Head-Fi.org › Forums › Summit-Fi (High-End Audio) › High-end Audio Forum › Audio-Gd Master 7 - Discrete Fully Balanced DAC (PCM1704)
New Posts  All Forums:Forum Nav:

Audio-Gd Master 7 - Discrete Fully Balanced DAC (PCM1704) - Page 67

post #991 of 3640

The I2S input is 150 ohm.

Interesting.... Why does my Master 7 RJ-45 jack connector board have 220 ohm resistors for I2S termination?

Edited by DACLadder - 12/1/13 at 8:01am
post #992 of 3640
I do not know and when I asked kingwa I asked for the m7 and the ref 10.32.
Same answer for both so now another misery. . I would like to ask Steve nugent but I would like to know about this first.

Are you saying the wire is not the termination to match or as I am starting to understand is that that is what we need to match. So with i2s in twisted pair format should be about 120 ohm termination resistance , and now you say yours and I a sure mine is the same is 220 double that ??? Why

Al D
post #993 of 3640

Steve's reply for a twisted pair was about 120 ohms imp. For the i2s. Termination so please explain for me . I only gathered info. From both parties and based on what you just stated I do not understand the perimeter. To follow.

I'm going nuts here...  Then why does the supplied Off Ramp RJ-45 terminating plug use 75 ohms for the MCLK signal?   And why doesn't the first scope capture look great with a 120 ohm resistor?  This exactly duplicates the 75 ohm plug but only with a different value resistor.   And one more thing to ask Steve.,...  If not used do the HDMI I2S outputs need termination for the RJ-45 I2S outputs to function normally - or can HDMI (or any other output) be left unconnected?


Since there is no spec for external I2S it seems one vendor assumes one thing and another something else which doesn't surprise me.  I'm just going the have to tune up the I2S interface with what works best for my setup.  I'll be sure pass anything significant along.

post #994 of 3640

Ask him we both own his product. He will answer you to and you seem to understand this better. Than me. There must be a reason. He will know. Also put up a post on audio circle. His forum . Pose the question and both him n others will give us answers . I would but the truth is I do not even know what a MCLK signal is . He did say that all the outputs are on at the same time also . So that is yet another mystery , although other devices we play with also do this . I rec my short blocks today. Heavy and really well made . I will open it and take some hi Rez pics and post them. Although if explained I am sure I will understand the imp matching dlilema . Right now I do not have any answers . Also the way he answered my question made me be leave there is no imp to match from his device , just the wire used to match the input of the dac.

Al D
post #995 of 3640
One more thing. I sent all the spec,s to Steve ahead of time , as well as asking kingwa too. So I am really puzzled in this and great full that you did the test .thanks . I never planed to even care. One mire thing in reading his forum. He and others prefer the spidif as the input of choice to the dac, another puzzle . Oh and I am planning on buying his over drive dac next week. And he does say that the dac can be used with his off ramp in i2s format. This is why he has an i2s input to HIS dac... Let me know

Al D
post #996 of 3640
Dear Albert,
The resistors is solder on the I2S kit bottom side, you can replace by other resistors, but low impedance request the source have better output drive ability.

More info
post #997 of 3640

The resistors is solder on the I2S kit bottom side, you can replace by other resistors, but low impedance request the source have better output drive ability.

Kingwa has some good advice.  75 ohms is quite a large current load for say a normal TTL logic driver.   The chip's output impedance would come into play and you may see stair steps on the rising/falling edges.  But it appears Off Ramp uses a healthy driver for MCLK at least.


Kingwa also chose a good logic chip for receiving the I2S signals (74VHC125).   These chips can tolerate a wide variation of input signal levels without latching up.


I wrote to Steve N. at Empirical.  Supplied waveforms.   I asked if all the I2S RJ-45 signals are driven the same way as the MCLK signal so if I chose 75 ohms (or something else) could I expect similar performance.   I explained what was going on and was just trying to tune up the I2S interface for best performance with the M7.


Here's a scope capture of the Off Ramp MCLK signal with no cable and 75 ohm resistor crimped to a RJ-45 plug.  The waveform looks good with a healthy 4V signal and clean edges.  Mild overshoot but nothing like 120 ohm or 220 ohm termination.  I would be very happy with this signal quality at the M7 end point.



Let's see what Steve N. at Off Ramp has to say.

post #998 of 3640
I recently got the 75 ohm termination plug-in for my GHz scope. Very expensive BTW. This new tool allowed me to see what I could never see before: exactly what is happening at the S/PDIF termination and the jitter there. This revealed that my current circuit, which is typical for transformer-coupled galvanically isolated S/PDIF, was holding-back the performance of both the Synchro-Mesh and the Off-Ramp 5. Bummer Not that the jitter was high, but it should be a LOT lower that what I am measuring.

In order to determine the cause of this, which is incidently very difficult to observe, I embarked on a multi-week prototyping of different pulse transformers and associated circuits with several Synchro-Meshes. I got help from the transformer designer too. Yesterday I had a breakthrough. I realized that I don't need to change to a custom transformer as I had anticipated. Some circuits changes and tuning will be sufficient.


