I`m member of this forum for almost two years so I think I have the right to comment:
Mikeks is one of those guys who know everything and he`s always right.
(Very interesting kind of people:smash)
I`ll be VERY happy if MIKEKS give us ANY of his amp designs!
Anyone seen something like that or I`m making mistake?
Mikeks is one of those guys who know everything and he`s always right.
(Very interesting kind of people:smash)
I`ll be VERY happy if MIKEKS give us ANY of his amp designs!
Anyone seen something like that or I`m making mistake?
ingrast said:
Rodolfo,
I agree!
On the other way I´m a bit tired to discuss further variants of Bob´s good and simple design from 1982, which also only improve the emitter follower.
But of course as always a improvement is possible, so lets see!
Just for learning more I´m much more interested to see some examples for overall feedback EC, like the ones you designed, but never show it here.
The only few examples with overall feedback EC in this far too big and therefore blind thread are from Tom and me.
So at least I still wait to see yours and also Jan´s implementation!
Some time ago, I try to improve the performance of a simple UCD type class D amp.
Although this design is not bad, it suffers like all switching amps under the deadtime and Rdson distortion.
Because of the strongly limited bandwidth one cannot improve it simply with negative FB or feedback EC.
Anyone thinks this statement is wrong?
So I try the use of "true" EC, which means Black´s old invention of the feedforward EC.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=81081
I found Stochinos idea in EW October 1994 page 818, where he use a small transformer as a power summer.
In the meantime I did some further simulations, which looks much nearer to a real implementation, with impressive results.
If I will find the time I will show it here later.
Regards
Heinz!
anatech said:Hi Bob,
I think we can leave this open in case someone has something constructive to add. I would like to thank you for a very instructive discussion!
Do you think this old horse has gone to the glue factory?
-Chris
This has been a very good thread, and it has helped to revive interest in and development of error correction circuits. While I think the level of activity has died down a bit, I believe that the ongoing work on EC that this thread has inspired will continue to result in additional interesting posts. Let's not forget that one of the most highly reviewed amplifiers in Stereophile, the Halcro, uses this technique. I expect more will do so as the technique becomes better understood. After all, it has been shown in several ways by many on this thread that EC is not expensive to implement, and certainly need not be confined to extraordinarily expensive amplifiers. I continue to dabble with variants of the EC circuit as well.
Bob
Hi, Mr. Cordell,
I made EC with bipolar outputs. After the VAS, I use 10V zener (where your mosfet amp uses 22V zener).
The amp uses the same voltage rail for front stage and output stage, +/-40V
I test the amp with load nearly clipping, the rails are dropping to +/-36V, but the output sinusoidal before clipping is only capable of +/-30V.
This means EC configuration that I make loses rail efficiency about 6V on positive rail and another 6V on negative rail.
How many volts of loss is your mosfet EC? How to roughly calculate the loss (if we use the same power supply for front stage and output stage) Is it connected to the 10V or 22V zeners that we use?
Thank you
I made EC with bipolar outputs. After the VAS, I use 10V zener (where your mosfet amp uses 22V zener).
The amp uses the same voltage rail for front stage and output stage, +/-40V
I test the amp with load nearly clipping, the rails are dropping to +/-36V, but the output sinusoidal before clipping is only capable of +/-30V.
This means EC configuration that I make loses rail efficiency about 6V on positive rail and another 6V on negative rail.
How many volts of loss is your mosfet EC? How to roughly calculate the loss (if we use the same power supply for front stage and output stage) Is it connected to the 10V or 22V zeners that we use?
Thank you
lumanauw said:
These losses are not unexpected for the EC architecture I use. Basically, the EC circuits need headroom to work. However, when applying EC to a bipolar output stage, a little less headroom will be lost because the turn-on voltage of the bipolar does not increase as much with higher current drive as it does with MOSFETs. This in turn means that the amount of error correction signal swing needed may be a bit less. However, the value you are using for your output stage emitter resistors will also have some influence on the headroom needed.
In general, I always say that if you are going to the trouble of making a very good amplifier (which you are if you are using EC), then you should always spend the small extra cost on a boosted-rail power supply for the input and driver stages.
Cheers,
Bob
Hi Bob,
I was experimenting with complimentary differential pairs a couple years ago to error correct in a bipolar output stage without knowing it had been developed before, or what is was called. No, I don't read much.
This thread has been a great way to answer questions on things I hadn't thought of. The developement is on the back burner at the moment. The material presented by many members has been very helpful.
