Before we start getting into new hardware, digital interfaces, software configuration, digital filters (and more), I wanted to run down a list of tips and tricks to make you more successful in your own high res DAC adventures. Many of these tips are spread around the web and found in excellent books (like R. Harley’s “The Complete Guide to High End Audio”). It was time to round the important ones up together in one place. I couldn’t think of a better preface for my 2nd review than the pitfalls I found in my own usage and how they can be avoided.
It’s easy to make mistakes when dealing with so many moving parts, so develop a system and stick to it each time you change out your op-amps. If the op-amp orientation isn’t clearly marked to your satisfaction, then mark the DIP end of the op-amp yourself. Having a bit of paint/ink on one end of the op-amp is much better than blowing it up! Trust me on that one… So let’s jump right into ways to ensure your I2S adventures go smoothly. Then we will talk about digital interfaces and what really separates the USB (Universal Serial Bus) +I2S (Inter-IC Sound) combo from the S/PDIF (Sony/Philips Digital Interface Format) crowd.
Keys to I2S Output & Input Devices:
- Always check the I2S pin arrangement, as I found out, the USB module’s (Amanero/XMOS) I2S pin arrangement does not always match your DAC’s I2S pin arrangement!! For this reason, I would not recommend using ribbon cables for I2S without checking this prior (on both DAC and I2S device). Even Leaf Audio included ribbon cable, that as I said, is useless without rearranging the pattern. Instead, look at other cables with 6 conductors, such as CAT5, CAT5E, CAT6, CAT6A for example, heck, even SATA cable works (has 7 conductors).
(Leaf Audio Amanero ONLY) I2S Header Looking from the USB input end (left to right):
GND | MCLK | LRCLK (or FSCLK) | BCLK | DATA | DSD On (<– Optional Pin, also called DSD0E)
(ES9038P DAC) Now the reference arrangement from the USB input end (left to right) is:
GND | DSD On (<– Optional Pin, also called DSD0E)| MCLK | LRCLK (or FSCLK) | BCLK | DATA
(Note: I tried my XMOS unit and the pinout arrangement is the same as above except there is no DSD On pin, so you skip that pin.)
XMOS Arrangement for ES9038Pro DAC:
GND | SKIP | MCLK | LRCLK (or FSCLK) | BCLK | DATA
So just having the DSD On pin in the wrong place wreaks havoc when trying to solve playback issues. Check your USB to I2S devices to ensure their pin patterns follow your DAC’s pattern “to a T”.
- Do not install a bunch of components into FooBar2000 that you don’t know how or what they do. As of Jan 2018, the Amanero only needs ONE component to be added in Foobar ( foo_input_SACD) for native DSD playback. Using foo_input_SACD, you can output either by ASIO (Audio Stream Input/Output) or WASAPI (Windows Audio Session API). Most guides recommend using WASAPI (Event) over WASAPI (Push). Add-ins like ASIO4All are also not needed for the Amanero anymore. The latest SACD plugin release for Foobar2000 doesn’t need anything except your Amanero’s (or other device’s) ASIO driver (from the Amanero.com website). You also don’t need foo_out_asio or a bunch of other components to slow down your PC and thus foobar/music. I have 32GB of RAM and I can still hear the effects of tasking processes/programs. During certain scenarios the audio is/can be affected by CPU & GPU load. Cutting out these add-ins also cuts down on a lot of settings that most people do not understand without using Google. It’s not always that your PC can’t handle the additional tasks, a lot of the time it is just the increased load on the power supply introduces more noise.
- Use a fast and reliable hard drive when doing objective listening tests (preferably a solid state drive (SSD)), also look into A/B listening components for foobar to help with your comparisons. What you really want is a partner to help you with this. You want the listener in one room completely alone with headphones/speakers. In the other room your partner will control everything from a HW perspective and only your opinions of the changes will be needed.Note: Hallman Labs uses a Samsung 860 EVO 1TB for storing music, that can be found for under $300. One DSD256 album is 11 GB in size, so you need something big and fast!! I remember when SSD was over $1 per terabyte!! The 860 EVO literally brings SSD to the limit of SATA-II (~500-600 MB/s) while still being a great value. The 500GB version is below $200! If you have M.2 or NVMe capability, go with those for better speed!
Serial ATA (SATA, abbreviated from Serial AT Attachment) is a computer bus interface that connects host bus adapters to mass storage devices such as hard disk drives, optical drives, and solid-state drives. Serial ATA succeeded the older Parallel ATA (PATA) standard
M.2, formerly known as the Next Generation Form Factor (NGFF), is a specification from 2013 for internally mounted computer expansion cards and associated connectors. It replaces the mSATA standard, which uses the PCI Express Mini Card physical card layout and connectors.
NVM Express (NVMe) or Non-Volatile Memory Host Controller Interface Specification (NVMHCIS) is an open logical device interface specification for accessing non-volatile storage media attached via a PCI Express (PCIe) bus.
- Check the dates of all the guides you read on DSD playback, (this “guide” was written in March of 2018). If you are reading anything that is before 2018, you can’t trust everything it says about “what is needed” for Amanero or XMOS DSD playback on Windows. Note that this can include hardware, but no such cases have been encountered here. One place I recommend from my own experience looking for answers is: https://diyaudioheaven.wordpress.com/digital/pc-software/foobar-2000-for-dummies/ and https://diyaudioheaven.wordpress.com/digital/pc-software/foobar-2000-for-dummies-part-3-new-experimental-sacd-plugin-v-0-9-x/ Those links will give you the essentials for getting your Amanero up and running with Foobar2000. (2nd link is more for those out there running something other than the Amanero that natively decode DSD)
- Get rid of that 3.3V power from the PC as soon as you can. I recommend SBooster’s VBus2 (or VBus1/VBus3) to stop that 3.3V line and ground right at the start, before it ever has a chance to get to the USB cable you are using to connect PC and DAC. (Note: PC is not an operating system, it’s the hardware device overall) I was considering all sorts of options on how to kill the 3.3V (or 5V for XMOS) & Ground from the USB cable and one included cutting the traces on the Amanero’s PCB. I then realized this wasn’t a great solution as all USB devices like it would need the same modification. When I came across the VBus2 it seemed to be exactly what I was looking for and it has been doing great since day one! This makes your USB cable purely a data cable and so without external power, the Amanero/XMOS will not turn on. Speaking of USB cables…
- You need a solidly built USB cable. I have one from AudioQuest that is a Forest. I prefer the USB cables that Ghent Audio is making, using excellent materials combined with excellent workmanship. I have been doing business with Ghent Audio for over 5 years now and I have never been let down by their cables. I always recommend people at least take a look first before deciding. I only bring them up because I believe Ghent Audio deserves some praise in the budget HiFI Audio Community. People try and give them crap because they are in Asia. Well, so is almost every other cable company. The tone is clearer than AudioQuest at ¼ the price, so there isn’t much thinking to do there for me. Note: this could be due to the short length of my Audioquest Forest vs. the 1.5m length Ghent Audio USB cable.
- In general, always keep your data cables as short as possible, especially the I2S cable/s. Robert Harley says he finds the best fidelity USB cables to be around 1.5 meters long. Increasing the length of a cable increases the capacitance, just like guitar cables, it will roll off some of the high frequency signal. If we were talking about analog cables (speaker wire, RCA, etc.), this rolling off of the high frequencies would create a warmer tone. USB is a data cable, so rolling off the highs will clean up high frequency noise from the data signal vs. rolling off highs from the sound. This should result in a “cleaner” digital signal with less floor noise and THD+N (Total Harmonic Distortion + Noise). Keep in mind with your I2S cables, you want all of them to be roughly the same length so there is little delay between each I2S signal. As long as you have them within 0.25 inches of each other, you should be fine.
- If your setup worked when you went to sleep and when you get back it’s not acting right, always check your I2S connections (unless you already soldered both ends). Next check your op-amps as well as your digital and analog interconnects. I2S connections have been the single most pain in my ass of anything else in the entire build. I recommend going with soldered connections, not pin headers. I have been looking into ways to permanently connect the I2S out to the DAC’s I2S in. I plan to use Belden 1303E CAT6A “CatSnake” for a permanent connection. You can get two meters for around $25. I picked this specific brand and model because it’s what Ghent Audio uses in their high end JSSG CAT6A cables! You can get a chassis mount Neutrik or Switchcraft CAT5 or CAT6 panel connector for under $20 and it should last forever. They offer varying outputs inside the chassis, so see what is going to work best for you.
