The SB3/SPDIF output thread.......

randytsuch · « **Reply #20 on:** 20 Sep 2007, 04:57 pm »

Quote from: justin_case on 20 Sep 2007, 04:49 pm

terminations with less parasitic inductance.

I could not see the pictures, then I tried copied the picture address' into a brower, and still could not see them.

Not sure if it is my work connection, or something with photobucket.
Maybe you could bring the pictures into an audiocircle photo album, and put it up that way?

Randy

tanchiro58 · « **Reply #21 on:** 20 Sep 2007, 05:45 pm »

Pat,

Very nice work of you. However, since most of us especially in my case that I am a DIYer (considering as an amateur) and do not clearly understand what components (resistors and capacitors) are used to modify with Newava S22083 which I have in hand in the SB3 according to your picture. I have not only a little understanding to your explanation but also hardly recognize the values of resistors you are using and what positions did you connect to. Therefore, would you please provide us some of the lists with exact values of the resistors and to which positions should these resistors be soldered. Thanks for your effort and time.

Cheers,
Tan

art · « **Reply #22 on:** 20 Sep 2007, 07:41 pm »

Well...only thing left to show is the output waveform. Items of concern are HF roll-off, and LF tilt. Here goes.....

Here is the Newava output:

Well, some LF tilt. I can cut it by about half with a larger cap. Who knows.....I may have a Black Gate collecting dust that I can get rid of.

But the SC output:

Oh, my....quite a bit of LF tilt. Unfortunately, adding C to the circuit did not help.

Yet another reason why I don't like them.

Why do I get the feeling that Mr. Case will have some words on this subject? Just remember, I did not design these transformers. Both have limitations. (As will any other......it is just a choice of which ones you can live with.)

Well, while we all await Mr. Case's reply, I am going to see if I can do anything about the SC. If it was my unit, I would call it quits. But Dan took a great leap of faith, sending his unit to some unknown goober in Texas, so I owe him my best.

But in closing.........remember what I said in http://www.audiocircle.com/index.php?topic=45330.msg406683#msg406683?

"A lot of you guys will be much better off by just taking out the output inductor, and leaving it alone."

I'm serious. This is much more involved than just swapping out capacitors and op-amps. Just because it is a piece of audio gear does not mean that you proceed at will, mucking with things that are beyond the grasp of the typical DIYer. You will notice that even engineers (I assume Mr. case has a similar background.......he at least seems to) can not agree on technical merits. But for those of you are brave, the info is here for you to do as you please.

Pat

justin_case · « **Reply #23 on:** 20 Sep 2007, 08:28 pm »

Well...only thing left to show is the output waveform.........

What is the frequency of the square wave you are measuring? What is the actual circuit you are measuring? (i.e. source and load impedance....what transformers?)

I assume Mr. case has a similar background.......he at least seems to)

mr. lower case or Mr. Upper Case?

BSEE 25 years Telecom design as well as (digital and analog) audio design (and redesign).

art · « **Reply #24 on:** 20 Sep 2007, 09:20 pm »

That is the SPDIF output. I hoped that I would not have to give the 'scope horizontal sweep rate, or 'splain why I can get 2 traces of the same waveform on my 'scope. (46B......) Only trying to show that it is not flat.

And of course..........it is 75 ohms!

I only hope it is at this point that you don't tell us that you are Pete Goodreau.

If you are........don't!

(Sorry for the typo...thought that I changed it.)

Source of LF tilt coming up. Making plots. Be patient.

Pat

art · « **Reply #25 on:** 20 Sep 2007, 10:03 pm »

May not be much hope for some of these, gang.

OK, for Mr. Upper Case, these plots were made on a H-P network analyser. Yes, it is a 50 ohm set-up. I could barely afford it, so I was not about to fork out for the 75 ohm s-parameter test set, which is nothing more than a 50 ohm s-parameter test set, with some matching pads inside.

(OK......I don't even have the 50 ohm s-parameter test set. Had to use a splitter and normalise the measurements.)

Anyway.........first up is the SC. I was not about to unsolder the one in Dan's unit, so I used the shielded one for this test. It is probably safe to assume that they use the same core, so LF response is the same. HF response will probably be lower in this one. Deal with it.)

Well, that 'splains most of it. The core limits the LF response.

The spec sheet says (I have it right here in front of me........):

Bandwidth extends from 25 kHz to 100 MHz.

Somehow, down 10 dB at 100 MHz isn't my idea of BW. At least they got the LF part right.

Oh, well.....

Here is the Newava:

A bit better on the low end. Much more on the top end. Both are important.

Just for grins, here is a Schott/Newava 22160. From my junque boit.

If it weren't for the even worse LF response, I would say they are the same thing. Even though they have nothing in common, construction-wise. Interesting.

Now the real mind boggler: The PE-65612. This one is a real pain in the @$$ to get right on the TDR. After the spike from the leakage inductance, the trace dips, and then rises up in a broad lump. Trying to get rid of that lead to my rethinking drive impedance.

Well, if LF tilt is important, this may be the one to use. Maybe some day I will figure out all the tricks to make it work the way I like. Hopefully, that will get rid of the crappy out-of-band response on the top.

