I just realized that this Headphone Amp section of the forum existed! My previous copy of this post was incorrectly put in the Solid State section.
I'm trying to improve my analog layout and circuit design - please provide feedback! I designed and laid-out a dual headphone amplifier based on nwavguy's O2 amp:
NwAvGuy: O2 Details
There is a ton of open space on the board, so clearly, this design can be miniaturized further. I wanted to make something nice for listening to music or watching a movie together in an airport / plane / quiet place.
One of the most notable differences is that there is only one 4556 output stage opamp per headphone output, meaning this amp can deliver about half the current of the O2 per channel. The thing I'm most concerned about are the thin-film surface mount resistors. I'm using Yageo 1% carbon thin-films, eg RT0603FRE07100RL and others from that series. I know others have used Susumu thin films and I'd be interested to know what the difference is between those and the ones I picked.
I was really excited about getting an initial prototype built, so I omitted some of the more practical features present in the O2 such as:
-switchable gain
-wall power
-low voltage battery protection: this device could become dangerous near the end of uneven battery discharge
-power LED indicator
-the ability to remove power from one of the headphone output stages (n/a to the O2)
The 3 PCBs I ordered from OSHPark are arriving tomorrow! For the next revision, I am considering using a single Li-Po cell and boost / inverting circuitry to generate the dual voltage rails. If proper filtering precautions are taken for the switch-mode converters, I think it will introduce minimal noise into the system as well as dramatically decrease the board space currently required by the batteries.
Bill of Materials
https://docs.google.com/spreadsheet/ccc?key=0AlUlVoMq1FqldFVGcDd6SzY4SGcwOGVFU01aUFpJb3c&usp=sharing
Schematic
https://docs.google.com/file/d/0B1UlVoMq1FqlQXNoekRCZ1hRZmc/edit?usp=sharing
Layout
Bottom layer is a ground pour
3D render
I'm trying to improve my analog layout and circuit design - please provide feedback! I designed and laid-out a dual headphone amplifier based on nwavguy's O2 amp:
NwAvGuy: O2 Details
There is a ton of open space on the board, so clearly, this design can be miniaturized further. I wanted to make something nice for listening to music or watching a movie together in an airport / plane / quiet place.
One of the most notable differences is that there is only one 4556 output stage opamp per headphone output, meaning this amp can deliver about half the current of the O2 per channel. The thing I'm most concerned about are the thin-film surface mount resistors. I'm using Yageo 1% carbon thin-films, eg RT0603FRE07100RL and others from that series. I know others have used Susumu thin films and I'd be interested to know what the difference is between those and the ones I picked.
I was really excited about getting an initial prototype built, so I omitted some of the more practical features present in the O2 such as:
-switchable gain
-wall power
-low voltage battery protection: this device could become dangerous near the end of uneven battery discharge
-power LED indicator
-the ability to remove power from one of the headphone output stages (n/a to the O2)
The 3 PCBs I ordered from OSHPark are arriving tomorrow! For the next revision, I am considering using a single Li-Po cell and boost / inverting circuitry to generate the dual voltage rails. If proper filtering precautions are taken for the switch-mode converters, I think it will introduce minimal noise into the system as well as dramatically decrease the board space currently required by the batteries.
Bill of Materials
https://docs.google.com/spreadsheet/ccc?key=0AlUlVoMq1FqldFVGcDd6SzY4SGcwOGVFU01aUFpJb3c&usp=sharing
Schematic
https://docs.google.com/file/d/0B1UlVoMq1FqlQXNoekRCZ1hRZmc/edit?usp=sharing
Layout
Bottom layer is a ground pour
3D render
Last edited:
Good work! There have been a couple of posts in the main O2 thread over the last year or two from people wanting a dual "his and hers" O2. One quick solution in the past was to stick two O2 boards back to back in a longer B2-160 case and use a Y 3.5mm splitter cable on the input. Yours is a more elegant solution though, especially if someone has phones that don't need the current from the paralleled NJM4556A.
Not having the battery low voltage protection and/or charging circuit would have been a problem, but not since you are going to switch to LiPo. Apparently it is possible to keep the DC-DC noise out of the audio. Take a look at JDS labs C5 if you haven't already:
JDS Labs - C5 Headphone Amplifier
He has a converter in there and has performed the same dScope tests that NwAvGuy did, plus an FFT at the switching frequency. Everything looks good.
One suggestion, since you have some board space. I see why you changed the NJM4556A input ground return resistors to 100K from 40.2K. You only have one half chip's worth of input bias current draw through each instead of two, hence half the IR voltage. And that 100K lets you shrink the coupling cap to 1uF keeping the corner frequency the same.