- Mechanical isolation of the device (prevention of vibrations from speakers and other factors) plays a big role when we are dealing with removable/temporary interconnections (like the I2S cables seen in the photos) and/or pieces in the DAC that reverberate (to echo a vibration). Even your discrete op-amps and heat sinks can reverberate! Try to cut down on vibrations as much as possible. In Part 2 of this review you will see the actual CAT6A that I decide to go with for a soldered in place connection. If you are still using the normal, round style female pin header, vibration is your enemy. This is what causes you to keep messing with your I2S cable and eventually having to repair your I2S cables and/or their pin/s. Those pins seen on my I2S cables will break 95% of the time if bent once and then straightened. You are never going to be able to break off the rectangular style pins. However, they are harder to work with and require a hotter iron temperature to properly bond. The rectangular style is also orientation specific and present more resistance vs. the small ribbon or cable that goes with the round pins (keep them short!).
- Check your temperatures and your voltages on the DAC’s power inputs. Especially if you detect too much heat from the LDOs (Low Dropout Voltage Regulators at 70C+). If the voltage is above 7.5-7.8+ Vrms (other words, ~7.5V AC) for the two digital inputs, pull the power immediately (it’s made for 6V). If the analog stage is much higher than 19V – 0V – 19V, pull the power. Those are roughly the power limits and in general this device shouldn’t use a lot of power. If you do the smart thing and buy the Amanero that fits inside of the DAC, (unlike me) everything is so much easier from a hardware standpoint (so do that if you don’t trust your soldering, pictured below). On those tiny “real” Amanero boards, I can’t even upgrade the capacitors! I prefer things I can tweak when it comes to sensitive sound hardware.
I lost the pictured Amanero unit above due to yet another I2S cable breaking and then moving onto the top of a capacitor, subsequently discharging into the Amanero through the I2S line. The DAC wasn’t even on, but apparently the cap was still holding a charge. So I ordered another Amanero and this time we are going to permanently secure the I2S lines so this never happens again.
Edit 3/20/18: Since we are going to permanently mount the unit, I got some really special upgrade capacitors for the 2nd Amanero. Rubycon Black Gate (Non-Polar) capacitors!!! These usually bring $40 or so EACH, but I managed to get 6 for a very reasonable price due to them being slightly used. I measured the ESR and both capacitors are around 0.030 ohm which is rated excellent for 100uf/16V! The rated ESR value for 100uf/16V is 0.7 ohm, so this is considered very very low ESR.
Edit 3/22/18: Belden CAT6A “CatSnake” arrives.
Oscillators (XOs) are important, but misunderstood by a lot of the Audiophile community (excellent read here on clocks/XOs). First, let me say that gold doesn’t mean anything, especially if it’s not even in/on a conductor! There are many types of oscillators that range in size, from ½ a postage stamp to something called DIP-14, the full size 4 pin metal cans (seen on mine). There is also the dedicated external oscillator, which is usually very expensive due to the use of rubidium (basically makes this an atomic clock)! Oscillators can be made various ways, but the most popular designs in the budget audiophile community still rely on actual crystal with a specific resonance frequency at a specific voltage. This voltage is what you will see in the graph below, the other factors make up the the TCXO and give us the best long term stable clock signal possible.
The most popular HiFi oscillators used are the TCXO (Temperature Controlled Oscillator) type, mainly due to availability and cost. This oscillator varies the voltage into the crystal (usually quartz) based on the temperature to ensure long term stability of the clock in a variety of environments (see illustration below). I found out from multiple people in the Audiophile communities that short term stability is what is most important to a DAC (see Crystek oscillators). Personally, I think this is debatable when you are going for bang for your buck. BMG, A Russian company, also began making modern precision oscillators that have pretty good reviews, but pricey at $80 a pair. For that price you could buy three of Crystek’s midrange lines or one of their high end. Trying to “hear” an oscillator change can only be done IF the previous setup contained moderate or more amounts of jitter. See more on TCXO function here.
Besides the normal crystal based oscillator we also have electronic oscillators that use properties of LC and RC circuits to create specific frequency vibration. If you want to know more about crystals and oscillators, check out this great article here on Electronic Products.
I found something worth noting about oscillators and being a US citizen during my HiFi hunts. Currently, if you need to buy a full-size 4 pin oscillator for the Amanero, nobody sells those speeds (12 MHz, 22.5792 Mhz & 24.576 Mhz); not Digikey, Mouser, or Arrow. Even the oscillators that Digikey/Mouser show in stock have 1-2 month lead times (waits). The only way to source new full-sized oscillators quickly is from Asia or Russia using Ebay or Audiophile forums.
Edit: Thanks to one of my Facebook readers, I now know that you can find at least one of the Amanero speeds in DIP-14 (12 MHz). It’s an OCXO design and cost less than $3.
You will see over 20 different op-amps tested here, discrete and normal. Op-amps are the most complicated building block in the entire DAC, aside from the primary 8 Ch DAC IC, the ES9038Pro. This makes op-amps extremely interesting pieces of hardware to examine, they range in size more than almost any other electrical component. Their performance characteristics are almost as detailed as a completed amplifier you’d see for sale. However, to look at raw output signals from op-amps and oscillators, you need a good enough scope. Lucky for us, scopes have come way down in price, even in just the last 5 years. A scope like the SPO (Super Phosphor Oscilloscope) 1202X-E would have been $1000+ 5-10 years ago, but the times, they are a changing! Below I have the raw outputs from the RCA jacks of the ES9038Pro DAC shown in both the time and frequency domains, all on one screen!! I prefer to use the dots method for the plotting FFT vs. the vectors like we see in the time domain (top half). This SPO has the ability to show colors which gives another axis of measurement! This is the first scope I have ever used that has this color functionality, that includes my old EE lab scopes at NC State Univ!
(Just noticed one of the channels was DC coupled while the other was AC coupled!)
To date I have spoken with many employees of Burson Audio who specializes in, but not limited to, op-amps. Burson Audio has also provided me with their Cable+ R2R cable and their Burson Play DAC that utilizes an older primary ESS DAC. This will give me an excellent comparison DAC to see the advantages of the ES9038 (or not). I want to thank Burson Audio for standing behind me and my work, even with delays. They never pressured me into releasing my review and even called what I do as “art”. I wouldn’t go that far, but I appreciate the confidence in my work. Orange Amplification reached out to me after seeing the initial Burson V5 Series review (you can find this in the site’s header drop down menu).
I heard of Sparkos Labs by word of mouth and purchased a set of the SS3601 single-DIP line at a discounted price for being a reviewer. I was pleasantly surprised by how these single-DIP op-amps sounded! They even give the Burson V6-OPA line a run for their money (although Sparkos Labs are more expensive). As I saw the review was finally lining up for a publishing sometime in March, I reached out to Sparkos Labs and got a set of their Dual-DIP SS3602. Their performance and price is in direct competition of Burson Audio, which is why I brought them in. I have spoken with Sparkos Labs President and he is going to be sharing these reviews. I just hope I can live up to the expectations! Part 1 is all about giving critical information and explaining a lot of the hardware used in the review. I wanted to thank Sparkos Labs for sending me out another SS3602 at no cost to me. This was to replace the one SS3602 I put in backwards and then saw a mini-fireball between the two PCBs! Since that happened I have marked my Sparkos Labs op-amps to easily show the DIP end, in order to avoid this from occurring again.
Edit 3/17/18: Sparkos Labs replacement SS3602 arrived, sound is excellent! It’s going to be tough deciding between these and Burson’s V6-OPA-D!!
To give our Black Gate-X powered Amanero a run for its money, I have purchased an upgraded XU208 XMOS design pictured below. This XMOS unit has full digital isolation for all signals, made obvious by the removed PCB planes through the center of the board. The ICs connecting the two halves of the boards are the digital isolators. This should give our Amanero with Black Gate-X capacitors a good run for its money.
“This is a high-performance USB digital audio interface, using the XMOS latest xCORE-200 series chips. this board is using self-developed core, low-jitter clock system, clock system used in Japan KDS low phase noise TCXO, frequency accuracy of up 2PPM, RMS JITTER is in 0.5PS less. FUN01 input using a standard USB2.0 interface with rich output interfaces, including XLR balanced output AES, RCA coaxial and optical output, with RJ45 socket output I2S interface. Power input DC 5V.
