O2 headamp output booster PCB - Page 3 - diyAudio
Go Back   Home > Forums > Amplifiers > Headphone Systems

Headphone Systems Everything to do with Headphones

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 24th October 2013, 12:28 AM   #21
agdr is offline agdr  United States
diyAudio Member
 
agdr's Avatar
 
Join Date: Sep 2010
Default good stuff -> opa140 / opa2140 / opa4140

Sergey888 - I finally got a chance to look into your OPA140/2140/4140 suggestion back in post #12. Now that chip is fantastic for this project! Great find. Still JFET input with pico-amps of input bias current, won't load the O2's pot or cause any significant drop across the 40.2K resistor, AND still has ultra-low output offset voltage. 120uV, just like the OPA827. Finding that combination is a real trick. I've found several op amps with low Vos, like the OPA211, but high (bipolar) input bias current. Then a bunch of FET input op amps that have high input offset voltages.

I see that TI is pitching the OPA140 group specifically for battery operatated equipment like this project, since it only has 2mA or so per amplifier. The NJM4556A in the O2 was listed at 9-12mA total, and the LME49600 in low BW mode is 7mA, so the OPA141+LME49600 would be 9mA vs. a maximum of 6mA per amp for the NJM4556A, only about 50% more. Much lower draw than the OPA827+LME49600 pair.

I see the noise figures on the OPA140/2140 are just a hair worse, 5nV/sqrt(Hz) vs. 4 for the OPA827, might not even be audible. A reasonable trade-off for increaed battery life. Same with THD+N. At 20kHz the 140/2140 it 0.0004% vs. 0.0002% for the OPA827.

Plus the OPA140, the single version, is pin compatible with the OPA827 and would work in this same board layout! In fact, I'll get a few in from Mouser and build one board up with OPA827s and another with OPA140s. The 2140 would save space and eliminate one chip, still SOIC-8, but I kind of like being able to use either the 827s (or 627's - those would work too for more $$$) or the 140s in the dual single layout.

So now there can be two different builds on the same board. OPA827+LME49600 for the "high performance" build with higher current draw but a tad less noise and THD, or the OPA140+LME49600 for the "high battery life" build.

You are right about C5 and C11. Right two traces but wrong place on the traces. Given the high bandwidth of the chips those were in case of very high frequency ocillations when the traces look like inductors. But isn't going to do much good over on the LME49600 end rather than the OPA827 end. I couldn't find a place to fit the caps, then saw that spot by the 49600, but didn't go back and think through the circuit. Oh well, the hope/expectation is those won't be needed anyway given the short, wide, traces.

I also settled on the decoupling method in fig. 3 in this Analog Devices app note on layouts for dealing with falling chip PSRR vs. frequency:

http://www.analog.com/static/importe...als/MT-043.pdf

Hence the (relatively) low ESR 47uF solid tantalums in the middle, with the COG ceramics up at the chip leads, plus the 4.7uF solid tantulums in parallel on the LME49600 leads from that chip's data sheet. Lol - after all that hopefully it will be stable. I'm going to drop 1nF and 2nF across the headphones for a reactance test.

Last edited by agdr; 24th October 2013 at 12:47 AM.
  Reply With Quote
Old 24th October 2013, 03:07 AM   #22
diyAudio Member
 
Join Date: Feb 2009
I didn't tell you that OPA140 is a brother of OPA1641 with, i guess, slightly different front end bias current and a bit lager area of input transistors. So all worries about how it sounds are relevant here But honestly I would not worry about.
I would not bother with C0G/NP0 cap on supply lines. Capacitor linearity is irrelevant here. X7R/X5R will work as good or even better, because you''l be able to place larger value capacitor, or use smaller capacitor with the same value and put it closer to a consumer.
  Reply With Quote
Old 24th October 2013, 04:52 AM   #23
diyAudio Member
 
Join Date: Feb 2009
agdr

LME49600 has an ugly peak around 100MHz. It look bad in combination amost with any opamp if you do not put any additional freq compensation around.
That is what it looks like with OPA140. It is not acceptable. It gets even worse if you use wide bandwight mode.
I'm playing with few different variants in simulator. Nor sure which on is the way to go so far.
Attached Images
File Type: png lme49600_****.png (79.6 KB, 248 views)
  Reply With Quote
Old 24th October 2013, 11:45 PM   #24
diyAudio Member
 
