[art]

I think what can help to distinguish between the two is that it would take a real big stray input inductance in the circuit, to reach that monster timing in the shot of Pat: tau is something like 20 nseconds?
As amplifier settling time, it's even quite nice.

In my case, tau is ~ 7nsec, and that is a 1.5 Ghz GBW amplifier.

I think we determined the horizontal scale was 0.05 uSec. So, tau is roughly 50 nSec.

So, from that, we can determine the amount of inductane.

L = t*2*Z.

In this case, 50 e-9 * 2 * 75 = 7.5 uH.

Which, doesn't sound right! Most SPDIF transformers should have a leakage inductance under 1 uH. Before anyone goes off on a tangent, let me test a known transformer, and see what it measures.

(Way too cold, to go into the garage, so hang in there!)

Pat

[art]

Ok, not trusting my Tek 'scope (that now seems to be on its last legs, I think the cold snap took its toll), I took said DAC, to the lab, and measured it, on the TDR.

Got a tau of 30 nSec. Which gives us 4.5 uH. Lower, but still horrible!

Still doubting the results, I took a known transformer, put BNCs on each end, and measured it.

Got a tau of 2 nSec. Which gives us 0.3 uH, which is the value given on the data sheet. So, looks like the results are believable.

So, as hard as it seems to be, yes, that spike is real, and it is really bad.

Moral of the story: if you use a transformer, with that much leakage inductance, it probably doesn't matter how good (or bad) the circuitry, after it is. That much leakage inductance suggests a very low BW, so I would guess all sources will sound as bad as the next.

No, it isn't "Bose". Just some other brand that begins with "B".

Oh, someone asked if it made a difference, when it was on or off. I can confirm turning it on did not improve it.

Pat

[Joseph K]

B.. B...

Baudio Bote used to apply strange looking transformers, very low Capacitance 

That much leakage inductance suggests a very low BW, so I would guess all sources will sound as bad as the next.

Good point. And brings up the question: what input stage would be up to the level (speed -wise) of some well known spdif interface,  made by some kind of cousin of yours?

How fast are the receivers, inside?

Ciao, George

[art]

Not sure what is inside. I suspect all it has is a ASRC, on the underside of the PCB.

Well......come to think of it...........there is a switch, on the front panel, to switch between AES/EBU, SPDIF (RCA) and TOSLINK. So, could be they stick a '8416, in front of it.

As to what I would build, if we were in that business............

Since I used MECL, way back in the 70s, I still have a fondness, for that power hungry stuff. Wouldn't be hard to make something similar, with some BJT matched pairs. Although, it is tough (but not impossible) to find good PNPs, that have a high f_t. I'll let you ponder why using NPNs would be inferior.

(And, no, it has nothing to do with lower capacitance, noise, or anything to do with linearity.)

(Remember: MECL worked on -5.2 V. No more hints.)

OK, for anyone still following this, and wondering what he is getting at........

In an earlier post, I made mention of how much attenuation, the DAC would take. If you put a linear amp, in front of the '8416 (or whatever), you would be able to stick a bigger Mini-Circuits Lab attenuator, in front of the box. (Other benefits, but those are harder to show. At least, on the gear I made those measurements on.)

Pat

[jneutron]

Oh, someone asked if it made a difference, when it was on or off. I can confirm turning it on did not improve it.
Pat

I asked, doing so confirmed that there was no active electronic reason for the reflection.  Simply stating that it is inductance, while correct, was not rigorous.  Your test guarantees the waveform in this case, was not that of a video input node bounce.

Thanks.

Cheers, John

[art]

You are welcome. Still can't believe it is that much. I have measured some bad DACs in my days. This one is one of the worst.

(Can't think of many that I was impressed by. At least, TDR-wise.)

The worst may have been one with a 5-pole LPF, on the input. Built by a "W" company, supposedly run by a bunch of telecom experts. (Not their most expensive one, luckily.)

Pat

[JoshK]

I think you should make and sell a DAC. 

[JoshK]

B's.... Bryston, Bel Canto, Benchmark, Berkeley ....I'd guess Benchmark

W's....Wadia, Weiss, Wavelength  ....I'd guess Wadia or WL

[art]

All those "B" are wrong.

Yeah, one of the "W". I can assure you it is not Wavelength, as all he does is USB. (Right?)

As for us...........we can not keep up with what we make now. Every time we get close to getting caught up, something hits the fan.

(Like missing 6 weeks, playing nurse, to our "customer service babe", after her brain surgery. Yes, this is true: not making it up. Not that I am qualified, or possess the right temperament. Her family was unavailable, so guess who got the job.)

Yes, we have plans for a DAC. But, I have a sneaky feeling the next product will be another USB-SDIF doohickey.

Pat

[art]

OK, curiosity got the best of me, and I took all the screws out, and looked at the PCB.

No wonder it has such a huge inductive kick: there is a HUGE ferrite bead, in line, with each leg of the transformer primary. Ditto for the AES/EBU input. From there, the output leads go to a '4052 chip, along with the TOSLINK output. That is where the selection is done.

After that, a CS8420!

Yep, quality stuff here. I think we can put this part of the tale to bed.

Pat

[tpaxadpom]

it would be nice to see the eye diagrams for 5m and 1m cables used. Making the rise time faster one has to worry about the extra bandwidth needed to carry the signal.
Great writeup Pat.

[art]

Not sure I follow, but there is not enough pulse dispersion, in either cable, to matter.

The BW, of the DAC, used as an example, is much more of a limiting factor.

Speaking of which.........anyone in favor of a thread to fix its inherent problems?

Pat

[mgalusha]

Speaking of which.........anyone in favor of a thread to fix its inherent problems?

Absolutely, I learn a lot reading your threads and as always, thanks for the great information. 

mike

[art]

I'll try to start on it, tonight. First thing is to get rid of the ferrites, and see where that leaves us. Next, either futz with a zobel, or change the transformer loading. Won't know until I get started.

Pat

[wakibaki]

...you can make a pad, built right on the input (or output) jack, of your unit, for the cost of a few resistors. Yes, there may be some squiggles, that will show up on our TDR. But, with the typical HC CMOS chips, most of you use, you won't see the effects.

So, in that case, if you measured your DAC input with a DMM and found 75R, that would indicate a satisfactory termination?

w

[art]

No, it wouldn't. That is what vector network analyzers are for.

Pat

[wakibaki]

The long pulses will be 44.1 kHz x 32 = 1.4112 MHz.

Pat

SPDIF data is organized into blocks consisting of 384 x 32-bit words (sub frames). These are nominally subdivided into 192.frames each of which consists of 2 sub-frames, A-channel and B-channel. The sub-frame comprises a preamble, 4 bits of aux. audio data, 20 bits of standard audio data, and 4 bits that carry other information such as user data and channel status information, for a total of 32 bits.

The block structure enables more complex user data and channel status information to be encoded by distributing bytes across a number of frames. The 4 bits are called the validity bit, the user bit, the channel status bit and the parity bit, which is used for error checking (provides even parity of the audio data).

There are 3 forms of preamble (x, y and z), all having zero DC bias. Preamble Z indicates the start of a block and the start of sub-frame channel A. Preamble X indicates the start of a channel A sub-frame when not at the start of a block. Preamble Y indicates the start of a channel B sub-frame.

Each sub-frame carries one sample, whether 16-, 20- or 24-bit.

The bit-rate for carrying CD-standard music is therefore 44,100*32*2=2,822,400 bits per sec, biphase encoded.

w

[wakibaki]

...with the typical HC CMOS chips, most of you use...

... the rise time is such that...

...you won't see the effects...

...if...

...you ... make a pad, built right on the input (or output) jack, of your unit, for the cost of a few resistors.

...there may be some squiggles, that will show up on our TDR. But...

...this only indicates that, at these frequencies (rise times) the discontinuity is insignificant.

Which means that the discontinuity represented by the RCA connector and a few cms of PCB track is also insignificant, and a network analyser is not required to assess the termination, a simple DMM will do, if you discover 75R.

w

[Joseph K]

He said : "long pulses" - 1.4112 MHz
Two posts later he said:
"there are also short pulses, at 2.822 MHz."

So, what is your point here?

[wakibaki]

The clock rate is 64 times the word clock, not 32.

w