- After a long period of development and tuning, and ultimately achieve the advanced digital interface level. Its positioning in the high-end interface, mainly uses the following advanced technologies:
- ground full isolation technology, can interfere with the PC completely from the ground and other front-end equipment, isolation technology uses high-speed chip rate up to 150Mbps;
- CPLD shaping technology, I2S signal isolation after the re-shaping; thus effectively eliminating the isolation chip to bring added jitter;
- Developed its own system clock, high-performance crystal KDS, and low phase noise and low jitter.
- Using the latest XMOS program: XU208, XU208 is the second generation XMOS chip memory is 2 time of U8.computing speed arrived1000MIPS, U8 is 500MIPS.
- Input port ： USB 2.0
- digital output port: AES/EBU OPT COAX I2S
- works voltage: DC5V
- PCB size: 126*77mm ( highest about 30mm)Specifications:
- Each output interface supports sample rates:
- PCM: 44.1KHz, 48KHz, 88.2KHz, 96KHz,
- 176.4KHz, 192KHz, 352.8KHz, 384KHz
- [ I2S out support full sample rate, S / PDIF supports up to 192KHz]
- DSD: 2.8 MHz (DSD64) – DoP, native
- 5.6 MHz (DSD128) – DoP, native
- 11.2 MHz (DSD256) – DoP, native
- [ I2S support all DSD format out, S / PDIF and AES / EBU support DSD64 DOP mode]
- Bits wide: the highest 32 bit over I2S output
- The highest 24 bit over S / PDIF”
Audio Codecs and Modern Decoding Technologies:
There are so many audio codecs being used to decode stereo that is now getting a little ridiculous. The majority of this review will be using DSD (Direct Stream Digital), FLAC (Free Lossless Audio Codec), Dolby Digital, and DTS. DSD is the highest fidelity of the 4, the latest released codec and requires proprietary hardware to decode natively. (Note: local DSD files are either found in a .iso or .DSF form) This is in contrast to converting DSD to PCM which is commonly seen (this is how FLAC works, Pulse-Code Modulation). It’s worth noting here, many people assume LPCM means Lossless PCM, it doesn’t (PCM is already considered a lossless format anyway). LPCM stands for linear-PCM, “LPCM is a specific type of PCM where the quantization levels are linearly uniform” PCM has its quantization level vary as the amplitude does. Note: quantization in audio, is assigning a number (called a word) to represent the audio signal’s amplitude ((Vpeak-to-peak)/2 or just Vpeak). The word length determines the systems resolution, SNR, DR, etc.
Two things exist in PCM and LPCM that determine the audio quality, the sampling rate (44.1kHz-384kHz) and the bit depth (16-32 bits, this is our “word” mentioned above in quantization). LPCM is most commonly seen offered on DVD and Blu-Ray disks in addition to DTS and/or Dolby. Some of the movies with the best audio have 5.1 LPCM offered instead of Dolby, which is almost always an upgrade in sound quality. DTS-HD MA and 5.1 LPCM are pretty close in quality, the track quality being movie dependent. I recommend trying all of the audio tracks offered on the disk and go with what you think sounds best. I’m not going to talk about everything offered for digital audio playback, in order to keep the scope of the content to what is relevant (I’m already pushing 5800 words).
Once you are done picking how to get the data stream from your disk or SSD, you need to pick a data cable to send it. Most DACs support Coax and/or Optical, then we have what I use, I2S (or IIS or I2S). This is equivalent to USB and most of the time to get an I2S output, you need a special device like the Amanero or XMOS. A lot of the higher end brands like Oppo, send I2S over HDMI without making it obvious to most consumers. Receiver companies like Marantz accept this I2S input (either over HDMI or CAT5/6), and only then can you natively decode a DSD file. DSD stores the signal as delta-sigma modulated, this is a sequence of 1-bit values with a minimum sampling rate of 2.82224 MHz. This sampling rate is also called DSD64, which is 64 times the sampling rate of CDs. Although CDs use 16-bit word size and DSD uses 1-bit, the speed of the sampling rate makes up for the smaller word size. DSD goes up to DSD512, but I have yet to find any sources for DSD512 (called Octuple-rate DSD), at all. The most common found in the wild are DSD64 and DSD128. I recommend using HDTracks to try out some high resolution FLAC/ALAC tracks before investing a ton of money into DSD.
