Sunday, May 3, 2009

DIY Silver Interconnects

Evaluating Different Designs

Correction 2009-05-04: I found some of the original cables tested back in 2006 and noticed that the one having the rectangular silver conductor only had 2 layers and not 4 as the cables that I currently use. This did not change any of the conclusions, but it did make the differences more dramatic. I have edited the text below accordingly. Thanks!


Background

The measurements described here were done in 06. However, a couple of days ago I run into the Audiogon thread listed below and that motivated me to write things down.

The approach described sets out to select a cable design that minimally affects the signal it carries. We adopted a simple and easily repeatable methodology.

Below are a few references which we found very useful. They in turn have other very useful links. I lost track of all the pages I visited, so I apologize for all the omissions, but these will get you started. These writings assume you are somewhat familiar with DIY interconnects so if you are not make sure you check some of the links.



In 2006 I got a new pair of speakers and decided to upgrade my cabling.

I do not use a preamp. I run the interconnects directly from my source to the amps. Therefore the cable runs I'm interested in are in the 8-10 ft range. I run 2 pairs of interconnects for each channel as the speakers are bi-amplified. At some point it might be interesting to try/test different cables for high and low frequencies.

Willing to give DIY cables a try, I set out to choose a design by measuring its effect on audio signals. The measurements are very easy to perform. They can be done by anyone with access to a 2ch digital oscilloscope and a wave generator (both can cost as little as $1000 new). The setup is simple enough that could be used as a lab exercise in high school or 1st year in college.

There is no complicated math or theory. The idea being that irrespective of the underlying phenomena what we care about is how it affects the audio signal. If high frequencies arrive later and are more attenuated than low frequencies, what we want is to compare these delays and attenuations for two cables over a range of frequencies.


Conclusions

Cable construction affects audio signals. This can be verified and measured at a very low cost with very accessible equipment.

A debate may still exist as to whether the effects can be perceived by a particular listener or on a given audio setup. However, the range of values measured does support the idea that some setups and people would be sensitive to these differences.

All cable configurations tested attenuated the highs to a greater degree than they did the base or the mid-range. That indicates that cables probably do not make signals brighter, but push the highs down to different degrees. If that is true, there are no bright cables, but only cables introducing different degrees of warmth.

Unfortunatly besides compressing the highs cables also store and release energy which at higher frequencies creates a destructive component off-phase with the audio signal. On a sine wave this looks like a delay. It grows as the frequency increases. In a square wave this will round off the corners and perhaps add some wigling. If you had a sharp peak I presume this would translate into it being rounded off and perhaps slightly shifted in time. Perhaps it is this rounding of sharp peaks which cause difference in imaging and other subtle sound changes.

Of the designs tested, the one that best satisfied the objective stated above was a 2 conductor stack (signal + return) of fine silver tape 2mm wide and 30AWG thick tempered dead soft.

Many variations did not get tested, some because I could not make the cable on my own, many because I did not think of them at the time, and a few because my lease on the kitchen table expired. However, here you have enough information to repeat and expand these tests on your own. If you do come up with something interesting please drop me a note and I'll link to you.


My personal choice

I currently use a doubled up version of the "2 x 30AWG x 2mm" cable mentioned above as a "4 x 30AWG x 2mm" in an 8 foot run to connect my source directly to 4 amplifiers in a shotgun configuration where speakers are bi-amplified. I’m also using the "4 x 30AWG x 2mm" cable as jumpers for each pole of an amp to the corresponding post in the speaker.


Other promising designs

The experiments below point to 1) minimizing the thickness of individual conductors as a way to postpone the introduction of phase delays in higher frequencies which push highs sideways mushing the signal; 2) increase the cross section (sum of all conductors) as a way to minimize how much attenuation is generated at lower, but more significantly at higher frequencies.

These ideas are not new, but the contribution of this article is that we measured it and are making the results available and repeatable.

Aligned with these 2 guidelines other promising designs would include a very large number of very thin conductors arranged in either a parallel or braid, as well as foil designs where you have fewer conductors which are also very thin but and somewhat wide.

Allen Wright’s vacuumstate.com silver foil seems promising. It is 50% wider and 5 times thinner than the 30AWG x 2mm I'm currently using. However, at almost 10 times the cost (per ounce) the option is probably not viable in my case. Still if this is what you want he is the person that has it. If I had shorter runs I would try it.

If you want to go the braid way, homegrownaudio.com offers some options like a 8 x 26AWG for interconnects and 16 x 22AWG for speaker wire. If you want to braid your own that may be doable by hand using some sort of Madurai see http://en.wikipedia.org/wiki/Kumihimo.

Someday I hope to braid a 40 x 30AWG interconnect and see what it is like.


Choices may vary

We did not test for it, but it is my assumption that as the cable length increases phase shifts will start to occur earlier and will continue to increase at higher frequencies. Signal attenuation will also increase at the top end. Conversely with short cable lengths these effects will diminish.

Therefore for short cable runs a single wire, or perhaps a small number of individually insulated wires may do the job.

For very long runs I suspect that phase shifts will be mostly a lost cause. The battle will probably be to get the signal attenuation, particularly in higher frequencies, under control.


The Experiment

We feed a sine wave to the cable and at the other end we place a load (we used a couple of resistors 5,000 ohms 2W for the interconnects and 4ohm 10W for the speaker cable. On the pole carrying the signal we attach 2 probes, one on each end and hooked these up to a small digital oscilloscope.

We then record differences in amplitude (signal strength) and phase shifts (delays) as we vary frequencies from 10Hz to 100,000Hz.

The scope and the wave generator can cost as little as 1,000 new. The cost in materials is about $200. Time to make the cables, perform the experiments and compile results should be around 40 hours.

Note that I was not able to go beyond 150mA RMS on the speaker cable because of my puny wave generator. However, it does provide us with the information that at higher currents the signal degradation is much higher. As you look at the charts below that should become clear. Therefore I conclude that where possible speaker cables should be shortened and that doing that at the cost of longer interconnects is a good trade off.

Nine configurations were tested. If I were to redo this experiment I would certainly include many more variations. Some of these are noted in the results section.




How to look at the charts, or at least how I have looked at them.

Signal attenuation is bad

The signal attenuation (Peak to Peak Cable Attenuation) is how much signal is lost over the traversal of the cable (6.5 ft) at the top of the sine wave for a given frequency. The curve for an ideal cable would be flat (constant signal loss irrespective of frequency), but it is not that important that it be zero. In the charts below this curve reflects how higher frequencies get progressive less energetic (more attenuated) when compared to their mid-range and base cousins.


Phase shifts are evil

The phase shift (Attenuation Phase Shift) is obtained by subtracting the signal from the end of the cable from the signal at the beginning of the cable. Then we measure the delay between the peak of the difference and the peak of the original sine wave. These curves reflect how information for a particular point in time arrives later for higher frequencies than it does for lower frequencies. This is probably much worse than simply a less energetic high because it distorts the "character" of the curve. It pushes higher frequencies sideways while leaving lower frequencies in their place.


Results

Teflon wins over Cotton



This seems to back the statement that Teflon sounds bright. Looking at the charts we see that Teflon interferes less with the higher frequencies than cotton. Cotton shifts phase earlier and to a higher degree than Teflon does. It also has a slight more aggressive rise on the top end attenuation. Therefore cotton may be perceived as being warmer, but it is in reality pushing down the highs.

Note that these conductors were from different sources. If someone wishes to improve on this experiment I would suggest: 1) measure wire with the cotton insulation; 2) strip the cotton from the conductor taking care not to bend or stress it and measure it again in mid air; 3) add a Teflon tubing jacket to it and measure it once more.

However, in this round and given our objective of less signal degradation, Teflon wins.

I should also point out that the results for Teflon are based on inserting a polished bare silver wire in a Teflon tube. There is much less contact between the wire and the sleeving than if you actually bought Teflon insulated wire which has Teflon more closely bounded to the silver. I did not test that so I can not say what kind of effect that would have, if any.


Thinner is better



Phase shifts increase dramatically with higher conductor diameter and signal attenuation also rises more steeply.

It would seem that smaller is better, except that as the conductor gets narrower and you try to pump more current through it (as in the case of the speaker cable) there is too much signal attenuation. The AWG 28 having a 20% signal loss in a 6ft run. However, things do not look that bad for interconnects using that same wire. They operate with a much lower current and are able to take advantage of the flatter curve and smaller phase shift while losing only 0.03% of the signal.

I believe that it is better to have a smaller phase shift and a flatter attenuation rather than having a lower attenuation in the base and mid-range just to look at it climb more steeply on the highs.


Braiding is good



Here we braided four 20AWG conductors and compared using only 2 of those (pink) against 2 parallel conductors (purple) and found that in itself improved the phase shifts mildly.

Then we hook up one of the extra conductors of the braid to the ground in order to use it as a drain wire (yellow) and find that using the drain wire negatively affects the phase shift a bit.

Lastly we unhook the drain wire and use the 4 conductor braid as 2 for the signal and 2 for the return (light blue). This drops the attenuation substantially as expected (double the cross section), but it also reduces the phase shift -- a nice bonus.

Thus braiding, at least the one above with 4 conductors, is good.

I would not extrapolate the results from braiding to twisting.

If someone wishes to improve on this experiment I would suggest: 1) measure how the distance parallel conductors are set apart affects the signal; 2) check if there are any differences when we float conductors on air, rest them on wood or rest them on carpet; 3) measure the effects of twisting and rates of twisting, which we did not do; 4) verify if the tightness of braiding and its stress affect the signal. 5) see if there is an point of diminishing returns as we add more conductors to the braid.


Double helix degrades signal (makes it warmer ?)



The double helix design (wrapping two conductors around a core in a configuration similar to DNA) did not offer measurable advantages over the parallel design. It increases overall signal attenuation and more so at higher frequencies (in part perhaps due to the increase in conductor length needed to cover the same distance in the cable).

Note that speaker wire in this configuration also has the detrimental effect of increasing phase shifts.

Depending on the system, this configuration may be perceived as adding warmth, but it is in reality trimming down the rest of the spectrum. In fact for interconnects this looks like a fair option to warm up a signal if that is what you are looking for. It does not interfere much with the phase shift when compared to the parallel configuration.

For cable management and construction the previous section indicates that braiding is a better option.


Flat is much better than round



Here we have a 4 wire braid of 20AWG and the 2 strip stack of 30AWG x 2mm.

The tape results are much better eventhough it only has about 1/2 of the cross section of the round wire. The tape has much flatter attenuation and significantly delayed and less aggressive phase shifts.


All together



IMHO the 2 x 30AWG x 2mm configuration is the winner when compared against all others tested. Its competition, as stated earlier, would likely be 1) thinner and wider tapes and 2) a lot of very small conductors in either a braid or parallel configuration. However, both of these are more expensive and/or more difficult to put together.


More tests

For those interested in expanding these experiments I suggest 1) how shielding with foil, tinned copper mesh and carbon fiber at different distances from the conductors affect the signal; 2) checking to see as if the number of wires in a braid grows we see degradation in the signal quality; 3) test for different metals like copper, gold and others vs. silver; 4) combine different conductor sizes and material individually insulated; 5) try different conductor combinations for signal and return.

If I had to pick one I would say that shielding might be the one to offer the most exciting results.

Thanks!