Pretty certain it never made JAES so it would be a preprint from AES Hamburg, Mar 81.
___________
My apologies Mr. Cordell.
My senile memory has Otala responsible for the funky waveforms instead of you.
It's exactly "the extent that such conditions occur with real program material" that is of interest to me. My guess (from looking very hard) is 0% probability .. but I'm willing to be proved wrong
Evidence please .. rather than speculation.
Rayma, if you look at the funky Cordell waveforms as well as jcx's (along with their carefully matched loads
), you'll realise real life percussion is unlikely to excite this._____________________
But to come back to Bob's 2nd edition, how about stuff that HAS got DBLT evidence of worth?
Earlier on, you dismissed using Class G (extra rails ala Carver cube & Hitachi Class H bla bla) to make supa high current Iq sensible (?).
But higher unclipped power is easily shown to be audible and preferred (in DBLTs bla bla). An extra set of rails will at least double the power output with the same thermal design, SOA devices etc (theoretically up to 4x sine wave) of the 'conventional' amp.
Meanwhile footling changes to Class XD, supa high current Class A bla bla only twiddle stuff below 0.01% THD .. for which NO evidence has ever been presented of worth .. as is the existence of the funky Cordell / Otala waveforms
I'm as interested in 1pp zillion THD20k as the next guy but ONLY if it can be achieved with LESS COMPLEXITY than most of the stuff presented on this forum.
SIMPLICITY also allows easier implementation of Class G etc.
________________
Another 'feature' that has DBLT backing is overload behaviour. I share your believe that for an amp about 2x50W 8R, the overload behaviour is probably the most audible difference between amps .. we discount the Golden Pinnae stuff that is marginally unstable on real speakers.
With a conventional PSU, an overloaded amp superimposes the 50/60 Hz sawtooth on the signal.
I conducted a series of tests on using a toroid with slightly LESS output voltage but the same size, regulation & cost for a 2x50W amp. There was slight but unanimous preference for this (even though it had slightly less power) .. but alas, marketing insisted we go for the bigger power spec.
Then there's Self who thinks seriously skew-wif overload behaviour is OK ... but I don't think he's conducted DBLTs
Today, SMPSs may be economical for even 50W amps and they can be designed to be 'semi-regulated' (rails will sag but remain 'clean' without da evil sawtooth). I think some of the Hypex PSUs are like this.
And we can incorporate PFC etc and be nice to our neighbours too
___________________
Another subject is soft clipping. You have already made contributions to this subject with Klever Klipper.
I think all this stuff deserves serious discussion in at least a separate section, cos they WILL make amps that SOUND BETTER.
_____________________
This is the DBLT stuff which I referred to earlier and there is good 'objective' criteria and backing for having this in your book.
The bandwidth limitation stuff is of course heretical to both objectivists & Golden Pinnae alike but I think you should at least mention it as the properly gathered hard evidence appears to be unanimous.
Last edited:
Self definitely covers the Oliver criteria and I think he just calls it optimally biased class B. I don't recall him citing Oliver as a reference, however.
Cheers,
Bob
I think we agree. I think I missed that you had been saying that the max current could equal the rail-to-rail voltage divided by the voice coil resistance. This of course is twice the max current the amp can deliver into a resistive load whose value is the same as the voice coil resistance.
Cheers,
Bob
Oliver's condition essentially takes the average error over the whole transfer curve and finds the degeneration voltage which gives an overall good result. If you exclude the class B region, then optimizing only the class A region results in a different optimal degeneration value.
Hi
I hate to beat the dead horse over lowering the emitter resistor Re for output EF stage and increase bias current to satisfy Oliver's condition.
If you look at D Self's book p249 Fig 9.19. It is a 3 output stages in parallel. The Re=0.1 ohm that Self has been talking in p279. This is his Class B which is Oliver's condition. He uses 215mA bias current as described in Table 10.2 in p279. He obviously they it's safe to use as low as 0.1ohm Re and not worry about burning the amp!!!
