[Goobley]

I built this amp about a month back and have been very impressed by its performance. I haven't however been able to listen to it as much as I'd like due to my main cans being away for repair. I've got them back now though and turned my amp on again today after a couple of weeks. While it's as good as I remember it (extremely good and perfect complement to my high distortion starving student variant), I still have a slight problem with the build and would like some advice.
My issue is with the time that the build takes to turn the relay on. It took 5.5 minutes with the standard configuration (I set my power supply up for +\-15V rails). My tests however show that offset is below 1mV within 10s of the amp turning on. the simulations I ran on the servo part of the amp indicated that the relay should have clicked in well under a minute (probably closer to 30s), but as this wasn't the case I decided to reduce C27 to 100uF today. Doing his has reduced turn on time to 4 minutes but I still feel that it's longer that necessary. So my question is, how long do your amps take to turn on and what can I do to speed mine up as I can easily measure that after 10-20s there's not going to be an offset issue. I was thinking of changing the resistor divider that biases C27, but would like advice before I do it as what's happening in reality doesn't quite add up to the sims and I didn't design the offset comparator (not quite sure what/if/whether the opamps are still configured as AND and if so what effect this has).
 
Thanks again for such a good design, it always amazes me the sound that comes out of such a small package.
 
Cheers.
 
(hope you're feeling at least a bit better by the time you read this waki).

 

[wakibaki]

OK, mine doesn't take that long to turn on.
 
Did you see the note I included in the package about the polarity + on C27 is on the wrong pin? This is the first thing to check. Most turn-on problems have been associated with this or the transistor that drives the relay.  If configured correctly the delay should be a few seconds, so there is no necessity to change any values. Electrolytic caps have quite a wide tolerance so there may be a little variation in delay times.I fixed the transistor a couple of iterations back, I'll fix the cap the next time I order boards, which will be soon.
 
I'm glad that you feel the amp's performance is good. Obviously the more people who report good results, the more popular the design will be. Which any designer would hope for, it's the ultimate verification that you did your job right. 
 
The performance is extracted at the cost of running some of the components quite hot. Engineering the cooling was the biggest challenge in the design next to grounding and these were in competition for PCB. Some components moreover require the removal of planes (copper) around them to reduce stray capacitance, the chips have such high bandwidth that these measures are required to reduce the chances of oscillation.
 
Unlike the TPA6120A however, no recommendation is made in the datasheets regarding special measures to prevent oscillation with the LME49600, such as resistors or inductors at the output, and no provision is made for fitting them.
 
This amplifier should compete with the O2 in every parameter except cost. It is capable of greater output voltage swing and delivers greater current.
 
I have been debating with myself whether any improvement could be made by going to a 4-layer board. I think it's already a very good compromise between cost, cooling and signal integrity.
 
I've been looking at buying a 3D printer. This would allow me to make a couple of big improvements. I could engineer a clamshell case with a heatsink.
 
Still no date for surgery.
 
Fred

 

[Goobley]

Yeah I checked that the capacitor was installed as per the note and I checked the transistors data sheet, the schematic and the the traces on the board. I did go with precision OPA277s instead of chopper amps for the servo due the the price difference allowing me to complete the entire build for under €100 (which does actually compete with the O2 on cost in Europe as far as I'm aware, I also ABed them with a friend's O2 and we both thought the 49600 was clearer and punchier).
 
With regards to the biasing of the relay transistor, after a few seconds the base sits at just under 0.65V, which is certainly very close to its minimum on voltage. However it remains at the same voltage when the relay is energised suggesting that it is sufficient. I'll give the board another in depth going over later but the only thing that comes to mind that's different to the schematic values is R46 (resistor in series with the diode that's parallel to the relay's coil). I used a 470R here as I seem to recall you mentioning that it was a) a non-critical value and b) supposed to match the impedance of the coil if possible (I used the IM03 which as I just found from checking the datasheet again has an impedance of 178R and not the 500R I thought it had. Must've confused the 03 I ave and the 23 you used... Gah silly mistake!). Any chance that this could be part of the problem? If the answer is nay then the board'll be going back under the loupe. I can't quite work out the purpose of R46 either, I'd imagined that the diode was there for flyback or something but Im confused now, it's not like the coil is polarised or anything...
 