But... the 100K resistor may give you more Johnson / thermal noise. I would suggest backing those resistor back down to 40.2K and put two 1.0uF's in parallel on each circuit to go back to the 2.0 (2.2) uF. Or just a single 2.2uF SMD cap if one will fit.
I would also suggest putting your gain resistors R6 and R19 in header sockets, the ones that NwAvGuy has listed at the bottom of his BOM. That would make it easy to change the resistors, or pull them out entirely for 1x gain, especially since you don't have a gain switch.
It is kind of slick that your design does away with the need for the output (1R) balancing resistors! Even lower effective output impedance from the amp. Since the NJM4556A are internally current limited they don't need the resistors to protect against the TRS jack/plug shorting, of course.
Not having the battery low voltage protection and/or charging circuit would have been a problem, but not since you are going to switch to LiPo. Apparently it is possible to keep the DC-DC noise out of the audio. Take a look at JDS labs C5 if you haven't already:
JDS Labs - C5 Headphone Amplifier
He has a converter in there and has performed the same dScope tests that NwAvGuy did, plus an FFT at the switching frequency. Everything looks good.
One suggestion, since you have some board space. I see why you changed the NJM4556A input ground return resistors to 100K from 40.2K. You only have one half chip's worth of input bias current draw through each instead of two, hence half the IR voltage. And that 100K lets you shrink the coupling cap to 1uF keeping the corner frequency the same.
But... the 100K resistor may give you more Johnson / thermal noise. I would suggest backing those resistor back down to 40.2K and put two 1.0uF's in parallel on each circuit to go back to the 2.0 (2.2) uF. Or just a single 2.2uF SMD cap if one will fit.
I would also suggest putting your gain resistors R6 and R19 in header sockets, the ones that NwAvGuy has listed at the bottom of his BOM. That would make it easy to change the resistors, or pull them out entirely for 1x gain, especially since you don't have a gain switch.
It is kind of slick that your design does away with the need for the output (1R) balancing resistors! Even lower effective output impedance from the amp. Since the NJM4556A are internally current limited they don't need the resistors to protect against the TRS jack/plug shorting, of course.
Last edited:
Hi agdr - thanks for the suggestions! I have a lot to learn about analog design and any feedback is very helpful.
I like SMD a lot. Once you are used to it, its much faster to populate SMD boards by hand compared to through-hole, which is why I tried to choose SMD components wherever possible. Also, if you ever decide to get the board manufactured by a third-party, it is much cheaper to assemble SMD boards because its easier for the assembly house to setup automated pick and place -> reflow. Unless you are getting a LOT of boards manufactured, you're basically paying for someone to individually hand-solder your through-hole components.
That said... SMD film caps are ridiculously expensive and generally hard to find. For example, this is the only 2.2uF film cap available at Mouser and it costs $3.33 in single quantities, and its a larger footprint, so its not really an option from a cost-perspective. If I chose to go with 2uF for AC-coupling, I'd do 2 of the 1uF caps in parallel as you suggested, which would add about $2 to the BOM parts cost (would need 4 additional caps). With the current PCB, I could probably stack the caps on top of each other.
Part of my design choice for the current 1uF configuration was influenced by this paragraph in nwavguy's O2 Details post:
I figured a 4-5mV difference in DC offset wasn't going to be a big deal. 5mV / 16ohm load = 300uA. If it ends up being an issue - I'll gladly change the design.
Its kind of funny how the parts cost of a design like this is dominated by the electro-mechanical components (switches, pot, connectors, batteryholder). These parts cost $13.92 of the $20.83 parts total, thats 67%!
I hadn't seen the C5! Thanks for the link. It looks like they're boost / inverting the Li-Po to greater than +/- 7V, then using the LDOs for their final output, which is a well-documented method for obtaining a high efficiency and low-noise supply. I'd love to see a picture of the disassembled C5
I like SMD a lot. Once you are used to it, its much faster to populate SMD boards by hand compared to through-hole, which is why I tried to choose SMD components wherever possible. Also, if you ever decide to get the board manufactured by a third-party, it is much cheaper to assemble SMD boards because its easier for the assembly house to setup automated pick and place -> reflow. Unless you are getting a LOT of boards manufactured, you're basically paying for someone to individually hand-solder your through-hole components.
That said... SMD film caps are ridiculously expensive and generally hard to find. For example, this is the only 2.2uF film cap available at Mouser and it costs $3.33 in single quantities, and its a larger footprint, so its not really an option from a cost-perspective. If I chose to go with 2uF for AC-coupling, I'd do 2 of the 1uF caps in parallel as you suggested, which would add about $2 to the BOM parts cost (would need 4 additional caps). With the current PCB, I could probably stack the caps on top of each other.
Part of my design choice for the current 1uF configuration was influenced by this paragraph in nwavguy's O2 Details post:
I figured a 4-5mV difference in DC offset wasn't going to be a big deal. 5mV / 16ohm load = 300uA. If it ends up being an issue - I'll gladly change the design.
Its kind of funny how the parts cost of a design like this is dominated by the electro-mechanical components (switches, pot, connectors, batteryholder). These parts cost $13.92 of the $20.83 parts total, thats 67%!
I hadn't seen the C5! Thanks for the link. It looks like they're boost / inverting the Li-Po to greater than +/- 7V, then using the LDOs for their final output, which is a well-documented method for obtaining a high efficiency and low-noise supply. I'd love to see a picture of the disassembled C5
Last edited:
Here is a picture of the innards of the C5: Search down for "reference level performance" and the picture is just below.
JDS Labs Blog | Premium Headphone Amplifiers
In your case you wouldn't get twice the output DC offset with 100K since you only have one half of the NJM4556A connected to each 100K resistor. In the O2 both halves go through each resistor, so that is twice the current draw and twice the I * R voltage which then gets reflected to the op amp output by the 1x buffer stage. So if the O2 amp has about 3mV of output DC offset with two NJM4556A inputs pulling through one 40.2K resistor, you should also have about 3mV with one NJM4556A input pulling through one 100K resistor. Half the current times about twice the resistance.
But NwAvGuy didn't really get into the Johnson noise issue with that resistor, probably since it is feeding a 1x voltage gain stage. At 1x at least the noise isn't amplified, just passed right through. He did talk about the Johnson noise at the gain stage, the 274R resistor, since that noise would get amplified by any voltage gain in that first stage.
So... 40.2K + 2uF would give you less noise that would in turn get sent though, 1x, to the output. The Johnson noise is just related to the resistance of the resistor (more is noisier) and the ambient temperature.
To keep the corner frequency at 1.8Hz with the 2uF instead of 2.2uF you would actually need 44.2K instead of 40.2k.
Search down for "reference level performance" and the picture is just below.JDS Labs Blog | Premium Headphone Amplifiers
In your case you wouldn't get twice the output DC offset with 100K since you only have one half of the NJM4556A connected to each 100K resistor. In the O2 both halves go through each resistor, so that is twice the current draw and twice the I * R voltage which then gets reflected to the op amp output by the 1x buffer stage. So if the O2 amp has about 3mV of output DC offset with two NJM4556A inputs pulling through one 40.2K resistor, you should also have about 3mV with one NJM4556A input pulling through one 100K resistor. Half the current times about twice the resistance.
But NwAvGuy didn't really get into the Johnson noise issue with that resistor, probably since it is feeding a 1x voltage gain stage. At 1x at least the noise isn't amplified, just passed right through. He did talk about the Johnson noise at the gain stage, the 274R resistor, since that noise would get amplified by any voltage gain in that first stage.
So... 40.2K + 2uF would give you less noise that would in turn get sent though, 1x, to the output. The Johnson noise is just related to the resistance of the resistor (more is noisier) and the ambient temperature.
To keep the corner frequency at 1.8Hz with the 2uF instead of 2.2uF you would actually need 44.2K instead of 40.2k.
Last edited:
Looks good!
If you go with the LiPo power supply your amp will have one big leg-up over the C5: more output current capability. He is just using a single op amp stage for everything, gain and output. I forget which chip at the moment but I think it had the typical 24mA or so maximum out into the typical 2K or so. I believe the lowest he has measured with the dScope was 150R or so (details are in that blog post of his). Yours probably will be good down to 32R or so load. I think NwAvGuy went down to 15R with the two NJM455A sections paralleled.
You will need 2.5x the current capability on your LiPo circuit though, vs. the C5.
Another suggestion just hit me. I just realized you are using the surface mount version of the NJM4556A, which only has 300mW of power dissipation from the data sheet. You would get 700mW by switching to the DIP, or 800mW with the inline SIP - I'm using the SIP in a project in another thread. Looks like you may have enough vertical height and board space for either one.
If you go with the LiPo power supply your amp will have one big leg-up over the C5: more output current capability. He is just using a single op amp stage for everything, gain and output. I forget which chip at the moment but I think it had the typical 24mA or so maximum out into the typical 2K or so. I believe the lowest he has measured with the dScope was 150R or so (details are in that blog post of his). Yours probably will be good down to 32R or so load. I think NwAvGuy went down to 15R with the two NJM455A sections paralleled.
You will need 2.5x the current capability on your LiPo circuit though, vs. the C5.
Another suggestion just hit me.
I just realized you are using the surface mount version of the NJM4556A, which only has 300mW of power dissipation from the data sheet. You would get 700mW by switching to the DIP, or 800mW with the inline SIP - I'm using the SIP in a project in another thread. Looks like you may have enough vertical height and board space for either one.
Last edited:
Thanks!
The C5 uses the SO-8 version of the OPA2227 according to the product picture.
I'm not too concerned about the limited power dissipation of the SO-8 package. Worst case power dissipation in the 4556 opamps will occur at 5.7V p-to-p, or 4.0Vrms because the amp has a bipolar 9V rail.
Power dissipation in the opamp driving ... at 4Vrms
For example, limiting the power dissipation in the amplifier to 300mW, you could drive a 32ohm load with 1.6Vrms.
(1.6Vrms)^2 / 32ohm = 80mW. Are there any 32ohm headphones that require >=80mW to reach full loudness (100-130dB)? I haven't found any, but if they exist, I'd like to know!
I used this appnote for the calculations http://www.ti.com/lit/an/sboa022/sboa022.pdf
The C5 uses the SO-8 version of the OPA2227 according to the product picture.
I'm not too concerned about the limited power dissipation of the SO-8 package. Worst case power dissipation in the 4556 opamps will occur at 5.7V p-to-p, or 4.0Vrms because the amp has a bipolar 9V rail.
Power dissipation in the opamp driving ... at 4Vrms
- 600ohm: 27mW
- 250ohm: 65mW
- 54ohm: 300mW
- 32ohm: 506mW
- 16ohm: 1010mW
For example, limiting the power dissipation in the amplifier to 300mW, you could drive a 32ohm load with 1.6Vrms.
(1.6Vrms)^2 / 32ohm = 80mW. Are there any 32ohm headphones that require >=80mW to reach full loudness (100-130dB)? I haven't found any, but if they exist, I'd like to know!
I used this appnote for the calculations http://www.ti.com/lit/an/sboa022/sboa022.pdf
Last edited:
Orthos can be quite demanding. Hifiman HE-6s are pretty notorious, for example. They measured 53 ohms @ 1 kHz and needed 1.0 Vrms or 20 mW for 90 dB SPL.
Those would still make it to a bit over 100 dB on 4 Vrms, though honestly I'd be more worried about distortion at this point. Poor ol' 4556. NwAvGuy didn't even spec the O2 to drive these (though people have apparently reported pretty good results), and that's with twice the output current available.
BTW, 4.0 Vrms != 5.7 Vpp (but rather 5.7 Vp or 11.4 Vpp).
Those would still make it to a bit over 100 dB on 4 Vrms, though honestly I'd be more worried about distortion at this point. Poor ol' 4556. NwAvGuy didn't even spec the O2 to drive these (though people have apparently reported pretty good results), and that's with twice the output current available.
BTW, 4.0 Vrms != 5.7 Vpp (but rather 5.7 Vp or 11.4 Vpp).
I hadn't considered this when I was building the circuit, but as it stands, the turn-on transient is an issue that needs to be addressed. There are differences in the rise time of the positive and negative rails that the battery protection circuit of the O2 guards against.
My mistake. It will be fixed.
My mistake. It will be fixed.
Last edited:
Hey congratulations, pretty good job! Looks very professional.
According to agdr's idea try pair of 4.7K or 10K(as a bleeder resistor) in parallel with electrolyte cap. to reduce the turn-off transient (if its there). As for turn ON transient I had used 47uF with 0.1uF & toggle switch instead of push-switch for my JRC4556 CMOY with no audible thump.
BTW does those smd JRC4556 gets little warm (no load). I faced thermal issues with smd JRC4556 at +/-7.5V. With +/-5V everything was fine.
Have you(or planning to) published BOM, build-guide, e.t.c. like NwAvGuy?
Last edited:
Thanks!
I could try wiring in the 10Ks for the turn-off transient - that would be an easy experiment. Interesting note about the toggle switch. The BOM is in the original post. I wont be doing a build guide for this revision since there are practical issues that make it inconvenient to use. Once I've finished the second revision, I will publish more complete documentation.
The NJM4556AM feel slightly warm, but I don't think its an issue. I used an IR thermometer and the IC temperature never got above 30 C during operation.
Last edited:
Hey Gastro, in the end what did you do with the transient?
Did you update the board or just let it be?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.