Alright, I don’t expect you to actually slog through the whole video. However, if you did get through the first five minutes or so then I’m fairly confident that you have a pretty good idea of what a Field Programmable Gate Array (FPGA) is now. But to recap, an FPGA is a collection of controllable logic blocks (CLBs) that are all interconnected via a high-speed bus. In it themselves these logic blocks don’t do much. But when strung together, they can be used to perform all kinds of interesting computational work. So let’s say you want to program your FPGA to emulate a general purpose microprocessor like the Intel-based one you are using right now to read this review. You’re in luck, you can! Or perhaps you want it to perform the same functionality as an Application Specific Integrated Circuit (ASIC), say like Bitcoin mining so you can make millions of dollars by the time you actually finish reading this review? No problem. Or maybe you just want to write your own Digital Signal Processor (DSP), like an ESS SABRE, and use it in your own custom DAC? You guessed it, you can do that too. And that’s the beauty behind using an FPGA – it’s like being given your own computational Lego set that you can piece together to create whatever kind of application your heart desires.
But the intrinsic flexibility of FPGAs over DSPs also has its downsides. First, DSPs have historically been a more power conscious choice over FPGAs since their silicon can be better optimized for their specific computational workload. Second, FPGAs are simply more complex devices than DSPs. For example, let’s say you are looking to create your own custom DAC. Then the simplest route would be to pick some off-the-shelf DSP, slap it into your circuit, and feed Pig Destroyer through it. Done. With an FPGA however, you would have to go about writing all the custom code to program all of those CLBs yourself. That is no easy feat since an FPGA is typically programmed in some kind of hardware description language instead of a high-level language like C or Java. And once your code complete, you then have to suffer through this elaborate testing cycle which includes timing analysis, model simulation, and various other verification methodologies just to verify that your code actually works. So if time-to-market is critical, an FPGA based product is probably a non-starter from the get go. Finally, FPGAs almost always have higher per-unit costs because they have more transistors, and thus a bigger footprint than your comparable DSP. Remember, an FPGA vendor doesn’t know how a potential customer will use their chip, which means they typically stuff in it a lot more functionality (read: more CLBs and I/O pins) than one might need for a certain application. DSPs on the other hand, again, know exactly how they are going to be used, and can leverage that fact accordingly.
So why the heck am I telling you any of this in the first place? Because I feel that in order for you to better appreciate a product by UK based Chord Electronics, you need to first understand a little about FPGAs. As you can see, designing products around an FPGA requires a lot more expertise outside of traditional analog circuit design, which is why most of the products we are all familiar with contain off-the-shelve third-party DSPs to handle D-to-A duties. But can an FPGA based product sound just as good, or even better than a traditional dedicated DSP based design? Read on and find out.
Striking The Right Chord
My first experience with Chord was with their now discontinued DAC64 product. Now don’t hold me to the fire since it was a long time ago, but I do vaguely remember that the DAC64 had two major issues. First, it had a problem with buffering whereby there was a noticeable delay between the moment you pressed play and the metal hitting your ears. And two, and most importantly, the DAC64 was expensive (~3k), or at least seem so at the time. But despite all that, I still walked away extremely impressed since the unit looked and sounded gorgeous.
Related Pages:
Audioengine HD6 Powered Speakers
Since then, Chord has come out with a number of equally sumptuous offerings. Probably their most well known product is the Hugo, an integrated headphone amp/DAC now based around Xilinx’s top of the line 45nm Spartan-6 chipset that has built a reputation for delivering reference level sound in a somewhat portable package. But again, like all FPGA based designs, it comes at a price. The standard Hugo is $2,195 list while its bigger desktop variant, the TT, tips the scale at around the $4k mark. Not outlandish for this kind of performance by any means, but certainly not priced for your budget conscious audiophile either. Chord of course understood this, and decided to set out and design a product that would target a more younger demographic that wants great sound but on the go. It had to be truly portable as well as play nice with all manner of smart phone. And most importantly, had to have an attractive price point. So what did Chord come up with? Meet the Mojo ($599).
Meet The Mojo
First off, the Mojo is indeed a true portable device. No, it’s not dongle tiny, but at about the size of a pack of Altoids, it easily fits in the palm of your hand. But despite its relatively small stature, it features a myriad array of I/O ports, including optical TOSlink (24-bit/192kHz) and 3.5mm coaxial inputs as well as not one, but two 3.5mm single-ended headphone outputs. And those outputs offer plenty of power too, featuring 35mW at 600 ohms and 720mW at 8 ohms with an overall output impedance of 75 mOhms. In English, I was able to comfortably juice my HiFiMAN HE-1Ks as well as my Audeze LCD-3s with ease. Chord’s lead designer, Rob Watts, explained to me that he went to great lengths to preserve the small signal accuracy by using dither and noise shaping so sound quality does not vary with volume setting. Analog purists will balk, but with a product like this where IEMs and other portable headphones are going to be its chief customer, perfect volume tracking is key, and the Mojo delivers in spades.
