Hello all,
I wanted an headphone amplifier for a long now, so decided it was time to build one on my own.
I was confused, because I cannot decide on which topology to build (cmoy, mini3, discrete output, etc...), so instead I decided to build a modular amplifier.
What does this means?
Simply, the amplifier is composed by a "motherboard", which comprises the two Lithium batteries, the battery chargers and I/O plugs (3.5mm jacks + USB connector); on the motherboard can be installed a daughter board, on which the "real" amplifier is built.
In this way I can switch as many daughter boards as I like, testing different topologies.
Main characteristics:
- Housed in an Hammond 1553B enclosure (I really like it!)
- REAL dual power supply, with two Nokia BL-6p 830 Li-ion batteries
- Integrated battery charger ics, with porotection and automatic battery disconnect circuit.
- Single USB power input, so you can charge the internal batteries via a single USB port.
- Alps RK09 volume potentiometer, with rotary switch (it disconnects both the batteries)
First of all, the motherboard schematic:
As you can see, using two isolated DC/DC converters I can charge and use two lithium batteries from a single USB port.
Next, the real thing:
Blue leds mean the batteries are under charge process; once full, a red led will be lit.
Today I built a daughter board, a really simple AD8397 stage, without any input/output capacitor or servo DC; it seems to work fine, no oscillation or whatever else, BUT the output DC offset change from a minimum of -0.7 mV to a maximum of 17mV (with volume potentiometer tapers moving from minimum to maximum): Do you think this is OK for my Sony MDR-V55 headphones?
Or do I have to add a DC servo circuit to cut off these DC offset?
This is the schematic I've used (C5 and C6 have been omitted):
And this is the real thing:
Please confirm me if the DC offset is OK or not.
Any opinion on my new amplifier?
Ciao,
Giovanni
I wanted an headphone amplifier for a long now, so decided it was time to build one on my own.
I was confused, because I cannot decide on which topology to build (cmoy, mini3, discrete output, etc...), so instead I decided to build a modular amplifier.
What does this means?
Simply, the amplifier is composed by a "motherboard", which comprises the two Lithium batteries, the battery chargers and I/O plugs (3.5mm jacks + USB connector); on the motherboard can be installed a daughter board, on which the "real" amplifier is built.
In this way I can switch as many daughter boards as I like, testing different topologies.
Main characteristics:
- Housed in an Hammond 1553B enclosure (I really like it!)
- REAL dual power supply, with two Nokia BL-6p 830 Li-ion batteries
- Integrated battery charger ics, with porotection and automatic battery disconnect circuit.
- Single USB power input, so you can charge the internal batteries via a single USB port.
- Alps RK09 volume potentiometer, with rotary switch (it disconnects both the batteries)
First of all, the motherboard schematic:
An externally hosted image should be here but it was not working when we last tested it.
As you can see, using two isolated DC/DC converters I can charge and use two lithium batteries from a single USB port.
Next, the real thing:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Blue leds mean the batteries are under charge process; once full, a red led will be lit.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Today I built a daughter board, a really simple AD8397 stage, without any input/output capacitor or servo DC; it seems to work fine, no oscillation or whatever else, BUT the output DC offset change from a minimum of -0.7 mV to a maximum of 17mV (with volume potentiometer tapers moving from minimum to maximum): Do you think this is OK for my Sony MDR-V55 headphones?
Or do I have to add a DC servo circuit to cut off these DC offset?
This is the schematic I've used (C5 and C6 have been omitted):
An externally hosted image should be here but it was not working when we last tested it.
And this is the real thing:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Please confirm me if the DC offset is OK or not.
Any opinion on my new amplifier?
Ciao,
Giovanni
Last edited:
I see the problem. 😱 You need ground return resistors on the non-inverting inputs so the op amp input bias current has somewhere to go. Otherwise it will just charge up those input coupling caps and produce a large input DC offset, which then gets reflected to the output (input offset x closed loop gain).
