Print

[art]

Ok...here it is....the SB3/SPDIF output transformer fix-it thread.

First off, I want to make some things perfectly clear:

This thread is intended for the education and edification of the AC members who are interested in learning something. If you are happy with the status quo of your unit, either in stock form or some mystical output kludge, then this thread is probably not going to be of interest to you. The handful of malcontents who want to kvetch, moan, gripe, complain, and just be an all-around PITA.......do us all a favour, and butt out.

Likewise, the content of this thread is not intended for commercial use. To the handful of garage shops, that modify stuff like this for more than it is worth, you need to ask yourself one question:

"Why am I reading a thread like this to learn how to make this stuff work better?"

If you need me to teach you how to, you are in the wrong business. Get a new job, and butt out of this thread.

As for what this thread will entail.......

I will be working on "fixing" the SPDIF output on the SB3. I will present a series of measurements documenting the procedure(s). This will be a strictly "by the meter" thread. At no time will I attempt to listen to the results. Dan (jhm731) will do that, once the unit is returned to him. I not only will not listen to it, but I have neither the computer h/w or s/w to do anything with it. Besides........I almost never listen to the stuff I design. It is too tempting for someone who has done this stuff as long as I have to get carried away by my knowledge of the design process. However, since this is an RF sort of thing, it really takes someone with my background to really do justice to this sort of mod. Ergo, we will be going only be the meter on this one. Afterwards, Dan will get the task of deciding it this is really a case where the meter rules or not.

Everyone who gets anything out of this thread owes him a really big "THANKS!!!!!!!!!", as he was the only one with enough chutzpah to serve up his personal gear to be the guinea pig.

OK, now that we have all the unpleasantness (and my bad attitude) out of the way, let's start.

First off, I looked at the stock output on the 'scope and the TDR. Oh, my.................the waveform shows some signs of HF roll-off, but the TDR...........UGH!!!!!!!!! Lots of inductance creeping in somewhere.





(BTW....some of the photos did not come out as well as I hoped. Would take someone with 3 hands to better some, and someone with better photography skills on some others.............and both on some others. Anyway, I think they are evident enough.)

The stock output is a simple enough circuit: 2 gates of a 'HCU04 driving a cheap ceramic cap, and a voltage divider. Bunch of PBC trace......and........oh, my........what is this............a nasty SM inductor. Presumably to keep EMI down.

Well, this is DIY country........sod EMI.........out comes the inductor!

Ah, much better!



Impedance is off a few ohms, but for starting point, good enough.

[art]

OK......now what some of you are really dying to know........what happens when you insert a transformer into the circuit.

First off, the one most here have been bamboozled into buying: the SC947-02.



Well, I think most of you know what I think of this. And that point was made in the last post.

Oh, before I forget.................the horizontal scale on this set of TDR shots are not the same as the thread where this issue was raised. Due to all the nasty inductance in the stock circuit, I had to change it. (Exact setting not important yet.........when it is, I will tell you.) The vertical is still 0.1 rho/div.

So.........what if you go to the same design, but without that stupid shield???



Well, since coupling capacitance goes up, the leakage inductance goes down. (Were any of you paying attention last time I pointed this out?) Anyway, the trace looks better. Not great, but better.

I think you guys need to seriously rethink your decision to use the shielded version. Let's assume that you buy into SC's claim of great CMRR and HF noise reduction.

On the RX end.

What HF noise are you going to eliminate on the TX end? And how does CMRR help on a single-ended drive? Anyone..................? I'm waiting.................

Well, while you ponder that one, Dan was kind enough to provide a Lundahl LL1572.



Oh, my......this is rather disappointing. Other than the fact it does not use a teeny-tiny ferrite core (that can easily saturate), I see no need to look at this one any further. And I am not sure why they chose a pinout that makes a decent RF connection difficult. As much as I bad mouth the SC, the do come in SMT. (Well, they could use pins 2 &3, instead of pins 1 & 4, but hey............better than Lundahl.)

[art]

OK.....so let's say we decide to use the unshielded SC as our starting point for the mod. How to do it best............?

Well, to Mr. RF here, the first thing to do is get rid of the mystery PCB trace, and go directly to the output jack. Here, we run into a problem:

Use the stock RCA (bleah!!!!!!!!) or go to a BNC.

Well, since a lot of you guys are going to stick with RCA, I will use it on this phase. If you look at the TDR trace, you will see a downward spike, due to the capacitance of the lower-than-75 ohm-RCA-jack. But, that is not all that bad, when compared to the leakage inductance spike of the transformer.

