Guys, if Class AB is better, there got to be a sound theory. And I would really like to hear it. Every single book and article I read talks about cross over distortion being the worst kind that is not related to the output amplitude. Even with Oliver's optimization, you just minimize the crossover distortion. Even Self said in the book that there is NO current setting to eliminate the crossover distortion....PERIOD.
So the only sure way to get rid of the single most offensive crossover distortion is run in Class A. Period.
As for optimizing the current. Even Self gave the optimal with 0.1ohm emitter resistor is 200mA per stage. I don't have the book right now, but I read it very clearly. This is what he so called Class B!!!
Yes, I am disgusted with Self's book, all the mistakes, naming the wrong figures, unclear in what circuit he was describing, calling running 200mA idle current as Class B. The final straw was I simulated his exact circuit in his stupid book and got worst distortion compare with all the designs I simulated. that's when I just threw his book aside. All his assertion in Chapter 9 and 10 without giving any proof as if we should just take his words. That got to be one of the worst book I read. All these mistakes in the SIXTH edition, someone needs to get fired editing that book. All these were described in one of my old thread complaining about the book. I pointed to every single issue on that two chapters.
So the only sure way to get rid of the single most offensive crossover distortion is run in Class A. Period.
As for optimizing the current. Even Self gave the optimal with 0.1ohm emitter resistor is 200mA per stage. I don't have the book right now, but I read it very clearly. This is what he so called Class B!!!
Yes, I am disgusted with Self's book, all the mistakes, naming the wrong figures, unclear in what circuit he was describing, calling running 200mA idle current as Class B. The final straw was I simulated his exact circuit in his stupid book and got worst distortion compare with all the designs I simulated. that's when I just threw his book aside. All his assertion in Chapter 9 and 10 without giving any proof as if we should just take his words. That got to be one of the worst book I read. All these mistakes in the SIXTH edition, someone needs to get fired editing that book. All these were described in one of my old thread complaining about the book. I pointed to every single issue on that two chapters.
Last edited:
I covered my complains in this thread:http://www.diyaudio.com/forums/solid-state/270285-how-design-sziklai-pair-ops.html
Thanks, I have to go find the book, I just threw it aside after I posted that thread and never look at it again. I tried to look for the book just now and can't find it!!!
Last edited:
Class A = build the XD or a high bias Vfet AB. The Vfet , by design "likes" much
higher bias (and runs hot).
Class AB BJT's run with the lowest THD at 50-100ma. That Xover distortion
can even be corrected by present fast input stages. Some VERY fast
input stages will almost totally correct it , leaving the amp in single
digit PPM territory.
After you leave Class A in a AB BJT , you are in "transient territory" anyways.
at 10W (with 150W peaks) , I don't see how you would hear that 4-5ppm
of residual Xover distortion ??
(your speakers would also be distorting a 100 times greater at this level 😱).
OS
Why 50 to 100mA? What is the magic about that? Do you have any article on this?
If I run in class A, why is MOSFET better? Seems like BJT should be better as the output impedance is lower.
If I run in class A, why is MOSFET better? Seems like BJT should be better as the output impedance is lower.
All manufacturers (who make $$) set up like this in the service manuals for
BJT (sanken , toshiba,, ON). My own experience - nothing to gain over 100ma
, except unwanted heat. Optimum gm vs Ic slope is what we are after here ,
a constant Z (even in the Xover region) - thus, the limited bias range.
Vfet's have a much wider range where the
slope is constant (very tolerant of high bias -wider "A" region).
OS
I found the book, I actually read that part before, I forgot the name. It is not simpler at all. You have to use a big CCS current sink to pull the current to keep the top output transistors on. I gave it a lot of thoughts, it's not a win.
I did this trick on IC opamp long before I read this book. It's just common sense to pull current to make it into class A. It works for IC opamp because you have no control on the bias inside. AND also because they are low current, usually you pull 5 mA, you turn the opamp into class A. There is no negative consequence for preamp IC opamp. BUT for power amp, it's a different story.
Example:
If you want 8W of Class A, you need to move the crossover to -8V. This mean the lower half of the pushpull cannot turn on until output goes below -8V. To do that, the CCS has to sink 2A!!! If the rail is 40V, the CCS sink has to sink 2A to -40V=80W. Then the upper transistor has to supply 2A!!! How is that saving power and lesser parts? If I am right, this will generate double amount of heat and you need even more transistors for the extra CCS.
It is actually worst. If I just use normal cross over at 0V. I only need to set the bias current to 1A. This will already run in Class A to +/-8Vpeak output. It's half the power dissipation than the XD.
Tell me if I am wrong. That's another thing I question Self that I did not even mention in that thread. I really did not give a lot of thoughts until you mentioned it. I just don't see it. I just did not think it was a win at the time and dismissed it totally.
Last edited:
I just read page 451 right hand column. Again, this remind me why I threw this book aside.
1) He made mistake talking about Fig 18.2b "the crossover point when Q2 hands over to Q4....." It is Q3 and Q4!!! Again, typo that confuse people. Q2 is the driver transistor. If you have mistake like this on almost every page, you cannot read the book unless you just trust his conclusion without thinking.
2) He is wrong that with the 1A displacement current, that Q4 is still on at idle!!! There is no way Q4 can be on.
Look at Fig18.2b. Without the displacement current, say Q3 and Q4 is biased to have 250mA quiescent current to meet Oliver's optimization. This means it drop 25mV across Re1, and 50mV total separating the emitters of Q3 and Q4( generated by the bias spreader). But if you pull 1A by the Displacer to -40V. The voltage drop across Re1 will be 1A X 0.1ohm = 100mV. With the Displacer that drop 100mV across Re1, Q4 is turned off completely!!! Self is wrong.
Yes, when I read any book, I do go through the calculations to verify, I never take the words from any book. I verified most of the equations from Cordell's book, he was spot on. That's the reason I consider Cordell's book is at the top.
So far, every chapter I read Self's book, they are full of problems. I stand by my derivation in the last post. That Q4 is off until the load draw the whole 1A from the Displacer. As long as Q4 is not on, the Displacer HAS TO supply all the current to swing the load. That if Self claimed he make the crossover point to -8V in page 451, he need the Displacer to be able to pull 8V from the load......Which is 2A!!!! Not 1A as the book claims.
1) He made mistake talking about Fig 18.2b "the crossover point when Q2 hands over to Q4....." It is Q3 and Q4!!! Again, typo that confuse people. Q2 is the driver transistor. If you have mistake like this on almost every page, you cannot read the book unless you just trust his conclusion without thinking.
2) He is wrong that with the 1A displacement current, that Q4 is still on at idle!!! There is no way Q4 can be on.
Look at Fig18.2b. Without the displacement current, say Q3 and Q4 is biased to have 250mA quiescent current to meet Oliver's optimization. This means it drop 25mV across Re1, and 50mV total separating the emitters of Q3 and Q4( generated by the bias spreader). But if you pull 1A by the Displacer to -40V. The voltage drop across Re1 will be 1A X 0.1ohm = 100mV. With the Displacer that drop 100mV across Re1, Q4 is turned off completely!!! Self is wrong.
Yes, when I read any book, I do go through the calculations to verify, I never take the words from any book. I verified most of the equations from Cordell's book, he was spot on. That's the reason I consider Cordell's book is at the top.
So far, every chapter I read Self's book, they are full of problems. I stand by my derivation in the last post. That Q4 is off until the load draw the whole 1A from the Displacer. As long as Q4 is not on, the Displacer HAS TO supply all the current to swing the load. That if Self claimed he make the crossover point to -8V in page 451, he need the Displacer to be able to pull 8V from the load......Which is 2A!!!! Not 1A as the book claims.
Last edited:
I really don't see it in the simulation. I happened to have done the simulation comparing 200mA to 100mA both with emitter resistor adjusted to satisfy Oliver's optimization ( dropping 26mV across each emitter resistor). Here is the graphs. The name might be misleading, but I did change the emitter resistor to achieve Oliver's optimization. I simulated 1KHz and 20KHz, with input voltage of 0.35Vpeak and 1.5V peak. 0.35Vin gives 7Vpeak out which is in Class A for 200mA bias.