DSD is literally the digital equivalent of SACD as 16/44.1 FLAC (PCM) is to CDs. If you are picking between Coax and Optical, pick Coax and get a decent Coax cable, usually 75 ohm or so is nominal (that is what my Canare based Coax is). Optical is the least desirable mainly due to how it handles the data and clock stream, it’s almost screwed from the get go. Toslink embeds the clock and the audio signal into a single optical signal that must be recovered at the output device’s input. The ES9308Pro employs some neat audio magic to make Toslink (Toshiba Link)/Optical better than usual, but go with Coax if possible. I’ll talk in more detail about this in part 2. Common ways to send I2S include, but are not limited to: HDMI, USB, Ribbon Cable, CAT5, CAT6, etc. anything with 5-6 conductors will usually work. (Once again) These are all data signals, so with our I2S cables we want to keep them as short as possible and optimized for high frequency transfer. I’ve talked about the skin effect before, but we’ll get into that also in part 2.
Amanero/XMOS 3.3V & 5V LT1963/A External Power Supply:
Here at Hallman Labs we are running USB A to USB B cables sourced from Ghent Audio and we are using a SBooster VBus2 to isolate the 3.3V line and GND from the noisy PC (full of fans). I am using a dedicated power supply board for the Amanero and/or XMOS that has a full wave rectifier, LT1963 LDOs (Low Dropout Voltage Regulators) with ELNA SILMIC-II & WIMA filtering/smoothing capacitors. Because of the VBus2, if you don’t give the Amanero dedicated power it doesn’t work.
This PSU board can handle AC or DC power in, but will require a heat sink depending on what you send to it. I send 12V 3A DC into it, well above the needed amperage and voltage, but much less than 120V AC. The bridge rectifier has a pretty standard loss from transforming AC into DC; that formula is Vdc = Vav = 2 * Vpeak / Pi and then you have to take into account what smoothing capacitors are used after rectification. Here we have a RH25 LA Nover 6800uF / 35V audio capacitor, the other three pairs of capacitors are downstream from each of their respective LDOs.
Digital Interfaces: What makes USB so special, why not just use Optical or Coax?
I am sure some of you are wondering why anyone would bother with I2S (which by the way, is pronounced “I squared S”) and USB if it can be so much trouble. Why is I2S any better than using coax or optical? What makes each of these different from one another? There are two types of digital interfaces, optical and electrical. Coax and I2S are electrical (they transfer a signal down a copper wire or other conductor), while optical is obviously, optical. Well, what is “optical” with regard to audio signals?
Optical sends light down a plastic or glass tube which have both the clock and the data embedded in the same optical signal. Coax cables can suffer from radiated noise if not made properly, which was one reason for the Toslink push in the 90’s. Toslink was also cheaper and easier to make than coax cables. Then why has Toslink/optical been phased out of a lot of designs over the last 10 years in favor of Coax? Toslink is considered the worst of the digital interfaces in mechanical connection/attributes, electrically (has the lowest bandwidth) and as a result of the former, sonically too. According to “The Complete Guide to High End Audio” by Robert Harley, “Toslink tends to blur the separation between individual instrumental images, adds a layer of grunge to instrumental textures, softens the bass, and doesn’t provide the same impression of “black” silence between notes.” Harley says although better optical cables can improve on these attributes, he recommends forgetting about Toslink/optical all together. Often I hear people confuse the terms S/PDIF and Toslink, S/PDIF is present in both coax and optical. Toslink is the digital interface (connection), S/PDIF is the transport method used in both.
Often I hear people confuse the terms S/PDIF and Toslink, S/PDIF is present in both coax and optical. Toslink is the digital interface (connection), S/PDIF is the transport method used in both.
There are the rare instances where optical sounds better because it’s not connecting/coupling the coax shield’s ground noise (found on any electrical interface) to your other pieces of equipment. If you are testing your optical vs. your coax input, don’t leave the coax plugged in while testing optical; you are unnecessarily creating multiple paths to ground.