1) I discovered that the impedance of the S/PDIF source is critical to getting low jitter from S/PDIF coax. Even 1 ohm off makes a big difference. This is why most manufacturers cannot get it even close and why so many users believe that S/PDIF coax is a broken standard. Most other products that I modded in the past did not even have 75 ohm output impedance, much less exactly 75 ohms.
2) Pulse transformers can be inconsistent in characteristics. They are wound with tape etc, so the spacings vary from one to the next. These differences result in ratio and impedance differences.
3) I believe based on anecdotal evidence that break-in of the device changes the impedance, so tuning it when the device is not broken-in might require a re-tuning later to optimize for jitter.

I am in the final stages of testing of the mod now. I still need to verify the repeatability of the mod and the effect of break-in. I will be posting some before and after jitter plots soon.

The impact for me and my customers is an inconvenience for both, but I believe that this one will be worth it. I expect this mod to cost about $200.00 for either product. Other companies would put it in the next product, but I believe this is so important that I have to get it out ASAP. This is another reason to stick with Empirical Audio.

Exciting development. More to come.

Steve N.

Here is a copy Nd paste.

I am glad you are doing the test,s . According to Steve what you are adjusting is critical .


Al D
post #999 of 3640

Steve N. from Off Ramp maker responded and said it is OK to use 75 ohm termination on all I2S signals.  Each of the four Off Ramp I2S signals are driven the same way so nothing special about MCLK.


Steve recommended termination directly at the endpoint chip per transmission line theory.  The RJ-45 jack connector board, where termination is currently located, is not the end of the signal path as there is still a section of ribbon cable to go to reach the I2S receiver chip on the M7 DSP board.  Adding termination at the very end point chip would be very difficult to do for the casual electronic enthusiast especially with no schematic or guidelines.   If termination at the M7 I2S RJ-45 connector board doesn't work out then building a special ribbon cable with terminations at the DSP header end is the next best option (remove the RJ-45 jack board terminations first).


But first will take the easy road and change the RJ-45 jack connector board terminations from 220 ohm to 75 ohms and see how well the signals look with the oscilloscope.  At the very least I believe they will look better than they do currently.  Then evaluate and go forward to something more exotic if needed.  75 ohm resistors on order so standby.....

Edited by DACLadder - 12/1/13 at 7:55am
post #1000 of 3640
Dear Albert,
The all our design, coaxial input impedance is 75 ohm.
The I2S input is 150 ohm.
Now you say there is a 220 ohm resister there and kingwa says it's 150 ??
post #1001 of 3640

Boy, I'm getting tired of writing up tests and I'm sure you are tired of me as well hogging all the ink.  But I have now have different termination on the four RJ-45 board termination locations.  I have a bunch of 120 ohm resistors and by luck 120 ohms in parallel with 220 ohms yields about 77 ohms.  77 ohms close enough for testing.  The waveforms look a LOT better but not perfect.   So let me listen to this for a while and get back in detail next week.


But I am now finding less variation sound wise with external I2S cables.  The short CAT 7 cable sounds very much like the 18" CAT 5.   Previously when I2S didn't sound quite right bass is weaker and the highs slightly fuzzy.  One day sound great and the next a little different.  Also for the past few days with the Master 7 on the workbench been listening to ADG Ref. 5.32 and USB32.  Didn't sound bad at all with very warm sound in comparison.  Now the Off Ramp 5 and Master 7 I2S are tuned up somewhat they kick the Ref 5.32 in the rear end.  The OR5/M7 sound is vibrant and clear - crystal highs, tight bass, and great vocals.


Again, I'll be listening to the new RJ-45 board terminations for a while to fully see how it works out long term.  And if anyone has a better approach I am all ears...

Edited by DACLadder - 12/1/13 at 11:24am
post #1002 of 3640
How many terminal points are there at the rj45 in the m7 ??
post #1003 of 3640
 ~~How many terminal points are there at the rj45 in the m7 ??

Four I2S signal plus common ground at the RJ45 jack board plus four locations for termination resistors.


It will be next weekend before I can posted the latest scope captures.   I am without the scope at the moment.   Plus am evaluating the (hopefully) final changes.  First moved the 75 ohm terminators to then end of the ribbon cable at the DSP board.  Not pretty but this sounds really good.   Also built a RG-179 coax cable (one for each of the four I2S signals) with termination again at the M7 cable end.  This removes the M7 RJ-45 jack board and CAT cable completely.  The coax cable sounds brighter with more silibance versus the previous ribbon cable termination and using a short CAT 7 cable.  So maybe too much of a good thing but has me curious.  Until next week....

post #1004 of 3640
I presume you used the individual cables to maintain a 75 ohm bal throughout ? Maybe it is not needed as the resister is the termination at the end point?
post #1005 of 3640

Here's another entry on the OR5/ M7 I2S cable saga and will try to be brief but give as much detail as possible.  And to repeat what I am doing here only applies to the Off Ramp 5.   A different I2S box will more than likely perform differently.  Also if you like the sound of the Off Ramp 5/ M7 as is then don't worry about it.  With all my testing I never got gross data errors that causes clicks, pops, or really bad sound.  Just had variation in sound quality when testing various CAT cables.   The sound also varied from day to day using the same cable so decided to put a scope to the signals to see what was going on. 


I am only posting the I2S MCLK (Master Clock) as this is the most important clock for jitter.   The other I2S signals are terminated in the exact same way in every case and look similar on the scope.  See page 65 for my initial scope trials using the standard M7 termination (220 ohms) and 120 ohm termination.  75 ohm was determined by trial and error to be a good match for preserving the I2S waveforms when driven by the Off Ramp 5.


Also using a different scope this time with higher bandwidth - 300Mhz now vs. 100Mhz a few weeks ago.  This new scope is revealing much more noise on the I2S signals.  This noise can introduce jitter so is not good for sound quality.


Edit: As a clarifier, MCLK varies in frequency based upon the data rate of the audio playback.  I used 192KHz files for my testing for maximum MCLK rate.  So MCLK with 192K is 24.576Mhz (192K x 128).  With 96Khz files MCLK is halved at 12.228Mhz.  But the important clock edges look the same with any clock speed.  


First, here is MCLK with the supplied Rj-45 resistors termination removed so no termination anywhere and using the short CAT 7 cable and supplied internal ribbon cable.  Running with no termination is not a good idea.  Also notice the added noise especially near ground.   Noise is now revealed by the higher bandwidth scope.


To pick up where I left off I ordered some 75 ohm resistors and first tried replacing the M7 stock 220 ohm termination on the backplate RJ-45 connector board.   The termination resistor locations are mid-span between the OR5 CAT7 cable and M7 internal ribbon cable.  The MCLK waveform below is measured at the M7 DSP board header (pin 5).   I also measured the signals directly at the DSP board receiver chip and see virtually no difference in signal quality so in all cases measure at the DSP header pin 5 for convenience.   As you can see in the scope capture below the square clock waveform is tilted slightly.  This indicates the clock signal is reflecting from the non-terminated end point and adding to the wave front rising/ falling edges and causing the tilt.  I listened to this setup for about three days and there was much less variation in sound quality when using the various CAT cables and lengths versus stock setup.  So a big step in the right direction but the noise riding on the signal is not a good sign.


OR5/ M7 MCLK 75 ohm termination on M7 RJ-45 board.  4" CAT 7 cable externally.  Stock ribbon cable internal.  1V/div vertical scale. 300Mhz bandwidth scope.



Next removed the M7 RJ-45 board terminations and moved termination near the end point by reworking the supplied internal M7 ribbon cable.  I sliced the ribbon cable end to expose the I2S signals and added a 75 ohm resistor on each of the four signal wires.  The other ends of the resistors all connect to GND at the DSP header connector pin 1.   Much better waveform shape with end point termination but the noise is still present. But this configuration sounded OK but with some silibance though in vocals that may be caused by the noise but getting better sound overall versus stock.  Not a bad looking signal except for the noise.


OR5/ M7 MCLK 75 ohm termination on ribbon cable end. 4" CAT 7 cable externally. Modified ribbon cable internal. 1V/div vertical scale.  300Mhz bandwidth scope


The final cable bypasses the M7 RJ-45 board, CAT cable, and ribbon cable.  Instead using 75 ohm RG-179 coax to directly carry the four I2S signals and terminating at the DSP header with 75 ohm resistors (see schematic diagram).  Again using the higher bandwidth scope there is much less noise showing up with pretty good wave shape and OK rise/fall times.  Signal levels are spot on for 3.3V LVTTL signaling.   And this setup sounds much different than all others tried.   Initially it seemed someone turned up the treble control as the high frequencies are clear with very little silibance if any.  I first rejected the sound as sterile but have grown to appreciate the fine detail in the resultant audio.  Nice harmonic structure of instruments with laser focused stage.  I may though need to tune my other system components like vacuum tubes to compensate slightly for the added high frequency clarity.  So am going to run with this cable for a long while but would like to ultimately see faster edge rise/fall times.


OR5/ M7 MCLK Coax cable with 75 ohm termination at the DSP header connector.  1V/div vertical scale. 300Mhz bandwidth scope



So in conclusion the M7 sound seems like it is affected directly by quality of the I2S signals.  The higher bandwidth scope is revealing noise on the I2S signal with a stock setup.  The last cable (coax) gets me closest to ultimate sound  and provides better shielding for noise. So am going to use this final coax cable setup for a while and until something better comes along.  It may be the best I can ever get but I am never going back to USB32.  Even with the original I2S cable setup I was getting better sound quality - just not consistently.   When dialed in the OR5/M7 sound is rich and vibrant.  Just needs to be a better way of delivering I2S signals from source to destination (vendors, we need a spec).  Hope this helps and if you have any questions please ask.  And if you like the sound of your system then enjoy it as is. 


I2S Coax Cable schematic


Edited by DACLadder - 12/16/13 at 6:20am
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: High-end Audio Forum
Head-Fi.org › Forums › Summit-Fi (High-End Audio) › High-end Audio Forum › Audio-Gd Master 7 - Discrete Fully Balanced DAC (PCM1704)