Alan,
You seem to be misunderstanding optimal ClassAB and the way it works. Both D.Self and R.Cordell expend a lot of effort explaining how it works and why it needs to be optimally based.
D.Self even invented a new name for it: "D.Self's ClassB".
Both explain that over biasing increases the distortion. R.Cordell an many others go to great lengths to prevent overbiasing as temperatures change inside the amplifier.
The ONLY way, in my view, to avoid all crossover distortion is to go to ClassA (a single ended ClassA just current clips).
That is not the same as overbiased ClassAB.
If the Output Current exceeds the ClassA limit, then a Push Pull ClassA amplifier transitions into ClassAB.
Finally, ClassA is a current issue.
You seem to be misunderstanding optimal ClassAB and the way it works. Both D.Self and R.Cordell expend a lot of effort explaining how it works and why it needs to be optimally based.
D.Self even invented a new name for it: "D.Self's ClassB".
Both explain that over biasing increases the distortion. R.Cordell an many others go to great lengths to prevent overbiasing as temperatures change inside the amplifier.
The ONLY way, in my view, to avoid all crossover distortion is to go to ClassA (a single ended ClassA just current clips).
That is not the same as overbiased ClassAB.
If the Output Current exceeds the ClassA limit, then a Push Pull ClassA amplifier transitions into ClassAB.
Finally, ClassA is a current issue.
Last edited:
It's not much , but I always see all my dozen very different input stages
"bottom out" into 10-20ppm at about 80ma class AB bias. Above 100ma -
not really bad -double that.
On the other end of this scale <30ma to class B (0) , things (THD) get bad.
If you scope the (-) ltp Collector you will see a larger glitch below 30ma -
NFB is "trying harder" to correct the X-over distortion (which eqauls >THD).
OS
There is no misunderstanding. I am looking for a Class A in low power range BUT with capability of giving high power transient when needed. You cannot get over 20W Class A amp that easy without paying a very steep price in heat and power. With normal listening level, you seldom go beyond 8 to 10W. So I go with the over bias Class AB just to give me PURE Class A 99.9% of the time at normal listening level.
I am looking for good sounding amp for everyday use, I am not looking for an amp that I can show off how low distortion even cranking up to ear splitting level. I design guitar amp, I know how loud is 5W. I design amps with power scaling that I can lower the power to 1W from 30W. It's still quite loud. I truly think people that claim they need over 50W are psychosomatic. You are designing an amp for home use, not for theater or stadium.
My design will give 8W of pure Class A and 100W of max RMS power. I have to see how hot it gets, I might even up the current to get 10W of pure Class A.
Doug Self invented a lot of things, his class B, his XD and all. In my book, his is full of it. It just angry me for someone to sell a book like this for $60. Don't readers check his stuffs? Never seen a book with SO SO many mistakes........in the SIXTH edition.
Last edited:
How is ppm relate to FFT graph? Is it if I have the worst harmonics is -100dB down from the fundamental, than the amplitude is antilog( -100/20)=10-5 = 10ppm?
If that is true, if you look at the graph I provided, at 1.5Vin(30Vpeak output), at 1KHz, it's about 10ppm, and at 20KHz, it's 25ppm. What is so wrong about using 200mA?
Regarding to IPS and VAS. Notice I never talk about them anymore? Any reasonable design will have distortion at least 20dB lower than any of the OPS. Why people even talk about IPS and VAS?
Regarding to your comment on making the amp as high slew rate and frequency response as possible to use GNFB to eliminate crossover distortion. I am new in this game, BUT from what I listen to Krell and Machintosh, they are clinical sounding. They both have all sort of fancy FB and all. My favorite amp is YBA and people laugh when they look at the schematic. It is my believe that there is more to it than just cancelling at the distortion. There is a certain organics about the sound the a lab grade distortionless amp cannot match.
I look at result, I spent a lot more time reading Nelson Pass/Threshold schematics even though his design is not really optimal in Cordell and Self's standard. But result speaks much louder than writing in the books. I don't think Pass ever claims his know better, he just design amps that people willing to pay tons of money for it......Not just one brand, he design more than one and people follow.