Join Date: Feb 2009
agdr

Probably the simplest way to make nice is to put around 15pF capacitor from output of opamp to its inverting input and 1k resistor from LME output.
If you want to make it second order, just replace 15pF capacitor with two 33pF in series and put something like 3.3k resistor from the middle point to the ground. But advantage is not as big, ~6dB in audio band.
  Reply With Quote
Old 25th October 2013, 04:38 AM   #25
agdr is offline agdr  United States
diyAudio Member
 
agdr's Avatar
 
Join Date: Sep 2010
Sergey888 - good find about the high frequency gain spike. Looks like the gain is heading back toward unity at 80MHz or so. I'll add the 15pF and 1kHz, circuit below. This V1.0 PCB is just a prototype anyway so I'll cut / jumper / hand-solder parts as needed to get those caps and resistors in the right place.

The oscilloscpe is 100Mhz, the probes 200MHz, and the LME49600 out to 110MHz or so in low BW mode. The day I sent the board in for fab I was thinking at one point "what are the chances that if it does oscillate, it does so in that window from 90MHz to 110Mhz that I couldn't see?" Pretty good chance, apparently. Murphy's law!

That is OK about the 1641. I'm still more of an objective person on the design than subjective. I see from that "related products" table on the first page of the OPA140 datasheet that TI is also putting it in the same general family as the OPA827, OPA132, and OPA209.

RocketScientist / NwAvGuy would probably consider the OPA140 family to be more of an "audio" chip than the OPA827. Or, said another way, the extra few dollars for the OPA827 being money down the drain for extra performance that doesn't matter for audio. The slew rate of either chip is many times what is needed for audio. As RS pointed out in one of his postings the slew rate of real world audio sources and signal processing chains isn't much above 2V/uS. The 3V/uS of the NJM4556A in the O2 was perfectly adequate. The higher slew in these two chips is unused excess. Would have a need for video but not audio. Same with the excessively large chip bandwidth, especially of that OPA827 and LME49600.

I've been thinking some more about the short connecting wires going from this add-on board to the DIP headers in the O2's NJM4556A IC sockets. Twisting the (now FET) input wire and the output wire together going to each DIP header would probably be a recipe for oscillation. I have some mini coax here that might be a better choice for the input wires, shield grounded at one end, or just twist the input wires of the two OPA140 chips together, the separately the outputs of the two together between the board and the DIP headers.

I haven't had much Google-searching luck turning up any mention at all of the OPA140 family being used in headphone amps, or even audio for that matter, but I finally found one today:

HeadAmp Pico Power: Pre-Order Thread - Page 8

the "pico power" headamp, a commmercial product. Sounds like he may be wrapping OPA140s around a discrete transistor output stage. He apparently has that amp designed to use batteries as a portable, so hence the choice of the low quiescent current OPA140, same as here.
Attached Images
File Type: png O2 output circuit.png (23.7 KB, 233 views)

Last edited by agdr; 25th October 2013 at 04:58 AM.
  Reply With Quote
Old 25th October 2013, 05:37 AM   #26
diyAudio Member
 
Join Date: Feb 2009
The impression I have that OPA140 is an "instrumental" brother of an audio opa1641.
These JFET input opamp have similar output stage to OPA211 1611 209 1661 1601 etc.
OPA827 is a different beast. At least it is not R2R output. OPA132 is something different too. If I'm not wrong that one was made in BB even before it became TI.

For "pico" power, if anyone interested, I can share schematic of this thing. It requires only single 9V battery. Consumes around 6mA and have pretty decent distortion level for such a small guy. I also have thoughts about making lower voltage version with R2R output which should work from 4 AAA batteries and have even lower current consumption.
I was trying to make it fit large heatshrink. It actually fits, if you ditch the connectors. Looks a bit ugly though

Last edited by Sergey888; 25th October 2013 at 05:51 AM.
  Reply With Quote
Old 28th October 2013, 02:14 AM   #27
agdr is offline agdr  United States
diyAudio Member
 
agdr's Avatar
 
Join Date: Sep 2010
Sergey888 - A first pass at V2.0. 15pF and 1k feedback resistors added. Had to use 0805 size SMD for the feedback parts and the 0.1uF OPA827 bypass caps. I switched decoupling values. The 1.0uF COG MLCC is now across the 4.7uF solid tantalum on the LME49600 rail bypass. The 0.1uF COG MLCCs are on the OPA827 rails. Good point about the X7Rs, for decoupling bigger should beat tempco and voltage stability. I may change those MLCCs to (bigger) X7R.

