Welcome! Need support, you got it. Or share your ideas and experiences.
Return to Planar Speaker Asylum
71.131.186.62
In Reply to: Re: Control of Maggies posted by morricab on October 11, 2006 at 08:35:09:
my comments are based on my experience with the Gilmore Raptor switching amps, which have no feedback yet have an output impedance below 5 milliohms.YMMV, especially with amps that employ massive amounts of global negative feedback. The uniformity of output impedance over frequency depends on the details of the feedback network, and some amps may deliver good measurements yet fail to deliver convincing sound.
Magnepans have driver interactions that are audible and identifiable when reduced. I was able to clean up the sound-stage clarity for mid-bass instruments by adding R-C filters to the outputs of the bass low-pass crossover sections. The stock design has an inductor as the last reactive element, so the bass panel sees a high impedance for resonances above the pass band. The R-C filter helps by damping these resonances.
Adding the Gilmore amps gave further improvements along this line, and at higher frequencies as well. This is why I believe that Magnepans need high damping factor amps, and that it is important to use amps that do not suffer from the means used to get the high damping factors.
Speakers can be "musical" while suffering from response anomalies that obscure certain kinds of detail. Most systems do not deliver all the sound-stage information present on good recordings, so a little loss of precision in less-obvious tonal bands is difficult to pick out with casual listening.
Follow Ups:
Interesting. I was under the impression that all Class D amps had to use negative feedback as fundamental to their proper operation? Do you know if that 5 milliohms is constant over the whole audible bandwidth?One of the problems I brought up in PHP a while back with high feedback amps is that their sonic character can change depending on the speaker they are connected to because the EMF from the speaker can find its way back to the input (through the feedback loop) and then get reamplified. This was pointed out by Otala in the late 70s early 80s and he called it IIMD (interface intermodulation distortion). The more reactive a speaker the worse the situation. So it if a Maggie is not so reactive (the panels themselves and the ribbon should be nearly resistive but the xover is pretty reactive I guess) then the character of the amp will not be so greatly affected.
If you look at the recent article in Audioholics they show pretty conclusively that a high damping factor is not necessary for good speaker control (from a frequency response POV). I agree that damping a panel resonance is a somewhat different case but my electrostats have the same problems in this regard and I have found that I get plenty of control from some amps with damping factors as low as 5. Some of the amps I have tried with higher damping factors made the bass sound too thin in fact.
I'll ask Mark, but he is busy with the Rocky Mountain show coming up next week, and it may be a while before I get hold of him.The statement, that no feedback is used, is taken directly from the Glacier Audio site. The specification for output impedance is at 1 KHz.
I looked there and didn't find that statement.
If you see a plot of the output impedance as a function of frequency you can usually tell immediately if there is negative feedback or not. If there is none then the output impedance is quite stable over the whole frequency range. If there is a lot of feedback you will see a sudden drop at higher frequencies.
used". It's in the bottom quarter of the page in reference to the Gilmore digital amplifiers.
Ok, now I see it but I wonder at this statement and why as a heading the words No Feedback and the explanation is no coupling caps used? Aren't these two different issues? Call me skeptical...
A full-bridge PWM amp can be built with a single Vdd supply. The output of the low-pass filter is the audio-frequency signal with the switching frequency artifacts removed. However, it is at one-half Vdd above ground.With zero audio input signal, the voltage difference between the (+) and (-) speaker terminals is zero. However, both terminals are at 1/2 Vdd above the input audio reference (audio ground). The amplifier case is generally connected to AC safety earth, and the audio ground is at a negligible DC voltage with respect to safety earth.
If feedback is to be used, this DC offset must be blocked. Just as in a multi-stage tube amp, the DC can be blocked by capacitors, use of a transformer, or balanced out by resistive coupling to a separate DC supply below audio ground. Each method has its own drawbacks. Capacitor coupling has the least of the drawbacks overall, but the capacitor sonic character is imposed on the feedback signal and necessarily colors the output of the amp. It would cause phase shift according to the natural output impedance of the switching stage and filter, but if this is small, the phase shift would be negligible.
As I understand the Glacier site statement ("No Feedback - No coupling capacitors used"), the Raptor amps do not use feedback, and avoid the use of coupling capacitors that would color the sound. I believe they do use capacitors in the low-pass filter, and I will ask Mark to clarify this when time permits.
Hi Al,
Well I understand, in principle at least, how Class D amps function and to my knowledge they all needed feedback, thus my surprise. I am not sure how they are filtering the ultrasonic crap but usually you need an inductor and/or capacitors to make a low pass filter. So there may not be a coupling cap in series but there is for sure an inductor (not necessarily better) at least.
I spoke with Mark about your questions. His intent with the Glacier web site comment was that the amps do not use global feedback in the audio domain, and that all the stages are direct-coupled. The only capacitors used are in the power supply filters and the output low-pass filters.These amps (including switching amps in general) use some form of digital control of the dark time, which is the time between one polarity of output transistor turning off and the other polarity turning on. Ideally, this time would be zero. However, due to the differences in available power NMOS and PMOS channel characteristics, including changes with ambient temperature variation, passive setting of the dark time results in too much distortion if it is too large, or the risk of output transistor destruction if it is too small and goes negative.
Thus, one could call this control a form of feedback. However, it is not applicable to the analog domain, and does not relate to the output impedance versus frequency characteristic.
The Raptor amp output impedance is specified at 50 milliohms (I may have been incorrect in another post), and actual measurements stay below 25 milliohms over the audio band.
either explain or correct this web page that states that the Raptor contains no transistors and no capacitors. It does not inpire credibility in his product.http://www.glacieraudio.com/new Glacier Audio Site/discussion2.htm
I appreciate this discussion between you and morricab. My take is that Mark is using a pretty standard Class-D (PWM) module is his product. The feedback is around the comparator and output section like everyone else does it. You can argue, reasonably that this is not the same as ordinary feedback in the strict analog domain sense.
What Mark appears to do, which is different, is that he replaces the analog front end (analog amplifying stage) that everyone else uses with a DAC/DSP/ADC front end. Pretty clever idea, because you can claim no feedback and you can also apply any frequency or phase correction in the digital domain to compensate for problems elsewhere. But I'm just guessing.
Cheers,
Graeme
Sorry about that.
Ok this makes more sense. THere are caps on the output as part of the HF rolloff. Not quite the same as a cap in series to block DC since the caps should be shunt to ground and the inductor(s) is in the signal path.I don't know if some other Class D amps are using "traditional" global negative feedback or not but the latest stereophile review of a tri-path amp shows a rise in distortion with increasing frequency, which is characteristic of global feedback. I would be interested to see some raptor measurements to see if it is possible to tell what the dead time control is doing.
This post is made possible by the generous support of people like you and our sponsors: