Build thread for ES9023 + JG Buffer boards (betatest)

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I used eight 0603 zero Ohm resistors to bridge the gap to the pads for the ES9023 that are spaced too far apart on the beta PCB's. I had a lot less trouble doing it that way than when I tried to copy Daniel's strategy using tiny snippets of wire. Don't worry, the flux will get cleaned off once I'm done with all soldering.
 

Attachments

  • _DSC0361.jpg
    _DSC0361.jpg
    124.2 KB · Views: 1,386
I seem to be just slower than the cycle-time-out of the FET tester, when I'm really moving along, maybe 1x in 10 I can get another FET tested before it goes to sleep. FET matching is something I worry about if we did a group-buy. If demand were anything like I think it could be, we'd be sorting FETs in 1000 lots, or around 2 solid 8+ hours days of FET grading, not including some breaks. I have an appreciation for how demanding the task could get. The up-side is if someone fries a FET or loses it, they would have the matching specs to ask for when requesting another. It's unclear yet what *I* will be left with after a few folks get first crack at the matched lots out of the 100+100 [I expect the matching to be further and further "off" unless I get more, which is likely]. There can be a pretty wide range of readings, FET matching seems like a must.

The 0R trick next to ES9023 is brilliant! I have kapton tape if I remember to send any to the USA beta people [remind me when we chat].

I have had ZERO time to get on the work-bench so no progress on the "NEG pin project". Parts for the test are here. No beta board built so either I test with my heavily modded HiFimeDIY UAE23... or carve out time to catch up and build the Curryman/JG. During the delay better and better-spec'd parts have been acquired for the main build.
 
Bin sorting FETs

> we'd be sorting FETs in 1000 lots, or around 2 solid 8+ hours days of FET grading, not including some breaks.

You are either extremely productive, or very optimistic. :)

Patrick

When I was at the FET measuring task it reminded me of a woman having a baby, I found myself saying "I'm NEVER going to do THIS again"!! And frankly I do not know if what I did is the best way to do it. I made rectangles on a landscape oriented piece of paper and wrote the Vgs and I numbers down, set the measured FET there, when a column was full I put clear shipping-tape over it so they can be cut out in small rectangles and packed in a baggie after comparing their readings. Technically I have only gotten READINGS for 200 FETs not decided how closely I will match them and whether they will be matched in pairs or quads. Pairs easier, but this is audio, some will consider matched quads a must? I bumped the piece of paper a few times....mixing up a number of FETs to then have to retest... The CAT jumped up to check out the FETs a few times...I managed that alright. I ordered 100+100 more FETs so there will be some nice matches for ME. When I might get the time and patience again, I cannot say. Taping the FETs down like that means a DIY has to get them harvested and not lose them or mix them up... a PCB would also be a great place to store them too. MORE time. I have the utmost respect for the people slaving over the JG filter buffer project; there's a lot more to it than people know. The FET readings are so wild and random that the matching step cannot be avoided. [Oh and during one of the tape-downs ~10 FETs got scrambled, I have those circled for yet another re-grade, cut them out/open and retest]. Just a hand full of Beta people could clean me out of good matches -- 4 boards worth at a time. 1000 lots, do I loan the DIY world many many hundreds?? Hoping they get sold? And what about the rest of my life?

On a cheerful note I found THIN FILM 0R 0603 jumpers Susumu YJP-1608-R001 and ordered 500x Mouser says they are thick film but datasheet indicates THIN. I read a nice explanation about Thick vs thin film R for audio, complete with pictures..."I get it". The list of beta people is small -- the longer you wait the better mix of goodies awaits you [well, in the USA]

Thin film Rs were also stocked for key Rs in the audio path.
 
Member
Joined 2010
Paid Member
Hello all,
after some time refurbishing my home office I am back on the DIY front:D
I just brought my second board to live, this time including the 1:4 MCLK devider. Man those FF are really damn small nevertheless everything worked out nice.
I am just working on the second version of the board. I had a lot of very fruitful discussions with some experts in circuit design and PCB layout and got some very helpful information how to improve the design (especially wrt grounding and supply bypassing incl. the special requirements of the XO).
I am looking forward to the results of your tests injecting external negative voltage at the NEG pin, however I will probably not follow this route for the next version for several reasons (e.g. layout disadvantages).
Another design decision I would like to hear your opinion is wrt regs. It would generally be possible to support several regs (e.g. LT1761, MIC5205), however this would compromise the board layout somewhat due to different bypass C configurations and external Rs for voltage adjustment of LT1761 (no 3.6V fixed). Thus I would personally tend towards the MIC here.
I will soon post a new version of the board for further discussions.

@grufti: very good work around! Looks much better than my wirework ;-)
@enjoybiking: FET matching for more than one hour, you must be crazy :eek: Where did you read Thinfilm for the 0R? However they are <0,3mOhm, thus noise won't be a problem anyway ;)

Have a nice day, Daniel
 
Last edited:
Susumu Thin Film

Resistors | Chip Resistor - Surface Mount | DigiKey

I could use 30Ga wire-wrap wire but the 0R would be cleaner so... "it's only money". A sliver of Kapton tape...

It was around 3hrs matching and I still ended up with ~~qty 10 that scrambled under the tape I used to tape them down, I must retest those. As you know, the readings are pretty wild -- they remeasure right, it's not measuring error, the FETs VARY from say a Vgs of 2.07 or so to up over 3.43 on the MMBFJ310 alone. I will send Mr Rogers {and S} matched quads. My strategy is to try and forget how much matching pains me, and match/score 200 more eventually...well, get the readings so I CAN bin them. I could hire out the matching too. Find someone on Valium?? :D

The Neg pin experiment, soon I hope -- family first. The plan is to set a multi-turn PCB POT 100R to ~30/70 ohms, inject -5V, and filter the wiper...and feed that to NEG. If it works, that source can live with my +/-15, +5, -5 on a perf board [well, I have a few premade +/- 15 boards from Thailand, they shipped with suffix "A" regulators] {as you know *I* stocked NJM7815FA, NJM7915FA series, I may switch to}.

I DID lay in a stock of the flip-flops if anyone needs them, I'll never use them? I probably have 100x. I have a number of 50MHz clocks and 225x 49.152MHz.. ya some people use a ferrite at the V+&EN to the clock. I suppose .01 to GND not a bad plan too.

If the Micrel 5205 will work {I have to look again} that is what I will use for 3.6V I have them. And I have over 800x LT1761 3.3

Time's been tight. Beta guys will be done before me.
 
Welcome back, Daniel. I'm getting close to having my beta version finished, so far so good. Let's hope it comes to life when I hook it up.

One question first: what are you going to use the 1:4 MCLK signal for? I assume that it's going to be a ~12.5 MHz clock signal after division.

I look forward to seeing/using your second version. I use ADP150 regs whenever I can. Between the LT1761 and the MIC5205 I have a slight preference for the LT. There is no denying that the fixed 3.6V MIC5205 is very convenient.

Layout and supply bypassing probably trumps the differences between the regs in a big way. I am very happy to hear that you went over the layout and grounding again. It's the hardest part of getting it right.

I have only three requests for the second version: please improve the silk screen and please consider adding pads for the 47k inductor damping resistor in the JG filter buffer that was suggested by abraxalito in the ES9022 thread. The third is more of a joke: please use the proper case size for the ES9023.
 
Last edited:
Member
Joined 2010
Paid Member
grufti, thanks for your feedback! I am quite sure everything will work fine.

The divided MCLK can be used for synchronous operation of source (e.g. SPDIF receiver) and ES9023. I have a miniSTREAMER with TE7022 USB to I2S and will test WM8804 based SPDIF receiver. Both need a 12MHz clock so I have a 48MHz on my boards.

Can you tell why you prefer ADP150 over LT and MIC? Due to one or several specs or because of subjective sound quality? I will have another look to all three datasheets.

Layout and grounding is indeed very important and to be honest there is some potential for optimization on the beta version ;) not at all bad, but can be improved...

The silk screen is on my list (you probably refer to the VIAS?). The damping resistor is already prepared and no further comment regarding the third - errm - silly thing :p
 
It's been some time since I compared the LDO reg datasheets, but I seem to recall that the ADP has the lowest noise of the bunch.

My USB input of choice is the Amanero. That one uses two clocks @ 22.5792MHz and 24.576MHz. I run the ES9023 asynch on a 50MHz XO. 48/12MHz makes a lot of sense for your configuration.
 
...if the artwork could tolerate lt1761 and ADP, and mic5205, so much the better. IN and out pins same? Locations of bypass or filter maybe not same? I have my NEG pin test to do yet, but I can always add or fly that V in {Kynar} if indeed that trick works and I/we could avoid the charge-pump noise. -3.4V easy to make. If the Neg pin PAD would have a V-in spot, and a .01 to ground....hmm. Hard to say yet.

VREG pads for optional R and C to GND, kind of given.

ODAC and maybe 1-2 other folks I've read, use a R in series with Aout R and Aout L 220R, one guy more like 400R... if and only if there is room, maybe you could drop in pads for 0R~220-440R? Unclear if anyone can HEAR the difference, or what it does to the SPICE simulation.

MCK probably nice flying that in from CM6631A 49.152 asynch-synch 8), well, I have many options. I have visions of this getting fed from Behringer DCX2496 channels too, regular I2S, pretty sure it will work.
 
Member
Joined 2010
Paid Member
...if the artwork could tolerate lt1761 and ADP, and mic5205, so much the better. IN and out pins same? Locations of bypass or filter maybe not same? I have my NEG pin test to do yet, but I can always add or fly that V in {Kynar} if indeed that trick works and I/we could avoid the charge-pump noise. -3.4V easy to make. If the Neg pin PAD would have a V-in spot, and a .01 to ground....hmm. Hard to say yet.

The pinout is identical or at least compatible but the external component arrangement differs. To suit all regs the board layout would have to be compromized. I'll try different options, but tend to share gruftis view at least for the 3.6V reg.

For the XO and FF supply the regs will not be too relevant anyway because all those linear regs are much too slow. Here the caps (type and arrangement) are more relevant and the regs simply provide correct DC voltage.

VREG pads for optional R and C to GND, kind of given.

ODAC and maybe 1-2 other folks I've read, use a R in series with Aout R and Aout L 220R, one guy more like 400R... if and only if there is room, maybe you could drop in pads for 0R~220-440R? Unclear if anyone can HEAR the difference, or what it does to the SPICE simulation.

The series R in combination with the cap to ground forms another low pass filter. In our case we have the LCR filter so it does not make sense to add another R here unless you want to disable the JG Filter (btw: could also be done since we now have R pads shunting the coils and we have the C to ground after the coils ;))
 
Beta PCB built up

Done! Almost all parts match the BOM value exactly, but size and type are sometimes what I already had lying around. This configuration is meant to run asynch on data from a USB port.

Now I'll have to come up with a power supply and then hopefully music.

Please let me know, if you see something that looks funky to you. The more eyes the better.
 

Attachments

  • _DSC0367_at50.jpg
    _DSC0367_at50.jpg
    363.6 KB · Views: 1,141
The 18.7K?

Done! Almost all parts match the BOM value exactly, but size and type are sometimes what I already had lying around. This configuration is meant to run asynch on data from a USB port.

Now I'll have to come up with a power supply and then hopefully music.

Please let me know, if you see something that looks funky to you. The more eyes the better.

R to ground just after the coils is confusing what value.
 
Thank you for checking things so closely. That resistor is a 17.4 kOhm Susumu thin film 0603. Mouser part #: 754-RR0816P-1742D24C. I'm not sure why it reads 24C, if I can make that out correctly. It measures 17.4 kOhm. It also happens to be a good match for my inductors. I picked two from the lower end of the distribution.
 
Thank you Daniel, Joachim, EUVL, Abraxalito, EssTech... the list goes on. No debugging needed.

Clean, clear, harmonious sounds right from the first second after power up. It will be some time before I get around to measuring its performance. Maybe I'll should save myself the effort and wait for the second version, which is coming anyway. There is absolutely nothing egregiously wrong with it. Very nicely done.
 
Member
Joined 2010
Paid Member
Great to hear that it worked out so well :)

Second version makes good progress. Please find below a teaser...
Will post some more information soon

GND is not routed yet. At least not on the board but only in my head ;)
 

Attachments

  • ES9023 + JG Buffer v1.01.jpg
    ES9023 + JG Buffer v1.01.jpg
    214 KB · Views: 1,064
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.