Like its bigger sibling, Mojo’s user interface is based on a spectral polychromatic approach, i.e. status and modes are indicated through color (You mean colour. -Dave). And as you can see, each color of the rainbow indicates a different sampling rate. Same goes for volume too, which cycles through various color schemes to indicate level. If you press both volume buttons together while the unit is on, then you will cycle through a couple of different brightness settings. Hold down both buttons while simultaneously pushing the power button and that will tell the Mojo to produce a 3V line-level output. Cute.
In terms of how effective the ROY G. BIV approach to user interface is, well let’s just say I didn’t think it was nearly half as bad as John Grandberg of Part-Time Audiophile made it out to be. No offense to John, but yes, it’s very quirky, and yes, I could definitely live without all the blinky lights since it makes the unit come off a bit childish. However, after using the Mojo everyday at work, I just got used to it. I mean at the end of the day, one button turned the volume up while the other turned it down. Really, that’s about it. In terms of its sampling rate indicator, who cares? Audirvana proudly displays the rate every time I hit play. Seriously, after about one, maybe two minutes, you’ll get the hang of it easily. More of an issue for me was the fact that though the Mojo features digital volume control, you can’t control it from the keyboard. That does suck. Granted, it is by no means the end of the world, but I do think hands free operation is ideal for a device like the Mojo.
The Mojo has two mini-USB inputs, one labeled for charge and the other for digital input. Yes, you can use the Mojo while it’s charging (obviously, you need two USB cables then), but the manual points out that you are going to wait for eternity to see the charge light turn off while shooting metal through it. In general though, you get 10 hours of headbanging for 4 hours of charge. That proved more than sufficient for my daily usage. And as the color scheme shows above, the Mojo supports ever sampling rate known to man, including 768kHz DXD as well as DSD64, DSD128, and DSD512 over DoP. Basically, this little black box covers all your bases and then some. There is even some talk about a future add-on module that will plug into these ports and offer Bluetooth, WiFi, and even a SD card reader to boot. Now that would be killer.
Tap Out
A lot of Chord’s secret sauce revolves around something they call the “Watts Transient Aligned” (WTA) filter. The main thrust behind the development of this custom filter is that our brains, not ears, are really responsible for our hearing, i.e. we hear with our minds. And it turns out that according to Watts, our ears are capable of detecting a 4us timing difference between incoming sounds while the CD format, due to its limited sampling rate (44.1kHz), can only retain down to 22us of timing difference between samples. And this effects how our brain processes the sound our ears collect. Worse still, the typical interpolation filter that is at the heart of every modern DAC can not reconstruct these timing differences accurately due to their limiting processing power and thus, limited number of filter coefficients they can implement. But because Watts harnesses the processing power of a full blown FPGA, he can implement an order of magnitude more taps than a convention FIR filter, which in turn results in a more accurate D-to-A conversion with respect to transients. And so the story goes.
To accomplish all of this magic, the Mojo uses the 28nm Artix-7 chip, Xilinx’s most power efficient and lowest cost FPGA to date. Speaking of power efficiency, though the Mojo gets warm during playback, I was always able to pick up my unit after several hours of usage with ease. Even more impressive is the fact that though the Artix-7 has significantly less processing power than the Spartan-6 found in the Hugo, Watts was still able to implement his WTA filter code with very little change. In fact, the code base between the Hugo and Mojo is relatively the same, with some minor changes to the Mojo to accommodate 768kHz sampling rates. The bottom line is that for almost half the price of the Hugo, the Mojo offers the same state-of-the-art design. Nice.
Most of my listening sessions were conducted with my Jerry Harvey Audio Roxanne CIEM through my Macbook Pro since that is my daily, on the go setup. Let me get this off my chest right now: the minute I swapped out my Geek Out v1 for the Mojo, my jaw dropped. I could not believe how much better the Mojo sounded over the Geek Out. And I do mean everything sounded better – bass extension, soundstage, imaging, you name it. It was one of those moments in this great hobby of ours that you rarely have, but always live for. In fact, it sounded so great out of the box, I found myself instantly going back to records that I knew by heart just to confirm what I was hearing was indeed real. So without further ado, here are my listening notes.