My impression of Self's book is he does not go deep into theory, more like he did the experiment and publish the optimal result and you follow like in Table 10.2. Never explain about Oliver's condition, just give you the current setting for a particular Re. I take that he actually tested the circuit in Fig. 9.19 and deem it's safe.
I was told 0.1ohm only works for single output stage, that current hogging can occur when I have multiple output stages in parallel and risk burning the transistor. But Self did it and published in his book. Now I don't know what to think.
I hate to beat the dead horse over lowering the emitter resistor Re for output EF stage and increase bias current to satisfy Oliver's condition.
If you look at D Self's book p249 Fig 9.19. It is a 3 output stages in parallel. The Re=0.1 ohm that Self has been talking in p279. This is his Class B which is Oliver's condition. He uses 215mA bias current as described in Table 10.2 in p279. He obviously they it's safe to use as low as 0.1ohm Re and not worry about burning the amp!!!
My impression of Self's book is he does not go deep into theory, more like he did the experiment and publish the optimal result and you follow like in Table 10.2. Never explain about Oliver's condition, just give you the current setting for a particular Re. I take that he actually tested the circuit in Fig. 9.19 and deem it's safe.
I was told 0.1ohm only works for single output stage, that current hogging can occur when I have multiple output stages in parallel and risk burning the transistor. But Self did it and published in his book. Now I don't know what to think.
Last edited:
Don't forget that the higher the bias current, the more drop across the Re, and the better the // stages will track.
Also don't forget there are several interrelated factors that come into play here. Looking for a simple, single 'truth' how to set it will most probably lead to continued confusion...
There's no substitute to doing it, playing with it, blowing up the occasional device and build your understanding.
Jan
Actually, Oliver's condition is much simpler than even that. Make the transconductance of the output stage at idle bias the same as the transconductance at high current being fed to the load. Alternatively put, make the gain of the output stage into a given load the same at zero current into the load as at high current into the load.
The criterion is best understood when considering ideal transistors, since details like emitter and base resistance in real transistors can muddy the waters. Also, the simple version of the criteria assumes that the P and N transistors are identical in their parameters.
When you operate at a current that puts 26mV across RE, 1/gm of the transistor equals RE.
Oliver's criteria ASSUMES that making output stage gm the same at zero load current as at high load current is the lowest-distortion solution.
Cheers,
Bob
Yes I agree. But Everyone told me it's a no no to use anything less than 0.22. I had to pull the 0.12 ohm and sit waiting for 2 weeks now for the 0.22ohm to arrive.
I have to say I never understand totally why I cannot use 0.12 as my original design. I am still dropping 26mV across the 0.12ohm just like if I use 0.22. Even Self said clearly in his experiments that it's the Vq that is important, not the bias current.
I keep thinking, thermal run away due to current hogging can be mitigated by using lower rail voltage.......which is obvious. I was planning to sacrifice max Class AB power by lowering the rail to lower the power dissipation. I was going to using as low as 25V to 30V rail to prevent current hogging.
If lowering the rail voltage can prevent transistors from blowing when using 0.12ohm emitter resistor, I really want to go back to 0.12ohm and run 200mA bias current for each stage so I get 8W of Class A power.
Last edited:
Alan,
So why don't you do what you thought was correct and test your understanding. Then when your 0.22r resistors arrive change them and the bias current and compare the results? As Bob and others have pointed out the best way is to actually try something and learn. Yes you may burn up a few devices now and then but it is the best way to learn, by doing rather than going by theoretical ideals.
So why don't you do what you thought was correct and test your understanding. Then when your 0.22r resistors arrive change them and the bias current and compare the results? As Bob and others have pointed out the best way is to actually try something and learn. Yes you may burn up a few devices now and then but it is the best way to learn, by doing rather than going by theoretical ideals.
I guess this is different style. I spent the first 15 years of my career staying on the bench even though I have my own office. I breadboard and build everything myself. But I move farther from doing hands on work and I find it more efficient as I can stay away and see the big picture better. I am new here, but my track record of first run success is almost 100% both in microwave amp and MPU design. I never have a degree in EE, everything was hands on before. BUT I really change my mind, I spent years studying deep into theory because in my experience it works. I got a 1.5GHz analog circuit on the Smith Chart plot by a network analyzer to match my simulation almost to the exact.