OK, Mr. Case, I yield the floor to you. I gotta start making dinner...........

Pat

mgalusha · « **Reply #26 on:** 21 Sep 2007, 12:56 am »

Another vote of thanks for your efforts Pat. And to Mr. Case as well.

jhm731 · « **Reply #27 on:** 22 Sep 2007, 05:02 pm »

Quote from: art on 20 Sep 2007, 10:03 pm

Anyway.........first up is the SC. I was not about to unsolder the one in Dan's unit, so I used the shielded one for this test. It is probably safe to assume that they use the same core, so LF response is the same. HF response will probably be lower in this one. Deal with it.)

Well, that 'splains most of it. The core limits the LF response.

The spec sheet says (I have it right here in front of me........):

Bandwidth extends from 25 kHz to 100 MHz.

Somehow, down 10 dB at 100 MHz isn't my idea of BW. At least they got the LF part right.

Oh, well.....

Pat-

In fairness to SC, it says in the material I sent you, that the SC947-02 is optimized for use at 110 ohms.

The unshielded SC979-03, which you now have in the SB3 is optimized for use at 75 ohms.

I'll email you that data sheet.

justin_case · « **Reply #28 on:** 23 Sep 2007, 09:11 pm »

SPDIF mod for Creative Labs USB SoundBlaster Live! USB DAC. Got rid of 4 ceramic caps in and 2 SMT inductors in circuit! I added 0.1 uF ERO polypro cap between the driver IC and voltage divider before the transformer. Replaced tiny cheap pulse transformer with Newava. The driver IC is actually the Cirrus Logic CS8406 which can be seen in far right of picture. I have only listened to the analog with with capacitor mods. It sounds pretty good and much better than stock. The designers have slowed the stock SPDIF circuit down so much for reduced EMI that I can imagine how bad it must sound. This might be the cheapest route for a decent SPDIF signal out of a PC.... I paid about 65 to 70 bucks for parts and DAC.

Justin

Likes to modify cheap stuff in case he breaks it.

art · « **Reply #29 on:** 24 Sep 2007, 09:00 pm »

Well, if you had a perfectly symmetrical, 50% duty cycle stream of alternate 1's and 0's, we would not have to. Be awful hard to listen to music that way, so............

Here is a typical SPDIF waveform, taken from my CD player:

When we have LF tilt problems, it has the effect of moving the decision point. Yep, just a little bit, but then, jitter is usually only a little bit as well. So, both are important. Perhaps Mr. Case would like to add his observations on that subject (instead of interrupting with pictures of modded sound cards).

For those who are curious about the modded circuit (I put in Dan's doo-hickey), it is virtually identical to the one used to produce the above photo.

"Whaddya mean 'virtually'? It either is, or isn't!"

Well, first off, it runs on +5 V, as opposed to +3.3 V. It has a differential drive, which allows me to isolate the transformer from the output with a bigger resistor. The resistors that drive the transformer are 1/2 the value of the one in Dan's. So, pretty much the same thing.

It originally had only a 1 uF or 4.7 uF non-polar cap. I raised it a bit! You will have to take my word for it that the difference between the 220 uF in mine, and the 160 uF polar 'lytic that I put in Dan's is insignificant.

Oh, perhaps I should point out..........I went back and put a 160 uF 'lytic in parallel with the 0.1 uF DC blocking cap on both outputs in Dan's doo-hickey. Just as I stated earlier, the Newava is helped a bit, but the SC stayed the same. (I had hoped that perhaps the cap really wasn't going to ground, but I did it over, and the same result: no change.)

Just to make sure I wasn't fooling myself by using the wrong drive impedance, I varied the drive Z in my unit from 0 up to 500 ohms. No effect on LF tilt. (Lots of effect on ringing, overshoot, and level. Obviously.)

So, I think that about wraps up all I can say about sticking transformers into your doo-hickey. Just remember.........what I said........take out the inductor.........leave it alone.

And add a transformer on the RX end.

I have a few more thoughts on modifying these doo-hickey gizmos, but I will start another thread on that. Won't be much, but no need to clog up this thread with any other stuff. Perhaps that would also be the place for some of you to ask related questions.

Happy trails to you..................

Pat

justin_case · « **Reply #30 on:** 25 Sep 2007, 08:51 pm »

Perhaps Mr. Case would like to add his observations on that subject (instead of interrupting with pictures of modded sound cards).

Or maybe he will spend all his pizza money allowance for the next year on a cheap digital TV. Tilt is a function of the primary inductance and the transformer load resistance. for a given primary resistance, decreasing the tilt calls for increasing the primary inductance with more windings. More winding cause more leakage inductance and limit the high frequency response of the transformer and increase the capacitance between primary and secondary and decrease noise isolation. using a core material with a fairly high permability to get decent primary inductance results in choice of ferrite material whose high frequecny characteristics limit its high frequency responce and distortion characteristics..... No free lunch? Not even a reasonably priced one. On top of that,most people design for the least capacitance which seriously compromises good signal performance. Designing a good transformer is like tug of war with 6 different ropes tied to a central knot.

http://i220.photobucket.com/albums/dd264/justin_case_2007/Xformer.jpg

ezkcdude · « **Reply #31 on:** 27 Sep 2007, 06:51 pm »

Quote from: art on 24 Sep 2007, 09:00 pm

And add a transformer on the RX end.

