Problem with SPDIF comparator (Elso)

[Fred Dieckmann]

* but were afraid to ask.

The input circuit show is not my creation but that of Scott Nixon who builds some of my favorite DACs (I have two of his DAC boards)

In previous life I designed digital interface devices and digital cables for the Audient Technologies products. I am a serious advocate of pulse transformers and they belong at the transport end of the interface where reflections due the impedance mismatch are absorbed before the round trip delay from the cable propagation time complicates things.

The design of a pulse transformer for this interface is a very complicated thing. a good pulse transformer shoud be flat from about 50KHz to 30MHz. 1 to 1 turns ratios are best since you can do bifilar winds and get more precise impedance matching and lowest leakage inductance. I was fortunate to be tutored tutored in this art by John Marshall who designed transformers for Schott and was an extremely knowledgeable designer. Almost all pulse transformers are designed for the lowest capacitance between winding (and hence the lowest common mode noise coupling possible.) This leads to some serious compromises in impedance matching at frequencies above a few MHz. This impedance mismatch is caused by the permeability changes and resistive losses in the core material as frequency increases. Also Reflections at lower frequencies are due to the primary inductance of the transformer.

I designed a product called the Tactic which was an external module containing a hand wound pulse transformer of my own design. I traded off an increase in capacitance for improvements in distortion and impedance flatness to several 10s of MHz. The Tactic also contained zoble networks for neutralization of leakage inductance, a few dB of intentional resistive loss (a pad often used RF interfaces) , and battery biased ERO polypropylene coupling caps. Oh yeah... it retailed for $350.

For the DAC end product was a module called the Audit that contained an analog devices monolythic differential video amp to buffer the digital cable from hysteresis and impedance mismatch from the DAC digital input. The digital interface is about as sensitive to component quality and cable microphonics as the line level analog interface and development of these products took over two years of computer modeling, listening test by several people, and TDR measurements. This is not simple interface and is the bottleneck for most Transport and DAC combinations. RCA connectors work very well for digital interface with a simple two element network soldered in the male connectors for better impedance matching. They sound better than many BNC connections that I have listened to. You can hear the difference in caps in this interface and small polypropylens of 100 volt ratings or less sounded best
all though I did hear a 100 volt RelCap Hematically seakel 0.1uF teflon that sounded very good but was about $45. A 1uF Black gate also sounded very good.

http://www.google.com/search?q=%22Au...UTF-8&filter=0

[jean-paul]

Hi Fred, thanks for clarifying. What exactly do you mean with two element network solders in the male connectors ?

I am surprised you don't have a problem with RCA/cinch connectors for digital interfacing as they are not designed for 75 Ohm usage

Regards,

Jean-Paul

[Fred Dieckmann]

The pretty shot path of a RCA connector can be aproximated by a capacitance. An L R network can be matched to pretty much null this impedance. I sold the design to a cable manufaturer so I don't want to be more specific than that. I used Cardas RCAs and I hear Keith Eichmann's Bullet Plug works even better for the male RCA but is difficult to terminate the shield the the connector ground.

[Elso Kwak]

Hi Fred,
Scott Nixon informed me:
"You don't need the caps in front of the inputs of the receiver, and the 2k should see the node between the hex sections and ground."
I don't understand this as I did my very best to translate the words of your old post into a schematic Philw asked for. Please advice.

Rather than using a resistor/cap network in the RCA connector I prefer using 75 Ohm BNC, Lemo or F-connectors at both ends of the cable.

Your explanation of the manufacture of the transformer is interesting but hard for the average DIYer to build or clone. Leaving us at square one: how to get the most out of the interface? Or asking the question what is the best ready built transformer for use at the transport end? For me the vagueries of the transfomer are a reason to omit it alltogether. I did hear a improvement when I omitted the original transformer in the Philips and Sony player. But of course these trannies aren't up to your standard....
Elso/not into SPDIF anymore
/not into transformers

[Fred Dieckmann]

I think Mr. Nixon advocates the use of another 2K resistor to act as a pull down to ground for the first hex inverter section. Might be worth trying but I have not used it.

The use of an UNBUFFER 74HC04 is required to make the first hex inverter into an amplifier. The 2K and 47K ohm resistors make the first Hex inverter look like an inverting amp with a gain of about 23.5. This allows 0.5 volts to overdrive the hex inverter/amp to acheive full logic transiston. The analog input signal is being conveted to a logic level digital signal by this stage The next hex inverter speeds up this logic transistion rise and fall times and buffers the first hex inverter/amplifier stage to drive the other logic. No capacitors are required at the schmitt trigger type inputs on the Crystal semiconductor inputs since they are now being driven at logic level inputs that swing between 0 and 5 volts. Do not use an inverter to drive the other reciever input as the unequal propagation delay to the receiver inputs due to the extra inverter to the screws up the sound. This circuit works quite well and was proof to me that Mr. Nixon is a talented designer. ( that and an amazing amout of excellent advice on audio design he has given me over the years! )

You would not believe the input circuits I have seen on some very famous and expensive DACs. I once very noticably iimproved the sound of a 5000 dollar DAC by removing half a dozen parts from the digital input circuit. Oh it had a nice 75 ohm BNC connector on the input, but input circuit didn 't even look close to a 75 ohms resistive input Jocko and I laugh are butts off when we see DACs with 50 ohm BNC connectors wired to input receivers with 4 inches of 110 ohm twisted pair wire. Impedance matched input? I don't think so! I have measured everything from 35 to 150 ohms on a "75 ohm" digital input. That and inductors, capacitors, relays, switches, and wire or printed circut board traces that were not 75 ohms; allowed me to sell quite few of my external input buffers to the owners of some very expensive DACs.


"resistor/cap network in the RCA connector" That might be because it is an inductor and resistor required for an RCA plug, Elso.......

The problem with BNC connectors is that some are made of pretty nasty materials compared to a good Cardas RCA connector. That and a typical RCA BNC adaptor that gets used when mixing RCA and BNC connectored equipment are the kiss of death for a digital interface. They sound really bad and I sold over a hundred custom adaptors made from Cardas and Canare connectors to lots of happy customers. Digital cable with an RCA on one end and BNC on the other cannot be reversed. Digital cables are often very audible for direction as the cable does not maintain the same caractistic impedance down its length and the termination impedance at the connector varies somwhat.

[Elso Kwak]

Hi fred,
I am very amazed Scott Nixon did not incorporate the 74HCU04/74HC86 frontend for the CS8412 in his DAC-Kit. The cost is minimal. And still is using the RCA for SPDIF input without the inductor and resistor network you so kindly corrected me on.....

[dhaen]

Slightly off topic, so please indulge me for a moment:

Why is there such an obsession with SPDIF?
I know it's the domestic common format, but this is DIY, we can do as we want

Doesn't anyone use balanced AES /EBU format with XLR's (at 5v P-P)?
This is a super format that goes for enormous distances with minimal deterioration. You can directly couple into the Crystal receivers, so there are a minimum of stages.

I use pulse transformers at each end, and standard 110 ohm balanced cable. The voltage waveform looks imperfect because of the pulse transfomers' reactance, but the current waveform looks just fine.

I have an adjective deficiency for describing sound, so I respectfully suggest you try it yourselves.

Thank you for your time..

Cheers,

[PhilW]

quote from Fred
>Do not use an inverter to drive the other reciever input as the
>unequal propagation delay to the receiver inputs due to the
>extra inverter to the screws up the sound. This circuit works
>quite well and was proof to me that ...

I'm still trying to get my head round this. Please could someone help me on this ...

Is Fred referring to the schematic earlier in this thread and saying that the second gate of the 74HC86 should NOT be used to drive the CS8412 differentially because it produces a bad effect by introducing unequal propagation delays?

Or is he saying DO use a 74HC86, because the alternative of doing it via an inverter produces the bad effect.

[tiroth]

You've got it right the second time round.

[Elso Kwak]

Originally posted by PhilW
quote from Fred
>Do not use an inverter to drive the other reciever input as the
>unequal propagation delay to the receiver inputs due to the
>extra inverter to the screws up the sound. This circuit works
>quite well and was proof to me that ...

I'm still trying to get my head round this. Please could someone help me on this ...

Is Fred referring to the schematic earlier in this thread and saying that the second gate of the 74HC86 should NOT be used to drive the CS8412 differentially because it produces a bad effect by introducing unequal propagation delays?

Or is he saying DO use a 74HC86, because the alternative of doing it via an inverter produces the bad effect.

Hi Philw,
I believe the second circuit I posted is the correct one as Fred meant. The trick with the 74HC86 is the inverted and non-inverted signal pass the same number of stages which is not the case if you use a extra inverter to obtain the inverted signal.
It was used in the Pass DAC D1 schematic to obtain inverted and non-inverted DATA!

John,
The schematic I made with the OPA603 and the AD8561 comparator is totally off AES/EBUnorm. It produces a 2.5V signal at the input of the comparator. I feel coax cable is better suitable and defined than 110 Ohm twisted cable. Jocko told me the latter is hopeless or words of the same meaning.