Building a headphone amp (LME49990 + LME49600)

[schwat]

Ever since I bought my Beyerdynamic DT990 250ohm cans I've been thinking about picking up an amp. However instead of paying $120 for one & letting someone else have all the fun of building it I decided to do it myself. My amp is based on the reference design from the LME49600 datasheet, I just replaced the LME49710 with the LME49990 for my voltage gain stage.

 

So far I've finished the schematic & breadboarded the first prototype. I'm actually listening to it on my desk as I type this up

 

Disclaimer: I am not an engineer. I am a technician. I am good at soldering & prototyping but but have very minimal experience designing things. This project is for fun & for learning. 

 

 

Reference design from LME49600 datasheet:

 

 

Schematic (updated 4/20):

 

 

 

Operation:

 

The input goes into a dual 10k Alps pot which lets me control volume. 1 pot controls both left & right channel.

 

The input is fed into an LME49990 ultra-low distortion, ultra-low noise op amp for my voltage gain stage. Currently have gain set to 11 (insert Spinal Tap joke) which gives me a nice bit of play on the volume knob when paired with my DT990s. With my ATH-M50s (38 ohms) gain of a 11 is a bit high, only have about 40% of the pot to play with before it gets too loud. When I layout the PCB I'm planning on adding a dip switch where you can switch in a couple different values to change the gain. 

 

Next the output goes to a LME49600 high-performance, high-fidelity, high-current headphone buffer. This stage is unity gain (gain of 1) and only serves to boost current (250mA typ according to datasheet). 

 

Everything is being driven off +/- 15V DC PSU. 

 

 

Capacitive vs DC coupled input:

I decided to go with a capacitively coupled input. The idea of an input capacitor is to block any possible DC offset voltage from an input source which would be amplified & coupled through to the output. This 680nF cap coupled with the 20k input resistor forms a high pass RC filter with a cutoff freq of 11.7hz so well below the freq response of my headphones. Did a quick unscientific test with DC coupled input and couldn't hear a difference. Will check and see with an oscilloscope at some point out of curiosity though. 

 

Here are some pics showing how an input capacitor can block DC offset. In the pics below the red signal on the oscope is the input signal measured at the output of the potentiometer & the green signal is the output signal measured across a 250 ohm load. Both are shown at the same voltage scale. The first two pics are both capacitively coupled & show that even with an offset on the input (red signal) in the 2nd pic there is still no offset in the green signal. In the 3rd pic its DC coupled & you can see that the offset on the input not only shows up on the output but is significantly larger because it is amplified just like our audio signal.

 

1. Capacitively coupled: Vin = no offset. Vo = no offset.
   
   
    
2. Capacitively coupled: Vin + DC offset. Vo = no offset.
   
   
    
3. DC Coupled (C1 removed): Vin + DC offset. Vo = DC offset.
   

 

 

Capacitive vs DC coupled output w/ rail monitor:

 

Since the output is currently not capacitively coupled you have to keep in mind a few things to prevent possible damage. If you lose one of your two voltage rails but not both of them you have the risk of your output signal having enough DC offset to potentially damage headphones. I'm going to look into a supervisor circuit that will cut both power rails if it detects either rail falling below a set reference.

 

The simplest solution is adding a capacitor on the output to block DC just like we did on the input. And just like on the input this capacitor will be combining with a resistor to form a high-pass filter for our signal but unlike the input we don't get to choose the value of that resistor. Instead it is determined by the resistance of the load (headphones).  

 

The formula for the cutoff frequency of a high pass filter is Fc = 1/(2πRC). On the input we chose a nice high value of 20k for our resistor so with a 680nF cap we got our cutoff freq all the way down to 11.4Hz. Since freq is determined partially by R*C that means with the lower resistance the higher the capacitance needed so we'll need to choose our cap based on our lowest possible load resistance. 

 

For a 250 ohm load I would only need a 33uF capacitor to get my cutoff frequency below 20Hz: 1/(2π*250*33e-6)=19.29Hz

 

For a 32 ohm load I would need a 250uF capacitor to get my cutoff frequency down below 20Hz: 1/(2*π*32*250e-6)=19.89Hz

 

So the trick will be finding a suitably large capacitor that has an acceptable ESR & doesn't affect sound quality. 

 

• This picture shows what could happen if you lose 1 of your two rails. In this example I've removed VEE from U3 & the output is railed at 13.8V DC. No bueno.
  
  
   
• But if we add that capacitor in between the output of our amp & our load it blocks all that DC voltage giving us 0 V on the output and happy headphones.
  

 

 

 

 

Still in work: 

 

Power supply stage. I'm currently using a 120VAC to 5VDC PSU salvaged from an old network switch that is fed into a DC-DC converter (also salvaged) that takes in 5VDC & outputs +/- 15VDC. However this DC-DC converter is quite expensive if purchased so it wouldn't be practical if I wanted to make more of these. So I'm going to build a supply using a fairly cheap transformer (~$8-11) & some LM7815 & LM7915 regulators. Should be able to keep total cost under $20 for the PSU & about $50 for the whole project.

 

Also considering going with an AC wall wart & just handling the rectification on board. Would let me get away with a smaller enclosure by moving the transformer outside the case which should help keep possible EMI noise down.

 

PCB. I'm also trying to decide how I'm going to go about making a board. If I can talk a few friends into pitching in I might do a small order of proper PCBs from a vendor otherwise I'm going to do it old school & do the toner transfer method & etch me some copper clad.

 

Inputs/Outputs. In the final design I plan on having both RCA  in and regular old 1/8" headphone jack. For outputs I'm thinking both 1/4" & 1/8" jacks. 

 

Enclosure. TBD after everything else is finalized.

 

 

 

 

Pics:

 

LME49990s are on the little surfboards & the LME49600s are on the big copper clad surfboards.

 

 

 

Salvaged power supplies:

 

 

 

Hand-carved TO-263 surfboard. Made with an exacto knife & some precision files and probably the biggest pain in the ass of this whole project. 

 

[schwat]

<reserved>

[ZetZet]

First

[schwat]

And just in case reserving this one too

[Shaoty]

*slow clap*

 

Nice reserve steal!

 

 

[ZetZet]

Nice reserve steal!

 

 

I will edit that post now to: "First" just because of you.

[Freaky_spider]

nice, even tho i dont even know what im looking at (i mean, the specific parts are 'n stuff)

[Assassin]

Skill Level = Wizard!

[MayflowerElectronics]

nwavguy?

[schwat]

nwavguy?

 

Haha I wish, he definitely knew what he was doing. Wonder whatever happened to that guy.

 

I'm not even an EE, just an electronics technician & my soldering skills far outweigh my design skills. I can do some serious wizardry with a soldering iron but I let the smarter people do the math

[Avratz]

You know I'm in for a board  B)

[Mcdoorknob]

That's pretty crazy stuff. Very impressive.

[EmoRarity]

Neat, you have the means to get us some measurements at any point? 

[mr moose]

What's all the caps in your supply line for?  And why 14uF total?

[schwat]

Neat, you have the means to get us some measurements at any point? 

 

After I get everything complete I'll see what I can get my hands on at work. I should be able to provide some measurements, not sure if I can get my hands on something to do proper characterization though. I'm actually surprised at how good it sounds just breadboarded. I originally had it spread across 3 breadboards & had a loose connection somewhere that was causing intermittent crackling but after re-wiring it nice and compact I haven't heard a hint of noise.

 

If I crank the volume up as high as it goes with no audio playing I can hear a tiny bit of hissing but with the knob at regular listening levels I can't hear anything. I honestly expected a lot higher noise floor considering it's on a breadboard. With any audio playing I haven't heard a hint of noise.

 

My design goals for this project are "sounds good to me" and "cheap" and so far so good. I'll be curious to see how it measures but this is really just for fun. 

 

What's all the caps in your supply line for?  And why 14uF total?

 

The caps are just the decoupling caps that go on all the power rails next to the ICs. The LME49600 datasheet shows using 0.1uF caps on the opamp & then 10uF caps on the LME49600. I just put em over on the side of the schematic to save space, in circuit they're connected right from the IC power pins to ground. Ends up 10.4 uF on each rail which seems to be plenty.

[mr moose]

After I get everything complete I'll see what I can get my hands on at work. I should be able to provide some measurements, not sure if I can get my hands on something to do proper characterization though. I'm actually surprised at how good it sounds just breadboarded. I originally had it spread across 3 breadboards & had a loose connection somewhere that was causing intermittent crackling but after re-wiring it nice and compact I haven't heard a hint of noise.

 

If I crank the volume up as high as it goes with no audio playing I can hear a tiny bit of hissing but with the knob at regular listening levels I can't hear anything. I honestly expected a lot higher noise floor considering it's on a breadboard. With any audio playing I haven't heard a hint of noise.

 

My design goals for this project are "sounds good to me" and "cheap" and so far so good. I'll be curious to see how it measures but this is really just for fun. 

 

 

The caps are just the decoupling caps that go on all the power rails next to the ICs. The LME49600 datasheet shows using 0.1uF caps on the opamp & then 10uF caps on the LME49600. I just put em over on the side of the schematic to save space, in circuit they're connected right from the IC power pins to ground. Ends up 10.4 uF on each rail which seems to be plenty.

Perfect, physically getting them as close to the IC package is optimal.  Although to be honest I have wondered if this only prevents interference for those one in a million situations.

[schwat]

Took the amp with me today when visiting family & friends for Easter to get some more opinions. Most people there had never had a chance to try out a nice pair of headphones in the first place much less paired with an amplifier so they were trippin out. It's always fun introducing people to new things & I'm pretty sure I sold a couple pairs of DT990s today. Practically had to pry my setup away from them when it was time to leave  B)

[Kevin_Walter]

My uncle used to be able to do this sort of thing. 

 

Anyways, nice work! Hope it works well. 

[schwat]

About 90% done w/ the layout for my first pcb prototype. This was my first time using Eagle PCB so it was a learning experience & I suck at layout in general anyway. This board is just for the amp no PSU since I'm going to be powering it off that +/- 15V DC-DC converter. This is going to be a single copper layer board with surface mount parts on the bottom & the through hole parts on the top. Both the input & output are on the front because I didn't have a choice because of the case & psu I'm using for this one. On the final version inputs gonna be on the back.

 

Currently board outline is 3.5" x 3.15"

 

Red = copper (bottom layer), green = holes, blue = jumper wires.

 

[schwat]

Test printed for fit against my case. Going to fit with tons of free space. Also did test fits for all the components, all my pads & holes line up perfectly. Just a few minor tweaks before v 0.1 is ready for etching.

 

I've also mostly completed research on my rail monitor idea. I've pretty much settled on an output relay that uses a time delay circuit before closing on power up & will open immediately upon power down. Once I get it prototyped & tested I'll post schematic for it too. It's not going on this first prototype board though.

 

As soon as boards are etched and wired up I'm going to see about getting some preliminary measurements.

 

 

[schwat]

Did a bit of tweaking to the layout to break up the current return paths for power & signal. Now the vast majority of the power current is returned through the upper ground plane & the input signal current is returned through the lower ground plane. 

 

Originally had all the return current running through the same path so this should be a tiny bit better for noise. 

 

Here's the final copper for rev 0.1.  Picking up chemicals for etching solution when I get off work and going to take a first shot at etching boards tonight.

 

 

[Glyptodont7]

Sick amp I think I will steal your plans cuz this is exactly what i am looking for

[Glyptodont7]

What are the specifications for you amp?

[schwat]

What are the specifications for you amp?

 

So won't have specs until final version when I can make some measurements. I've talked to some of the guys at work & found a DSA I can borrow so I should be able to do a full characterization with THD+n & all that good stuff. Hopefully will have preliminary measurements for this first prototype early next week but it depends on if everything works when put together (fingers crossed) & when the DSA will be available for me to borrow.

 

This is my first real "design" project, as a technician I mostly do prototyping and testing of designs other people make so this is quite a learning process for me. This is my first real layout too, I've only done layout for a couple very simple breakout boards that only had 2 connectors & some wires before this. Mostly doing this as a learning experience & for shits n grins, not shooting for any particular performance goals besides "sounds good" &  "reasonably cheap". 

 

However the specs for the parts I've chosen are pretty good so assuming my layout isn't just absolutely terrible for crosstalk & noise then overall performance should be pretty decent. Even with it just mocked up on a breadboard it already is meeting my goal of "sounds good" so I doubt it could be any worse on a PCB but I'll find out for sure here shortly. 

 

Specs for my components:

 

1) LME49990 (Voltage gain stage)

• Datasheet

2)​ LME49600 (current buffer stage)

• Datasheet
• 

[schwat]

Etching PCBs using toner transfer method:

 

Materials:

• Photo paper
• Copper clad board
• Clothes iron
• Sharpie (permanent marker)
• 2 plastic containers (large enough to fit the board)
• Muriatic acid (found at hardware store in pool chemical section. Also called hydrochloric acid)
• Hydrogen peroxide (3%)
• Acetone
• Fine steel wool
• X-acto knife 
• Water
• Latex or nitrile gloves
• Safety goggles
• Pin vise drill & drill bits (if doing through hole design)
• Beer (optional)

 

 

1. Printing the design
   ​
   First step in this process is to print your design out on either glossy photo paper or paper from a magazine. This must be done on a laser printer since it requires toner, not ink. Make sure to set the quality to as high as it will go so it prints as dark as possible. Cut the finished print to just a bit larger than the printed area & smaller than your piece of copper clad. 
   
   Design must be printed mirrored or it will be mirrored once transferred. 
   
   
   
    
2. Preparing the copper clad board
   
   Next you need to get the copper board as clean as possible & completely free from oil. It is a good idea to wear gloves at this point to prevent getting oil from your hands onto the board. 
   
   First wipe the board down with acetone then give it a good scrubbing with fine steel wool until it shines & give it another final wipe with acetone so it's as clean as you can get it.
   
   
   
    
3. Transferring the design from paper to copper
   
   Get your iron & turn it up as hot as it will go and make sure the steam feature is disabled. Take your photo paper and put it face down on your copper clad board. Holding one corner in place with something heat resistant (paper towel bunched up works) start ironing it from a corner & slowly work your way out. Apply even pressure and move the iron in small circles taking extra care to not wrinkle the paper & preventing air bubbles. 
   
   This should take ~5 min or so, you want to really iron the heck out of it so that everything sticks tight to the copper.
   
   
   
   
   
    
4. Removing the paper
   
   Toss your board in a bowl of water & let it sit for a bit. The amount of time this takes can vary pretty wildly I've found. After the paper starts to bubble up & look transparent start rubbing it off with your fingers like you're removing a sticker from something. Make sure to rub all the residue off the bare copper but it is OK if there is some residue left on the parts with toner.
   
   
   
   
   
   
    
5. Retouching before etching
   
   At this point you may notice some areas where toner didn't transfer 100% perfectly and this is OK. Go back and color in any light sections with sharpie so it is good and dark. Sharpie resists the etching solution just as well as toner so you can even use it to draw on traces or make labels. 
   
   In these pics you can see the areas where I went over with sharpie. The board on the left is dry which is why the paper residue on the toner is so white. Again this is totally ok & that board actually came out the best of all of them.
   
   If you put on too much sharpie just use an x-acto knife to scrape off the excess.