I am pretty sure Mr. Cordell will agree. We both design IC before, try and see is an expensive proposal for IC. We go on theory a lot before committing to real circuit. That's the reason I like his book so much. he explain why instead of just do it.
JMHO
Alan,
All I am saying is that you thought you were correct going by theory and reading all that you had at hand. Now build the circuit and see if theory matches reality. I think what throws off many theoretical plans are the unknown parasitic values that you can only test by actually building the circuit and testing. Nobody said you have to burn up an entire circuit to find out that theory is just that many times. Many builders of amplifiers will attach a light bulb to the outputs before they would ever attach their speakers, just in case there is an error. All learning in my eyes takes making mistakes. If you never make a mistake I would say you learn little except to follow someone else direction and you are just copying a proven design.
I have worked in R&D in the plastics field for many years, some of the best learning was when something went wrong or I tried something that others said was impossible and I proved those theoretical answers wrong. I have had to prove things more than once to the chemist at Bayer Material Science that they said couldn't be done or would never happen. Reality doesn't always follow predictions. I have done work on solid fuel rocket parts that I was told could not be made, we did it and it worked, we were paid to try something and made it work.
I know electronics are not the same as working with polymers but many of the same lessons by trial and error are the best teachers.
All I am saying is that you thought you were correct going by theory and reading all that you had at hand. Now build the circuit and see if theory matches reality. I think what throws off many theoretical plans are the unknown parasitic values that you can only test by actually building the circuit and testing. Nobody said you have to burn up an entire circuit to find out that theory is just that many times. Many builders of amplifiers will attach a light bulb to the outputs before they would ever attach their speakers, just in case there is an error. All learning in my eyes takes making mistakes. If you never make a mistake I would say you learn little except to follow someone else direction and you are just copying a proven design.
I have worked in R&D in the plastics field for many years, some of the best learning was when something went wrong or I tried something that others said was impossible and I proved those theoretical answers wrong. I have had to prove things more than once to the chemist at Bayer Material Science that they said couldn't be done or would never happen. Reality doesn't always follow predictions. I have done work on solid fuel rocket parts that I was told could not be made, we did it and it worked, we were paid to try something and made it work.
I know electronics are not the same as working with polymers but many of the same lessons by trial and error are the best teachers.
I think electronics is the same, I agree with what you say. Actually I got offered a contract to design something that people don't think it can be done. They want me to give it the best shot. I have to design a transimpedance amp that can measure 4pA to 0.1% precision in less than 5uS!!! That's when theory runs out and have to try and see. pcb is coming back this week, I am crossing my finger whether it's a beauty or a beast.
Thermal issue is really my weak point as I never have to deal with this in my career. So I have to listen here and be more caution. Really depends on you guys.
Obviously not Alan. Vsup does't figure in current hogging, at least not in the 1st, 2nd, 3rd place.
How do you think current hogging works?
Jan
Last edited:
-2mV/deg of Vbe, so the hotter it gets, the lower the Vbe and more voltage drop across Re that cause more current, then it get's hotter, draw more current until it burns.
this is true whether you use Re=0.5 and run 50mA or 0.12 running 200mA. It should be all about the heatsink capability and rail voltage. Am I missing something?
Last edited:
Gee, I thought everything Doug wrote was highly suspect. You mean you are starting to read his book now? Read as in consider and understand.
0.1 ohm works with devices semi-matched as they would be from the manufacturer in the quantities used in making amplifiers. Otherwise you need a higher Re.
They used 0.05 ohm emitter resistors in many Carver amplifiers. Food for thought.
-Chris
-Chris
0.1 ohm works with devices semi-matched as they would be from the manufacturer in the quantities used in making amplifiers. Otherwise you need a higher Re.
They used 0.05 ohm emitter resistors in many Carver amplifiers. Food for thought.
-Chris
-Chris