I spent a few hours yesterday studying the THD + noise graphs in the datasheets for the OPA827, LME48720, LME49600 (which the sheet measures looped with the LME49720), the LME49990, and RocketScientist / NwAvGuy's THD + N measurement of the O2 with 6Vrms swing. Interesting stuff. From the datasheets the THD+N at 1kHz with the OPA827 wrapped around the LME49600 should be nearly identical to that with the LME49720. Both about 0.00003% with a 3V swing, 15V rails, 600R load. The LME49600 input impedance will be nowhere near 600 of course, but it just provides some apples-to-apples numbers from the datasheets.

I had to puzzle a while over NwAvGuy's O2 amp THD+N measurements with the 6Vrms swing. Those are just amazingly low THD+N he measurements for the NJM4556A chips in the O2 headphone amp with that large of a voltage swing. The low level of THD+N sure doesn't pop out at me from the little datasheet graph NJM provides in the 4556A datasheet, especially on the 20kHz end of things. I guess the datasheet THD+N measurement being done at 30dB gain is making the difference. At any rate it bodes very well for my ODA amp version with those parallel 4556A's. His posted measurements were verified when Tyll Hertsens re-did them with his Audio Precision tester last year.

THD+N snippets are LME49600 (with LME49720) into 32R, LME49720 into 600R, OPA872 into 600R, LME49990 into 600R. All +/-15Vdc rails and all at 1jHz. At 3Vrms swing the LME49990 is around 0.00001%, the others around 0.00003%. Note the X axis on the LME49600 is watts, so it lets go at 1W = about 5.6Vrms swing. Although the LME49990 has the best THD+N number of the bunch, the OPA827 has the best number that also has gauranteed maximum input offset voltage of just 125uV. The maixmum of the LME49990 is listed as 1000uV. 1000uV = 1mV would still beat the 3mV of the O2 amp, but I would really like to get it down to around 100uV is possible, and do it without using a DC servo. The OPA827 should fit the bill.

Last edited by agdr; 28th October 2013 at 02:32 AM.
  Reply With Quote
Old 28th October 2013, 05:04 AM   #28
diyAudio Member
 
Join Date: Feb 2009
agdr

There is a chance that the distortion performance will be layout limited. Let's hope it will not be the case.
  Reply With Quote
Old 30th October 2013, 03:48 AM   #29
agdr is offline agdr  United States
diyAudio Member
 
agdr's Avatar
 
Join Date: Sep 2010
Default It fits!

The boards for this project are back already, with one screw-up. The O2 amp case is 100mm wide, not 80mm. What a dumb mistake to make. Depth swapped for width.

But regardless the board answers the DPAK size question. The two LME49600's fit just fine soldered down in the top slot of the O2 headphone amp case. I wasn't expecting that. Even without the board being the full width I can tell both DPAK's will fit by the proximity to the O2 parts underneath. Isn't even a tight fit either, between the DPAKs and the top of the case.

The good news is that extra 20mm of width that I forgot is over the output jack and AC power stuff on the O2 headphone amp, a good place to avoid with the OPA827 input lines.

The photos below show:
* the board
* thinly solder coating the PCB DPAK pad and DPAK heat sink with a wide chisel tip on the iron
* Heating the DPAK pad and heatsink while pushing down (yep those are pliers, all grounded and on an anti-stat mat)
* clearance to the case top while inserted in the top slot
* the header that goes in where U3 is on the O2 headphone amp, and a closeup. The wires would be shorter for actual use in the O2, of course, this is just for test since
I botched the PCB width.
* PCB bottom, all the V- rail nail to the top heatsink with the thermal vias. The blank notch sits over the metal endcaps of the O2's two 220uF electrolytic capacitors.
* revised layout and circuit with Sergey888's feedback network and the correct 100mm board width. Even was able to increase the size back to 1206 SMD from 0805.
Attached Images
File Type: jpg IMG_2076.JPG (193.3 KB, 75 views)
File Type: jpg IMG_2078.JPG (120.6 KB, 85 views)
File Type: jpg IMG_2079.JPG (112.4 KB, 81 views)
File Type: jpg IMG_2080.JPG (124.2 KB, 67 views)
File Type: jpg IMG_2083.JPG (141.8 KB, 75 views)
File Type: jpg IMG_2084.JPG (172.1 KB, 87 views)
File Type: jpg IMG_2085.JPG (104.8 KB, 74 views)
File Type: jpg IMG_2086.JPG (139.2 KB, 64 views)
File Type: png O2 output PCB OPA827 LME49600 layout lyr 1_4.png (143.0 KB, 61 views)
File Type: png O2 output PCB OPA827 LME49600 circuit.png (40.2 KB, 48 views)

