Craig Sawyers

That is a good thread - thanks for the link deep. Pity eric's TDR measurements are defunct, but the thread is 6 years old after all. To quote Borbely from AudioXpress 5/02, about his design for the all-fet line amp design: "My goal was to be able to drive loads down to about 100Ω in pure Class-A, also allowing it to be used as a headphone amp. To be able to drive 100Ω in Class-A with up to 10V RMS, you need to have a bias current of 70mA." So this would definitely be able to drive 50 ohm cable (so 100 ohm total load for 50 ohm cable) at up to a whopping 10V RMS. Note that to handle that the 50 ohm load resistors would have to be at least half watt - but since the typical maximum you would be putting into an amp is 1V, low power resistors are fine - or even surface mount. I've been intending to build this beast for a while, if only to test my belief (which I now realise that KG shares) that matching impedances should render the analogue link cable insensitive.

That is really interesting. Erno Borbely went down the same route with his designs and kits before he retired. Enough drive capability from FET output stage to drive from a 50 ohm source down a matched cable and load. I makes the connection almost totally impervious to cables - just high quality 50-ohm cable is needed.

That is the problem in a nutshell. RCA's are ubiquitous. And when BNC's are fitted, or cables are BNC fitted more often as not they are 50 ohms, not 75. The difference is easily seen - 50's have insulation around the pin and earth, so when they engage the whole connection is encased in insulation. With 75's there is no insulation at all - the pin and socket mate in air. It is only because I am anal about interfaces that I've gone to the lengths I have in changing things to be impedance matched throughout, and done some supportive measurements.

I'm not into turning - cabinetmaking (just recreational) is my thing. The cocobolo will probably be used in things like lids of jewelry boxes, storage boxes etc, or small cabinets. I'm a follower of Krenov's philosophy of buy wood. Then buy more wood. Then look at it for a long time until a piece suggests itself. Ebony and blackwood I use for contrasting wedges in through tenons. Against a paler wood (and most wood is paler than those!) it is a superb effect. One of the more disappointing woods to work with is purpleheart. Looks absolutely stunning bright purple when first cut, but oxidises slowly to a kind of muddy brown.

So I did some formal tests of impedance mismatch at BNC 75 ohm and RCA connectors. Measurement system was a Tektronix 7S12/S6/S52 time domain reflectometer - it measures reflected signals from impedance discontinuities. It has a resolution of about 10mm. Set up was a 50-75 ohm matching pad, 75 ohm BNC terminated cable. The other end of this cable was either terminated in a 75 ohm BNC plug/socket with a 75 ohm termination, or Neutrix RCA plug/socket with 75 ohm termination The pics show the results. BNC gave a small reflection blip 200 picoseconds long, corresponding to a -9 ohm drop. So for 200ps there is an impedance of 66 ohms. The shortness of this event means that it is entirely insignificant, even at 192kHz sampling. RCA however gave a much more significant 600ps reflection, corresponding to a -23 ohm drop. For 0.6ns there is an impedance of 52 ohms. In fact if you look at the curve in detail you can see a double dip, where the Tek gear picks up slightly different impedances for the plug and the socket. Since it is likely that both ends of the cable are misterminated in this way, several bounces back and forth will be necessary before the reflection dies away, and since the data stream is pulse length encoded, line resonances from these reflection transients will add a noise and jitter inducing effect. Craig

You betcha Grahame!

Ice cold in Ewok. Oh shit that dates me on so many levels.

Damn right

That'll be superb once turned. I'm always amazed when a grotty piece of wood completely changes character once cut. That is part of the joy and grief - what is inside? Make a guess and hope you get the cut right. I hit a goldmine recently at my local wood and tools shop. He'd just got a shipment of Cocobolo turning blanks, about 8-10 inches square and around 2 1/2 inches thick. I got to them first! I rooted around and picked up the best figured pieces - about 15 in all, some smaller, some larger. Cost me £140 for the lot, which is pretty fantastic value. Also picked up some ebony and african blackwood too.

^Jeez. Makes you blood run cold - and want to give the guy that caught him a big slap on the back!

Nice looking build.

That was awesome - watched the full 40 minutes of it. Filmed in IMAX, and would have been awesomer in that.

Mont Saint Michel in Normandy. There is a panel of the Bayeux tapestry that shows William the Conqueror's army coming near to disaster in the quicksands you can see in the picture around 1000 years ago, with Mont Saint Michel sort of floating above the army.

The only thing to watch is that the BNC needs to be insulated from the chassis.

I have to say that I had not come across this beauty before - but it certainly looks the part.

That would certainly figure - but I'd sure like to know if that is the way it went for certain. It is surprising that Neutrix don't advertise the fact that their connectors are 110 ohm either.

Interesting. The characteristicimpdance of an XLR are fixed by the pin spacing and diameter, and the dielectric constant of the material between them - it essentially looks like a short length of parallel conductor transmission line. Putting in the pin dimension and spacing (between pins 1 and 2) for an XLR, and the dielectric constant for polyimide (the insulator) into the full equation for characteristic impedance of parallel conductors - I get 109 ohms! I'm not sure of the history behind choosing 110 ohms as the AES/EBU standard, but it seems a odd cooincidence that the (calculated) impedance of an XLR is the same value. Unless I have made a mistake in the calculation, AES/EBU via XLR's seems to provide an impedance-matched system.

And that is a fundamental statement that I agree with 100%

That is really sweet! And hats off to KG for keeping plugging away in the teeth of a cracked knob

I didn't manage to find any 110 ohm XLR's (and I have looked!) - have you got a part number or link?

^ Astronomy picture of the day, http://apod.nasa.gov/apod/ , yesterday.

OK - the CD transport is enormously modified (by me), and includes a high stability clock to de-jitter the data stream. That drives the output via an OPA623 wide bandwidth op-amp, with 1ns rise time and precise 75 ohms output impedance. That runs via RG179 75 ohm coax internally to a 75 ohm BNC connector. So it was in that state for years, so no further mods needed. The digital interconnect is RG302 (the big brother of RG179) again wth 75 ohm BNC's. The mods to the Octave was replacement of the RCA with a 75 ohm BNC. The chassis BNC's that I am using are the insulated ones, and are the same diameter as the hole in the Octave - so it is simply a remove and replace job. And then I replaced the pulse transformer with a much better one, which is a bir more of a fiddly mod, because the size and pin-out is different. Craig

The whole connector issue in digital audio cabling is a mess. The thing with AES/EBU is that is far less susceptible to common mode noise in long cable runs, typical of recording studios and professional installations. It is actually only partly compatible with SPDIF, because the preamble data is different - so using a 1:1.2 transformer to match impedance to a 75 ohm single ended won't necessarily work. I find it astonishing that there is so little focus on connector impedance though. With 192kHz sampling, typical in mixing desks, the frequency of the AES/EBU serial data is about 25MHz, and to maintain good square and jitter free edges needs a bandwidth to at least 100MHz, and really 200MHz, and rise times of a nanosecond or so. So matching the connector impedance is rather a big deal! And using XLR is just about as daft as using RCA - both standard practices are trying to spoon broadcast TV bandwidth through connectors designed for audio bandwidth. Even with vanilla CD data at 44.1kHz sampling, the serial data rate is 5.6MHz and so need at least 50MHz bandwidth to get good edges. So about twice the frequency of the maximum shortwave band frequency at 30MHz - via an audio connector.