So, here are some photos of running a mini 75 ohm coax, right to the RCA jack. The coax is either RG-179 or -187.........I forget, as I bought this stuff surplus >20 years ago. Both are roughly the same, so it really does not matter. Likewise, you could use some Belden 9221. (I think that is the p/n......have not bought any in 15 years.) It is a very flexible coax, cheap, even when they make you buy 100'. Drawbacks are the foamed insulation may melt rather easily, for DIYers. The HF loss is a bit more, but can be put to use in the right application.

[art]

Ok.........next step. And here is where the problems really crept in.

I decided to remove the TOSLINK, as no one in their right mind would use that. I planned to put a small driver IC there, go to a Newava, and then to a BNC.

Sounded like a good idea.

Too bad it did not work like I wanted.

The output had too much noise for my satisfaction. No amount of filtering the supply line, bypassing the chip did anything to help. I had ferrite beads........resistors.......in the +3.3 V rail........Black Gate caps as bypass. Etc., etc., blah, blah.

So, time to regroup. After staring at the unit for way too long, I gave up and decided to come back another day. To no avail. Then it hit me...............

I am going to cut the trace from one of the output gates, and use that to drive a Newava. Whatever noise that is present on the supply line or ground reference will affect both circuits the same. This will allow for a much closely matched test pair. One will have SC/RCA......the other Newava/BNC. I may make a few "minor" circuit changes, as I am Mr. RF.

No.....I am not going to state what, if any, changes there will be. No need to skew Dan's perception before he even hears it. Let's just say that it will be as close a comparison between a SC with a RCA and and Newava with a BNC as possible.

With maybe........just maybe........Mr. RF doing some little tricks of the trade.

Only problem is that Mr. RF is out of Newavas today. More on the way from The 'Key...........

(OK, I stepped on the last one that I had..........what a klutz!)

Ciao!
Mr. RF

[JoshK]

Thanks for doing this. 

[GBB]

Thanks for doing this. 

Art - my thanks as well.

I do have one question / comment.  Is there any chance we can convince you to put BNC connectors in place for both the SC transformer and the Newava transformer?  A comparison of an SC transformer to an RCA output jack vs. a Newava going to a proper 75ohm BNC seems unfairly weighted against the SC.  Even if the transformers were equally good, the connection to the RCA would probably make it sound worse.

Thanks,
---Gary

[art]

Thanks for doing this. 

See.....that wasn't that hard, was it? All it took was for one guy to say "Thanks!"

And one to be very brave......

As for the output:

Remember, this is not my unit. It will cost Dan another $9 (my cost) for each BNC I add. So, I think it is his call. However, I would be curious to know how many guys who are futzing with theirs will stick to RCAs, or are ditching them in favour of BNCs. May not affect Dan's decision, but I would like to know.

OK........today's technical query to ponder...........

Dan was able to provide the data sheet for the -947 transformer. For those of you who also have one, think about this:

The leakage inductance is 350 uH, and the interwinding capacitance is 3 pF with the shield left floating.

Which I should have pointed out more than once that all the measurements of that part were in that configuration. See previous comments on why I hate shields, and why I would not even attempt to connect it to ground.


So.......if anyone remembers how to calculate nominal impedance...........why do you think this part (or really, any of them, including Schott and Newava) are designed to work into a 75 ohm load??????

Yes, it has to look like 75 ohms to the outside world, but...........internally..........hmm.. .....Mr. RF puts forth a brain teaser.

Enjoy!

Pat

[justin_case]

Drive it with a lower input impedance.... say 50 ohms... and use a 25 ohm build out resistor on the output. Don't forget a good 0.22 uF input cap for the transformer. A physically small 63 volt metalized poly-something would be great.

To be more specific and in the right thread.... the 75 ohm output driving the pulse transformer should now driving a 0.22uF coupling cap into a 150 ohm resistor to ground. This now looks like 75 ohms in parallel  with 150 ohms which is 50 ohms and is AC coupled. Pulse transformers will distort with any DC, even less than a millivolt. The primary of the pulse transformer is across the 150 ohm resistor (in parallel with the transport output impedance) and its secordary output impedance now looks like 50 ohms for a 1:1 transformer. Add 25 ohm resistor in series with one leg and you now have 75 ohms out. The output level will be reduced though but most DACs should still work with less than the half a volt standard level for SPDIF signals.

Justin

[art]

The last few days, while I have been suffering with the flu...........

I had some thoughts: How to give Dan the most bang for his buck, and show you guys some tricks that you won't find anywhere else.

Well, thanks in part (he adds, sarcastically) to the latest poster to this thread, it came to me.

First, let's review some stuff. (Almost has the sound of some boring class you used to have..........difference being that you will actually learn something.)

Most of you believe you only have the following options:

Stock......as in Picture #2. Yuk.

Take out the EMI-lowering inductor, as in Pic. #3. Looks good, but no galvanic isolation.

Put in a transformer.........stuck right on the output of the stock voltage divider, and get Pic. #4. Also yuk.

OK.....how do we make "Yuk" into "Damn, that looks good"?

First, the RCA jack has to go.

Second, use the unshielded SC transformer instead of the shielded one, as in Pic. #5.

Still......leaves something to be desired. So, now what?

Well, just in case anyone is still reading this and did their homework on characteristic impedance, you have come the conclusion that none of these parts are really designed to work at 75 ohms.

Sure.......they are all 1:1 transformers, but so what? Has anyone stopped to think that maybe..........just maybe........they really work better at some other impedance?

Well.....ok, at least one person has. So, on to the next stage.

I'll forego the process for now, but most of these transformers work better in the range of.........oh.....say 150-300 ohms. Some higher........some lower.

So, what if we do as our mystery poster suggests:

Use the same parts, just change the order.

"HUH?"

You guys assume.......and with 99.99% of you not being engineers (which is fine.....we have too many unemployed old fart telecom engineers as it is.......don't need any more), that all you have to do is stick a 1:1 transformer on the output of the right value voltage divider, and "voila!"......instant transformer output.

Nope. The transformer should operate where it was designed to operate. In the case of the unshielded SC transformer, we are talking >300 ohms.

So.......what if we take the same circuit, and move the 100 resistor to the output side of the transformer.

Well, you get this:



Gee.......that almost looks useable. (Wish I could say the same for the alleged "auto-focus" on this camera!) Same parts (as in #5), sans RCA jack.....much different results. Imagine that. (OK........I used 100 ohms on the output instead of 102, because that is what I had. And I cut one of the drive gates loose, in preparation for adding a Newava. But you get the idea.) The output Z is a bit low, but if we raise the resistor on the input side.....up closer to what it really should be, we get this:



There we go!

Here is what the hook-up looks like:



Eventually (when I get over the flu), I'll get the right resistor, instead of using a 249 ohm in series with 40-something. All I had on hand. You can not see the 0.1 uF film cap, used for the DC block. It is under the transformer. One end is soldered into a ground via, the other to the transformer. I used a Vishay 1826 series (which is no longer made.........), because.............yep........that is what I had on hand.

BTW........prying up that sticky foam tape stuff is a bit of a hurdle. It is really strong, and there are SM parts on the PCB to be careful of. You can see their outline. Pry it loose carefully.

Ok, the "no fee lunch department" speech.............

We can only raise the input resistor so much on this particular device, because it works on 3.3 V. Raise it up to where it probably should be, and we will have an output voltage that might be a bit too low. So, this circuit should put out around 0.4 V p-p, with a 75 ohm load. We will know once I put it all back together. For now, let's just concentrate on getting the impedance to look right.

[art]

Ok........back to my beef with SC.

Most of us are aware that they have an AES paper, where they show how much better their parts are than the next guys. They parlay that into more sales.

But................how many here have actually seen their test jig? I know of only one person (who found it somewhere on the 'Net), and he says "Yeah, you're right. They have all kinds of damping networks and other stuff hanging all over it.

Why, you may ask?

Well, because they know it was not intended to work at 75 ohms! So they built a jig that allowed it to operate where it was at its optimum.

Hey, can't blame them. Can you? Nope, any of us would have done the same thing.

But wait........if the jig is optimised for theirs......(and I have shown that there are ways).............then that means...............yep!......it can not be optimsed for Brand X. Because Brand X (and Y and Z) have different amounts of interwinding capacitance and leakage inductance. So, they will have to operate under different parameter, in order to show their optimum performance.

Are you going to go all that work to make your competitors' stuff look good? No. Especially when the goal is to make yours beat the pants off of theirs? No.........!

So, without a level playing field, who is say who has the best part?

Well, marketing is just marketing, as far as I am concerned. That AES paper is just marketing. And damn good marketing, it seems.

How 'bout this for "good marketing".........

Let's use National Semiconductor as an example:

How many guys here would buy their amp chips if they did not provide good datasheets and app notes? Not many. So, they show you how to hook it up......warn you "Yeah, you can use it without a zobel network, but we don't advise it", etc.

Good marketing.

So, how friggin' hard is it to provide a 2-, or 3-, or 4-resistor schematic to show how to use it the right way? Not very hard from where I sit.

OK, to be fair, none of their competitors provide that info, either.

None of them brag about how much better their stuff is, either.

Oh, I know..........most folks will say "Well, it was intended for engineers to use. Not hobbyists. Real engineers would know all of that."

Well, bull! Most engineers that I know would not think to play games with different drive schemes and/or impedances. Yep, they would do the same thing damn near everyone here would do. Just look inside whatever gear you have at your house!!!!!! Most, if not all, is poorly designed . Some engineers who make fairly well-known stuff will say that these transformers work best at really low impedances. What are they smoking?

The bottom line is...............sometimes you actually need to have a detailed knowledge about what you are doing. I am not going to berate any of you guys for wanting to improve your doo-hickey. But this is one case where blindly trudging forward can actually end up taking you backwards a step or two. A lot of you guys will be much better off by just taking out the output inductor, and leaving it alone. Unless you are brave enough to risk messing the unit up, you might want to stop there. Yes, transformer isolation will help. But unless it is done properly, it is something that I would advise against.

My goal is to show there are right ways and wrongs ways to do this. Do it the right way, and you can improve your sound. Even using a part that I don't like, you can get good results. But.....you have to know what you are doing when it comes to RF. And SPDIF is RF. You can not stick a piece of silver-plated microphone cable into the chain, and expect it to work better. Just because "Hey, this stuff sounds good at audio........it has to be just as good at SPDIF!"

Wrong. Yet, I see that thinking all the time. Even from folks who should know better.

Well, draw whatever conclusions you like.

My conclusions:

Use a 1:1 transformer only.
Do not use a shield.
Use BNCs, not RCAs.
Drive the transformer with the right impedance. 3.3 V operation may require some compromises.

Next....we all sit and wait for my order from The -Key.

Later...........

Pat

[randytsuch]

Thanks for the info Pat.
I have seen threads like this before, at diyaudio, and later at Jocko's forum, but I don't remember ever seeing somebody do the work like this, with TDR pics. 

Really helps to see how much better you can make it.

Randy

[art]

Just so the DIYers don't get that feeling that they are wallowing in despair, all alone, doing everything the wrong way.......

On our first D/A boxes, I loaded the secondaries with 22 ohms or so. I wanted the bases of the input transistors to see as low a Z as possible. Maximising BW was the goal. It had a build-out network on the input, so the leakage inductance hump didn't mess things up.

When I started making SPDIF drivers, I made a lot of the same mistakes. I did the obvious dumb thing, and matched both the primary and secondary to 75 ohms. I wrote a Mathcad program to come up with all sorts of variations on the network values.

Then I had this huge hump to deal with..............and it would not go away.

Eventually........by the time we were out of that segment of the business......I figured out why that big hump was there. Wrong impedance. I was able to get rid of some of it by using some other tricks, but it never completely went away.

I learned. Took a few years. And I had the RF background. Sometimes, you get too hung up on one tiny detail, and end up making a muck of the whole works.

Pat

[Builder Brad]

Pat,

Thank-you for putting the time and effort into this "project"

I was concerned that you were going to get upset by the minority here with fixed view on this subject.

Out of interest do you feel that the RX end requires the same level of attention?

Brad

[art]

Of course! But this is for a specific application. We would need a similar guinea pig for an RX experiment. I have no idea what any of you guys are using for this sort of application. We stopped making D/A boxes over 10 years ago.

StereoPhoole discovered that they had jitter.............

The market collapsed.

Pat

[justin_case]

I'll forego the process for now, but most of these transformers work better in the range of.........oh.....say 150-300 ohms. Some higher........some lower.

No.............. then the primary inductance gets become significant and your waveform
drops the low frequency end. The transformer should show good matching from about 50kHz to 35 MHz. Leakage inductance generally increases with primary inductance. It is easier to find and/or design a good pulse transformer for 50 ohms than 150 ohms.