Chord Electronics Mojo
GearFeb15
Tags
amp
chord
dac
mojoRelated Posts
Review: Audioengine D3 DAC and Headphone Amp
Review: AudioQuest Dragonfly Red and Dragonfly Black
Poly Rhythms
The Basic Schiit: Modi Multibit and Vali 2 ReviewShare This
Chord Electronics Mojo
Posted by Alex on Feb 15, 2016 in Gear | 33 Comments
Introduction
Alright, I don’t expect you to actually slog through the whole video. However, if you did get through the first five minutes or so then I’m fairly confident that you have a pretty good idea of what a Field Programmable Gate Array (FPGA) is now. But to recap, an FPGA is a collection of controllable logic blocks (CLBs) that are all interconnected via a high-speed bus. In it themselves these logic blocks don’t do much. But when strung together, they can be used to perform all kinds of interesting computational work. So let’s say you want to program your FPGA to emulate a general purpose microprocessor like the Intel-based one you are using right now to read this review. You’re in luck, you can! Or perhaps you want it to perform the same functionality as an Application Specific Integrated Circuit (ASIC), say like Bitcoin mining so you can make millions of dollars by the time you actually finish reading this review? No problem. Or maybe you just want to write your own Digital Signal Processor (DSP), like an ESS SABRE, and use it in your own custom DAC? You guessed it, you can do that too. And that’s the beauty behind using an FPGA – it’s like being given your own computational Lego set that you can piece together to create whatever kind of application your heart desires.
But the intrinsic flexibility of FPGAs over DSPs also has its downsides. First, DSPs have historically been a more power conscious choice over FPGAs since their silicon can be better optimized for their specific computational workload. Second, FPGAs are simply more complex devices than DSPs. For example, let’s say you are looking to create your own custom DAC. Then the simplest route would be to pick some off-the-shelf DSP, slap it into your circuit, and feed Pig Destroyer through it. Done. With an FPGA however, you would have to go about writing all the custom code to program all of those CLBs yourself. That is no easy feat since an FPGA is typically programmed in some kind of hardware description language instead of a high-level language like C or Java. And once your code complete, you then have to suffer through this elaborate testing cycle which includes timing analysis, model simulation, and various other verification methodologies just to verify that your code actually works. So if time-to-market is critical, an FPGA based product is probably a non-starter from the get go. Finally, FPGAs almost always have higher per-unit costs because they have more transistors, and thus a bigger footprint than your comparable DSP. Remember, an FPGA vendor doesn’t know how a potential customer will use their chip, which means they typically stuff in it a lot more functionality (read: more CLBs and I/O pins) than one might need for a certain application. DSPs on the other hand, again, know exactly how they are going to be used, and can leverage that fact accordingly.
So why the heck am I telling you any of this in the first place? Because I feel that in order for you to better appreciate a product by UK based Chord Electronics, you need to first understand a little about FPGAs. As you can see, designing products around an FPGA requires a lot more expertise outside of traditional analog circuit design, which is why most of the products we are all familiar with contain off-the-shelve third-party DSPs to handle D-to-A duties. But can an FPGA based product sound just as good, or even better than a traditional dedicated DSP based design? Read on and find out.
Striking The Right Chord
My first experience with Chord was with their now discontinued DAC64 product. Now don’t hold me to the fire since it was a long time ago, but I do vaguely remember that the DAC64 had two major issues. First, it had a problem with buffering whereby there was a noticeable delay between the moment you pressed play and the metal hitting your ears. And two, and most importantly, the DAC64 was expensive (~3k), or at least seem so at the time. But despite all that, I still walked away extremely impressed since the unit looked and sounded gorgeous.
Since then, Chord has come out with a number of equally sumptuous offerings. Probably their most well known product is the Hugo, an integrated headphone amp/DAC now based around Xilinx’s top of the line 45nm Spartan-6 chipset that has built a reputation for delivering reference level sound in a somewhat portable package. But again, like all FPGA based designs, it comes at a price. The standard Hugo is $2,195 list while its bigger desktop variant, the TT, tips the scale at around the $4k mark. Not outlandish for this kind of performance by any means, but certainly not priced for your budget conscious audiophile either. Chord of course understood this, and decided to set out and design a product that would target a more younger demographic that wants great sound but on the go. It had to be truly portable as well as play nice with all manner of smart phone. And most importantly, had to have an attractive price point. So what did Chord come up with? Meet the Mojo ($599).
