O2 headamp output booster PCB

Hi agdr

Glad to see your progress with the booster PCB!

Knowing you are O2 lover, decided to show you what could have been done in O2 design.
Basically this amplifier contains 5 NE5532 opamps per channel. It is made in wonderful dead beetle style.
Driving 3Vpk(+9.5dBVpk) into 56Ohm load it has THD around 0.000025% for 1kHz and something like 0.00006-0.00007% for 10kHz, see screenshots attached.
 

Attachments

  • pcb.jpg
    pcb.jpg
    106.5 KB · Views: 413
  • 2k.jpg
    2k.jpg
    107.4 KB · Views: 397
  • 3k.jpg
    3k.jpg
    116.4 KB · Views: 378
  • 4k.jpg
    4k.jpg
    96.5 KB · Views: 356
  • 20k.jpg
    20k.jpg
    115.2 KB · Views: 354
  • 30k.jpg
    30k.jpg
    116.1 KB · Views: 121
  • 40k.jpg
    40k.jpg
    106.3 KB · Views: 123
  • 50k.jpg
    50k.jpg
    110.7 KB · Views: 137
Sergey888: That is a seriously good bit of point-to-point wiring work! Very neatly done. I've done a lot of point-to-point wiring on perfboards with IC sockets but never have tried right to the chips like that. Did you score the copper clad board yourself to break it up into those sections or did it come that way?

Well those THD results are very encouraging for the ODA amp with the 6 NJM4556A sections in parallel.:) At 56R load you would have around 10mA peak per chip section, well withing specs for your NE5532's. I'll bet your design sounds fantastic!

Turbon: will do! I'll add you to the interested list. I ordered 20 boards for $140 total, fab + DHL express, so that works out to around $7 USD per board. I'll send you a PM once I'm satisfied with the testing on the board.

Here is some good news. I did a post about the board over on Head-Fi. MisterX came up with a pointer to a SIP header at Digikey I didn't know about. It has DIP socket sized pins at both ends (dual "pluggable" header) and an extension post in the middle. Which means that it may be possible to simply solder two DIP8 sockets upside down under the tab on the O2 booster board here, so that they sit right over the U3 and U4 sockets on the O2, then just insert that row of SIP pins in both to connect. Quick and easy interconnect between the two boards! I'm going to order some of those headers from Digikey and give it a try.

If that works then I'll also include a set of SIP headers with the PC board and add $1 or whatever they work out to be.

If there are any parts in this project that are hard to get outside the US at Farnell, RS, etc, let me know. I could consider having those available too, to ship along with the board.
 
Last edited:
Did you score the copper clad board yourself to break it up into those sections

Yeah. It is very easy to do using just a scalpel blade and a ruler. If you want to go brutal use a bayonet :)

Well those THD results are very encouraging for the ODA amp with the 6 NJM4556A sections in parallel.:) At 56R load you would have around 10mA peak per chip section, well withing specs for your NE5532's.

Actually I was truing to made somewhat different point. I was trying to show that using literally the same components, but reorganizing feedback, it is possible to achieve significantly better results. Just using parallel opams, like it is done in O2, is very hard or impossible to get THD that low.
 
Yeah. It is very easy to do using just a scalpel blade and a ruler. If you want to go brutal use a bayonet :)

Lol! :D I'll keep that technique in mind for rapid prototyping.

Actually I was truing to made somewhat different point. I was trying to show that using literally the same components, but reorganizing feedback, it is possible to achieve significantly better results. Just using parallel opams, like it is done in O2, is very hard or impossible to get THD that low.

Ah, I see! I agree. In V2.0 of the ODA I'm going to fit in your feedback network suggestions. I also want to work in your suggestion of small resistors on the op amp power leads to extend the PSRR to higher frequencies.
 
Last edited:
Sergey888 - I just had an idea. :D I'm guessing that if LME49990 chips were substituted into your NE5532 prototype the distortion & noise levels would go even lower. Extremely low, since for a 3Vrms swing into 600R the LME49990 datasheet graph shows 0.00001% THD+N already at 1kHz unity-gain and about double that at 20kHz.

I essentially lost interest in this one

http://www.diyaudio.com/forums/headphone-systems/237041-massively-parallel-lme49990-headamp.html

due to the prospects of sodering 16 SOIC-8 chips, although as a couple of posters noted that apparently can be done with solder paste and a toaster oven.

