This review page is supported in part by the sponsors whose ad banners are displayed below

Tech talk with Roland Krammer. "First of all the PCB is completely new. That's because half the old boards developed an eventual short between VCC and GND. Now our multi-layer board is made by Ilfa of Hannover, final board assembly by Taube Electronic of Berlin. These are two small hi-tech companies I can rely on because they already worked with us at Unitel on boards for Pandora International Ltd. That was a 26-layer PCB with a 1.500-pin ball-grid array Xilinx FPGA and assorted devilish stuff. I have tested five CFA-1.2 prototype boards for a long stress period with zero failure. The standby power supply now also is a switch-mode variant which supports 115-230VAC like the main PSU. We changed the display backlight to amber because it turned out that white LEDs have a life span of only about 1000 hours whilst amber does 100 times that. Sadly modern marketing isn't honest about this stuff so we had to learn the hard way. The CD input now accepts a max input voltage of +16dBu for machines with an output voltage over 2V. To compensate, our setup menu can cut this by up to -6dB on this input.

"The defeatable tone controls are now limited to bass and treble. The controller software is more mature and the bridge controller has more gain. The CFA-1.2 runs hotter as we'd say with a studio mixing console and the sound is tangier and livelier. Little things learnt over time were adjusted like the surge protection varistors. The standby/active LED is now a perfectly intuitive red/green. The main power supply was bumped to 320 watts and is no longer from Traco but Mean Well. This 87% efficient supply outputs 36VDC at 8.8A of current. Its global input tolerance is 88-264VAC at 47-63Hz. We added a variable out and the former banana-only terminals changed to NextGen WBT. There's even an optional full-metal remote from Bent Audio if someone prefers that to our standard plastic-body wand.


"Some people wonder about DC speaker protection. They should remember that the CFA-1.2 is a bridged amplifier with 18VDC on the outputs. This results in a max 10mV offset between the +/- terminals or at the most 12uW. The power supply is protected against a continuous short with fold-back regulation. The output stage itself will withstand brief shorts but not continuous. The circuit design is very robust. We tested it into 3 paralleled low-impedance speakers for 4 weeks continuous and measured 1Ω DC resistance. During this test the CFA-1.2 only got slightly warmer than normal.
"As for relays, the phono stage switch between MM and MC uses one as does the input multiplexer, volume control and switch which activates the power supply. When asked whether our output transistors run in class A, I reply that this is a current feedback amplifier and as such difficult to compare with conventional complementary emitter followers. It's a very different circuit concept with very different behavior. With common emitter or collector followers as used in 98% of all cases one talks of class A or B which differ only in bias current. In class A the quiescent current is so high as to run its NPN and PNP output transistors in their linear region. The main difference in a current feedback amplifier is that one of the transistors is always without current whilst the active one is already in its linear zone. That latter part mimics class A but fails to meet the full technical textbook definition. Yet for the same reason current feedback efficiency is higher than with a collector/emitter follower.

CMA-1 monos

"About our choice of output devices, we began work on the CFA in 2000 and production in 2006. At that time we only had Mosfets available for our circuit. That makes it harder than bipolars to match N- and P-channel devices. The essential parameters for Mosfets are their gate threshold voltage and forward transconductance. Between 2006 and 2012 Mosfets improved to the extent that the difference of gate threshold voltage between P/N-channel parts was down to 200mV. Forward transconductance grew from S=9 to S=>19. Hence the sound of the CFA continuously improved with these better parts.

Freshly machined chassis for the CD player

"Output impedance is <0.06Ω. For the fixed and variable outputs it is 27Ω up to 1MHz. We don't have a volume control bypass so for use with a preamp, it's best to use the CFA at max volume. Relative to THD and a circuit sans global NFB, I would say the following. A Darlington transistor circuit combines two transistors such that we get higher gain than the single transistor provided but pay for it with reduced bandwidth and longer lag times. With current feedback a NPN bipolar transistor connects to a P-channel Mosfet and vice versa for the complementary half. This creates one compound transistor with higher gain, higher linearity and higher bandwidth. One could never build a linear amplifier with just a Darlington transistor. That would generate >15% distortion at 5 watts. With current feedback one gets 0.14% THD. Such circuits thus behave different from conventional circuits with rising distortion. The ratio of harmonics remains consistent, with only a slight increase at higher frequencies. In our case it's 0.1% 2nd order, 0.05% 3rd order and 0.04% 4th order.

"Warm-up time to my ears is about 15 minutes. Standby consumption is a negligible 15mW. Our electrolytic capacitors are derated for more than 5000 hours at 105°C. Some worry that switch-mode supplies are harder on their parts due to their rapid switching. It's true that much is required of capacitors in such a supply but that's mainly on ESR (the internal resistance) and low terminal inductance. In linear supplies such capacitors must compensate for far higher ripple current—several amperes at 50/60Hz—which puts a far higher load on the caps. In our case it's a few 100mA at 100kHz. The real difficulty with switch-mode supplies is how to attenuate their switching transients. The design appeal of linear power supplies is their simplicity. SMPS are more complicated. That said, the SMPS industry has now had more than 20 years of development under its belt. And the undeniable advantage of switching supplies is regulated output voltage. We learnt 16 years ago how much sound quality gets sacrificed by unregulated supplies. We were thus faced with the choice of using a linear supply with a voltage regulator; or a SMPS. An analog power supply with regulator is a waste of energy. We thus opted for the switcher and haven't look back."

About the forthcoming CFA-2, Roland had this: "The CFA-2 line will include the same hardware but with a really good heatsink and some improvements for a little more power. Above all though we want to satisfy the measurement freaks and reduce harmonic distortion to move deeper into so-called class A operation. With this line we also want to split into different components like a pre and power amp. We developed completely new amplifier technology with overall feedback now but the same sound characteristics, just higher precision and adjustable damping factor. My partner is a purist and into feedback-free systems. In his opinion high damping factor only creates problems. Hence I was very pleased when after the first test run he agreed that the high damping factor sounded better than the low. When I say it's new technology I'm serious. It's really a new but logical consequence of the CFA platform which essentially consists only of the power transistors at relatively low distortion. Now I have improved the front end a lot but (and this is big) I don't need high gain as would be the case for conventional input stages. The power buffer already generates 15dB of voltage gain. My input stage doesn't have to compensate for all the nonlinearities of the output stage, just reduce its 0.14% THD. With my 3MHz bandwidth that's simple yet retains all the established sonic benefits of the CFA design platform."