Sonore MicroRendu

The microRendu's hardware was no small task to design. It took John Swenson over a year to get it right. This often meant getting new boards produced with the smallest of tweaks to eke out the final ounces of performance. In fact very close to the data of launch, Jesus decided to throw away all the newly delivered boards because of a single design change. This change could have been made to the existing boards after the fact, but this team is all about perfection. So, out went the "production" boards and a new order was placed.

At a high level, the microRendu consists of a tiny processor module (System On Module) that's directly connected to a carrier board. The processor module contains an i.MX6 chip with a dual core processor and RAM. The processor module is attached to the carrier board via two 80 pin headers. It's this carrier board combined with custom software that separates the men from the boys and turns the microRendu into a true audiophile class component.

The carrier board contains the regulators, oscillators, USB port, and Ethernet port. Let's start with the Ethernet input and work our way to the USB output. The microRendu contains a 10/100/1000 Gbps Ethernet interface. This interface is limited to 470 Mbps due to the internal i.MX6 bus. Audiophile needn't worry about this "limitation" because 470 Mbps is still hundreds of Mbps more than is required for even the highest resolution audio files. The microRendu features signal conditioning, signal isolation, and EMI suppression on this Ethernet input in part by using a radical power network with multiple regulators between the power to the Ethernet PHY and the power to the USB subsystem. These regulators have a very high power supply rejection ration or PSSR. The PSSR is used to describe the amount of noise that can be rejected from a source of power. Readers familiar with commercial motherboards built to hit the lowest price point will understand this is a huge difference because those cheap boards don't contain much isolation between the power to the Ethernet PHY and USB subsystem. This extensive design may be responsible for some of the network immunity or lack of sensitivity I've found with the microRendu. No matter what I do prior to the Ethernet input of the unit, the sound remains the same. Even using CAT7 shielded cables that break the inherent galvanic isolation of Ethernet by using connected shields on both ends.

The carrier board features a very low jitter oscillator that feeds the hub chip, PLL, and clock network that has anything to do with the USB subsystem. The other on-chip oscillator is used to drive the processor and memory. This is where the software customization comes into play. The design team was able to shut off the processor module's internal oscillator circuit and externally clock the chip from the much better oscillators on the carrier board. Just like externally clocking a DAC, Sonore changed the reference clock of the PLLs to point to the external clock that's fed with the low jitter main clock.

The microRendu has extremely low ground noise due in part to its design and linear regulators, but also because everything not used for audio purposes has been eliminated or completely shut off. There are many noisy processor circuits not simply unused, but totally shut off.

The USB output of the microRendu is equally as special as anything else contained on the carrier board. Most, if not all, commercial motherboards contain extremely noisy DC to DC converters and switch mode regulators. Thus, even though an expensive linear power supply may be used on the outside, the power signal is going through a gauntlet of garbage once it hits the motherboard on its way to the USB output that feeds power to the USB DAC. It's like running a linear supply though a terrible switching supply in order to feed one's DAC. This isn't the case with the microRendu. The incoming power goes through a linear regulator on its way out the USB port and on to the USB DAC. This ultra clean path completely avoids switching regulators.

In addition the design of the microRendu's USB architecture generates a completely new USB data signal and is highly optimized for signal integrity and impedance matching. To quote John Swenson, designer of both the microRendu and USB REGEN, "The microRendu does contain a circuit which is essentially an improved REGEN. There is no need to add an external REGEN between it and a DAC."

The microRendu requires between 6 and 9 volts of power. During its approximately twenty-five second boot time it peaks at about 0.4A and during regular playback settles in at about 0.2A. Using the forthcoming Sonore 7V power supply provided for this review by Sonore, the microRendu uses 1.4 watts during playback.

One of the beauties of the microRendu's design is its' separate power supply domains. The individual supply domains receive the appropriate regulation scheme for their functions. The processor domain uses a high quality switching regulator since it requires low voltage at high current.

The PLLs that generate clock signals for many different systems use a single ultra low noise regulator, while the USB subsystem uses three ultra low noise regulators.

One additional note about the hardware design. One of the first items I noticed upon receiving my unit was the SD card. This card is required, as it's loaded with the operating system. I figured that storing the OS on FLASH (eMMC embedded MultiMedia Card) or NVRAM would have been a better option. It's a good thing I didn't attempt to design the microRendu because my figuring was a bit off. According to John Swenson, "The i.MX6 has three memory subsystems, the DDR3, which we need to use for the main memory of the system, a very small simple, low power SD card subsystem, and the generic everything else memory subsystem. The later is what you use for NVRAM, flash chips etc. It is a large complex system designed to run very fast. This uses a lot of power and generates a lot of noise in the chip. "

The SD card controller is slow, low power and generates very little noise, and on top of that has its own power supply pins on the chip which cuts down even more on the noise it generates. So by using the SD card rather than something like NVRAM I can drastically cut down on the noise in the chip. There are also things like SSDs, but they all need some form of high power bus to talk to (SATA, PCIE etc), which would mean I would have to turn on those subsystems.

On the reliability front, I have actually found that using on board FLASH or NVRAM is actually less reliable. I have worked with several embedded boards over the last few years that have had flash chips, that have had problems far more often than ones that run straight off an SD card. I think it has to do with where the controller is. With SD card the flash controller is built into the card, the software doesn't have to know anything about that. The inexpensive flash chips used with these systems do not have a built in controller, they require the OS to deal with the issues specific to flash memory. Linux has some good code for this, but if something happens with the kernel during runtime, it is very easy for the flash to get corrupted. I had one board that if power went out during boot the flash was guaranteed to be corrupted."

The SD card simply clicks into the microRendu and sits there without requiring any user intervention. If the OS is somehow corrupted or there's a problem with the unit, a new SD card can be placed into the slot very easily. I like this option much better than sending the unit back to Sonore to get re-flashed if onboard solid state storage was used.