That is definitely not true. What is true is that the circuit works exactly as promised. As can be shown in other ways using simulators or practical measurements.
Oh, that may be the way it should be - but I'm not interested.
And I'm certainly not complaining about the THD curves of the actual core above. Let's see what we can get out of the given QUAD405 circuit?
HBt.
That's the big and all-important difference. So I didn't cheat and used this bias diode. I also don't need automatic offset correction ..!
CL +6dB
RL 4Ohm
We'd better find out what the problem with the core is at power levels above 20W RMS.
RL 4Ohm
We'd better find out what the problem with the core is at power levels above 20W RMS.
My last post on the core of the QUAD405, before I bring the thread back to the feasibility study using operational amplifiers, as per the title.
Iq_Q7 < 1.5mAdc
Iq_Q8 > 30mAdc
A horizontal curve with 6m% THD and equal distribution of the components from k3 onwards is nothing to sneeze at. But being able to deliver only 24 real watts with +/- 50Vdc rails is really poor.
As soon as we increase the base bias voltage of Q7 and Q6, in the good faith that this (alone) could have a positive effect, we are mistaken. The ensemble quickly falls apart /tilt.
With the Z1,Z2,Z3,Z4 (bridge) balancing you can play very nicely - and with MC12 you can observe the function well, which already works very well with the dynamic ac mode. No question, it works and you can see the result of the intervention on the THD quite easily and quickly if you want to handle it that way.
But, and now I have to agree with @wahab again, the circuit is interesting as it is, but it was never the yellow of the egg. It's actually a totally messed up design. Nevertheless, somehow a Quad405 has to be convincing in everyday use, otherwise ..!
Back to the OP-Amp studies, that's more promising than bringing an ancient dinosaur back to life.
Bye,
HBt.
👽
Iq_Q7 < 1.5mAdc
Iq_Q8 > 30mAdc
A horizontal curve with 6m% THD and equal distribution of the components from k3 onwards is nothing to sneeze at. But being able to deliver only 24 real watts with +/- 50Vdc rails is really poor.
As soon as we increase the base bias voltage of Q7 and Q6, in the good faith that this (alone) could have a positive effect, we are mistaken. The ensemble quickly falls apart /tilt.
With the Z1,Z2,Z3,Z4 (bridge) balancing you can play very nicely - and with MC12 you can observe the function well, which already works very well with the dynamic ac mode. No question, it works and you can see the result of the intervention on the THD quite easily and quickly if you want to handle it that way.
But, and now I have to agree with @wahab again, the circuit is interesting as it is, but it was never the yellow of the egg. It's actually a totally messed up design. Nevertheless, somehow a Quad405 has to be convincing in everyday use, otherwise ..!
Back to the OP-Amp studies, that's more promising than bringing an ancient dinosaur back to life.
Bye,
HBt.
👽
Last edited:
This bias diode didnt exist in the first version of the 405, they added it in the 1978 second version with the result of increasing
the VAS current as well as the negative rail OS bias, the result is a 28mA Iq for the driver and 5mA for the power transistor,
so the lower side went totaly class AB, the other effect is to get the upper device close to conduction since it was previously
-600mV reverse biaised wich is morphed to 60mV forward biaised with this diode.
So you didnt cheat but Quad did on several points.
As for doing tests with an opamp this will be a different configuration unless you have access the op amp compensation,
but this will be closer to the principle schematic since the actual Quad is different from the simplified scematic that is supposed
to explain the FFWD.
Interesting, perhaps we can clarify this point in the new thread initiated for this purpose. I am always interested in clarification.
neuer Faden
HBt.
Explanation / Statement
I had an age-related obduracy, I'll just call it that now (it's still embarrassing), which led to a completely stupid misinterpretation, namely the assumption that 0dBW is correctly 1W, but 20dBW should miraculously suddenly only be 20W. This is of course complete nonsense from an old grandpa. 20dBW refers to the reference of one watt, i.e. 10^(20dBW/10) equals 100W.
So everything is fine with the forefather 405. It works according to the CD principle and delivers remarkable results.
😎
6m% at 100W with 8Ohm !
I had an age-related obduracy, I'll just call it that now (it's still embarrassing), which led to a completely stupid misinterpretation, namely the assumption that 0dBW is correctly 1W, but 20dBW should miraculously suddenly only be 20W. This is of course complete nonsense from an old grandpa. 20dBW refers to the reference of one watt, i.e. 10^(20dBW/10) equals 100W.
So everything is fine with the forefather 405. It works according to the CD principle and delivers remarkable results.
😎
6m% at 100W with 8Ohm !
Walker s as well as Wireless World article are worthless because both are assuming that there s as much
NFB around the OS and VAS at audio frequencies than at DC.
