A version of an O2 Desktop Amp (ODA)
V2.1 of this ODA project PC board is now available. The materials for V2.1 of this ODA project are at the same Google Drive link immediately below, except at the folder 80x160mm -> 7_9_2014 V2.1. One of the sub-folders has a description of the changes between V2.1 and V2.0. The V2.0 build photos in posts #278 - 290, #301, #302, and #368 still apply since there were so few differences betweeen V2.1 and V2.0. I have a full set of V2.1 build photos posted at the V2.1 Google Drive link, with a text file for each photo section that talks about any differences between the two versions.
V2.0 of this ODA project PC board is now available. The part of this thread for V2.0 starts at post 272. For all the files go to the Google Drive link here: Go to the 80x160mm folder -> 3_19_2014 V2.0.
Go to the 80x160mm folder -> then the folder "ODA 3_19_2014 V2.0 fabricated" . There you will find:
* Layout and schematic in both png and PDF format. The PDFs can be zoomed up as large as you want to clarity.
* Bill of Materials (BOM) - the stuff you need to buy from Mouser or Digikey. Mouser has all of it but the LT LDO regulators, which you have to get from Digikey. Mouser is out of the pot at the moment but Digikey has those too.
* Gerber files. These are the board layout files you can send to any board fabrication house to have a run of your own boards made. I've been using Seeed Studio in China but OSH park is another good one.
* Build instructions. Plus I've posted a step by step photo log of the build in this thread.
I would advise using this V2.0 rather than any of the past versions at this point since it adds the DC output offset null feature. If you are working on a past version board, please be aware of the one marking error, C51 has polarity reversed ont he board marking. Be sure to flip it. That error is fixed in V2.0.
Please note: the ODA designed evolved over time from the first post in this thread, below, to what is shipping now in V2.0. The current version uses 3 NJM4556AL chips per channel in parallel (6 op amps total per channel), no longer uses an OPA627 at all, uses LME49990 chips for the gain stage, and many other changes. You may want to start with post #272 first to get up to speed on the current stuff, then go back and read some or all of the earlier posts for historical background. :)
************************************************** ************************************************** *****
************************************************** ************************************************** *****
NwAvGuy/RocketScientist never did get around to publishing his O2 Desktop Amplifier (ODA) before he disappeared last year. For fun I've taken a few guesses on what might be in an ODA and whipped one up. :D
Below is an LT Spice sim of the ODA circuit on the top half, one channel, and the regulator O2 circuit on the lower half for comparison. From the plots the two outputs are virtually sitting on top of each other, green and blue, which they should be. The next two below are the schematic and about 2/3 of the way through a layout. Once the Chinese PCB shops open up in a week after the holiday I'll get one fabbed and stuffed. Once I finally get a working design I'll post the layout in the wiki section here for anyone who might want to DIY it. Since RocketScientist released his O2 under the gnu license, and this design is a derivation, I'll list it as being covered by the same gnu license for public DIY use.
This goal of this ODA design is to maintain the same basic design philosophy that NwAvGuy/RocketScientist used, but to try to bump up the performance slightly in most areas. Plus add some features that people have posted about and a few of the O2 mods I posted.
I know that one of RocketScientist's goals with the O2 was to make it cheap to be readily affordable. I've kind of thrown that one out here with an eye towards some improvements. :) The OPA627 chips alone are $20 for the low end and $30 each for the high end. I don't have a dScope or AudioPrecision analyzer, so the final result here has not been properly measured, which of course was sort of the whole point of the O2, so fair warning.
Here is a summary:
Best new stuff:
The 3.5mm jacks are higher quality Switchcraft jacks, with actual springs inside and not just springy metal, another thing a lot of folks have posted about (bad jacks). I've used vertically oriented jacks to make better use of the vertical dimension and leave more front panel space. I've included 0 ohm surface mount jumpers on the back of the board to disable the grounds on the input jack switches if using the RCA jacks - no more cutting board traces as with the O2 to add the external jacks. The design has a companion top board that slides in the top slot of the B3-080 chassis, upside down. That board has a Neutrik 1/4" output jack and RCA jacks that can be wired for input or preamp out. The top board also has space to mount a ODAC.
New stuff on the front panel now includes a rotary 4 position gain switch. Gains are 1x, 2x, 4x, and 6x. There is a bass boost switch for boost on/off. The resistors can be selected for any level of boost, such as 3dB, 6dB, etc. Another one of the O2 mods I posted. The power LED is now on the back panel, one for each supply rail to help diagnose dead power supplies. The unit has a (pico)fuse now too on the AC input and the power switch shuts off the entire power supply now.
I've included the output relay mod I posted for the O2 which delays the headphone output turn-on by about 2 seconds, then switches the relay out quickly when the power is shut off to prevent thumps.
The rest of the new stuff is in the middle. The gain stage now uses an OPA627 with a feedback loop wrapped around 2 NJM4556AL (SIP 8) chips in parallel on each channel, for a total of 4 op amps in parallel now on the output. The pot in the middle had to go to allow the feedback loop (DC) to function. The pot and coupling cap are now on the amp input, with the cap value adjusted up 4x to account for the 10K pot.
The new arrangement should nearly null out (around 300uV vs. 3mV for the O2) any DC offset on the output of the amp, given the exceedingly low 100uV max input offset voltage of the OPA627 and its tiny 5 picoamp input bias current, which results in nearly zero IR drop across the input resistors. This also means the pot won't have any substantial DC through the wiper even though there is no longer the coupling cap on the wiper as with the O2. Just another way around the silent-pot design issue. :)
The cap is still there feeding the pot though to block incoming DC from the source. Highly not recommended to bypass that. I've made sure the frequency response still matches what RocketScientist's design had on the low 10Hz end.
A by-product the the new negative feedback loop design should be even slightly lower distortion. RocketScientist had pointed out that the distortion figures of the output chips swamped that of the NJM2068 input chip. Essentially made the output chips the limiting factor on distortion. They were just a necessary evil to get the high current. The NFB loop should lower that output chip distortion a small amount. The OPA627 has excellent distortion numbers.
The design uses both surface mount and through hole components and has parts on both sides of the board, so it probably isn't a novice level of soldering type of DIY build, unfortunately.
So the aim is to get better figures but not better listening experience? Carrying on the 'objective' philosophy but to the next level?
I'd be a bit worried that by sticking 4556s inside the feedback loop of the OPA627 stability is going to be compromised. Also is the 627 being chosen for its figures? If so, which figures?
The OPA627 is chosen primarily for the DC parameters, having both a very low input offset voltage and extremely low input bias current (FET input), along with having good AC parameters for audio. The project originally started out trying to reduce the DC offset of the O2. I posted a servo using the OPA627, then realized I could just put it in a loop around the buffer. TI actually shows this application of the 627, to reduce a current buffer's ouput offset, in figure 14 on the datasheet:
www.ti.com/lit/ds/symlink/opa627.pdf (opens PDF)
I took a look at the OPA211 which comes in a dual package (power pad, which I hate to solder) and also has low input offset. But that has bipolar inputs and still has a hefty input bias current. Having the FET input on the OPA627 lets me directly connect it to the pot wiper without a cap and solves that issue (pot turning noise due to DC current through the wiper). Also means very low IR drop in external input resistors due to input bias current, making the imbalance in impedance looking out the inverting and non-inverting inputs (a pot, which will be varying) insignificant. Interesting to note that the leakage current through the pot wiper coupling cap on the O2 should be about the same, or even more, than the 5pA of input current through the OPA627 input. So the two different designs really should be equivalent.
I've left the 220pF feedback caps from the O2 in to roll the chip's gain off to unity above the audio band. Turns out 220pF was still a good value for this chip.
On the subjective vs. objective thing, I'm in the camp that an amp has to sound good as well as measure well. I do believe an amp can measure well and sound bad. I'll post my subjective thoughts once I get a board stuffed and have a listen.
As for why someone would want 4 paralleled NJM4556s vs. just a single LME49600, BUF634, or even an OPA551 ddpak, beats me other than that seemed to be the logical extension of the O2 to an ODA. :) The ddpak parts make it a bit harder to build a compact amp due to the square inches of heat sink foil needed. I prefer chips that can be heat sinked, or in the case of thse SIPs do a good thermal transfer job themselves. The SIPs are rated at 800mW vs. 700mW for the DIPs in the latest version of the data sheet:
I actually laid out an all-smd version first but decided it would be too hard for folks to solder. Then I laid out an all-through hole version that used two 100 x 80 boards in a B3-160 chasis, with the amp board in front and power supply board in back. After pondering that one a bit I decided it was too physically big. That led to this version with the SIP parts to save space.
BTW, I left the OPA627 as DIPs in sockets rather than the SOIC version due to the senstive FET inputs. That way all the soldering can be done with the socket, then the chip inserted with anti-static mat and wrist strap later. I figured the odds of toasting the $30 chip with hand assembly would go way up with the smd version.
I should also say that I'm not selling anything and have no plans to sell or do a group buy of PC boards. I'll post the layout and PCB gerbers once I get the design verified.
Interestingly, TI has recently provided an EMI resistance report for the OPA627 - here : http://www.ti.com/litv/pdf/sboz016a
Its showing around 30dB rejection of RF in the most sensitive band around 40MHz. 30dB figure means that for every mV of RF entering it will give about 30uV of offset variation.
Thanks for the link! I'll give that a read.
Lol! :) Hey jcx, I would value any thoughts you might have on the ODA design and layout. You are pretty good at this stuff. :)
I've made one change today. Although the op-amps inside a package are going to be well matched, no so much between packages. So in addition to the 1R resistors on each output op amp output I've added another 1R going from the tie point on those on each chip to the next chip. So (1R || 1R) + 1R = 1.5R on each package, which should yield an output R in parallel somewhere around 0.75R.
I don't see any consensus for translating "a better listening experience" to amplifier hardware
but there are some tricks for improving linearity - even assuming the proposed chips
I would try Class A push-pull bias of the output op amps ~20-30 mA total to ~ 1/2 the combined packages power rating
the 50-60 mApk Class A range would drive most headphones over 100 dB SPL before leaving Class A - and the the amp would still be able to push 200+ mA peak in Class B if required by the load
"Class A" should at least be worth some expectation effect listening experience enhancement - toss in Nelson Pass "First Watt" musings and point to the GedLee Metric to add the Guru factor too
and Class A operation helps reduce distortion from poor power, gnd layout - which definitely needs some work
preventing RF/EMI power line problems in an objective manner requires a definition of the “problem” - threat environment and sensitivity of both circuits and audibility consequence
verification requires expensive equipment and some knowledge of how to use it
Looks to me (from the 20kHz THD plot) that the linearity problems are down to the NJM4556's undegenerated input stage. Also there's no PSRR nor open loop gain/phase plot provided, I think some data could be inferred from the 40dB gain reponse plot though with reasonable assumptions being made.
|All times are GMT. The time now is 06:06 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio