BBS: My new headphone amp using TI's new opamp for headphones OPA1688

[doors666]

I made a headphone amp using the new TI opamp OPA1688 for headphone amplifiers (THD+N, 10 mW, 32 ?, 1 kHz, 109 dB; Iq - 1.6ma, 50mW, 10Mhz, 8V/us, R2R). Good thd numbers for even 32 ohms phones, low quiscent current, enough power to drive even low impedance phones, rail to rail output.. its screaming for a pair of 9v batteries.

I am using my BBS preamp board for this. It uses two gain stages with 10k log pot in the middle. The second stage has switchable gain which can be used if one has headphones of varying impedances like 32 and 600 ohms, but I am using the second stage as a buffer as all my headphones are low impedance. Output impedance is not switchable though. So its gain-pot-buffer. First stage has 2x gain, I dont listen to very high volumes anyway so I kept it low. The output voltage of portable devices is all over the place, from 0.3 to 1.2v rms. I plan to use it for travel times with ipods or mobiles. The two channels are independent of each other and just coexist on a pcb, so crosstalk should be minimal.

It runs on two 9v batteries. For caps I am using 470uf panny FM and 4.7uf x7r smd cap. All smd parts are 1206.

With this design, simulation shows a THD of 0.000051% while doing 1.07v rms into 32 ohms and 2nf for headphone cable capacitance. Thats about 35mw power.

I used a plastic box chassis because the Input Output connectors body is not insulated and of course its cheaper and lighter and I could make very decent holes with a pen knife and small files. Insulated headphone IO jacks are hard to find. I probably should trim the wires a bit though, some of them are quite long.

There is no protection from loss of rail as of now if a battery disconnects or fails, that will expose the headphones to half the rail voltage. I could add an output cap but it needs to be non-polarized which I dont have, also I am not too keen on that output cap as the dc offset is very low anyway (around 1mv IIRC) Will probably need to make a small ckt to handle this on a perfboard later. I just have to be careful for now.

Sound is really good, this is quite a decent opamp. Sounds much better than the cmoys that I had earlier with ne5532 and 2132. This chip should do well in cmoys also.

 

[linuxguru]

Thanks for the heads-up - it has fairly decent specs and a reasonable price. The OPA1622 seems to have better drive capability at 145 mA vs. 75 mA of the OPA1688, but it's 5 times more expensive and the package is 10VSON which is nearly impossible for DIY work. I also need to check out the OPA1688 as a general-purpose small-signal and line-level dual opamp as an alternative to OPA2134, LM4562, etc., at a very competitive price, although with modestly high voltage noise at 8 nV/sqrt(Hz). So far, the only one that is decisively better sounding is LM6172, but it is very fast and a bit tricky to stabilize - while the OPA1688 may be much easier to stabilize due to its modest GBW and slew rate.

It looks like a very good all-round opamp at a reasonable price - like a 21st century version of the NE5532. Let's see how it goes.

 

[doors666]

Thanks for the heads-up - it has fairly decent specs and a reasonable price. The OPA1622 seems to have better drive capability at 145 mA vs. 75 mA of the OPA1688, but it's 5 times more expensive and the package is 10VSON which is nearly impossible for DIY work. I also need to check out the OPA1688 as a general-purpose small-signal and line-level dual opamp as an alternative to OPA2134, LM4562, etc., at a very competitive price, although with modestly high voltage noise at 8 nV/sqrt(Hz). So far, the only one that is decisively better sounding is LM6172, but it is very fast and a bit tricky to stabilize - while the OPA1688 may be much easier to stabilize due to its modest GBW and slew rate.

It looks like a very good all-round opamp at a reasonable price - like a 21st century version of the NE5532. Let's see how it goes.

yeah, the 1622 surely needs a dedicated pcb at the least. with 10 pins and no leads, very small and thermal pad at the bottom needing soldering, it makes it quite impossible for diy use. It still will make for a killer preamp where it probably wont need a heatsink, does 132 db at 1khz into 600R with 3.5vrms out. That enable pin could have lots of uses like low voltage detection in batteries, rail loss, startup/shutdown mute circuit, standby mode etc.

Lm6172 is stable in this pcb, atleast in my preamp version with around 2k loading. I tested it with my 200mhz scope and its all good, no external compensation required. I have used this without ever realizing that it has 50ma current capability, I guess i just ignored that part of the datasheet as I was only making a preamp that time. I will give it a spin with the hp amp also. I also want to try the ad8397 due to its 310 ma current into 32 ohms.

 

[doors666]

Time for some pictures:licklips:

schematic diagram:

Testing with junk headphones:

PCB Backside:

Fitted in a plastic box:

With ipod and koss porta pros:

With ipod and realistic pro 30 orthos:

 

[doors666]

Just tested with my scope with both opa1688 and LM6172, both are stable with 33R + 1.1nf load at 20hz, 1khz and 20khz. Output was about 1.7v peak which is about 50mW. No compensation cap was used.

 

[tcpip]

Just tested with my scope with both opa1688 and LM6172, both are stable with 33R + 1.1nf load at 20hz, 1khz and 20khz. Output was about 1.7v peak which is about 50mW. No compensation cap was used.

Are you seeing any drop in output amplitude as you move across the frequency range? Sometimes, headphone amps can't keep the amplitude flat at high frequencies with low-impedance loads, I'm told -- something to do with output impedance of the amp. So, getting low output impedance is apparently an area of concern for these amps?