But I was thinking today that it could be laid out on a much larger PC board, or even better two idential boards stacked, to put more space between the chips for easier DIY soldering. With two stacked identical boards each could have their own power supply, making the thing true dual mono. It would be fun to add your feedback circuit to that and see just how low distortion could be pushed in a modern chip amp with paralleled outputs of non-power-buffer (no LME49600, BUF634, etc) op amps. :) And use the pricey low tempco 1206's (lower distortion) as per that set of slides by Hoffer.

The output should not be TRS but something isolated like mini XLR or Hirose. Lol - up to the user to figure out how to rewire their headphones to get rid of any common ground. Same with inputs, should be RCA. The target headphone of the amp should be a "normal", modern, fairly sensitive 32R-or-so headphones, that only need a slight bit more voltage gain or curret buffering over just pass-through from the source, not the headphone models that are two sigmas out on sensitivity or impedance that require a large voltage swing or massive current buffering. Only headphones that can stay within the 3Vrms sweet spot of the LME49990 datasheet's distortion graph conditions.

With those limitations on the type of headphones the power supply voltage regulator requirements shrink. Instead of TO-220 there are some very low-noise SOIC-8 or SOT-23 dual rail vregs out there, as I recall, that would fit right on a dual stacked board arangement. Then add your resistors to the power supply pins on each LME49990 chip for better high(er) frequency PSRR and isolation. The datasheet shows the PSRR on the LME49990 going south at around 200 Hz. Ideally the filter should kick in about there.
 
Last edited:
If there is an aim to get lower distortion, LME49990 is not the only option. It is possible to significantly improve an opamp linearity just adding some gain afterwards. Working with small(er) signal makes opamp more linear. The logic is simple - the lower output voltage swing - the lower current input stage should provide to recharge a miller capacitor. Lower current from differential stage means it works on the shorter section of the transimpedance curve, thus it is more linear like.
It will be easier to implement this if you ditch the idea of multiple parallel opamps on output.

I do not share your obsession with dual mono thing :D You can get very good channel separation using proper layout technics.

If you remember, a while ago i was showing one of my designs where I was getting less than 0.1ppm distortion driving 3VPk into 15Ohms at 1kHz, which is <0.00001%. And it was much better with higher impedance loads.

As for resistors, reasonably good thin film resistors will do the job. Don't have to be top of the price line.
 
Last edited:
Sergey888: That is a seriously good bit of point-to-point wiring work! Very neatly done. I've done a lot of point-to-point wiring on perfboards with IC sockets but never have tried right to the chips like that. Did you score the copper clad board yourself to break it up into those sections or did it come that way?

Well those THD results are very encouraging for the ODA amp with the 6 NJM4556A sections in parallel.:) At 56R load you would have around 10mA peak per chip section, well withing specs for your NE5532's. I'll bet your design sounds fantastic!

Turbon: will do! I'll add you to the interested list. I ordered 20 boards for $140 total, fab + DHL express, so that works out to around $7 USD per board. I'll send you a PM once I'm satisfied with the testing on the board.

Here is some good news. I did a post about the board over on Head-Fi. MisterX came up with a pointer to a SIP header at Digikey I didn't know about. It has DIP socket sized pins at both ends (dual "pluggable" header) and an extension post in the middle. Which means that it may be possible to simply solder two DIP8 sockets upside down under the tab on the O2 booster board here, so that they sit right over the U3 and U4 sockets on the O2, then just insert that row of SIP pins in both to connect. Quick and easy interconnect between the two boards! I'm going to order some of those headers from Digikey and give it a try.

If that works then I'll also include a set of SIP headers with the PC board and add $1 or whatever they work out to be.

If there are any parts in this project that are hard to get outside the US at Farnell, RS, etc, let me know. I could consider having those available too, to ship along with the board.

I wouldn't mind a complete kit :).

Regards
 
dodog - will do. I'll put you down on the list. Assuming all continues to go well with testing I'll send out PMs when the time comes. OK to cancel then too - I already have the boards. :)

Turbon - I might as well make the parts for the O2 modifications on the board available too, at cost, since they only add up to $3 or so as I recall.

Sergey888 - good thoughts! Yeah, the dual mono is more of an audiophile thing, I think. In fact I've always thought that just the opposite about channel separation, adding crossfeed, would be a good addition. I've never added crossfeed in hardware though since it sounds like a lot of software programs can do the same on the digital source end of things.

Good thoughts too about paralleling, or lack of it. Assuming 15mA drive from a single op amp into 32 ohms, that is around 0.5 volt swing which may be enough for a lot of sensitive modern cans and IEMs. My K550's only need 0.040V or so for normal levels.

I'll have to go back and take a another look at that design you posted. :)
 
I am up for a board and parts also if you have one.

You are added. :)

I've put together a spreadsheet with everyone interested listed here:

https://drive.google.com/file/d/0B67cJELZW-i8dG1YcHpxbktPWkE/edit?usp=sharing

Please let me know if anyone spots an error!

I oredered 20 boards and I'll be putting 14 up for grags, again contingent on further testing working out. I still need to build up and test the modifications I put on the board, and get those header pins in from Digikey and test that new interconnect idea.

Assuming all goes well with testing I'll PM eveyrone when the time comes to work out shipping and move the details over to the vendor thread.
 
very nice job - congrtulations

I scared because I am not sure to be able to know how to solder the square chips on the top , most of my attemps are the resistor of the CMOY.
Is it a risky soldering ?

I have got an other question :D I have got an ODAC (amp+DAC). This modification also work for it ?

thank you
 
anatopeck: If your ODAC is mounted in the battery area of the O2 then the booster board would fit just fine. The booster board goes around the battery area and shouldn't interfere with the ODAC.

Good question about the surface mount soldering! :) I'll bet there are several people who have built O2's wondering the same thing. Well, there are two ways to look at it. If you haven't done any surface mount soldering before then the surface mount parts on the booster board will definitely be a new challenge. On the other hand, if you want to learn some surface mount soldering, this $7 board is one of the cheaper ways to get your feet wet. :D

What I would recommend is getting a couple of the boards and just use the first one for soldering practice. 1K ohm thick-film surface mount resistors in the 1206 size at Mouser are 10 for just 36 cents ( $0.36 USD, Mouser #71-CRCW1206-1.0K-E3), practically nothing. The surface mount version of the NJM2068 chip (Mouser #513-NJM2068MD) is the same size as the OPA140s on the booster board and are just 60 cents each. I'll see if I can find a 5 leg DPAK part the same size as the LME49600 to recommend that would be cheap for practice soldering.

So in other words just use the first board to solder a bunch of 1K 1206-sizeed surface mount resistors and NJM2068 surface mount op amps for practice, to get the hang of it. Then go give it a try with the real parts on the second PC board.

One thing that I consider essential for surface mount soldering is a Panavise to hold the board while soldering, or equivalent. Something like the Panavise Jr if the base is screwed down to something to hold it steady:

PanaVise Products, Inc.

PanaVise Model 201 "Junior" Miniature Vise: Amazon.com: Industrial & Scientific

or the larger Panavise 350 version which is what I use and would highly recommend. The base on the 350 is heavy enough to stay put all by itself and doesn't have to be screwed down to a table:

PanaVise Products, Inc.

PanaVise 350 Multi-Purpose Work Center: Amazon.com: Industrial & Scientific

Without a Panavise or the equivalent to hold the PC board I would definitely not recommend trying surface mount soldering. The problem here is that you need a 3rd hand. First you put a little solder on one of the two surface mount part pads (like a 1206 sized resistor or capacitor). Then grab the part by the sides with tweezers, heat that pre-soldered pad, and then stick one end of the part in there while you make sure the other end of the part is aligned with the pad on the other end. You have the solding iron in one hand and the tweezers in the other, so that leaves you one hand short unless you have something like a Panavise holding the board by the edges. Once that end is soldered you just over and solder the other end of the part, but again you have the solder in one hand and the iron in the other, leaving you one hand short to hold the board unless it is in a Panavise.
 
Last edited:
anatopeck: a soldering iron, or ideally soldering station, that puts out at least 80W is one more thing I would consider essential for soldering the O2 booster board. I have an 80W xytronic station that works just fine for the LME49600s. I also have a 45W xytronic stration and can say from experience that 45W is not enough, nor would be a 30W or 40 soldering iron.

The problem here is that you are essentially soldering to a heatsink. Your iron needs enough wattage to overcome that or the solder either won't melt over the full surface of the LME49600 metal tab, or the chip could be dead due to overheating by the time you are done.

A cheap-o soldering gun from the hardware store may also work for the LME49600 tab. I haven't tried it. But it may also be too much heat and lift the foil trace off the PC board and/or cook the chip if left on too long.
 
Last edited: