jcx said:
Stan Curtis, the designer of the successful Rotel producs, once in an interview said that his particular way of manipulating the amp response to the dynamics of the speaker impedance greatly contributed to that succes. Reviewers were irritated that his amps didn't measure particularly well but boy they loved to listen to them!
BTW your link above appears to be dead.
Jan Didden
http://www.diyaudio.com/forums/showthread.php?postid=152392#post152392
word did the wrong thing with the paste - but I need the spell checker
word did the wrong thing with the paste - but I need the spell checker
janneman said:
Hi Jan,
Now THAT idea is interesting! (Do you have any idea what or how he was manipulating?)
Why do we design amps to have flat responses, when speakers' impedances and frequency responses are so variable?
Speakers are said to be the weakest link in the audio chain, by far. Their plots for impedance and response versus frquency are practically frightening. Yet I almost never hear anyone talking about doing much about it, at the _amplifier_ end of the cables. But, since speakers apparently have inherent limitations that are difficult to "design out", and amplifiers are very flexible in comparison, it seems to make sense to use the amplifier to compensate, rather than wait for perfect speakers to be designed.
I suppose that a flat-response amp is the best _compromise_, if it is to be usable with "any" speakers, which have characteristics that are unknown, when the amp is designed.
But maybe we should also be thinking about designing amplifiers that can sense the speakers' characteristics and automatically adapt to them.
(Ha! I just thought of a name for those "adaptive" amps: "Chameleon". They can change their "color" to match to the speakers. 🙂
There "must" be an easier way to do that, than to have something like a spectrum analyzer and maybe a network analyzer in every amp (not to mention HOW the amp might then be "adjusted"). [Or... maybe there's not. :-( ]
I recall, from my first course in "Probabilistic Signals and Systems" (Good 'ol EE302, at Purdue, back in 1978!), that it is possible to inject a short burst of a very low level of pink noise into a system, WHILE the system is operating, and use the output to calculate the system's transfer function. (This was suggested, at the time, for use with large industrial processes that couldn't be shut down just to take measurements needed by feedback-control-system designers, for example.) That might be quite complex to do, too. But maybe there's some way to radically simplify it, or one of the other methods, for a narrower purpose.
Alternatively (and since such adjustments would probably be "a one-time thing", for most amps), maybe there should at least (or in the meantime) be some "manual" way to set some frequency-dependent shaping of an amplfier's characteristics, even if it were rather "coarse", as long as it could possibly give a significant improvement in many cases. It would be "nice", for that, if there were several known categories and ranges of shaping that might be needed, due to known characteristics or limitations that might be common to certain types of speakers. Or, is an ordinary "equalizer" good-enough? Obviously, I don't know enough about any of this and am just "shooting from the hip". But it might be interesting to do some research. (And I'm sorry of all of this has already been done, or discussed-to-death, before.) However, and I guess this seems obvious, I HAVE noticed significant differences in spice-simulated amplifier performance, when the speaker (or speaker cable) spice models were changed. And of course that was at the amplifier output (and internals), not just the speaker's output. And I could then change the amplifier design to compensate. At the time, it was a bit frustrating, since I didn't know exactly WHICH load to optimize an amplifier for, if I was trying to "push the envelope", even though I usually could back it off a little and get better _average_ performance, i.e. for more different types of loads. That actually reinforces the point: It should be possible to have a better "best" performance level, if a design could be optimized (_OR_ adaptively changed, "in the field") based on somewhat-more-specific knowledge of speaker characteristics, rather than being designed to work with a broad range of speakers (as Stan Curtis has apparently already verified).
Just some "food for thought".
(Sorry about getting off-topic.)
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
gootee said:
I know of at least one active studio monitor manufacturer (I don't recall who but a quick google should spot it) who includes a calibration microphone and software to compensate overall response.
In theory it is possible to pick up the system response including room acoustics at least for a selected floor spot, synthetize the reverse transfer with a DSP and insert it before the amplifier so as to get a fairly flat overall response.
As is usually the case, results may vary depending both on execution and ease of use, but we may be seeing more of this kind of stuff in the near future.
Rodolfo
ingrast said:
Hi Rodolfo,
It could get quite interesting. Maybe someone will devise a way to use tiny wireless microphones that we could wear on each of our ears, and have the DSP try to optimize what we hear.
(On a lower-tech note: ) I was also wondering if using a pair of remote-sense wires, from the speaker terminals back to the amp, might be able to do some good.
(Maybe slightly more-interesting: ) And maybe some microphone (or cone-motion sensor) wires could go from the speakers back to the amp, too, to try to get the speaker inside a feedback loop.
I know that that's been done with woofers (at least), already (much-discussed, here, too), with tiny motion sensors on the speaker cones themselves. And Analog Devices just came out with a "50X better" tiny 3-D accelerometer, I think. So maybe that will be useful. (Check out the "MEMS" devices, at analog.com. Inertial guidance [and other coool stuff] is cheap and easy, now!)
With enough "good-enough" sensors for feedback, almost anything seems possible, now.
Since technology has advanced so much, maybe I should go back to designing automatic-weapon-delivery systems. It should be even more fun than it used to be. 🙂
It's Friday evening, here: Cheers!
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
To be at least a little on topic:
Silbersand makes active speakers which have motion sensors and feedback systems on all drivers, not only the woofer. The driving amps run in current mode (VCCS) not voltage mode, derived filters are used for xovers, and other interesting features.
http://www.johannes-krings.com/uploads/krinks2/702/2/firmenprofil.pdf
DSP'd room correction... one of the more successful aproaches comes from Lyngdorf. They take an average of the room response at various positions, not only the actual listening position. This gives way smoother correction. True (and perfect) single-point-only correction works great in theory but we have two ears and usually don't listen with our head in a vice.
http://www.lyngdorf.com/index.php?option=com_content&task=view&id=21&Itemid=43
Combine this with Bongiorno's Trinaural Processor or a simlar vectorizer for driving three speakers and you will have a stereo setup hard to beat, if you have your room treatment right in the first place.
- Klaus
Silbersand makes active speakers which have motion sensors and feedback systems on all drivers, not only the woofer. The driving amps run in current mode (VCCS) not voltage mode, derived filters are used for xovers, and other interesting features.
http://www.johannes-krings.com/uploads/krinks2/702/2/firmenprofil.pdf
DSP'd room correction... one of the more successful aproaches comes from Lyngdorf. They take an average of the room response at various positions, not only the actual listening position. This gives way smoother correction. True (and perfect) single-point-only correction works great in theory but we have two ears and usually don't listen with our head in a vice.
http://www.lyngdorf.com/index.php?option=com_content&task=view&id=21&Itemid=43
Combine this with Bongiorno's Trinaural Processor or a simlar vectorizer for driving three speakers and you will have a stereo setup hard to beat, if you have your room treatment right in the first place.
- Klaus
KSTR said:
Klaus, I wish you lived next door to me, so we could talk. But the next time I'm in Berlin, I'm going to knock on your door, with spirits in hand. (The last time I was in Europe was in 1975, so you might not have to worry, too much, for a while. 🙂
- Tom
Well,I have a DEQX that lets me first measure and auto-correct the response. I have a difficult room but this thing cures it all. Expensive though.
Jan Didden
Jan Didden
myhrrhleine said:
There is no EC in the output stage mentioned here. Schematic:
http://www.diyaudio.com/forums/showthread.php?postid=1296735#post1296735
(ufortunately a bit OT both here and there)
PMA said:
<sigh> the curse of multiple open windows.
My apology.
I was asking about those circuits which use EC.
I've only seen it with 1 basic transistor.
I was wondering about current source loads or similar.
anyway, sorry for the oops.
PMA said:
This looks remarkably good. Remind me again how much idle current are you running?
Best,
Bob
Bob Cordell said:
The iddle current of the measured sample is 1.55 - 1.6A.
Power supply +/-27V (2 x 10 000uF). Max. output amplitude below clipping +/-22.5Vp. IM measured at +/-18Vp into 8ohm.
parametrized open-loop EF dist./harm. vs bias graphs
(this might rather belong to the SPICE thread, but I didn't want to disturb the current specialist discussion going on there. Also, this is an expansion of a detail posted in this thread about two weeks ago, page 63, ...)
Inspired by the work of Mr.Cordell (see post #1574), I simulated an open-loop push-pull EF stage, using andy_c's models for the OnSemi ThermalTrak BJT's and made a set of 1kHz distortion/harmonics (THD, H2...H15) vs. emitter R bias voltage graphs (21 in total), running bias from 0.5mV to 50mV in 0.5mV steps.
One parameter was the emitter R, stepping the complete E6 range from 0.047R to 0.47R. Another parameter was drive voltage, I chose 0.5V, 3.0V and 15V, to cover a range from small signal to large signal. Supplies were +-20V and I used 4.7R base stoppers, load was 4 ohms.
The NPN and PNP bias was generated individually to make any output DC offset small and with an extra control loop I generated the corresponding base voltages for the stepped emitter bias voltage.
I also checked LTSpice's sine generator distortion, which settled at approx -163dB THD and around -180dB for the harmonics with a reasonable time step. I found the need to switch to the alternate solver which is more precise, otherwise the sim would hang.
Please find the data attached below in the .ZIPs (6 chunks). Included is also the sim file and the BJT models. Also available as a complete .ZIP (and with all graphs in one PDF) which I can email on request.
I hope this is useful. I personally got some insight on how the harmonics tend to shift with different parameter changes. Maybe they do so even in real circuits...
- Klaus
chunk 1/6
(this might rather belong to the SPICE thread, but I didn't want to disturb the current specialist discussion going on there. Also, this is an expansion of a detail posted in this thread about two weeks ago, page 63, ...)
Inspired by the work of Mr.Cordell (see post #1574), I simulated an open-loop push-pull EF stage, using andy_c's models for the OnSemi ThermalTrak BJT's and made a set of 1kHz distortion/harmonics (THD, H2...H15) vs. emitter R bias voltage graphs (21 in total), running bias from 0.5mV to 50mV in 0.5mV steps.
One parameter was the emitter R, stepping the complete E6 range from 0.047R to 0.47R. Another parameter was drive voltage, I chose 0.5V, 3.0V and 15V, to cover a range from small signal to large signal. Supplies were +-20V and I used 4.7R base stoppers, load was 4 ohms.
The NPN and PNP bias was generated individually to make any output DC offset small and with an extra control loop I generated the corresponding base voltages for the stepped emitter bias voltage.
I also checked LTSpice's sine generator distortion, which settled at approx -163dB THD and around -180dB for the harmonics with a reasonable time step. I found the need to switch to the alternate solver which is more precise, otherwise the sim would hang.
Please find the data attached below in the .ZIPs (6 chunks). Included is also the sim file and the BJT models. Also available as a complete .ZIP (and with all graphs in one PDF) which I can email on request.
I hope this is useful. I personally got some insight on how the harmonics tend to shift with different parameter changes. Maybe they do so even in real circuits...
- Klaus
chunk 1/6