The Objective2 (O2) Headphone Amp DIY Project

[RocketScientist]

Thanks for the compliments. What ultimately matters is how long the period is between the op amps going unstable and the 220 uF caps gasping out their few remaining electrons into the headphone load. And your sim shows both rails are way down by the time that happens.

It's also worth pointing out such a rapid change can only happen in the real world if a battery is somehow disconnected in use. And for anyone building the full amp with the B2-080 case, that's not very likely to happen if they secure the batteries properly. The batteries would just gradually die which is a less stressful condition for the circuit to manage.

I measured the actual transients at the headphone outputs and worked backwards from there to optimize the power management circuit (without going overboard in terms of cost and board space--it could certainly be improved with those constraints removed).

It's much like the rest of the amp. I tried to aim for "well past good enough" and that was especially obvious with the power management circuit. The goal is not to damage any headphones if there's a battery problem and, from my tests, I've met the goal. I really tried to keep the O2 as "minimalist" as possible to keep the price, size, and DIY complexity all to a minimum.

[agdr]

Here is a suggestion. I'm kind of a reliability freak and I just know what is eventually going to happen to the Vgsmax = 25V mosfets over time operating at Vgs=24 +/-10%. I see some of the Sanyo datasheets actually putting it in writing these days, to the effect of "operating the device near max parameters, even if not exceeding, can result in shortened life. Contact your representative".

I would suggest adding 15V zener clamps between gate and source. That would allow the use of more standard - and available - Vgsmax=20V parts, plus would keep the Vgs a good 5V under the max. Still works, as the plot shows.

True some current would now flow through the zener, adding to battery drain, but it is only around 11uA for batteries or 33uA for the 12V rails. I'm not certain how much of an issue zener noise injection into the rails would be in this configuration - something to consider.

Another interesting way to handle the Vgsmax issue is make the mosfets double as cap multipliers when operating. Swap the two mosfets (n channel for p), flip them around to be source-follower, then swap the positions of r6/c3 r7/c4, swap the comparator outputs going to the gates, then adjust up the capacitor and down the resistor to maintain the same rc constant. A gate-source zener would still be needed since when the comparators toggle the output caps would still be charged up to the rail, with the other rail now on the gate. May also be an issue if the amp output is shorted with exceeding Vgsmax.

[agdr]

Whoops - brain hiccup in the above about the cap multiplier option. That isn't going to work with the mosfet Vgs threshold dropping the output voltage down significantly. They would have to be BJTs and that would introduce significant base current draw. Nevermind...

So the sole suggestion then is the addition of the 15V gate-source zeners in the existing design.

[RocketScientist]

Originally Posted by agdr

So the sole suggestion then is the addition of the 15V gate-source zeners in the existing design.

Yeah, that's bothered me a bit as well. But, technically, the MOSFETs I've specified are within spec. The diodes drop the 12 volt rails a bit which further helps. And there's nothing to cause transients much above the rail voltage.

The other good news is I've run MOSFETs rated at 20 Vgs(max) with zero problems--both P and N channel. For the first prototype that's all I had in my parts collection so that's what I used and they worked flawlessly. That might be just luck, but it's at least encouraging the Vgs(max) spec doesn't seem super critical.

The reality is the PCB layout is really tight. The PCB drawing doesn't show the routing, but when you consider all the traces, there's just isn't much room without making other compromises. Besides zeners, a resistor divider will also work and part of the dividers are already in place.

I'll look at the layout again and see if I can find some room. But, at least based on everything I know so far, it's not likely to be a real world problem even if it's not ideal. With the diode drops, and regulator tolerances, the design is still within spec.

[agdr]

I was wondering if there would be enough space for 2 more diodes. I'm sure the circuit will work fine in the short term - maybe even go years without a problem - it is just a long term reliability thing. Hopefully you can find a way to wedge in the diodes, or resistive divider. The divider is a good idea. Hadn't thought of that.

[RocketScientist]

I have a suspicion because the gate current is limited to 88 microamps worst case by the large 270K resistors the gates might be protected from damage even if the junction does start to break down. Power MOSFETs, except for ESD, are rather beefy and robust devices. That gate area is relatively large. But I don't know enough about the semiconductor properties to know for certain if 88 microamps is too low to damage the gate. But that would help explain why the circuit worked flawlessly even with 20V parts.

It's also worth noting C16 and C21, in combination with the 270K resistors, form a low pass filter that will tend to isolate the gates from any spikes or brief transients.

[agdr]

Good question. Seems that even just static voltage near or above max will eventually do semiconductor parts in. But the gate filtering is bound to help.

I think part of the problem is "binning" during post fab test by the manufacturers. There is just no margin left there anymore in any parameter. All the "better stuff" gets moved to a higher grade part classification. Unfortunate that there are not Vgsmax=30V mosfets. That would be enough headroom. Maybe there are - I'm pretty out of date on what is out there anymore.

[qusp]

Originally Posted by RocketScientist

@qusp, there certainly are ways to do Li-xxx batteries relatively safely. But even LiFePO4 has issues. I had a defective LiFe prism cell puff up and explode before it was ever used, charged, or connected to anything. It didn't catch on fire, it just built up pressure internally from a manufacturing defect and made a mess. I did some searching online and it turns out that's not uncommon. Many have blamed it on abusing the battery but it happens even with brand new unused batteries. Apparently if the wrong impurities get into the the wrong places when they're made they turn into ticking time bombs.

oh for sure, anything that can put out 14v x 140 amps (1960W and probably more into a short) and that can be unstable under some conditions, is a power that must be respected. i find it quite disturbing the trend of people on this forum and other audio forums just using trickle chargers for them. just because its proven to be pretty safe chemistry and nobody on here has had one explode in their gear from doing it yet, doesnt mean it wont happen. even just briefly accidentally shorting/grazing the terminals of a single 3v3 cell while soldering a pack together was enough to scare the crap out of me.

I went to great lengths to keep this as safe as possible, since it will go into my coat pocket sometimes. they are just such fantastic/pretty much perfect power sources, with that sort of current, low noise and output z of less than 10mOhms i couldnt resist, considering the needs of my portable are around 300-350ma i didnt really have many choices if i wanted to be able to carry it and have it last more than 1 or 2 hours

I use legit A123 cells only, they are the only brand of lifepo4 i have any real experience with and theyve been flawless, theyre rep rides on reliability due to the sector they are in and they charge a premium for it.

And I'm not aware of any HPA suitable through hole charger controller/power management IC for a bipolar Li-xxx supply. They're all surface mount.

yep, the only things on my board that are through hole are the relays, the toslink connector, pot for the ADC/volume, plus one of the connectors and only those because i already have the parts from previous projects. the regs i'm using are not even supposed to be hand soldered, but i made it work with a bit of creativity with the pads and vias. i dont really understand why everyone is so scared of smd tbh, i find it easier.

What's really weird to me is FiiO essentially cloned the AMB Mini3 with their new E11 amp. Except instead of a 9 volt battery they used LiPoly and a DC-DC converter. The could have very easily made it a bipolar DC-DC converter as in the E9 and the amp would have 0 volt referenced inputs and outputs--i.e. a real ground. Instead, however, they used a single DC-DC converter and the same inefficient expensive OPA690 virtual ground op amp that seriously degrades the Mini3's performance. Why?

yeah i dunno, seems strange to me also, solves a lot of problems when you can just treat everything as if it was a regular bipolar layout, grounding is just so much cleaner and you dont ever have to worry about a situation where the signal produced cannot be sunk properly, because by its nature if its been sourced it can be sunk.

[Arius]

Very nice work and excellent documentation. Thank you for sharing!

For the objectives stated, the part/design choices are well thought out.

Unfortunately, the same objectivity will put off a lot of the readers (to their own loss). Their idea of DIY involves changing regulators, caps, opamps, resistors, connectors, wires and reporting on the perceived improvements.

[RocketScientist]

@agdr there are some 30 V MOSFETs but they tend to be expensive high voltage versions with unacceptably high Rds(on) in this application.

@qusp thanks for your added comments. I agree they make nice low impedance power sources. And I think the round cells are far more stable than the flat prism LiFe cells. I'm curious if you have a solid source for real A123 cells? From what I understand nearly all the ones on eBay are Chinese clones (some of which apparently work reasonably well). And the manufacture supposedly refuses to sell them to distributors as they're not protected cells and they don't want the liability.

@Arius, thanks. I agree it puts some off. In the comments to the first article it didn't take long for some to post comments about upgrading the op amps, power supply, connectors, etc.