As you can see it the actual Quad miller loop doesnt enclose the IPS while it is supposed to do so in Walker s explanation.
Wworld at least pointed this fact as well as the hint that their own computation did hold only at DC or VLF,
that is at a frequencies at wich :
1)The miller loop is inactive since the cap Z is infinite while the output inductance Z is equal to 0.
2) The GNFB is hence at its peak 106dB, hence the totality of GNFB and local NFB enclose both VAS and OS because of 1).
1) and 2) amount to the same assumption than Walker s wich is precisely that there s 106dB NFB around both the OS and VAS
at all frequencies.
At audio frequencies the miller loop reduce the GNFB by its own local NFB amount, so the only way to have that much error
correction well above the audio band is that this loop enclose the OS.
As a conclusion it s not current dumping that work but the enclosement of the OS within the miller loop, that is,
to increase the NFB around the OS.
NFB around the OS and VAS at audio frequencies than at DC.
As you can see it the actual Quad miller loop doesnt enclose the IPS while it is supposed to do so in Walker s explanation.
Wworld at least pointed this fact as well as the hint that their own computation did hold only at DC or VLF,
that is at a frequencies at wich :
1)The miller loop is inactive since the cap Z is infinite while the output inductance Z is equal to 0.
2) The GNFB is hence at its peak 106dB, hence the totality of GNFB and local NFB enclose both VAS and OS because of 1).
1) and 2) amount to the same assumption than Walker s wich is precisely that there s 106dB NFB around both the OS and VAS
at all frequencies.
At audio frequencies the miller loop reduce the GNFB by its own local NFB amount, so the only way to have that much error
correction well above the audio band is that this loop enclose the OS.
As a conclusion it s not current dumping that work but the enclosement of the OS within the miller loop, that is,
to increase the NFB around the OS.
Last edited:
And that's exactly what I've been waiting for, because this is the only point of contention.
Please explain the "Miller loop" using equivalent equivalent-circuits, with sketches and math.
Please explain the "Miller loop" using equivalent equivalent-circuits, with sketches and math.
You could had googled wikipedia, that s a frequency dependent NFB from output to input of an inverting circuit using a cap,
that s akin to a low pass filter set apart that there s a gain, improperly called VAS in power amplifiers since output/input = voltage/current = (trans) impedance.
https://en.wikipedia.org/wiki/Miller_effect
that s akin to a low pass filter set apart that there s a gain, improperly called VAS in power amplifiers since output/input = voltage/current = (trans) impedance.
https://en.wikipedia.org/wiki/Miller_effect
What you want to sell us as a Miller loop is the so-called (local) voltage feedback by means of capacitive reactance. This includes the aforementioned 120pF capacitor - and this does not include the dumper, i.e. our C & B class push-pull output stage.
unfortunately I have such a strange university degree, on it we can read: of electronics, communications and electrical engineering.
I know what we're talking about here; rusty, yes I am and otherwise shaken, but ...
HBt.
Great interjection,
unfortunately I have such a strange university degree, on it we can read: of electronics, communications and electrical engineering.
I know what we're talking about here; rusty, yes I am and otherwise shaken, but ...
HBt.
Of course, this refers to the circuit of the QUAD405 and not to the super-smart WikiPedia thingy.
I'm asking you personally,
not me, not a book, not an article, not the internet - I'm just playing dumb according to the motto.
Because perhaps everything is based on a misunderstanding - and this needs to be cleared up.
Absolutely correct, agreed.
But don't forget that we can also accept another completely equivalent idea, we can permanently transform back and forth.
regards,
HBt.
So you didnt need to ask the question, but certainly that you can deduct that the so called Walker s balance has this
strange property :
The product of the miller cap Z by the inductance Z should be equal to the product of the FFWD resistance by the FB
return resistance, let s call the two latters product A and assume them as constant, in the schematic these are 47R and 500R,
so A = 47 x 500 = 23500.
Let s call the miller cap impedance Zm and the inductance impedance Zi.
Then the circuit is balanced if Zm x Zi = A.
You can see that if i set Zm to 10^(10) R the relation still hold as long as Zi is such that the Zm x Zi product is equal to A.
This amount to an extremely small cap and a corresponding extremely high inductance, and since it work both ways
i can as well set Zm at 1R and adjust Zi such that the product of the impedances is equal to A, of course the amp wouldnt
work in both cases, so they did an optimisation without even taking account of the constraints.
You ll understand that Walker s demonstration and Wireless world article are not rigourous maths when it comes
to actually characterise the circuit, they made both a DC anaysis but there s no frequency domain analysis in both cases,
their analysis assume that the available GNFB is the same at all frequencies, wich is wrong unless the OS is well enclosed
inside the miller loop, and that s the reason of the low THD, not counting that the OS is in class C only for the upper part
wich is not stated in any analysis.
strange property :
The product of the miller cap Z by the inductance Z should be equal to the product of the FFWD resistance by the FB
return resistance, let s call the two latters product A and assume them as constant, in the schematic these are 47R and 500R,
so A = 47 x 500 = 23500.