The discontinuity from RCA plugs looks more like a lumped sum (capacitor) than an transmission line impedance mismatch and can be fixed pretty well with a zoble network (about 8 turns of wire round a 1/4 watt 75 resistor) in series right at the RCA jack or plug. Transformer leakage inductance best fixed with a series RCs right across the transformer windings. Looking at reflections at the source of 20 foot 75 ohm cable allows you to perform a good TDR with a 100MHz or greater scope. I like test with a single with the same spectral content as what the circuit will be used with. Even one BNC to RCA adaptor will screw up the sound of a SPDIF digital interface. Stay BNC or Zoble compensated RCA for the whole interface. Cable with BNC at one end and RCA at the other are compromised since most digital cable sound better in one direction than the other.

[Builder Brad]

Hi Pat,

I am using a Paradisea DAC, fitted with one of the Sc trannies, I would happily send this to you for testing, although I am in the UK. Should be in Vegas early December  , but it may be easier for other AC members to help out with this.

Brad

[art]

I'll forego the process for now, but most of these transformers work better in the range of.........oh.....say 150-300 ohms. Some higher........some lower.

No.............. then the primary inductance gets become significant and your waveform
drops the low frequency end. The transformer should show good matching from about 50kHz to 35 MHz. Leakage inductance generally increases with primary inductance. It is easier to find and/or design a good pulse transformer for 50 ohms than 150 ohms.

The discontinuity from RCA plugs looks more like a lumped sum (capacitor) than an transmission line impedance mismatch and can be fixed pretty well with a zoble network (about 8 turns of wire round a 1/4 watt 75 resistor) in series right at the RCA jack or plug. Transformer leakage inductance best fixed with a series RCs right across the transformer windings. Looking at reflections at the source of 20 foot 75 ohm cable allows you to perform a good TDR with a 100MHz or greater scope. I like test with a single with the same spectral content as what the circuit will be used with. Even one BNC to RCA adaptor will screw up the sound of a SPDIF digital interface. Stay BNC or Zoble compensated RCA for the whole interface. Cable with BNC at one end and RCA at the other are compromised since most digital cable sound better in one direction than the other.

I don't know if you are agreeing or arguing..........but here goes:

Yes, it is better to design a transformer to operate in the 50 ohm range, but the fact is most of them don't work right there. Which is why your scheme of driving the primary of a Newava with 150 ohms or so (you have to take into account the impedance of the drive gate into the equation), than to drive it with 75 ohms, as the typical voltage divider scheme does. Proof upcoming in next post.

Yes, once you get that part right, you can cancel out the leakage inductance with a zobel network.

BNC all the way, in my world. Yes, you can compensate for a RCA, but if you do.......adding in those BNC-RCA adaptors muck things up.

I think we agree; we just express it differently.

Pat

[art]

OK.........if my memory serves right.....it sometimes does.........they work best at around 170 ohms or so. Only thing I had close was a 249 || 562 ohms. Terminate secondary with 44 something (what I had on hand), and stick around 41 ohms in series.


Gee, that sounds a lot like what Mr. Case said. Right????

OK, gang......here it is:



So........of all the transformer schemes, which looks best? (You can comment as well, Mr. Case.) I think your idea does. Even if you do not agree with my reasoning for doing it that way.

My proof: Which looks worse?

The ones with lots of leakage inductance, driven by 75 ohms. That is which ones.

OK.......something I need to 'splain.............

You will notice that there are 2 positive-going spikes on the last photo. The first one is the added inductance where the centre conductor breaks out of the coax. While it may not mean diddly-squat at SPDIF frequencies, it shows up when you use a TDR with a BW >1 GHz. Overkill? Maybe.

Here is a picture of what I mean:



So.....after 8" or so of coax, there is another similar spike. Part of that is the same effect.....and the rest........yes, Mr. Case......the leakage inductance. Which can be wiped out with a zobel. Maybe the right one will compensate for both.

So.....for those brave enough to attempt this mod, here is what it looks like, using the parts I had around today. It is easier to see the 0.1 uF DC blocking cap in this picture. (Lower right corner of the transformer.)



OK, all for now, unless there are questions.

Pat

[TomW16]

Wow!  I have a soldering station and now a little bit of knowledge.  How dangerous am I    Thanks for the step by step of what you did and why.  Much of it went over my head but I have a yearning to learn more.  Thanks Pat.

Cheers,
Tom

[justin_case]

Terminations with less parasitic inductance. Tin coax sheild with solder 360 degrees around cable, trim with flush cutters, and terminate to sheild with as short a lead as possible. I cut center conductor pin very short and solder the center conductor as close to the sheild as I can.