-Chris
I was experimenting with complimentary differential pairs a couple years ago to error correct in a bipolar output stage without knowing it had been developed before, or what is was called. No, I don't read much.
This thread has been a great way to answer questions on things I hadn't thought of. The developement is on the back burner at the moment. The material presented by many members has been very helpful.
-Chris
janneman said:
Jan,
now I read it again (December1994,page1037)!
Hmmm.... I had to admit that I be mathematically a nerd!
For me its important to understand HOW something work,
the mathematically work I had to leave to others like you or G. Stochino, in my case helps a simulator.
BTW, I try fruitless to find the mentioned patent!
Heinz!
sam9 said:
There are actually two approaches I have used. The first is to wind a few turns of small-gage wire around a torroid power transformer. This is pretty easy. I often get about 0.8V/turn on a torroid power transformer, so just 20 turns for each of the two new secondaries gives me about 16V rms on each one, which is usually enough.
The second approach is to just use a small additional 10-15 VA small transformer with two isolated secondaries.
In both cases, I connect these secondaies in series with the main high current a.c. secondaries and then have a dedicated bridge rectifier to rectify and produce the boosted rail voltage.
Cheers,
Bob
Hi all
sorry not quite on the thread, but modification of a commercial transformer is not to be recommended. In the UK the power line is 240V (sorry, the electricity companies would say 230V) and more chance for things to go wrong.
I've mentioned in another thread that I prefer a couple of caps and booster diodes to boost the voltage to the pre-stages of power amps.
This is to be recommended for bipolars and mosfet stages - very often, if the VAS/CCS is not the source of the clipping limit, no phase inversion occurs and amps which might misbehave seem to behave better if the output stages do the clip limiting.
Bob has mentioned that transformers don't cost much in the US - with the recent hike in copper maybe booster caps are now cheaper too?
cheers
John
sorry not quite on the thread, but modification of a commercial transformer is not to be recommended. In the UK the power line is 240V (sorry, the electricity companies would say 230V) and more chance for things to go wrong.
I've mentioned in another thread that I prefer a couple of caps and booster diodes to boost the voltage to the pre-stages of power amps.
This is to be recommended for bipolars and mosfet stages - very often, if the VAS/CCS is not the source of the clipping limit, no phase inversion occurs and amps which might misbehave seem to behave better if the output stages do the clip limiting.
Bob has mentioned that transformers don't cost much in the US - with the recent hike in copper maybe booster caps are now cheaper too?
cheers
John
That's total BS. Quit trying to make people avoid DIYing more on a DIY forum! Most transformers in my projects were the cheapest I found with sufficient core size, and then I completely rewound the secondaries, which allows me to perfectly customize for an exact voltage, put in auxiliary windings for say control logic or fans, add electrostatic shielding between windings, a flux band to deal with any leakage, etc. If you leave the primaries alone there's no problem here.john_ellis said:
An EC model.
Having mulled over the assertion that true cancellation cannot occur with Hawksford's EC arrangement unless you have infinite loop gain, I run a SPICE sim. of the arrangement below with realistic dominant singularities (complex and real) and established that, contrary to Lipshitz/Vanderkooy, loop transmission at balance is indeed equal to error extracted, as shown here.
The output stage error in this case is the additional gain of 3 introduced by E1.
S1 and S2 are merely differential amplifiers representing the summers previously introduced here.
Run AC analysis, and plot the following expression:
-1/(1-1/(2*(I(Vi1)@1*V(x)@2-V(x)@1*I(Vi1)@2)+V(x)@1+I(Vi1)@2))
It is recommended that the logarithmic scale rather than the dB scale be used for the vertical axis in the Waveform Viewer.
Having mulled over the assertion that true cancellation cannot occur with Hawksford's EC arrangement unless you have infinite loop gain, I run a SPICE sim. of the arrangement below with realistic dominant singularities (complex and real) and established that, contrary to Lipshitz/Vanderkooy, loop transmission at balance is indeed equal to error extracted, as shown here.
The output stage error in this case is the additional gain of 3 introduced by E1.
S1 and S2 are merely differential amplifiers representing the summers previously introduced here.
Run AC analysis, and plot the following expression:
-1/(1-1/(2*(I(Vi1)@1*V(x)@2-V(x)@1*I(Vi1)@2)+V(x)@1+I(Vi1)@2))
It is recommended that the logarithmic scale rather than the dB scale be used for the vertical axis in the Waveform Viewer.