Let s call the miller cap impedance Zm and the inductance impedance Zi.
Then the circuit is balanced if Zm x Zi = A.
You can see that if i set Zm to 10^(10) R the relation still hold as long as Zi is such that the Zm x Zi product is equal to A.
This amount to an extremely small cap and a corresponding extremely high inductance, and since it work both ways
i can as well set Zm at 1R and adjust Zi such that the product of the impedances is equal to A, of course the amp wouldnt
work in both cases, so they did an optimisation without even taking account of the constraints.
You ll understand that Walker s demonstration and Wireless world article are not rigourous maths when it comes
to actually characterise the circuit, they made both a DC anaysis but there s no frequency domain analysis in both cases,
their analysis assume that the available GNFB is the same at all frequencies, wich is wrong unless the OS is well enclosed
inside the miller loop, and that s the reason of the low THD, not counting that the OS is in class C only for the upper part
wich is not stated in any analysis.
without your "Miller-Schleife" OS not in
with your "Miller-Schleife" OS completly in
fully dumped, both OS BJT are in the game
But in the power range from 16W upwards, the so-called dumper has long been in full operation - and as soon as the dumper (i.e. the EF-1) is working, the picture changes to a plain EF-2 push-pull output stage, regardless of whether it is quasi-complementary or 100% complementary. Moving along our timeline of the instantaneous value, everything will eventually converge to a normal negative feedback system.
The only interesting point is the zero crossing of the signal curve, i.e. the range around 50µWrms to 60mWrms. Modulation products are also of interest. But to send a TMC into the race at just under 25Wrms is eyewash.
#
If we want to be honest, we have to analyze the circuit differently. In the three examples above, I use the cascode, which is why everything is so low and horizontal.
with your "Miller-Schleife" OS completly in
fully dumped, both OS BJT are in the game
But in the power range from 16W upwards, the so-called dumper has long been in full operation - and as soon as the dumper (i.e. the EF-1) is working, the picture changes to a plain EF-2 push-pull output stage, regardless of whether it is quasi-complementary or 100% complementary. Moving along our timeline of the instantaneous value, everything will eventually converge to a normal negative feedback system.
The only interesting point is the zero crossing of the signal curve, i.e. the range around 50µWrms to 60mWrms. Modulation products are also of interest. But to send a TMC into the race at just under 25Wrms is eyewash.
#
If we want to be honest, we have to analyze the circuit differently. In the three examples above, I use the cascode, which is why everything is so low and horizontal.
schaue hier
As soon as C6 is in the race, the available amount of beneficial negative feedback is of course no longer constant over the audio frequency range.
Without C6, we are up to 120dB OLG.
Can we seriously condemn Mr.Walker and his engineers for their perspective? Everything depends on the point of view and the model conception!
This little picture shows the situation quite clearly. Let's delete everything that disturbs us in the signal zero crossing, namely the dumpers.
I tested with a cascode, the difference is negligible.
The only valuable improvement, with the constraint of using the original circuit, is to halve the IPS CCS current
and to eventually set the miller cap at 100pF, but this wont reduce H2 as to get it at the same level than the other
harmonics, so far as much as i like the CFA amps, wich were my first builds, for simple and capacitive coupled amps
i discard them when it comes to very high linearity, PSRR and so on.
As for Walker no doubt that he was a very skilled enginer and he managed to do a lot with a surprisingly low
component count, but those truck loads of NFB were wasted linearizing a highly non linear OS, i would say that
if TMC had been explored in the 70s he would had implemeted it in subsequent versions of the 405 all while
using AB class with low Iq, wich he got close with the 405-2.
The only valuable improvement, with the constraint of using the original circuit, is to halve the IPS CCS current
and to eventually set the miller cap at 100pF, but this wont reduce H2 as to get it at the same level than the other
harmonics, so far as much as i like the CFA amps, wich were my first builds, for simple and capacitive coupled amps
i discard them when it comes to very high linearity, PSRR and so on.
As for Walker no doubt that he was a very skilled enginer and he managed to do a lot with a surprisingly low
component count, but those truck loads of NFB were wasted linearizing a highly non linear OS, i would say that
if TMC had been explored in the 70s he would had implemeted it in subsequent versions of the 405 all while
using AB class with low Iq, wich he got close with the 405-2.
Last edited:
It gets exciting from 2mWrms power consumption; by the way, you should examine the snot once metrologically. This was probably already done in the 1970s.