by Mike OBrien on February 29, 2012
WHAT GOES INTO USB DIGITAL CABLE DESIGN?
First, this is a layman’s discussion, no equations, no statistics. There are plenty of good technical documents on the web covering the USB standard, electrical requirements, and testing.
Digital data is composed of square wave pulses having fast edge transitions used to transmit data (1’s or 0’s). How accurately the data is transmitted is a function of the transmitter/receiver circuitry and connecting cable. Equalization of the interfaces (adding defined poles and zeros to the circuitry transfer functions) minimizes reflected waves in the cable and increases noise immunity of the system. Keep in mind that data integrity is always compromised during a transmission. This can be attributed to circuit timing and voltage variations (jitter), external noise, EMI, and physical cable limitations. As speed goes up, Bit Error Rate (BER) tends to go up along with it. Commercial error rate test instruments are available for testing device transmit/receive Bit Error Rates (BER). They provide predetermined types of data maladies (random jitter, sine wave noise interference, sine jitter, etc) but they don’t test cables directly. Cable effects must be evaluated as changes in performance of a complete system. It is difficult to assign a meaningful number to cable integrity since BER is inherently a statistical measurement. Any data system uses error correcting codes to reduce the effects of corrupted transmitted data. Since this “check and correct” step is an algorithm, it requires time, which implies data must be buffered, corrected, re-buffered and then clocked out. Too many errors can cause time delays in the output data stream. One way to maximize data integrity is to minimize waveform distortions in the transmission link. Sonic affects due to lost or contaminated data will depend on both the host and the peripheral device which is why a USB cable may “sound” different when used with different sources and DACs. Sometimes lower bit rates will actually sound better than high resolution source because fewer data errors occur during data transmission process. Matching data rates can be critical, especially with lower end equipment.
So anyway, how do we design a digital cable? Well, let’s think about what we need to do. Even “digital cables“ are actually high speed analog transmission lines as far as the electrical signal is concerned. The digital cable affects data transmission by distorting the square wave pulse, reducing the amplitude, causing phase (time) delay, and coupling noise into the data stream. This ultimately leads to detection threshold errors on the peripheral side of things which translates into lost or incorrect bits. So our goal is to maintain signal integrity at high frequency (fast edges translate into high frequencies, remember our old friend Fourier). The first step is to optimize cable physical geometry, including placement of conductors, shielding, and grounding. Transmit/receive termination admittance, cable distributed capacitance/inductance, and EMI/EMC are considered in the data channel optimization. Power channel wiring employs shielded oversized conductors to reduce loses in longer cables. Both channels have 93% silver plated copper braided shields. One version also includes a 100% coverage metalized Mylar shield in addition to the braided shield (helps with EMI/EMC requirements but didn’t affect sound in our tests). Separate shield for data and power channels reduces BER by decoupling power supply noise from the data channel. A very important aspect of any cable design is the material choice, both conductor and dielectric. Our material selection is based on years of experience in the aerospace industry where signal integrity and low noise are paramount. Employing the proper dielectric (optimized dielectric constant /dielectric absorption) makes a huge difference in waveform integrity (and it’s NOT Teflon like most people think!). Minimization of tribo-electric effects is also a major consideration when choosing a dielectric material for use in an acoustically active environment. We have also investigated a variety of conductor materials ranging from cheap Cu/Fe to 5 nines silver. If you think all the RG standard cables are alike, guess again! There is a huge disparity in performance even though they have the same characteristic impedance. Interestingly enough, we found that optimized impedance (distributed RLC) has a much larger effect than material choice for this application. Even though the single crystal copper and pure silver may offer lower noise (very noticeable in low level analog) the effect is much less pronounced. Would a silver conductor be better? Yes, but marginally. Any time a pure metal is employed as a conductor the inherent noise will be lower. Unfortunately the wire must be soldered to some other metal so intermetallics will form (bad for noise), plus the connector may have several plated layers of dissimilar metals. I highly recommend looking into the metallurgy of soldering, brazing, and welding, you’ll be amazed at all the bad “stuff” that forms in a connection, pretty scary!
Anyway, the upshot is that we feel we have struck a balance between cost and performance by taking into consideration physical construction and material properties. Our acid test is to have the “old analog guys”, non audiophiles, and musicians listen for differences in musical presentation. Here’s the catch; they don’t know what’s being changed during the test, only that something MIGHT change. Eliminating the psychological bias is critical when detecting perceived difference. Several theoretically superior cable designs ended up being disasters! After all, engineering is obtaining the best possible solution within the bounds defined by available resources and cost. It is possible to stumble on a decent design quite by accident, as they say, even a blind squirrel finds a nut sometimes, but we prefer to apply some science to find ours!
I hope I did not turn potential readers off with my technical overview, but instead brought to light some important variables in cable design that we have incorporated in our YFS designs. We are trying to let people know we didn't come up with our designs by "playing around" with different cabling/ shielding variants like some of the other guys. We used SCIENCE!
-MOB