|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
67.60.98.130
In Reply to: RE: Further Juli@ Follies... posted by GStew on October 13, 2011 at 18:17:53
BACKGROUND...I've spent much time over the last few weeks planning and executing some of the extreme JulI@ digital modifications I began outlining above. My direction has gained focus and changed somewhat from above:
1. Improve the power fed to the key processing chips.
2. Reduce EMI & RF interference produced on the card & fed into it from the computer.
3. Improve the signal processing.
SIMPLIFIED DAC...Before starting this inquiry, I first simplified my downstream processing by replacing my fairly complicated AK4399-based DAC (6 regulators, dual-mono output stages) with a fairly simple one based on the ESS ES9022. This is a very fine, low-cost ($2!) DAC from the same people who created the ES9018 chip used in Twisted Pear's Buffalo DACs and the latest Oppo Blue-Ray players(among others). It is an I2S-input only chip (like the AK4399 and many other modern DACs, but unlike the ES9018) with an onboard output stage. Only a single power supply of +3.3v is used for both the digital processing and the output stage. It uses an onboard charge-pump to create -3.3v to allow its output stage to provide a full 2v level.
With only one voltage feed (most high-end DACs such as the ES9018 and AK4399 have at least one for the digital processing and one for the analog processing, if not more) and very minimal options that can be selected, I thought this was a good way to minimize the variables so I could better hear the impact of the Juli@ mods.
Here are a couple of pictures of my initial implementation of the ES9022 DAC card (provided by EUVL on DIYAudio) mounted on a baseline-modified Juli@ digital section. The Dexa regulators are used as pre-regulators here:
Here is the same card with the Dexas bypassing the on-card monolithic regulators:
BTW, some have suggested that a very good implementation of the ES9022 (and its replacement, the ES9023) will better a merely-good implementation of the ES9018, so I do not feel I am limiting myself by going with this $2 chip. (see the last few posts in this thread: http://www.diyaudio.com/forums/digital-line-level/151846-anybody-using-new-ess-vout-dac-es9022-5.html). It certainly competes with my AK4399-based DAC and with the Dexa regulators bypassing the on-board ones, I'd rank it as slightly better.THE MODS SO FAR...
Before going on to what I've done so far, you might want to go back to my first post in this thread and revisit what I've done in the past to improve the Juli@ digital sections.
So far on this round, I have worked mostly with my first area of focus, improving the power feeds:
1. Replace my traditional input filtering using 2 1000uf Black Gate Standard capactors with a combination of conductive polymer capactors and high-value ceramic capactors. These capacitors are both designed to work effectively to higher frequencies than the BGs. Also, I installed them in a manner to minimize lead length inductance to preserve their high-frequency capabilities. I did this by installing them at the bypass caps at the input to the 3.3v regulator on the back of the Juli@ digital card. These bypass caps (labeled "BCxxx") are helpfully positioned as close as possible to where they are needed, so they are better mounting points for these add-on caps, especially coupled with very short lead:
2. Move the Dexa regulator I have traditionally mounted on a heatsink about 6"-8" from the Juli@ and mount it directly on top of the 3.3v Dexa on the JulI@ card. This again reduces inductive high-frequency loses, both for the 5v chips on the Juli@ and at the input to the 3.3v Dexa.
3. Feed both the 5v & 3.3v Dexa regulators separately from my 11.5v raw DC source instead of having the 5v Dexa feed the 3.3v one. This provided greater voltage headroom to the 3.3v Dexa, but eliminated the advantage of feeding it from a well-regulated 5v source AND the advantage of the short connection from that source provided in #2 above. To gain some of this back, I added more conductive polymer caps at the input of the 3.3v Dexa.
Pictures coming later!4. Install additional bypass capacitors at the power input to each of the key chips. While the stock Juli@ is well-designed in that it has the small bypass "BCxxx" capacitor on the back of the Juli@ cards at each of these locations, recent chatter on DIYAudio suggests further gains with higher values.
While I haven't measured the BCxxx caps, I suspect they are between .1uf & 1uf. The caps I added are 100uf. *** Update *** The BCxxx capacitors already on the Juli@ are about .5-.6uf each. ***Note that I haven't yet installed additional bypasses across BC10, BC11, and BC12. Since they have other components directly next to them, it is much harder to install the caps here and not cause a short. I have practiced on an old Juli@, but plan to wait until I get through most of the other mods to do these just in case I damage the card when I do them
SONICS?Step 1, replacing the add-on power supply filtering caps provided a very nice improvement in detailing, dynamics, & bass power & definition. Also instrument solidity and texture was improved.
Step 2, provided more of the same, but with the addition of greater instrument separation & layering.
Step 3 was originally disappointing as I did not add more power supply filtering at the input of the 3.3v Dexa. I could hear greater precision, but dynamics and bass took a step backward. Redoing it with the added caps at the 3.3v Dexa input brought back the dynamics & bass and made it an acceptable alternative. Since step 2 & 3 are alternatives, not a direct linear upgrade, I need to go back an forth a couple of times to see if I can identify the differences between them and decide which I prefer. I suspect long-term I will like 3, but I may try an additional Dexa (probably set to 7.5v or so) before the 3.3v one to see if I can get the best of both worlds.
Step 4 added more detailing & separation, but was a smaller improvement than the others so far.
CERAMIC CAPACITORS?
One caution I have on all of these mods so far is my heavy use of ceramic capacitors. I remember Marsh & Jung's Picking Capacitors and how ceramic caps were some of the worse measuring & sounding caps at the time. But I see that several things are different today:
- Back then, audio circuits did not involve much if any digital processing. Today, digital sources are the standard. And building a good digital circuit is different than traditional analog circuits (and this drove the creation of surface mount components, among other techniques). Digital circuits have to be well-behaved in the megahertz to gigahertz ranges. And large, long-leaded teflon, polyprop and polystyrene capacitors just don't work well in these ranges. Making a good digital circuit means doing things a LOT differently than making a good analog circuit.
- Ceramic (and for that matter, electrolytic) capacitors have gotten MUCH better over time. While I wouldn't use the same components in both digital and analog circuits, the small, short or no-leaded ceramic and high-frequency electrolytic caps I'm using here are now good choices and the same components you see in other high-end digital circuits. Also, they are the same type of components already on the Juli@... Just more and better of them.
I did detect a bit of harshness when I first put in any of the ceramic caps... And that has diminished as they have broken-in. With a couple of weeks of time on steps 1 & 2 by the beginning of this previous weekend, the sound was very smooth and not harsh at all. Adding the components I used in steps 3 and especially 4 brought some harshness back, but now a day or two later, smoothness is returning.
WHERE TO NEXT?
My next steps are to continue with improving the power feeds and reducing EMI & RF interference produced by the card. These will include:
5. Remove chips not needed for I2S connectivity... The TOSLINK output, the digital input optocoupler, the SPDIF input/output AK4114, and the SPDIF buffer/selector. I only have a concern about the AK4114 and whether the card will continue to operate with it disabled, so I will try disabling it first on an old, sporatically-functioning Juli@ (one of the first I modified).
6. Reduce RFI & EMI pollution by shield the chips using a stick-on ferrite material and installing ferrites on the power supply lines (Thanks hfavandepas!). I also plan to try conductive foam as a absorber of RFI/EMI, as championed by Elizabeth on the Tweak Asylum.
Then I want to improve the signal processing:
7. Upgrade the on-board clocks using the Fidelity Audio Micro Clock which can be purchased with the two clock frequencies needed for the Juli@. While I don't expect these clocks to equal the higher-priced upscale clocks, I do expect to hear a strong hint of what they can do for the sound.
And finally, complete step 4 by going back & install the additional bypass caps I deferred.
AND BEYOND?By the time I have gotten to this point, I hope to have information on devices I can use to provide galvanic isolation at the I2S interface.
Then my question will be: Will I be stuck using the Juli@ at the GA-H55M-UD2H level or will the H61 motherboards with the non-native PCI slots work ok for an audio setup using the Juli@?
Greg in Mississippi
Everything matters!
Edits: 11/14/11 11/15/11 11/15/11 12/02/11Follow Ups:
Hi,
Great job! You are going as crazy as me!
Regarding remove of chips the 74HC125 need not be removed. Alternatively, just lift pin 14 Vcc (the pin above the "D" of DGND1). Check its connection to the reg.
http://www.datasheetcatalog.org/datasheet/SGSThomsonMicroelectronics/mXrytxt.pdf
74HC125 is driving the Dig trans and optical output. Without power supply, the trans will no longer work, hence no need to remove. The optical output will not be working too, but it will still draw a small amount of current.
Edits: 11/20/11
Jack,
In any case I did just what you suggested for the 74HC125. BTW, the comments made (partly by you) in the thread about a year ago on a tube SPDIF buffer thread here helped me understand how that part of the Juli@ card works... thanks!
I also removed the TOSLINK connector (by un-soldering it) and the digital input optocoupler (by just twisting it off with pliers, since it is an easily-obtained part if I wanted to undo the change).
Immediate results (no break-in time for something like this!) seemed like a slight increase in HF clarity, but was not large. I have higher hopes that un-powering the AK4114 will produce a larger improvement, but I still need to confirm that it's removal won't stop the card from functioning in I2S mode. I have an older, somewhat sproratic Juli@ that I'm setting up to test this and I hope to do that test sometime this week.
I am still getting a little of what I identify as the ceramic cap signature. They've been running in for about a week (maybe 150 hours total on-time)... If it persists over the next several days, I'll try disconnecting them. If that removes what I'm hearing, I'll try different caps there (small organic polymers next). Got any suggestions for a good low-inductance local digital circuit bypass?
A related question here... back when you first posted here about your extreme mobo mods, you'd suggested that you were looking at simplifying the process by just cutting or disconnecting power pins instead of doing full chip removals. Now in your latest posts on removing chips, cutting power pins is not mentioned. Is this is a bad option for eliminating chips on a motherboard? Not that I am afraid of pulling chips, just lazy!
TIA!
Greg in Mississippi
Everything matters!
Greg,
I don't think AK4114 can be removed, it should be responsible for I2S output. See p.29.
http://www.asahi-kasei.co.jp/akm/en/product/ak4114/ak4114_f04e.pdf
Ceramic cap never sound good to me. Oscon caps do a better job than ceramic, Siemens Sikorel is even better. I also like Philips or BC but they are no longer available. Panasonic caps (those 105 *C, series M or FM, body in blue and word in gold colour)do a decent job although they aren't as good as the above or Black Gate.
For GA-H55M-UD2H, the following can be removed by cutting of their legs off and removing legs by solder,those smaller ones like video chips are too small to be cut off (I've tried):
- IDE controller next to the P24 connectors. However, the regulator below the roll of 3 SATA connector must be removed by desoldering.
- the sound chip at the corner
Better SQ improvement can only be done by removing the video chips however.
I just found it more neat and tidy using hot air gun to remove chips. Now I can heat up the chip directly with a temperature controlled hot air gun (just bought) and pick up a chip with a sharp nose plier. Tried on small chips like the video one on unused boards with full success without tamper with any other components.
Good luck
Jack
Comes from the AK4114 in this case. Envy 1724 also has I2S output (sorry, I don't have a link for the datasheet, not sure it's available openly on the net, but if you send me an email I can send it).My notes from my original Juli@ follies on the sources of the I2S lines are as follows:
J3 Pin 1 -> 1724 PSDOUT[0] ( I2S data out) -> AK4358 SDTI1 (I2S data in)
J3 Pin 5 -> 1724 PSYNC ( I2S Word Clock ) -> AK4358 LRCK (I2S Left/Right Word clock)
J3 Pin 7 -> 1724 PBCLK ( I2S bit clock out ) -> AK4358 BICK ( I2S bit clock in )
J3 Pin 9 - > Xilinx on the digital board -> AK4358 MCLK ( I2S Master Clock in )
J3 Pin 11 - > Xilinx & AK4114 on the digital board - > AK4358 PCN (Reset)
I suspect the AK4114 provides I2S from a SPDIF input to the 1724 which then passes it onto the AK4358 DAC. But as I'm killing the SPDIF input for my purposes, I'm hoping that doesn't matter.
That doesn't mean that the AK4114 doesn't provide the I2S to the Envy 1724, but with the clocks directly connected to the 1724, I suspect it is the one that provides the I2S.
But I2S out still may not work without the AK4114 available. We'll see!
On the caps, I hope to try Oscons there a little later in the week UNLESS the ceramics mellow out a bit more. I'll let you know on that too. Still, even with the ceramics, the sonics are quite nice... just not as sweet in the highs as I'd like under all conditions.
And thanks on the GA-H55M-UD2H chip removal info. See my questions about the Q67-based board in another thread... maybe the last native PCI option?
Thanks again!
Greg in Mississippi
Everything matters!
Edits: 11/21/11
I tried un-powering the AK4114 digital audio interface chip on the Juli@ last weekend by lifting the power pins. After this was done, that Juli@ no longer functioned correctly. It caused a requested resource error.
I guess it doesn't matter where the I2S originates... Juli@ needs the AK4114.
OTOH, the rest of the findings last weekend were VERY positive. Will post an update later this weekend.
Greg in Mississippi
Everything matters!
Can I find a PCI card similar but better than Juila@?
Or even if there is one... especially if you are willing to modify the card.
My opinion... Few if any sound cards are designed to provide a high-quality SPDIF & I2S feed for use in a dedicated, high-end, tweaky music like ours. The Juli@ uses older, but still good chips. No other card I know of provides such easy access to the I2S feeds and removal of 90% of the stuff we don't need by just lifting off the analog portion. And it takes well to modification... and is inexpensive enough to do so with impunity.
I'm interested in hearing of alternatives. But I think that most of the development today is around USB-connected interfaces... and the EXAU21, the Luckit.biz XMOS-based unit, and the upcoming Twisted Pear unit are the best bets at this time in this realm... but it will change at time marches.
Greg in Mississippi
P.S. Listening to the Juli@ digital section with mostly Oscon additional filtering caps instead of large-value ceramics. Interesting contrast... Both are good and each has it's strong points. I'm going to delay my next report a few days to give this card (it was unused) and it's mods time to break in a bit and also swap back in the ceramic-bypassed card with a couple of tweaks to try and reduce it's negatives while keeping it's strong points. Couple of teaser photos attached...
Everything matters!
Jack,
So I've run the Juli@ digital card with the additional ceramic cap bypasses in for a couple of weeks... and while it's improved, I'm still not thrilled by it. There's a bit of 'confusion' to it and a very slight sense of 'harshness' in the highs on some recordings.
Very curious about this as these are all deep-digital processing chips... the controlling CPU, the PCI/Audio controller, and some helper chips. Looking at a lot of the commercial, even high-end commercial DACs & digital playback units (SACD/DVD/CD/SD players, etc.) out there, they generally had some number of ceramic caps as PS bypasses. And the Juli@ already has all of those ceramic 'BCxxx' caps.
Darned... it makes a LOT of sense to me to use very low-inductance caps mounted in a way to keep that inductance low around these deep-digital processing chips. And the ceramics seemed to fit the bill AND be the industry standard.
So my next experiment will be to do another Juli@ digital that's identical to this one except these bypasses will use some small Oscons... I have some 33uf/16v that will work well here and are small enough to fit without a lot of hassle.
Continuing my research, I went looking for situations where people used caps other than ceramics for digital chip bypasses... and found these:
http://www.diyaudio.com/forums/digital-source/186245-dac-2496-ak4393-dac-kit-cs8416-ak4393-5532-a-5.html#post2622112 ("PPS film (again if you can get them) bypassing on the digital electrolytic's and those massive power filter caps")
http://www.diyaudio.com/forums/digital-source/142562-microsd-memory-card-transport-project-post2594576.html?highlight=pps+capacitor#post2594576 (for putting non-SMD PPS caps on the regulator outputs on their SD Card Player)
http://www.diyaudio.com/forums/digital-line-level/79452-building-ultimate-nos-dac-using-tda1541a-81.html (he also has some interesting regulators that filter both the rail and the ground... very curious about these!)
And looking closer at the AckoDAC & the RAKKDAC, I now suspect they both also use some SMD poly caps around their digital circuits, at least on the digital side of the DAC.
Of course, a number of them also use some tantalum caps there too (as does the Jul@ around the AK4114 digital input-output transceiver and the ADC & DAC & output stages on the analog card). I'm not ready to go there yet, but I don't have a problem with PPS SMD caps.
Now I'll also take the one with the ceramics, pull them off, and put on 1uf PPS caps (which I just happen to have a small supply of here).
I also found some people who swear by using BG NX caps as local bypasses, at least around DACs. I agree that's a good bet, but most people won't be able to duplicate that and I'm trying to come up with mods that can be duplicated with easily-available components, so I'm not planning to try my limited supply of those no-unobtainable caps on my Juli@ digital section.
I still owe some replies on this thread to Uncle Leon on his latest capacitor work and to fork. I hope to get to them this week.
Later!
Greg in Mississippi
P.S. Another interesting poly cap trick is discussed in this thread: http://www.diyaudio.com/forums/digital-line-level/196474-ess9018-try-new-try-more-3.html where Bunpei replaces the exotic TP shunt regulators for the Buffalo-II's AVCC with simpler & more pedestrian LT1763s (Actually a Toshiba TAR5SB which is a near equivalent) & a 100uf/800 polyprop cap from CDE on the output & likes it a LOT in comparison!
Everything matters!
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: