Headphones Amp

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
In conditions when you have a limited power supply of 12V (how did you get + 20dBu of possible output voltage?)
The +20dBu is from the output transformer of the tube stage. The semiconductor headphones amp will not be powered from the HT but from the 12V heater supply
I would suggest using rail to rail headphone opamp chips like ct
https://www.ti.com/lit/ds/symlink/o...89841&ref_url=https%3A%2F%2Fwww.google.com%2F
That looks to be just the job. It even provides balanced inputs.

Cheers

Ian
 
I would suggest using rail to rail headphone opamp chips like OPA1688.
That looks to be just the job. It even provides balanced inputs.
No it doesn't. It's just an op amp. Figure 49 is just a typical balanced input arrangement.

There's nothing wildly different about OPA1688 specific to headphones. It looks like they just added a bunch of stuff to the datasheet about driving headphone loads. They just characterized using the op amps natural current limiting to handle the issue of different loads. But the data is convincing. Clearly OPA1688 would make a very good headphone amplifier that will limit current as desired.

But for a single supply you still have the issue of DC coupling.
 
No it doesn't. It's just an op amp. Figure 49 is just a typical balanced input arrangement.
Indeed but most of the other op amp based topologies I looked at would have required another op amp to do that
There's nothing wildly different about OPA1688 specific to headphones. It looks like they just added a bunch of stuff to the datasheet about driving headphone loads. They just characterized using the op amps natural current limiting to handle the issue of different loads. But the data is convincing. Clearly OPA1688 would make a very good headphone amplifier that will limit current as desired.
It seems to me there must be some significant difference between driving a 600 ohm load (5532) and a 32 ohm load (1688) - of course it is current drive capability. 1V rms into 32 ohms is 30mW and also 30mA. 30mW into 600 ohms is just 7mA. 5532 at maximum swing into 600 ohms (+-15V supply) is only 15mA. I am by no means an op amp expert but it seems to me that high currents directly into low impedances must lead to inefficiency in the class B output stage which in turn leads to potential thermal problems. Staying with the 1V into 32 ohms example the pp voltage is 2.828V. If I have a 12V rail the output transistors must dissipate (12 -2.828) x 30 milliwatts = 275mW. To be honest I would not be at all happy about dissipating that much power in an IC. I think I would rather use a discrete complementary pair tacked on the end of an op amp.
But for a single supply you still have the issue of DC coupling.
Fortunately I don't plan on using dc coupling

Cheers

Ian
 
If I have a 12V rail the output transistors must dissipate (12 -2.828) x 30 milliwatts = 275mW. To be honest I would not be at all happy about dissipating that much power in an IC. I think I would rather use a discrete complementary pair tacked on the end of an op amp.
I think I have to challenge that. If you put an SOIC-8 on a 2-3 square inch area of copper on top and below stitched together with lots of vias, that makes a nice heat sink. Probably better than just using bulker transistor packages. Certainly it could handle a 1/4 Wp.
 
While possibly 'not cricket' you could consider a small DC-DC convertor (possibly 3 Watt module) and get yourself a symmetrical plus and minus 12 /15 Volts or even a cheap miniature mains transformer with dual 12 Volt secondaries and use it as a voltage doubler (and chop off the old mains pins!).
Paralleling 2 or more sections of NE5532 (using the scheme shown on the earlier post with perhaps 200 Ohm output resistors for each op amp section used so perhaps 2 op amps per channel. One section as balanced input then 3 'halves' paralleled.
A single 5532 section is getting into clipping (current limit) driving 0dBu into 33 Ohms I happen to remember.
 
While possibly 'not cricket' you could consider a small DC-DC convertor (possibly 3 Watt module) and get yourself a symmetrical plus and minus 12 /15 Volts or even a cheap miniature mains transformer with dual 12 Volt secondaries and use it as a voltage doubler (and chop off the old mains pins!).
I had considered the latter but I try very hard to keep poer supplies outside of my mixers!
Paralleling 2 or more sections of NE5532 (using the scheme shown on the earlier post with perhaps 200 Ohm output resistors for each op amp section used so perhaps 2 op amps per channel. One section as balanced input then 3 'halves' paralleled.
A single 5532 section is getting into clipping (current limit) driving 0dBu into 33 Ohms I happen to remember.
Paralleled 5532s seems to be a recurring theme. According to my calculator, the 15 ohm emitter resistor drops 0,65V at around 40 something mA. Three in parallel would be interesting.

Cheers

Ian
 
I think I have to challenge that. If you put an SOIC-8 on a 2-3 square inch area of copper on top and below stitched together with lots of vias, that makes a nice heat sink. Probably better than just using bulker transistor packages. Certainly it could handle a 1/4 Wp.
Hmmmm, I don't doubt that with that amount of copper it would handle 250mW but that is a lot of real estate that cannot be used for tracking.

Cheers

Ian
 
Indeed but most of the other op amp based topologies I looked at would have required another op amp to do that
As Bo mentioned, any opamp can offer differential inputs and deliver an unbalanced output.
However, the final arrangement should consider the nature of the source (bal /unbal) and the nature and placement of the volume control.
Unless you consider a quad volume pot or monitoring in mono, I believe you will separate the debal function and the power amplification.
To be honest I would not be at all happy about dissipating that much power in an IC.
Neither would I, but it seems TI are not shy about it, although they restrict the output power to 30mW into 16R (the recommended max power for earbuds), which is about 5mA.
In the absence of data, it's quite possible the current limit could increase when the rail voltage decreases
For higher impedance loads, the available power would increase linearly until it's limited by the rails. Admittedly, 30mW may not be enough for your typical DJ.
Fig. 49 in the datasheet shows an application that delivers 50mW into 32 R with +/-5V rails.
Simulation shows limitation at 70mA peak with +/-5V rails, and no clipping at 2Vpk into 32R (60mW). Of course simulation is not real life; I think thermal limits are not taken into account.
I think I would rather use a discrete complementary pair tacked on the end of an op amp.
That is a much simpler alternative and is known to work quite well. However, I'm not sure it exists in kit form.
Fortunately I don't plan on using dc coupling
How do you intend to do that? Are you using an smps with a virtual ground?
 
Last edited:
How do you intend to do that? Are you using an smps with a virtual ground?
Why would he need an SMPS?

Seems to me a simple voltage divider / cap virtual ground would work perfectly fine. All of the output current is coming or going through the load. Even a non-inverting with gain wouldn't budge Vg. So OPA1688 with voltage divider / cap virtual ground into two 1000uF 10V caps and you're done. With caps you don't really have to worry about startup. Easy. Done.
 
Why would he need an SMPS?

Seems to me a simple voltage divider / cap virtual ground would work perfectly fine. All of the output current is coming or going through the load. Even a non-inverting with gain wouldn't budge Vg. So OPA1688 with voltage divider / cap virtual ground into two 1000uF 10V caps and you're done. With caps you don't really have to worry about startup. Easy. Done.
Pretty close to what I was plannig.

Cheers

Ian
 
As Bo mentioned, any opamp can offer differential inputs and deliver an unbalanced output.
However, the final arrangement should consider the nature of the source (bal /unbal) and the nature and placement of the volume control.
Unless you consider a quad volume pot or monitoring in mono, I believe you will separate the debal function and the power amplification.
I am not worried about CMRR in the connection to the phones amp input so there is no need for a quad pot. A single pot between hot and cold of each channel will suffice.

Cheers

Ian
 
Seems to me a simple voltage divider / cap virtual ground would work perfectly fine. All of the output current is coming or going through the load. Even a non-inverting with gain wouldn't budge Vg. So OPA1688 with voltage divider / cap virtual ground into two 1000uF 10V caps and you're done. With caps you don't really have to worry about startup. Easy. Done.
Does it mean that the headphone will be connected between opamp's output and virtual ground point which sits at V/2 and is made with two resistors and two big caps?
 
Does it mean that the headphone will be connected between opamp's output and virtual ground point which sits at V/2 and is made with two resistors and two big caps?
No. If you have a cap in there you can source / sink from / to the regular host 0V (which is the negative rail from the prospective of the op amp). When sourcing current it's 12V > output > coupling cap > voice coil > host 0V but when sinking current it's voice coil > couping cap > output > host 0V. So it's a little odd in that the coupling cap is actually what sinks current from the voice coil on negative excursions. But I don't think there is any consequence for doing that. Whatever works.
 

Attachments

  • AA_HeadphoneGizmoUserGuide_2102151.pdf
    1.4 MB · Views: 17
No. If you have a cap in there you can source / sink from / to the regular host 0V (which is the negative rail from the prospective of the op amp). When sourcing current it's 12V > output > coupling cap > voice coil > host 0V but when sinking current it's voice coil > couping cap > output > host 0V. So it's a little odd in that the coupling cap is actually what sinks current from the voice coil on negative excursions.
Pretty contradictory with what Ian wrote: "Fortunately I don't plan on using dc coupling"
 
Back
Top