Last edited by agdr; 30th October 2013 at 04:06 AM.
  Reply With Quote
Old 1st November 2013, 01:00 PM   #30
agdr is offline agdr  United States
diyAudio Member
 
agdr's Avatar
 
Join Date: Sep 2010
Default 125mW into 8 ohms

It works. With the OPA140 chip looped around the LME49600 the O2 amp's output DC offset is just 24uV vs. 3.4mV using the O2's standard NJM4556A chip. In a worst-case scenario load test I hooked it up to an 8ohm speaker. The circuit was able to push about 125mW into the speaker. The heatsinking worked better than I expected with the LME49600 get just sligtly warm while dissipating about 1.3W at that point. Can't do that with an NJM4556A chip!

I soldered an OPA140 based loop on one channel of the board and a OPA827 on the other to compare. I had a little accident with the OPA827 end so I don't have those numbers yet. Who knew the brown bar on one end of solid tantalum SMD chips was positive rather than negative?

The photos below are:

* First two show the board still fits just fine in the O2 amp's top slot with the rest of the parts loaded. The 47uF 35V solid tantalum capacitors are about the same height as the LME49600 but they clear the case top with room to spare.
* DC test setup. The input is left unconnected so the jack grounds both channels.
* DC output offset voltage of the OPA140 + LME49600 on the left (24uV) vs. the other channel in the O2 with the NJM4556A chip left installed (3.4mV). A 99.3% reduction!
* Next 2 photos are an AC test with music going in. I'm using fairly sensitive 114dB/V headphones here, so the 28mV (rms) shown represents normal listening levels and 64mV(rms) very loud levels. Still just 190uV instantaneous DC output offset voltage on the OPA140+LME49600 channel vs. 3.4mV on the NJM4556A channel.
* Next 3 photos are a speaker test setup! My test headphones present a rediculously low load at 28mV(rms), so I wanted to try a load that would push the LME49600 to the limit. That black tower speaker is an 8ohm unit. I didn't whip out the scope, but I could hear distortion starting at about 1.3Vrms. 1Vrms into 8ohms would be 125mW and a current of 1/8 = 125mA (rms). That leaves 11V across the LME49600 with the O2's +/-12V power rails, so it dissipates 11 * .125mA = 1.37W. Only got slightly warm at that level. I'm impressed at how well the thermal vias are working. Only 16mV of DC offset into the speaker at 1.01Vac (rms).
Attached Images
File Type: jpg IMG_2088.JPG (151.3 KB, 55 views)
File Type: jpg IMG_2091.JPG (214.8 KB, 83 views)
File Type: jpg IMG_2093.JPG (205.3 KB, 76 views)
File Type: jpg IMG_2096.JPG (115.9 KB, 43 views)
File Type: jpg IMG_2097.JPG (202.6 KB, 38 views)
File Type: jpg IMG_2098.JPG (114.7 KB, 40 views)
File Type: jpg IMG_2099.JPG (204.6 KB, 43 views)
File Type: jpg IMG_2106.JPG (115.1 KB, 46 views)
File Type: jpg IMG_2101.JPG (155.7 KB, 36 views)

Last edited by agdr; 1st November 2013 at 01:12 PM.
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Zen headamp. PCB, my ver. Stream Pass Labs 43 21st May 2014 08:28 AM
FS - Fully populated & working O2 headamp Lil Knight Swap Meet 0 16th January 2013 02:30 AM
Grommes S35 booster amp output? Dill Pickle Tubes / Valves 1 9th October 2012 07:23 PM
WTT O2 headamp PCB for CCS PCBs regal Swap Meet 0 13th October 2011 03:42 AM
FS: PCB's for the Morgan Jones tube headamp Nisbeth Swap Meet 0 19th October 2004 07:43 PM


New To Site? Need Help?

All times are GMT. The time now is 02:33 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2