So just drop a 100K-or-so resistor from each non-inverting input to ground and you should be in business. Ideally that resistor should be the parallel combo of the resistors on the inverting input for lowest offset voltage, 820 ohm in this case (1K || 4.7K), but that would likely load the source too much going to the pot wiper.
In AMB's mini^3 that is the purpose of the 100K resistors to ground, R2L and R2R. In the O2 amplifier the 40.2K resistors on the inputs of the NJM4556 op amps serve the same purpose. One of the mistakes people have made in building the O2 is not getting those resistors soldered well and the same thing happens. The 2.2uF coupling caps get charged up by the op amp input bias current and a large DC offset results
Last edited:
Thank you for the answer.
Unfortunately, I have to disagree.
Actually C5 and C6 are not present, the amplifier has no capacitors in the signal path.
Moreover, the connector SV3 on the daughter board (female header) snap onto the SV2 connector (male header) ont the motherboard: nd, as you can see, there is the volume potentiometer, bringing the input to the ground.
I can measure -0.7 mV when the volume pot tapers are turned to ground (non-inverting inputs at virtually 1KOhm to ground), which rise to +17mV when the volume pot is turned completely clockwise (10KOhm + 1KOhm = 11 KOhm to ground).
Maybe I should really add two capacitor on the inputs, and then add two resistor after that on the non-inverting inputs, to level the + and - inputs.
Could I increase the feedback resistors from 1K/4K7 to 10K/47K (or even better, to 10K/68K to increase the gain)? Any drawback with this Op-amp?
This way I could use 4 10K resistor on every single op-amp input; input impedance as seen from the source would became 5K (in the worst case), but this should not be a problem.
By the way, I'm just using the amplifier connected to my notebook (the one I'm using now to write this), listening to Eurythmics's Sweet Dreams album: I have to say, amazing improvement from the previous configuration (direct notebook output to headphones).
Only bad thing, I tried it with my mobile phone (HTC Desire): not too bad from the point of view of sound (considering the poor source), but unfortunately I can hear strong radio interference from the phone... need to shield the enclosure from the inside (I have copper tape for exactly the purpose at work) and shorten as much as I can the interconnection cable (now I'm using the only one I have at hand, 1.5 m long).
Whoops! I was just going by the schematic. I agree, that can't be the problem if no coupling caps.
Hmmm.... I just looked up the input offset voltage for the chip and 2.5mV is worst case. Multiply that by the 5.7x closed loop gain and = 14.25mV. The input bias current is 200nA worst case. With the pot all the way up to 10K and about 820R on the other op amp input that is about 9K un-cancelled by the differential input. So 200nA x 9k = 1.8mV. Adding that to the 2.5 for 4.3mV x the gain 5.7x = 24mV worst case if I've done the math right. So it looks like that 17mV might be in the normal range for the circuit.
You can go to 10K/47K for the feedback resistors. I think that adds a bit of resistor noise, but probably not enough to matter.
I like the amp design! 🙂
Hmmm.... I just looked up the input offset voltage for the chip and 2.5mV is worst case. Multiply that by the 5.7x closed loop gain and = 14.25mV. The input bias current is 200nA worst case. With the pot all the way up to 10K and about 820R on the other op amp input that is about 9K un-cancelled by the differential input. So 200nA x 9k = 1.8mV. Adding that to the 2.5 for 4.3mV x the gain 5.7x = 24mV worst case if I've done the math right. So it looks like that 17mV might be in the normal range for the circuit.
You can go to 10K/47K for the feedback resistors. I think that adds a bit of resistor noise, but probably not enough to matter.
I like the amp design! 🙂
Last edited:
Everything you said is correct, I think.
Perfect explanation.
I can I can live with those (maximum) 17mV of DC offset; by the way, it seems the DC offset does not affect the sound in any way, it seems to me clear and powerful.
Insted, I think the amplifier is showing now the limits of my Sony MDR-V55 headphones, slightly bass-oriented and with a little lack in the middle range; I already read about it, but didn't experiment myself, without a real amplifier.
Or, it is showing the limits of the part of my media library which has not been ripped with lossless format; some 128kbps MP3 are particularly bad.
Or, again, it is showing the limits of my sources: my phone and my laptop.
I'm already building a 192Khz/24bit SPDIF DAC (CS8416+AK4396), I'm really ansious to try it with my amplifier and some lossless files...
Returning to the schematic, another modification could be to enclose the AD8397 op-amp into a closed loop with another, low offset, op-amp (like LT1115 which I have).
I should keep some gain on the AD8397, and add some more gain on the other amplifier.
The proposed schematic is attached.
It simulates amazigly well, with vanishing distorsion.
And gives an output DC voltage in the order of few picoVolts...
Any dual op-amp (LT1115 is single) I can use?
Ciao,
Giovanni
Attachments
That is a really good idea. I've been messing around with something similar lately, for exactly the same reason, a OPA627 in a loop with the paralleled NJM4556 unity-gain buffers in the O2 headphone amp. Just like you say, it simulates very well to drop the 3mV or so DC output offset of the NJM4556 down to that of the 100uV offset OPA627. Kind of an alternative to a DC Servo! I even have a partial layout done of just those parts and a power supply to test it.
I didn't know about the LT1115, but from a quick look at the datasheet it looks really good for your purpose. The one hiccup I know about in doing this is if the buffer, or second op amp in the chain, is slower than the first, then a more complicated feedback arrangement is necessary. But the LT1115 appears to be unity gain at 40Mhz and the AD8397 at 69Mhz so you should probably be OK. The OPA211 comes in a dual surface mount version and may be worth a look. It is another precision DC type op amp with some reasonable AC specs for audio:
http://www.ti.com/lit/gpn/opa211 (opens PDF)
The AD8397 is known to be a bit "cranky" as Tangent or AMB called it, with oscillation issues in some designs. With the two-stage design you could run the 8397 at its minimum allowed gain of 2 for maximum stability, then let the other chip do the rest of the gain, like you say.
Last edited:
I've read it, but actually I build an "not-so-good" circuit (using thru-hole components and wires running around it) and it work perfectly, without any (visible) oscillation.
Sincerely, I would not consider this chip so "cranky"... of course, if you know what to do.
I mean, there's few things to keep in mind when you use it, but if you know what you're doing, I can see no problem using it.
Actually, I really like it.
However, thank you for the answer.
Let's suppose we are running the amplifier on 3.7+3.7V (two fully loaded batteries).
I could reach a maximum output swing, from the AD8397, of about 3+3=6Vpp on a 40Ohm load (empirically tested).
Moreover, the LT1115 is not rail to rail, so at such a voltage it can give me out about 3Vpp (theoretically, it can swing to about -1.5V less than Vdd, but be conservative).
Again, I must leave some gain on the AD8397, because with unity gain you loose the rail-to-rail capability (as reported on the datasheet, page 11).
All that said, I would fix the AD8397 gain to something around 4; and I'd like to have a total gain of around 8.
Setting a gain of 4 on the AD8397 I can drive it to full swing (3V+3V= 6Vpp) with a input signal of 1,5Vpp (absolutely feasible with LT1115); and the preceeding op-amp (LT1115) will have a "real" gain of 2.
Quite nice, isn't it?
If you Want to simulate it, please find HERE the LTspice simulation file ad the AD8397.MOD model.
Below the schematic.
Next week I will try to build a second daughter board with the proposed schematic; I must say, I'm having big fun with this new amplifier and the idea behind it, to be able to play with every amplifier topology I can think of.
Last stupid idea: I have a USB dongle with a CM108 chip, and which sound quite fine, I could extract from it the internal PCB (which is really, really small) and place it on another (the third) daughter card togheter with a buffer...
this would allow me to use the amplifier on my laptop, directly from the USB, avoiding the internal DAC which is really, really bad (I'm currently using the external USB card between it and the amplifier).
Problem is, how can I switch then between USB DAC and analog input jack?
Ciao,
Giovanni
Attachments
Well that sounds like a good design! 🙂 I like how you have the inner and outer feedback/compensation loops on the two chips. Should help with phase shift.
Last edited:
- Status
- Not open for further replies.