% and will not get lower at higher bias currents. Also the spectrum analyzer will not give a better (cleaner) picture at high levels of bias current. So Quad was (is) right! For practical reasons, stability of the circuit, 10mA is a good value for the bias current.
Joost Plugge
That seems to suggest a better bias circuit could be worth looking at with 303 . Maybe one extra resistor to aid tracking at the Vbe collector perhaps ? Not bad for 1967 , my guess would be started circa 1962 after Tobey and Dinsdale and better than it or Leak / Bryan . I prefer it to the valve Quad . Like two sisters it becomes apparent the other one who gets less admirers is the better one . Quad always claimed in blind tests they sounded identical . Perhaps so in a narrow group of tests . Ask both for current and 303 wins ( hum modulation of PSU in class A , not completely cured by op amp type circuit of 2 x EF 86 and 2 x KT66 argued not to be true LTP by most , like saying Goose is not a Swan ..... ) .
Some interesting stuff here ( HC Lin link )
http://ochtedete.free.fr/!-amplis/quasi/QuasiComp/QuasiComp.PDF
My best attempt to freehand draw an exponential curve . 2nd a little down so not excessively euphonic . Quad 303 1 kHz 45 watts 8R .
I find I agree with a lot said here . Note he is not a fan of even order distortion .
http://sound.westhost.com/valves/thd-imd.html
I find I agree with a lot said here . Note he is not a fan of even order distortion .
http://sound.westhost.com/valves/thd-imd.html
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Just a side note about hearing thresholds ---- especially when quoted from decades in the past -->
1 percent. It may be 1% when listening thru 5-10% distortion THD/IM from typical speakers. Theer are some pretty low distortion speakers now and the electrostatics (Quad new versions esp) and electrostatic headphones would show the detectable distortion threshold from the elctronics to be much lower than 1%.
Tests i did on myself show my threshold is (depending on freq, spl etc) to be in the range of .01 to .1% thd as a general statement.
A side note to this side-note: Amps would need to be 5-10X better.
A side to the side to the side-note: the entire electronic chain - measured as a System - should be 5-10X better for the entire system to be truely transparent. It is, after all, the entire system we listen to/hear. In that case, how low would the distortion be for each component in the system to have its distortion so that the entire system is transparent.
Thx-RNMarsh
1 percent. It may be 1% when listening thru 5-10% distortion THD/IM from typical speakers. Theer are some pretty low distortion speakers now and the electrostatics (Quad new versions esp) and electrostatic headphones would show the detectable distortion threshold from the elctronics to be much lower than 1%.
Tests i did on myself show my threshold is (depending on freq, spl etc) to be in the range of .01 to .1% thd as a general statement.
A side note to this side-note: Amps would need to be 5-10X better.
A side to the side to the side-note: the entire electronic chain - measured as a System - should be 5-10X better for the entire system to be truely transparent. It is, after all, the entire system we listen to/hear. In that case, how low would the distortion be for each component in the system to have its distortion so that the entire system is transparent.
Thx-RNMarsh
Last edited:
With all due respect RNMarsh, but I find 0.01% of THD audibility to be incredible and VERY hard to believe. It flies in the face of literally everything I have ever read about THD threshold of audibility.
It could, however, be something else that changes with volume and/or frequency, that you are hearing, which also shows up as a by-product as slightly higher THD, such as changing settling times, rate and shape of THD decay curve, etc.
It could, however, be something else that changes with volume and/or frequency, that you are hearing, which also shows up as a by-product as slightly higher THD, such as changing settling times, rate and shape of THD decay curve, etc.
Simon,
D. Self was just commenting on post #1747. But in context to what is being said besides something about the book not much of any technical discussion for some time.
ps. This thread seems to be like the Blowtorch thread, it has taken on a life of its own without much to do with the original intent.
D. Self was just commenting on post #1747. But in context to what is being said besides something about the book not much of any technical discussion for some time.
ps. This thread seems to be like the Blowtorch thread, it has taken on a life of its own without much to do with the original intent.
Thanks,
I did splurge on his book and there are to me still some undeveloped issues. Ordinarily I would mention that privately by Email but my account is down at the moment.
So perhaps we could get onto the subject of power supplies, as I have been working on that for a few years now.
To begin with diode noise, there are two issues with diode noise, switch on causes large surge currents that can resonate with the inductance in the power transformer and also cause harmonics in the AC line from the partial cycle charging time. This was recently addressed by Morgan G.
The other issue that has been known for a while is switch off noise.The abrupt interruption of current flow causes the energy that was stored and now has no place to dissipated often radiates. This is often treated by RC snubbers across each diode.
The frequency range of turn off is hundreds of kilohertz to megahertz. The turn on can be thousands of hertz. I hadn't seen anyone addressing both issues.
ES
I did splurge on his book and there are to me still some undeveloped issues. Ordinarily I would mention that privately by Email but my account is down at the moment.
So perhaps we could get onto the subject of power supplies, as I have been working on that for a few years now.
To begin with diode noise, there are two issues with diode noise, switch on causes large surge currents that can resonate with the inductance in the power transformer and also cause harmonics in the AC line from the partial cycle charging time. This was recently addressed by Morgan G.
The other issue that has been known for a while is switch off noise.The abrupt interruption of current flow causes the energy that was stored and now has no place to dissipated often radiates. This is often treated by RC snubbers across each diode.
The frequency range of turn off is hundreds of kilohertz to megahertz. The turn on can be thousands of hertz. I hadn't seen anyone addressing both issues.
ES
Pick what ever number that makes you happy. BUT you can not get past the point I made with any number you like or believe. The .01% is a guess at where someone may be limited... the .1% is not a guess. The speakers I used are not as low distortion as I have available today;
You/they are trying to determine threshold by listening thru a large amount of speaker distortion. Older speakers maybe more than modern speakers of low distortion. The masking of the detectable threshold is at work. When you lower the masking, the threshold of detection is lowered as well.
If you rely on distortion thresholds from old old writings it is even less credible... old tube amps and limited bandwidth speakers - all with higher distortion than we can reproduce today. try your own listening tests with an arbitrary waveform generator and manipulate the harmonics etc with very low distortion speakers etc and see what You get. use that number for your basis.
Then measure the complete System and compare.
Thx-RNMarsh
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.