All this is second-hand knowledge in my case -- I've just read the ODAC/O2 pages a few times.

(Full disclosure: I am a fan of NWavguy's approach to design, like I am a fan of Doug Self, Lynn Olson, Linkwitz, and a few others.)

 

[doors666]

Are you seeing any drop in output amplitude as you move across the frequency range? Sometimes, headphone amps can't keep the amplitude flat at high frequencies with low-impedance loads, I'm told -- something to do with output impedance of the amp. So, getting low output impedance is apparently an area of concern for these amps?

All this is second-hand knowledge in my case -- I've just read the ODAC/O2 pages a few times.

(Full disclosure: I am a fan of NWavguy's approach to design, like I am a fan of Doug Self, Lynn Olson, Linkwitz, and a few others.)

I actually didnt check for that, will do and let you know.

Just went through the o2 schematic, design is quite similar to mine in many respects, didnt like a few things though. RF filter is too high at 2.6Mhz, and there is a heavy compensation cap of 220pf. Input impedance could have been made higher, its 10k. and then there is a 2.2uf cap between the two stages which is the biggest offender, I dont see any polyprops on the board. You can use a jfet input opamp to reduce the dc offset from the gain stage and omit that cap. 1688 and 6172 both have better specs than 4556 and can easily replace it.

 

[tcpip]

Just went through the o2 schematic, design is quite similar to mine in many respects, didnt like a few things though. RF filter is too high at 2.6Mhz, and there is a heavy compensation cap of 220pf. Input impedance could have been made higher, its 10k. and then there is a 2.2uf cap between the two stages which is the biggest offender, I dont see any polyprops on the board. You can use a jfet input opamp to reduce the dc offset from the gain stage and omit that cap. 1688 and 6172 both have better specs than 4556 and can easily replace it.

Valid points. But if you get around to reading his criteria of what constitutes a high performance headphone amp, it makes for interesting reading. Measuring any design against that list is worth the effort, IMHO. I learned a lot from that ghost-who-walks. He publishes full spec of the O2 after running through all his tests. Very few headphone amp brands (even big ones) seem to have the courage to do this, choosing to go with subjective fluff. Basic suggestion: don't just read his schematic, read his blog posts, if you find the time. Good stuff.

I have the Etymotic ER4-PT and the ODAC+O2. Would have loved to listen through your amp. But I'm in Bombay, so it may not happen soon.

 

[doors666]

Valid points. But if you get around to reading his criteria of what constitutes a high performance headphone amp, it makes for interesting reading. Measuring any design against that list is worth the effort, IMHO. I learned a lot from that ghost-who-walks. He publishes full spec of the O2 after running through all his tests. Very few headphone amp brands (even big ones) seem to have the courage to do this, choosing to go with subjective fluff. Basic suggestion: don't just read his schematic, read his blog posts, if you find the time. Good stuff.

I have the Etymotic ER4-PT and the ODAC+O2. Would have loved to listen through your amp. But I'm in Bombay, so it may not happen soon.

I have read most of his articles long ago. probably its time i did it again.

its not a question of courage, not everyone can afford the equipment. I remember reading somewhere that the distortion analyzer he uses costs 50K USD, even small companies will have a tough time purchasing that kind of gear, I will have to sell my home to buy that.

 

[tcpip]

I remember reading somewhere that the distortion analyzer he uses costs 50K USD

WOW!!!! :sad:

On eBay, Audio Precision audio analysers seem to be selling for roughly USD 15K. I had been under the impression that you can use something costing USD 10K.

 

[doors666]

those at 15k are used not new.

 

[johnysamajhdar]

Cool Amp Mate...Cheers...!!!

 

[doors666]

Are you seeing any drop in output amplitude as you move across the frequency range? Sometimes, headphone amps can't keep the amplitude flat at high frequencies with low-impedance loads, I'm told -- something to do with output impedance of the amp. So, getting low output impedance is apparently an area of concern for these amps?

I checked with the scope at 20, 1k, 20khz. Its fine. The output impedance is usually set with the output resistor, which in this case is 3.3R, about a tenth of the headphone impedance i am using.

 

[tcpip]

I checked with the scope at 20, 1k, 20khz. Its fine. The output impedance is usually set with the output resistor, which in this case is 3.3R.

For some reason, NWavguy seems to believe that even 2 Ohm is too high.

 

[doors666]

For some reason, NWavguy seems to believe that even 2 Ohm is too high.

O2 is designed for 16-600 ohms, so for 16 ohms use it needs to be less than 2 ohms, still following the about a tenth rule. I am using 3.3R because i dont have any 16 ohm headphones, if I had, I would have used a 1.5R resistor.

Some resistance at the output helps with stability while driving capacitive loads, headphone cables could have some seriously nasty capacitances, people use 5m long cables also.

There was a time when 120R output impedance was considered a standard.

 

[tcpip]

Yes, I can see that the low Zout is a trade-off -- you lose stability and safety to gain flat measurable performance.

 

[tcpip]

Hey, I found the link to NwAvGuy's testing page on his blog. His equipment costs of the order of $10,000, not $50,000, new.

The page makes for interesting re-reading.

 

[doors666]

Oh, he uses dscope III, I thought he mentioned somewhere that he has the ap-sys 2700 series. Even dscope is quite expensive..

 

[tcpip]

On reading his page closely, I am inclined to correct my comment... It is possible that he does have a $50k workbench. Other than dScope, which is probably $10k, he has four oscilloscopes. And sundry other gear....