Hold Everything! We need to talk a little about jitter, it’s extremely important in digital audio. Jitter is usually measured using an oscilloscope in the frequency domain (FFT function). There is also such thing as a “phase-noise analyzer” that can analyze high frequency clock signals too. Phase (angle) is always RMS in jitter measurements, this is because they are integrated over a specific frequency band, say 20 Hz to 100 kHz. The reason jitter is a problem is due to the fact when you have too much, you will start getting high bit-rate errors (“BER”). The amount of BER that makes it through into the audio you hear is based on how well your chipset rejects jitter, in my case the ES9038Pro. I think it’s worth throwing into this little section on oscillators and jitter what ppm means and why you should care. (Hint: has to do with aging of crystals)
“Frequency stability is a measure of how much the oscillator’s output frequency potentially changes during operation due to a change in temperature. If the frequency drifts beyond what the application expects, timing errors are likely to occur. Frequency stability is expressed in parts per million, or ppm, relative to a nominal frequency over a specific temperature range.
Oscillators use quartz crystals cut at different angles during manufacturing to produce different temperature responses. Common XO temperature-stability ratings include ±20 ppm, ±50 ppm, and ±100 ppm. A lower ppm means that the output frequency is more stable over the given temperature range.
..An XO may have excellent frequency stability over temperature, but this measurement is only relative to the nominal frequency that it provides at room temperature. So initial accuracy error can be quite large for some devices, such as SAW oscillators, and must be taken into consideration.”
Comparing interfaces are a great way to see how well your DAC does at rejecting jitter. The bigger the quality difference between each of your interfaces, the worse your DAC is at controlling S/PDIF jitter. This is one of the huge upsides for USB and using add-in boards like the Amanero, we are able to isolate the clock on the Amanero and provide the DAC precision timing with minimal jitter from dedicated XOs on the USB modules like the XMOS and Amanero.
USB and its implementation in the DAC is as important in a modern system as your speaker or coax cable. Robert Harley says that USB audio devices vary the most in quality than any other source device, from beyond reference quality (what we shoot for) to terrible quality. This is usually due to poor implementation either at the DAC or at the USB module. The majority of your DACs today will be using USB 2.0, which has a bandwidth of 480 Mbps. Some of the older HiFi DACs, like my Aune T1, use USB 1.0. There is an easy way to tell if a DAC uses USB 1.0, they are always limited to 24 bit 96 kHz. One benefit of using USB 1.0 is that you don’t need a driver for it to work, even in Windows 10! So this equals less headaches on the PC end, but at the cost of quality and compatibility with modern codecs. In all of the S/PDIF interfaces, the source component signal IS the master clock for which our DAC must lock onto. Anytime you see the DAC say NLOCK, this is what it is talking about, it was unable to lock onto the master clock needed.
The clock must be recovered from the audio signal itself, this is the most important step of decoding an S/PDIF signal. This is where we separate the Men DACs from the Boy DACs! To make it even more difficult, the embedded clock is usually varying, so there is no turnkey solution without quality loss. Even the famous PLL circuit can’t solve the issue without introducing its own problems into the signal/s.
Originally, USB was used Synchronously, meaning the clock is generated by the computer and everything happens in step, like dominoes. This clock is noisy, high in jitter and the last thing we want as our master clock for an audio application! Quickly the audio industry realized the answer was to run the audio signal clock independently of the PC/USB clock, thus asynchronous USB mode was born. This turned the USB interface into a high fidelity interface nearly overnight (if the interface is correctly implemented at both ends).
Many of these terms are commonly listed on sale ads without a full explanation. It was for that reason that I wanted to talk in detail about the data stream from PC to DAC. It’s one of the interesting things in DIY audio right now, figuring the best way to get USB to I2S then to the DAC while maintaining maximum quality. I would discourage anyone from investing over $300 into a Pink Faun PCI-e card that provides I2S over HDMI. In 2013 (when it was released) it was worth it, now there are too many other options for a fraction of the price while maintaining audio quality.
One other piece of gear you will see tested out here in Parts 2 & 3 is the AudioQuest “Jitterbug”. A very interesting piece of equipment at a very reasonable price, about the same cost as the VBus2. However they do completely different things, although the Jitterbug claims to help the GND and voltage lines. With the VBus in place it will only do its job on the data lines, which is what I intend to do with Jitterbug (running it with the VBus2).
Edit 3/22/18: I have been testing the Jitterbug not with the DAC (yet), but with an external hard drive (HD) that kept giving me intermittent connection issues. Surprisingly the Jitterbug stopped this problem. I also tried it with an XMOS that kept giving me random loud tone/beeps at 3-4 minute intervals. With the Jitterbug in place the XMOS wouldn’t beep, but after about 5 minutes it would mute the sound. Swapping the cable in and out would bring the sound back, but it confirmed the XMOS I have has an issue. So far, I am impressed with the product for less than $50.
The Problem: All computing devices—laptops, smartphones, Network Attached Storage devices (NAS drives), media servers, etc.—inherently generate a significant amount of noise and parasitic resonances. Additionally, computers contribute a considerable amount of RFI and EMI pollution onto the signal paths—all of which can easily find its way onto your USB cables and into your audio system. This noise and interference has many negative effects. Noise-compromised digital circuitry increases jitter and packet errors, resulting in distortion that causes a comparatively flat and irritating sound. Noise-compromised analog circuitry also damages the sound’s depth, warmth, and resolution.
The Solution: JitterBug’s dual-circuitry measurably reduces unwanted noise currents and parasitic resonances. It also reduces jitter and packet errors (in some cases, packet errors are completely eliminated).
The Result: Clearer and more compelling sound, music, dialog, etc. A better audio experience.
- Use one JitterBug in series (in line) with any computer and USB DAC (digital-to-analog converter).
- Use an additional JitterBug in parallel with the first for improved playback performance.Use JitterBug with other locally connected USB devices, such as hard drives, printers, and cameras, to effectively reduce audio interference.
- Use JitterBug with USB-enabled network devices, such as routers, NAS devices, and streamers.
- Use JitterBug with mobile devices into audio systems in the home and in the car.
WhatHiFi says “Nine times out of ten we’d rather have a Jitterbug in our system than leave it out. If it fits your current set-up, we’d certainly recommend taking one for a spin – it’s a clever, audio-improving critter.”
When I found I could get one new in the box for less than $50, I had to give it a shot. It’s one of those devices that falls right into the niche market this site caters too and it’s light on the wallet which means more people are likely to want one. If this device really does increase fidelity it will be a no brainer upgrade for anyone using USB audio (especially from a PC). I can already say it has decreased packet errors for a HD that was struggling to stay connected for extended periods.
But Keith, I thought we couldn’t hear above 20 kHz, why isn’t 44.1 kHz sampling and 16 bit word length enough?
It all comes back to the Nyquist theorem (also called the sampling theorem), which holds true in the digitization of audio signals (Analog to Digital Converter, ADC vs DAC). He tells us that if the sampling frequency is at least twice as high as the highest audio frequency we want to be encoded, we can call this lossless sampling. The 44.1 kHz signal was chosen had multiple reasons, but one of them was that it was just over 2x the supposed human hearing limit. 44.1 kHz/16-bit is what you see/hear on all CDs. The other reason had to do with frequencies of TV video signals and their frequency back then.
CDs were originally thought to sound “hard, glare overlaying in instrumental timbres and flattening of the soundstage” (R. Harley). What they forgot when they picked 44.1 kHz is how a digital filter works. First let me say this, the digital chain (connections from input to output) from recording all the way to playback in your home, is a series of digital filters. Each of these filters is degrading the sound quality to some extent. I am not going to go into full detail on digital filters yet. There is just so much to say and this is one of the tidbits that started the snowball that lead us to high res audio. In short, digital filters cause ringing (steeper slope = more ringing) and “energy smearing”, the filters used to decode 44.1 kHz were always steep. They had to pass the entire audioband to 20 kHz, then somehow between 20 kHz and 22.05 kHz the signal was asked to be attenuated by 120 dB, the slope to achieve this is huge!! When you bringing in digital filters this is notoriously bad for causing ringing. Ringing is a term we will visit in greater detail later on. Ringing and jitter both exemplify the nuances of working with digital audio systems and their interfaces.
Producers realized that the higher the sampling, the gentler the slope could be and the less ringing we see on the scope and hear (in the ear). This isn’t the only reason 44.1/16 isn’t “enough” for the purist, but it will leave us with a good spot to pick up on in part 2. Keep in mind, “high resolution audio doesn’t automatically equal good sound, it only means that the conditions are here for good sound”.
Wanted to say thanks to my editor Jonathan P. Only a select group of people could be asked to edit an article like this and I’m lucky to have a friend like Jonathan willing to take the time. I hope this article helps to shed some light on some of the more mysterious parts of digital audio and modern interfaces. In part 2 we will see head to head discrete op-amp competition!