[tanchiro58]

Those yellow caps look very big, won't the inductance be too high  ?

Those caps are taken from the MHDT DAC. That are the only one I got of the value 2x0.1mF in hand for right now (testing how they sound in this circuit). Astoundingly these caps sound so good. I know this is creating too much inductance but I did not have a choice. Resistors are carbon composite (maybe non-inductive) and those wires are well sealed. Anyways I love the sound more than before with Newava by itself in the SPDIF output. 

[art]

Where do I start with you guys............?

First, Mr. UK, I did not say that it was you who said to use a 33 uF cap. It was, in fact, another guy. But more about this later.

Ok.....caps.......

You can't use any ol' cap for coupling. As someone has pointed out, tight layout and small size are important. Big film caps, with long leads, aren't going to cut it.

So, what kind of cap should you use?

Well, that is a good question. Choices could be crappy ceramic, small radial film, large axial film, regular electrolytic, fancy electrolytic (Black Gate, OsCon), and some nasty combinations of any of these.

As "audiophiles", I can hear most of you saying "BIG FILM CAP!!!!!! They sound really great in my.....blah, blah.....blah unit."

Won't work here. Too much inductance.
Ceramic: I would not use any type of ceramic anywhere that is not C0G/NP0. Unless you know what kind you have, stay away from them.

So, that leaves small film or 'lytics of some sort. Small values like 0.1 uF may lead to excessive LF tilt, which can lead to decision point shifting. 'Lytics, well if they are physically small enough (as in low ESL), they should work better.

"But how are they going to sound? Won't they sound worse?"

Possibly. Let's just say that they will sound "different". As "audiophiles" you will automatically assume that because it sounds one way (or the other) that it has to be because those caps sound the same way in audio circuits.

Wrong. Yes, they will all probably sound a bit different. Why????? Because they will have different amounts of ESL, and lead inductance. That alone will make things sound different. So, just because it sounds different only means that it sounds different. There is no way to correlate how it sounds to the way it sound in audio circuits. Although I know that a lot of you will.

Those long, flying leads that you guys have, mystery wire all over the place, RCA jacks, and all the other construction techniques that are not suitable for RF will mostly likely have a larger impact than your choice of cap. Especially when it has long leads. You can not use audio construction methods and rationale here. Won't work.

OK, it will "work", in terms of music coming out of the far end. Hell, you already have that. Unless you are going to do it right, then only take out the ferrite bead and leave the damn thing alone. (I already said that more than once.) I showed a TDR trace of what this thing will look like with just the ferrite removed. I think that I can safely say that some of you will make things worse, judging by the questions that you are asking and your comments. Just some, not all. Some seem to "get it". Others don't. Nothing personal, just my observations.

OK.....more on caps.This time: bypass.

Well, yeah, you can stick an OsCon across pin 14. Or maybe even a Black Gate (some don't look too bad at HF, hate to say.) But here is where you are going to have problems:

If you remove the ceramic cap on the PCB now, what is going to bypass the really high frequency crud? And you will have HF crud. The clock works at 11 MHz, so you will have at least the 7th harmonic to get rid of. So, now we are talking about a cap that is effective to at least 80 MHz or so.

Those fancy 'lytics won't do anything there. Sure, some will work at 11 MHz, but none at that frequency. Especially when you factor in lead inductance. (Just how close are you going to be able to mount them to pin 14??????????????)

So, the only thing that works is those nasty SMT ceramics. Yeah, I hate 'em as well, but what will work as well? At 80 MHz: nothing.

So, it almost has to stay. But wait.......now you have really opened a new can of worms.

You have a (ceramic) cap at pin 14. Some wire, or should I say "inductor" going to a 'lytic. It has some ESR, ESL, along with its capacitance. Which makes a very nice parallel resonant circuit.

What does that mean to all you non-EE types?

Simple. It means that at some frequency, your bypass circuit will actually be less of a bypass circuit. That frequency will depend on the lead inductance etc., but you can be damn sure that it will happen at a high enough frequency to render your bypassing totally useless.

And here is what is worse about it:

The lower the ESR of the cap, the worse it will be! Low ESR means a large "Q" of the resonant circuit, which means at that frequency, it will really be useless.

Two ways around that. (Well, 3 really, if you count not using the 'lytic.)

1.) Put some series R in the lead to the cap. Lowers the Q, makes the resonance not as sharp. Either that, or use a cap with high ESR.

"Doesn't that negate the reason for using that cap?"

YES! You are catching on.

2.) Stick a ferrite bead between the 'lytic and the ceramic bypass at pin 14. The ferrite will be out of the circuit at LF, so the big 'lytic will do its thing, and stabilise the rail voltage. At HF, it will help things out, up until the point the ferrite isolates it. The ceramic does all the work at that point.

3.) Yep, don't stick one in unless you have the engineering skills and equipment to measure what it is doing.

Now, mind you, I am not saying that an OsCon by itself won't work. It will. It just works a lot better when it is part of the initial design. Where you can make damn sure that it is close enough to be effective.. In this case, you have the 11 MHz crystal in the way. So much for short leads. You could take out the 12 MHz crystal. That might work. Oh, wait..........some git stuck a pulse transformer there. Well, if you were willing to accept longer leads to it..........well, you get the idea.

So, you guys make your own choice(s) on how to proceed. All I can do is tell you the technical pros and cons.

These matters are a subject of great debate in engineering circles, where no one has to worry about "how it sounds". The DIY forums have lots of debate on this subject, as now it does matter how it will sound.

There is no easy answer. Just remember that everything that you will do will have good and bad effects. Since almost none of you are EEs, then my advice is to do as little as possible.

(I guess in theory you could pay someone to do this. But based on some of what I have heard......the guys doing this have no more clue than the average member here. Just the chutzpah to pretend that they do. Another subject best left for some other time.)

Pat

[tanchiro58]

Pat,

Good to learn your lesson.

Would this Wima 0.22uF/100V metallized polyproplene be good enough to not having or less inductance in the SPDIF output? Thanks in advance.



Tan

[tonyptony]

Unless you are going to do it right, then only take out the ferrite bead and leave the damn thing alone. (I already said that more than once.) I showed a TDR trace of what this thing will look like with just the ferrite removed.

Pat, I'm curious as to whether you think the circuit mod would actually sound better than the L8 bypass mod, given the relative similarity in TDR traces between the two (I know the bypass trace has a nasty downward spike, but I'm not sure I understand what that would do to the signal). I'm not a DBT'er (not in the least) but at this point I have to wonder how much better the circuit version can sound over the simple bypass. Enough to make it worth it, given how carefully it has to be done?

[Nuuk]

Having more or less completed the new case, I was able to use 'standard' 75 ohm coax cable for my SPDIF mod.

A very short length of wire now connects the input pad of C66 (originally the 0.1uF cap)  to R63 (249R resistor) and the leg of the traffo is soldered to the the output of R63) The DC blocker is soldered to the other side of the primary and then to ground. That's about as compact as I can get it.

I just couldn't get that pesky little 107R to solder to the output side, hence the temporary use of a non-magnet/non-inductive 100R resistor in that position until I buy another SMD type. As you can see, I made sure that the coax was well  secured so it can't tug at the traffo. That involved removing the right channel phono socket but I won't be using analogue out anyway.

The result is the best that I have had from the digital output of the SB3 so far. I tried soldering that 47uF BG NXQ across the 220UF cap (immediately above the phono socket in the picture below) and that made a small improvement too.

[Nuuk]

This morning I replaced the 0.47uF film cap with a 100uF lytic (Panasonic ECA) and the difference is 'night and day'!

I have been comparing the results with those I got using the Transporter as a transport. Although the Transporter has been returned to its owner, I no longer miss it, although it would have been educational to compare the two side by side. 

And a big thanks to Pat for this mod! 

Tonyptony:

I never tried only removing the inductor (L8) but I can't believe that it could be this good, so providing that you follow Pat's instructions to the letter, I can thoroughly recommend this mod.

[art]

There is no way to positively say which one is better by how much. A lot of that depends on the cable and how crappy the SPDIF RX stage is. (Almost all are dreadful.)

If you have a good cable, and input stage, this mod will be noticeably better. In many cases, just yanking out the ferrite will provide enough of an improvement that anyone might be able to live with.

If in doubt......just take out the ferrite. Cheap, easy to do, almost no risk involved, significant improvement.

Transformer.....better, costs $$, takes talent, patience and attention to detail. Need to convert at least one end of cable to BNC. (Which should be the objective anyway.)

Redesigning the input stages is beyond the scope of this thread, and may be beyond the abilities of most readers.

Pat

[Bill@LakeGeorge]

I had ART do the spdif mod on my SB3 and after letting it settle in for a week or so I can only WOW.  I also got one of Pat's coax cables and a Channel Island VDC SB power source.  With all my music ripped to my PC.  I will never go back.  Pat was great to work with and turned this around for me in 2 days.

Thanks Pat

[crooner]

Ok, I'm giving in...

Ready to yank out my Scientific Conversion transformers on both Squeezebox (TX) and DAC (RX).

If I mess up with my SB again, I'm thinking of using a little circuit board for the transformer and necessary caps to make the thing look better and also to avoid the dreaded "flying cables". Should be fun!

Again, thanks guys for the great thread. With a big special thanks to Pat for his contributions.

[art]

If you are going to all the work of making a PCB. make it mate directly to the BNC jack. Better chance of teaking output RL.

You may even consider adding a zobel. I do not have the values, but they could be worked out.

One interesting thing to try in your scheme would be to use build-out resistor that would make the output Z of the chip to be 75 ohms. Drive a piece of 75 ohm mini-coax, and then put a resistor at the transformer to up the resistance to the required value.

Just a thought. Never tried it, but it could work well. Of course, some nerd with some fancy gear would have to measure it.

Pat

[crooner]

Excellent idea Pat!

I'll do some research on Zobel networks.

Attaching the BNC directly to the PCB is a natural.

Gotta put my ingenuity to work now! 

Oh, and in regards to the DAC side of things, my DSP-7000 has two available inputs. I converted the obsolete Glass AT&T input to coax using the Scientific Conversion trannie and BNC connector.

The remaining input is standard 75 ohm terminated coax. Factory (not modified).

I can rework the first input to accept the Nevawa or modify the second input keep both transformers in the DAC.

If you are going to all the work of making a PCB. make it mate directly to the BNC jack. Better chance of teaking output RL.

You may even consider adding a zobel. I do not have the values, but they could be worked out.

One interesting thing to try in your scheme would be to use build-out resistor that would make the output Z of the chip to be 75 ohms. Drive a piece of 75 ohm mini-coax, and then put a resistor at the transformer to up the resistance to the required value.

Just a thought. Never tried it, but it could work well. Of course, some nerd with some fancy gear would have to measure it.

Pat

[art]

Option 1.) Since you have 2 inputs, make one Brand X, and the other input, the decent brand.

Option 2.) Once you establish that the other brand is obviously better, despite tons of gibberish written to the opposite, "fix" that input with a better input stage. Then you will really have something to talk about.

(Find someone who knows something about video op-amps to help you........)

Pat

[crooner]

I listened critically to my system tonight and heck I still think my digital front end, the way I have it configured, sounds terrific to my (admittedly) tin ears...
And yes, I tried hard to listen to obvious abnormalities that would justify taking the thing apart again!

I'm thinking of several factors in play here:

1- My DAC internally reclocks the digital data, so perhaps it is impervious to the less than optimal 75 ohm connection (with the Brand X trannies at both TX and RX).

2 - The other modifications I've done to the DAC (Black Gate NXs in the output stage, among others) are somehow masking the deleterious effects of the Brand X implementation in the SPDIF.

3 - Ditto for the additional mods I've performed on my Squeezebox (12A power supply with HEXFREDs, ERS Stillpoints sheets, silver wiring, etc).

I have no doubt, however, that reworking the SPDIF from what I have now will be an improvement. Whether it will audible enough to warrant risking a few lifted traces on the fragile SB board remains to be seen!

Option 1.) Since you have 2 inputs, make one Brand X, and the other input, the decent brand.

Option 2.) Once you establish that the other brand is obviously better, despite tons of gibberish written to the opposite, "fix" that input with a better input stage. Then you will really have something to talk about.

(Find someone who knows something about video op-amps to help you........)

Pat

[art]

How does it reclock?

In theory, fixing the input makes the job of cleaning up the clock easier. Some schemes need more help than others.

Pat

[crooner]

Hi Pat,
From the EAD's manual:

"Digital Flywheel reclocking circuit is a state-of-the-art, propietary, dual time constant phase locked loop, utilizing a precision voltage controlled crystal oscillator (VCXO). The electronic equivalent of a massive rotating flywheel, Digital Flywheel smooths out instantaneous time-base fluctuations (i.e., jitter), achieving a ten-fold jitter reduction..."

Maximum Jitter (at 20 kHz) is specified as less than 10 picoseconds rms.

The DSP-7000 III was introduced in 1995, so that jitter figure was quite low then

How does it reclock?

In theory, fixing the input makes the job of cleaning up the clock easier. Some schemes need more help than others.

Pat

[art]

OK, analogue PLL. Didn't think that crowd used SRCs. Yeah, it would help some.

What is the SPDIF RX chip in the front of this thing. Something by Crystal, right?

I remember that they did use that AT&T fiber stuff. Done all wrong, just like all the other ones. (I once told one of their cronies in the industry what they did wrong. He wasn't amused. Oh, well.)

Pat

[tonyptony]

Hey I use that AT&T fiber stuff with my DSP-7000 Mk.III ! How's it all messed up, Pat?

[crooner]

Yeah, EAD used the industry standard (back then) Crystal CS8412 as the RX chip.

The digital filter on version III is the Pacific Microsonics PMD-100 with HDCD capability. It was added to an additional daughterboard containing the Digital Flywheel circuitry. The DAC chips, BTW, are PCM63P-K's.

While I never got to use the AT&T stuff, it was forgotten after only a few years. I figured it never really offered an improvement over 75 ohm coax.

OK, analogue PLL. Didn't think that crowd used SRCs. Yeah, it would help some.

What is the SPDIF RX chip in the front of this thing. Something by Crystal, right?

I remember that they did use that AT&T fiber stuff. Done all wrong, just like all the other ones. (I once told one of their cronies in the industry what they did wrong. He wasn't amused. Oh, well.)

Pat

[art]

Lemme guess.........you are using a 1 meter pigtail.

As for the EAD, with a PMD-100, and PLL, it has the promise of a good unit.

You may have to redesign 1/2 of it, but a good starting point.

Pat

[crooner]

As for the EAD, with a PMD-100, and PLL, it has the promise of a good unit.

You may have to redesign 1/2 of it, but a good starting point.

Pat

Hi Pat,

I agree about the PMD-100 plus PLL. Excellent combination. EAD took a lot of crap from Stereophool with their earlier interactions of the DSP-7000. The HDCD capable chip plus re-clocking circuit took the thing to the next level.

I wish I had the level of expertise to rework the unit to make it even better. I'll be content just to improve the SPDIF a bit, I guess!