Then I suspect that you Icepower having huge damping factor at low frequencies (about 4000) overdamped your speakers making very lean bass? |
Orpheus10 - I have Rowland Model 102 based on the same Icepower modules as mentioned Bel Canto S300 (and M300). It is very smooth and transparent (nice liquid midrange) but also very revealing. With new speakers (Hyperion HPS-938) I can finally hear smooth natural sounding sibilants. Rowland makes very nice preamp - Capri in the same case, designed as a set.
Here is review of the set: http://www.audioenz.co.nz/2007/rowland_capri-102.shtml |
Afc - Ref1000 can drive 2ohm delivering 40A. Size of power supply has nothing to do with it since 2" toroidal transformer at 100kHz can pass as much power as 10" toroidal transformer at 60Hz. In addition voltage doesn't drop since it is line and load regulated. Traditional "linear supply" needs a lot of caps because of voltage drop and necessity to filter out 120Hz (difficult to filter out and audible) vs. SMPS high switching frequency (non-audible and easy to filter out). A lot of caps in parallel have a lot of inductance - not desirable since it is in series with the speaker. SMPS is in reality class D amp (origin of class D) and got bad rap only because of cheap computer power supply implementations.
"A Class D amplifier has one output stage fully "off" for half the cycle"
High efficiency of class D is not caused by switches being OFF. Speaker is always connected by Mosfet switches (there is no OFF) but direction of current changes. Icepower uses single supply and full Mosfet bridge reversing output while Hypex uses half bridge connecting output to positive and negative supply. Efficiency comes from the fact that Mosfets in ON state have very low resistance (in order of 20mohm).
Replacing in class D analog voltage with analog time (duty cycle) allows to get rid of nasty problem of output transistor nonlinearity that in class AB is corrected by deep negative feedback that causes transient intermodulation (TIM) and exaggeration of odd harmonics (unpleasant bright sound).
I don't understand mentioned output voltage offset of class D amps. It has as much offset voltage as any other class amp without output capacitor (it uses feedback/servo to zero output). Self zeroing is probably responsible in my Rowland for 5Hz (and not DC) bottom frequency range.
|
Pugstub - It might be related to fact that Icepower's full bridge places half of supply voltage (30-40V) on both speaker terminals. Non-powered speaker sees only voltage between terminals (difference) but ground referenced powered speaker might not tolerate such common mode voltage. They also mentioned in Stereophile that some electrostats don't work with Icepower for the same reason. Half bridge schemes like Hypex don't have this problem and can be even bridged while Icepower cannot be (already is). |
"With medium efficiency speakers the MEASURED peak voltage is above 100v."
- I don't understand it. What speaker efficiency has to do with the voltage. Icepower 1000ASP voltage is regulated always 80V and since it is bidirectional it becomes +/-80V.
Switching power supplies generate high frequency noise that is easy to filter out and non audible. For this reason Rowland uses switching supply in Capri preamp and he is very noise conscious guy (he used batteries in preamps).
Linear power supply is also switching - current comes in narrow spikes of very high amplitude producing high frequency noise in addition to 120Hz noise (difficult to filter out).
Now - In order to become 1/4 wave antenna for 100kHz SMPS frequency wire/cable has to be 2460ft. Below 1/10 of the wave (984ft) antenna practically doesn't work. Small amount of the amp's carrier frequency (1%) appears on the speaker wires but at 500kHz it needs 197ft of wire to become 1/10 wave antenna. |
Mike - Problem with switching power supplies is that most of them are designed to save money and space and not to improve performance. REF1000 for instance in improved version REF1000m has additional power supply with a lot of capacitors on separate board. I believe that with the same resources (cost and space) better PS can be achieved in switching than linear technologies.
I remember that there was an argument that high frequency can modulate audible frequencies on nonlinear (at this range) transducer - tweeter. Argument is valid under condition that tweeter membrane can move at 500kHz - not likely.
Larger Icepowers have a little lower carrier frequency than small ones and therefore lower -3dB frequency (50kHz vs. 60kHz). It is because stronger Mosfets are slower. Wide frequency range is necessary to provide smallest phase shift in audible range. Even with 60kHz amplifier shows about 30deg phase shift over 5Hz-20kHz.
This is first avalanche of class D amps (remember first SS amps?). Next generations will have faster Mosfets, higher frequency range, better power supplies and better noise filtering. I see it as technology of the future.
Magfan - this Zobel, can be improved as well. In Icepower it is simply common mode choke two filtering capacitors and one resistor. This Zobel network is responsible for time/power limitations and available power at higher frequencies. In practice musical power is only few percent of peak power (unless one listens to sine-waves)and high frequency energy is minimal. Should I pay for larger Zobel toroid to allow high frequency power that would make me deaf? Thanks for the link to IRF primer. I read opinion that full bridge sounds like very good tube amp while half bridge sounds more like high quality class AB amp (less tubey). |
Coxhaus - I'm surprised since I know audiogoner who drives with Icepower speakers with nominal impedance of 1 ohm.
Isn't Spectron a class D amp as well? There is probably more to it than just impedance alone. |
Mapman - you're right about low input impedance. My Rowland has extra board (with THAT1200 intrumentation amp) to increase impedance to 40k. |
"The ASP1000 module is worst in this respect being what....about 1/2 db down at 20khz w/4 ohm load?"
Magfan - you're right (just checked datasheet graphs) but 8 ohms have too much down and up by 1/4dB starting in less than 10kHz. My Rowlands 200ASC is a little better - it dives down about 1/4dB but comes back to 0dB at 35-40kHz. Phase shift over 20kHz is also smaller - effect of higher carrier frequency.
Mapman - I was turning everything OFF but now I have Furman Elite 20 PFI conditioner with non-sacrificial over and undervoltage protection and keep everything ON all the time. It clamps at much lower voltage than traditional MOV protection. |
"I'm convinced the better Class D amps are those that pay special attention to the power source on-board."
Mapman - That would explain better reviews for BC Ref1000mkII
vs. original Ref1000. Same Icepower module (1000ASP) with built-in SMPS but new version has additional supply module/board with bunch of electrolytic capacitors. |
It is also possible that new generation is just only a little bit better (like BC Ref1000M with extra power supply caps) and "new generation" allows some critics to reevalute (read: save face) since others scored them much higher. For instance REF1000 was evaluated bu Martin Colloms for 10.5 points while Krell got 130 - that's less than 1 of 10. Ref 1000M is Stereophile class A. Sound like two completely different amps but in reality both use exactly same Icepower module with built-in SMPS. |
6550 - The problem here is that SMPS is small. It is hard to believe that something smaller can be better. It is line and load regulated and basically the same thing as class D amps that win Stereophile class A classification.
It is difficult to change traditional way of thinking (you being example of it) but some great companies (like Linn) used SMPS for class AB amplifiers for a long time. New $48k Rowland class AB amp uses SMPS. As for being better suitable for the car since it is always used there - you can make the same argument for transistors (tubes for home, transistors for cars). I'm pretty sure it was the case at the beginning of transistor amps in audio.
Why not to listen first and learn what is inside later? |
Psag,
Class D and SMPS operate on the same principle (PWM). Class D was invented when designers of SMPS showed that response to reference voltage is so fast it can play the music. |