With regards to heat, mine doesn't really get hot. I'm driving 250R cans so obviously less current is being solicited from the buffers than with lower impedance phones, but the hottest component the LME49990 barely makes my finger tip tingle (and I had it one for at least 4 hours straight yesterday afternoon). I'd say this two layer board is pretty much spot on, and as others have done in the thread we can always stick on little baby heatsinks ourselves. It is fair to say that a case designed for this amp would be very cool though... Particularly as the half finished MDF box I have it in at the moment is not the most elegant. My woodworking skills leave a lot to be desired.
 
Cheers

 

[wakibaki]

You know it fits the same box as the O2? A box is shown on page 6. You do need a drill press and needle files to make it fit. There are sources for the box on page 11 and thereabouts.
 
A lower resistance relay in the coil means the relay is pulling more current. This probably means that the delay circuit takes a bit more time to deliver the current. You could try reducing rhe the value of R41 currently 390k by my diagram. This will pull up the voltage at the base of the resistor, make more current available and charge the cap quicker. Try 270k.
 
Fred

 

[Avro_Arrow]

I noticed my relay sticks.
I had the same problem of the amp taking forever to kick in.
I found that if I waited for ten seconds, the gave the relay
a flick with my finger that it kicked in right away.

 

[Goobley]

Yeah, I had seen the box, but I'm not overly keen on it. I may still end up going that way though as my Dad has all the kit I'd need to make it fit (might have to get him to do it or I'll end up destroying the anodising or something).
 
I changed my R41 yesterday, unfortunately I didn't have any SMD values between 120k and 390k in the house so I've paralleled 2*390k in series with 100k (giving 295k). The relay now engages within 5 seconds at which point the offset is around 5mV in one channel (worst spike I measured on cold turn on was 8mV - there are plenty of amps that run consistently with more than that!). This causes a very slightly noticeable pop, but is hardly gonna cause any damage. Thanks for you help in getting my build up and running!
I also scoped the output yesterday looking for the possible oscillation that was mentioned up-thread somewhere. Nothing was visible with my 60MHz scope and I'm not even convinced that any of the chips get more than 20-25°C over ambient which is a real win! The 7805 gets a bit toasty but nothing that really worries me, however I finally box it I want to keep ventilation holes in the top.
 
I'm gonna be going off to uni next school year with this amp and I have a question that someone with a bit of knowledge of electrics in the UK should hopefully be able to answer: All my electrical devices will have to undergo PAT testing for my accommodation, do you have any idea what class this amp comes under? I'd guess that since it's floating and runs from low power 15VAC it's probably class 3. If it's not then I'm gonna have a bit of trouble because I can't claim that it's double insulated, so I'd have to earth the case or at least all exposed metal. I'm also not using a proper wallwart but instead a boxed toroid with fuse that I'm somehow going to have to make pass commercial wiring requirements O.O And that's before I get to my point to point valve amp...
I'd be really grateful if anyone knows their stuff well enough to offer me some advice on this.
 
In the meanwhile I'm gonna chill hear listening to this really sweet piece of engineering and curse the fact that I have school tomorrow!
Cheers,
Chris

 

[G.Trenchev]

Hello guys,check this out.2 buffers per channel in parallel and lots of filtering.It outs 2 watts without effort.Zero mV offset.
 
 

 

 
Also 0.03 ohm output resistance!

 

[Goobley]

I've almost finished the casing on this amp using a slightly taller box so I don't have to worry about the height of the to220 packages. However I have some hum since putting the amp into a metal enclosure, the noise level increases a fair amount if I touch the pot, but I can't really see a way to ground the pot. Any ideas?

If I short the barrel of the ac adapter to the case then the noise disappears. Should I keep it isolated and then connect a wire from the input jack to the case (like the O2's designer recommends for his amp)? I'm not worried about a slight but of noise when I change the volume but I'd like to get rid of the permanent noise as it wasn't there when I had it in a dodgy wooden box.

Cheers

 

[wakibaki]

It's difficult to know for sure, but I've always thought that the lead from the AC adapter to the amplifier is a potential cause of problems, hence there are ferrites just after the power socket put there to mitigate any problems arising. Evidently they are not sufficient in this case. You could try passing the wire through a clamp-on or toroidal ferrite positioned as close to the box as possible. You can take the wire through a number of times to increase the effect.
 
I have had no issues when using the smaller case.
 
If connecting the input jack to the case as in the O2 works, then that is an option. Since shorting the barrel of the AC adapter to the case works, then I would be tempted to take this straight-line solution. You can find a soldered connection to the barrel on the board.
 
As far as the pot is concerned, I would expect the body to be in galvanic contact with the case when the retaining washer and nut are in place since drilling the hole in the fascia usually exposes bare metal. It's possible though that the front panel is not truly connected to the rest of the case, because most aluminium parts (smart-looking ones that is) are anodized. The coating produced by anodizing is definitely non-conductive, and may need to be scraped away in some places to ensure that there is a galvanic connection between the parts of the case, I can't tell though, because I haven't had a chance to inspect your case. If the AC adapter barrel is shorted to the case, and there is galvanic integrity throughout the case, then the pot body will be grounded.
 
Since the amplifier actually is quiet in some configurations, my recommendation is that you should fiddle around with it until you get the hum to go away. I appreciate that this is not a very scientific-sounding solution, but we're just really concerned with results here.
 
w

 

[Goobley]

It's done.
 
I just connected from the power jack to the case by connecting a jumper to the side of the connecter and clamping it between the case and rear panel. The amp is silent all the way to the noise floor of my usb powered DAC I did a quick RMAA and all it really shows is that my kit isn't good enough to quantify the performance of this amp, it just measure's the same as my DAC: ( https://www.dropbox.com/s/ysyx6qhhzcl646i/LME49600.pdf ). I'm running at 4x gain and decided to do the face of the amp in nice thin leather. Glamour shots below:
 

 

 

 

 
Thanks for the cool project waki, and the others who got involved

 

[NuClear235]

I am starting my project of semiportable headphone amp with OP1611 + LF411 as DC servo + 2x buffer paralel LME49600 /channel. Powered by 6x AW cells 750mAh -12V and +12V.
http://www.lighthound.com/AW-RCR123a-Protected-750-mAh-Battery_p_114.html
I hope it will play some time long- 750mAh is not much!
One relay as undervoltage (3.0V/cell) switch off device for protect deep discharge accus.
Second relay delayed switch on headphones to output for protect crack to headphones after start.
Blue ALPS pot, output JACK 6.3 mm and Neutrik XLR NC4 (pseudo balanced).
I hope this is way for my Senn HD800...

 

[wakibaki]

Very nice, NuClear235.

How are you charging the batteries?

w

 

[NuClear235]

I am RC modeler, so I have model hobby charger for Li-Ions. I think is OK constant voltage (6x3.1V) with limited current (750mA) supply too. But modell chargers are very cheap and very universal.
Here is some solving with accus too:
http://www.head-fi.org/t/586042/what-i-did-next-with-texas-headamp-chip-a-digital-control-desktop-portable

 

[wakibaki]

I am an RC modeller too, but an external charger is not always what is wanted.

You have found my TPA6120 amplifier. I have built 2 of these with protected li-ion batteries. One set of batteries has survived being left on continuous charge, but one set died, I don't know whether the batteries were at fault, or if the charger is wrongly adjusted. The batteries are from dealextreme.com, I buy them for my LED flashlights, I have had a few fail. Anyway, be careful if you use the charging circuit, do not set the end-of-charge voltage too high. Unfortunately I am ill, and I will probably not be able to follow up this issue.

w

 

[NuClear235]

I am using service connector to every pack. Is perfect for balanced charging and for diagnostic of every cell separate. On my picture the long one.