I know you have a long standing thread about slew master, CFA and your 200V/uS design. I am still new, there are too many things to learn. I have to pick and choose what to learn first. I can only based on my listening experience, that Krell and Mach are sterile and YBA comes alive. So I pick the conventional way. In fact I am very intrigued by the Haraga design that uses pseudo current output that react with the complex load of the speaker, totally away from the ultra low output impedance, ultra linear with RESISTOR load. Maybe with time, I might understand more.
Last edited:
Your statement about the higher THD amps less clinical nature will be put to
the test with CCS less bootstrapped 200ppm creation.
I fully understand there might be some merit to a "dirty" approach.
I also understand the power supply , grounding/layout , and even the source
are the other big factors.
I don't believe in 1,000$ cables , silver wire , and elevated cabling..
Simulator , scope and ears will suffice.
OS
the test with CCS less bootstrapped 200ppm creation.
I fully understand there might be some merit to a "dirty" approach.
I also understand the power supply , grounding/layout , and even the source
are the other big factors.
I don't believe in 1,000$ cables , silver wire , and elevated cabling..
Simulator , scope and ears will suffice.
OS
alan0354,
if you are looking for class a with bipolar outputs and don't need high very high power, have you considered building nelson pass' 20wt class a project from 1977 with updated semiconductors?
mlloyd1
if you are looking for class a with bipolar outputs and don't need high very high power, have you considered building nelson pass' 20wt class a project from 1977 with updated semiconductors?
mlloyd1
I have to look into that, too many things to learn. Do you have link for the schematic?
Thanks
I really have no idea what is better, I can only judge based on what I listened to before and pick the first battle. I am not saying at all that I disagree with you. I just need to pick my first battle and I picked the YBA route. I have to get one amp going first and then see what happens. When I have more experience, I might agree with you, but it will take time.....after this first amp.
Layout is everything. I spend so much time on the layout of the pcb. Particularly these are not low frequency circuits with high slew rates. Even when I simulate my amp that is nothing fast like your Slew Master, the closed loop corner frequency is over 1MHz, close to 2MHz. I use high current to get the slew rate.
I don't believe in expensive cable and silver wire and all. But I do believe in thick and big cables. I actually tested with my wife by adding one pair of Monster clone cable at a time, We could hear improvement until after the 4th pairs. I used cheap 12 gauge speaker cable from OSH, but I use 4 pairs for each side.
I think I talk too much on forum already, it's time to put the rubber on the road. The OPS board is going to come back in about a week and I am almost ready to send out the IPS/VAS board. It's just a complementary IPS with darlington VAS. This should give the highest speed as it's push pull VAS. I run 12mA VAS to overcome the slew rate limit and with a 3EF OPS, speed should not be the bottle neck. I don't even feel like experimenting with Blameless and others for now as they are all so much lower distortion than the OPS. They are all good enough.
Last edited:
Damn!!!! I just found out I lost the big order from Allied Electronics. UPS claimed they delivered last Friday but I never received it. This is a big order of $230. 35 each of the MJL1302 and MJL3281, 16 of the 10,000uF 63V caps and the other smaller caps!!!! Now UPS has to open an investigation and I don't even know how this ends.
Don't order from Allied Electronics if you can. They are slow.....very slow. Take two weeks before they ship out. They have not ship out my second order yet and they charge double of Digikey. I ordered two package from Newark after I ordered from Allied. One has to be shipped from Britain and I received them few days ago already. Digikey got to break the record. I always received from them in 3 days!!!! Digikey is a little more expensive, but their shipment is cheaper and FAST.
Don't order from Allied Electronics if you can. They are slow.....very slow. Take two weeks before they ship out. They have not ship out my second order yet and they charge double of Digikey. I ordered two package from Newark after I ordered from Allied. One has to be shipped from Britain and I received them few days ago already. Digikey got to break the record. I always received from them in 3 days!!!! Digikey is a little more expensive, but their shipment is cheaper and FAST.
- Status
- Not open for further replies.