I have a few more thoughts on modifying these doo-hickey gizmos, but I will start another thread on that. Won't be much, but no need to clog up this thread with any other stuff. Perhaps that would also be the place for some of you to ask related questions.

Happy trails to you..................

Pat

Pat, thanks, for doing these experiments. I came by way of the Slim Devices forum (I'm sure there will be more of us). My DIY DAC has a Newava S22083 on the input to the SPDIF receiver (CS8416). I'm just wondering, since I don't have a spectrum analyzer - just a lowly Tektronix 465B, would your "tilt" measurements work with this? If there are some tests I can do with my scope, could you give me suggestions, for example, which traces to check? And what to look for? Here is a pic of the PCB near the Rx:

art · « **Reply #32 on:** 27 Sep 2007, 09:26 pm »

I am not sure what you are asking. (Or why.) This thread is specifically about adding a transformer to the SPDIF output on a Squeeze-thingie. If you want to discuss another topic, it should be discussed in another thread.

No spectrum analyzer was used in any of these tests. The "tilt" measurement that I believe you are alluding to was made on a lowly Tek 465B 'scope. Gee, same as what you have.

I don't know what I am supposed to glean from what appears to be a PCB layout. It is not germane to this discussion.

Pat

justin_case · « **Reply #33 on:** 29 Sep 2007, 05:54 pm »

- just a lowly Tektronix 465B, would your "tilt" measurements work with this?

Time domain reflectometry... a scope picture of the tilt ( function of the low frequency limit) of a pulse waveform in the time domain. (Volts Y-axis.... microseconds X-axis) None of this has anything to do with spectrum analysers (amplitude Y-axis frequency X-axis)

The 465B is a fine scope and very useful for tool for SPDIF measurements. You have to know the fundamentals of electronics before making useful measurements and interpreting their relationship to the quality of the SPDIF circuit. PCB layout for RF range circuits is a very advanced topic,way beyond the scope of discussion for a forum. The shown layout has too much stray capacitance and low impedance transmission line sections. Some resistive loss pads, spit loading between primary and secondary and elimination of the ground fill around the transformer would help... but the specifics are beyond the level most people would be comfortable discussing on a forum. Ground fills (not correctly designed ground planes) look pretty but are disastrous for many circuits. They generally suck for audio circuits.

http://www.ultracad.com/prenticehall/prentice%20hall.htm

http://www.sigcon.com/bookHSDD.htm

art · « **Reply #34 on:** 29 Sep 2007, 06:29 pm »

Have you been reverse-engineering any of our stuff? Sure sounds like it.

Either that, or you know what you are doing. In that case, I will leave you to help our resident ME learn this stuff. I'm gonna watch football.

Pat

justin_case · « **Reply #35 on:** 29 Sep 2007, 08:22 pm »

I'm gonna watch football.

.... the pastime where TDR means Touch Down Replay.

or you know what you are doing ....

It could happen...after a few decades of experience, it's called engineering. Sometimes I even read a book or two on a given subject before reinventing the wheel in the lab.

Reverse engineering should be called the reverse of engineering because you often wind up copying someone else's mistakes or trying to adapt a circuit to a situation where it doesn't work very well. Doing it right is what puts the fun in knowing your fundamentals.

jhm731 · « **Reply #36 on:** 18 Oct 2007, 09:53 am »

Received my SB3 back from Pat (thanks for all your efforts) with the SC & Newava pulse transformers installed(Toslink connection was removed).

Initial listening indicates both outputs are a big improvement over the stock unit.

Will do some additional listening/break-in and try and determine which output sounds better.

randytsuch · « **Reply #37 on:** 18 Oct 2007, 03:51 pm »

Quote from: jhm731 on 18 Oct 2007, 09:53 am

Received my SB3 back from Pat (thanks for all your efforts) with the SC & Newava pulse transformers installed(Toslink connection was removed).

Initial listening indicates both outputs are a big improvement over the stock unit.

Will do some additional listening/break-in and try and determine which output sounds better.

Please keep us posted and

Thanks for letting Pat play with you SB

Randy

tanchiro58 · « **Reply #38 on:** 18 Oct 2007, 06:17 pm »

Quote

Will do some additional listening/break-in and try and determine which output sounds better.
Quote

jhm731,

It would be helpful if you open it up and share with us how to install the resistors and Newava pulse transformer (at what positions of the pcb on digital part of SB3) according to Pat modification after comparison A and B between two SB outputs. The pictures Pat provided were not so clear in this thread. Thanks for your kindness, effort and time.

Regards,
Tan

art · « **Reply #39 on:** 20 Dec 2007, 05:09 am »

I haven't made a post in this thread for almost 3 months. Yet, when I look at the traffic logs, I see where there are still lots of folks reading this every day. I suspect that many of them are being linked to here from some other discussion, on some other forum.

Anyone care to clue me in?

Thanks,
Pat