Meet The Mojo
First off, the Mojo is indeed a true portable device. No, it’s not dongle tiny, but at about the size of a pack of Altoids, it easily fits in the palm of your hand. But despite its relatively small stature, it features a myriad array of I/O ports, including optical TOSlink (24-bit/192kHz) and 3.5mm coaxial inputs as well as not one, but two 3.5mm single-ended headphone outputs. And those outputs offer plenty of power too, featuring 35mW at 600 ohms and 720mW at 8 ohms with an overall output impedance of 75 mOhms. In English, I was able to comfortably juice my HiFiMAN HE-1Ks as well as my Audeze LCD-3s with ease. Chord’s lead designer, Rob Watts, explained to me that he went to great lengths to preserve the small signal accuracy by using dither and noise shaping so sound quality does not vary with volume setting. Analog purists will balk, but with a product like this where IEMs and other portable headphones are going to be its chief customer, perfect volume tracking is key, and the Mojo delivers in spades.
Like its bigger sibling, Mojo’s user interface is based on a spectral polychromatic approach, i.e. status and modes are indicated through color (You mean colour. -Dave). And as you can see, each color of the rainbow indicates a different sampling rate. Same goes for volume too, which cycles through various color schemes to indicate level. If you press both volume buttons together while the unit is on, then you will cycle through a couple of different brightness settings. Hold down both buttons while simultaneously pushing the power button and that will tell the Mojo to produce a 3V line-level output. Cute.
In terms of how effective the ROY G. BIV approach to user interface is, well let’s just say I didn’t think it was nearly half as bad as John Grandberg of Part-Time Audiophile made it out to be. No offense to John, but yes, it’s very quirky, and yes, I could definitely live without all the blinky lights since it makes the unit come off a bit childish. However, after using the Mojo everyday at work, I just got used to it. I mean at the end of the day, one button turned the volume up while the other turned it down. Really, that’s about it. In terms of its sampling rate indicator, who cares? Audirvana proudly displays the rate every time I hit play. Seriously, after about one, maybe two minutes, you’ll get the hang of it easily. More of an issue for me was the fact that though the Mojo features digital volume control, you can’t control it from the keyboard. That does suck. Granted, it is by no means the end of the world, but I do think hands free operation is ideal for a device like the Mojo.
The Mojo has two mini-USB inputs, one labeled for charge and the other for digital input. Yes, you can use the Mojo while it’s charging (obviously, you need two USB cables then), but the manual points out that you are going to wait for eternity to see the charge light turn off while shooting metal through it. In general though, you get 10 hours of headbanging for 4 hours of charge. That proved more than sufficient for my daily usage. And as the color scheme shows above, the Mojo supports ever sampling rate known to man, including 768kHz DXD as well as DSD64, DSD128, and DSD512 over DoP. Basically, this little black box covers all your bases and then some. There is even some talk about a future add-on module that will plug into these ports and offer Bluetooth, WiFi, and even a SD card reader to boot. Now that would be killer.
Tap Out
A lot of Chord’s secret sauce revolves around something they call the “Watts Transient Aligned” (WTA) filter. The main thrust behind the development of this custom filter is that our brains, not ears, are really responsible for our hearing, i.e. we hear with our minds. And it turns out that according to Watts, our ears are capable of detecting a 4us timing difference between incoming sounds while the CD format, due to its limited sampling rate (44.1kHz), can only retain down to 22us of timing difference between samples. And this effects how our brain processes the sound our ears collect. Worse still, the typical interpolation filter that is at the heart of every modern DAC can not reconstruct these timing differences accurately due to their limiting processing power and thus, limited number of filter coefficients they can implement. But because Watts harnesses the processing power of a full blown FPGA, he can implement an order of magnitude more taps than a convention FIR filter, which in turn results in a more accurate D-to-A conversion with respect to transients. And so the story goes.
To accomplish all of this magic, the Mojo uses the 28nm Artix-7 chip, Xilinx’s most power efficient and lowest cost FPGA to date. Speaking of power efficiency, though the Mojo gets warm during playback, I was always able to pick up my unit after several hours of usage with ease. Even more impressive is the fact that though the Artix-7 has significantly less processing power than the Spartan-6 found in the Hugo, Watts was still able to implement his WTA filter code with very little change. In fact, the code base between the Hugo and Mojo is relatively the same, with some minor changes to the Mojo to accommodate 768kHz sampling rates. The bottom line is that for almost half the price of the Hugo, the Mojo offers the same state-of-the-art design. Nice.
Time To Get My Mojo On!
Most of my listening sessions were conducted with my Jerry Harvey Audio Roxanne CIEM through my Macbook Pro since that is my daily, on the go setup. Let me get this off my chest right now: the minute I swapped out my Geek Out v1 for the Mojo, my jaw dropped. I could not believe how much better the Mojo sounded over the Geek Out. And I do mean everything sounded better – bass extension, soundstage, imaging, you name it. It was one of those moments in this great hobby of ours that you rarely have, but always live for. In fact, it sounded so great out of the box, I found myself instantly going back to records that I knew by heart just to confirm what I was hearing was indeed real. So without further ado, here are my listening notes.
I’m a huge Rammstein fan. YUGE! On November 12, 2010, after 10 years of not setting foot on US soil, Rammstein played to a sold out audience of over 18k people at Madison Square Garden in New York City. I was there. It was incredible. I’m still trying to find myself on the Blu-ray disc they just released of the show! In addition to this live concert set, these crazy Germans also released the mother of all vinyl boxsets that contains their complete discography remastered for black 180g vinyl. My wife bought me this boxset for my Christmas present (she’s clearly a keeper) and I have been enjoying it ever since.
Like so may other US fans, I became a fan of the band when I first heard “Du Hast” on the radio off of 1997’s Sehnsucht. Their magnum opus however is probably that record’s follow up, 2001’s Mutter, which of course I could hum in my sleep at this point. Listening to a needle drop of it on the Mojo was really something magical. Not only are the dedicated vinyl masters an order of magnitude better than their crushed CD counterparts, but sound absolutely glorious on the Mojo. John Darko over on DAR calls the Mojo’s sound piquant. I could not agree more, though on this side of the globe I would of said alive. Everything sounds and feels sonically more alive through the Mojo. Listening to “Ich Will,” it was impossible not to ride the cymbals with my head, as their reverb sounded airy and real. No muddiness, no sense of awkward treble decay – they just sounded natural. Same is true for the low-end as well, as bass extension is equally ungodly. I really believe that this might be the first time that all of my Roxanne’s 12 drivers were actually being put through the ringer! Call me impressed.
As another test, I threw my Roxanne’s into iFi Audio’s iDSD for a little ABX comparo (I tried my darnest to level match as best I could). With the Roxanne’s at least, I preferred the Mojo almost every time. They both seem to be able to eek out all the little sonic nuances these vinyl masters have to offer, but the Mojo’s airy, spacious presentation just put it over the top.
Staying with our wax motif, I decided to bring out an oldie but goodie, and one that I’ve used in the past as review source material, a needle drop of Bolt Thrower‘s 2005 epic Those Once Loyal. It is still my favorite BT record to date and one that continues to get heavy rotation. But for this listening session I decided to switch ears and plug in my HE-1K instead for some more Mojo/iDSD ABX fun. Detail retrieval was still top notch through both, with tracks like “The Killchain” and “Anti-Tank (Dead Armor)” roaring out the gate in piquant fashion. What I found most fascinating however, is that though the iDSD seemed to eek out more mid-bass thump, it was clearly at the sacrifice of clarity. For example, on the track “Entrenched,” the immediate groove filled buzz of Ward and Thompson’s guitars sounded ostensibly clearer through the Mojo. Same was true with bass, as Jo’s presence seemed more integrated into the whole on every track. Don’t get me wrong, the iDSD was no slouch either, and with its robust power supply, easily kept up. But through the Mojo, the music sounded bigger, bolder, and again, just more alive.
As my in-depth review of Dystopia revealed, I think this is the best record Sgt. Mustaine’s Lonely Hearts Club Band have released since Youth. Regardless, one of the unfortunate aspects of Megadeth‘s latest offering is that its production is quite poor, with the whole album pushed to near inches of its sonic life. I am still hoping for a FDR version to be released like the last two, but so far, no dice. Ergo, the album is excellent fodder for the Mojo to gauge how it handles overly compressed material.
Truth be told, I was a bit fearful that Mojo’s infatuation with accuracy and wide soundstage might actually prove to be to its detriment when confronted with low DR material. Let’s just say my worries were put to rest rather quickly, as the Mojo sounded remarkably civil despite the fact that the recording I was pumping through it was not. Sure, while listening to “Post-American World” or “Conquer or Die,” I cringed at the muddiness of the guitars and the complete lifelessness of the drums. But that was clearly not the Mojo’s fault, and in fact, I thought it worked its magic as as much as it could, pulling out Ellefson’s bass out of production oblivion as well as giving Adler a little bit more oophm where say the Geek Out would simply sound competent. The Mojo time and time again offered big sound regardless of the source material’s production value.