Merlins SSG Headphone amp

[Tom Tech]

Hello to all HA-self-builders,

I have been lurking around here for some time, getting ideas and reading about all the great amps people make and I have been thinking about a portable headphone amp that will be simple, small and good ;-)

I guess that the amp that I have in mind does not fall in the category "super-duper-high-end", it's more a good quality amp with some handy extra features, read on 😉

For the headphone amp I thought along the lines of :

A) an amp with either a good quality OpAmp that can drive headphones directly
like an OPA 1612 or OPA 1688 when headphone impedance is not too low.

or

B) an good (audio) OpAmp with added buffer to drive the phones.

I have made headphones amps like this before with an AD823 opamp and BD139/BD140 transistors. I will look for other transistors now, since I will make an all SMD version. For the opamp I have chosen an SO8 footprint and for the transistors I'm looking at SOT89 types that will get a good area of copper for cooling so I can still choose a class A / class A-B amp.
There are many dual opamps in SOIC8 available, so plenty of room for experiments.

(see schematic)

Instead of using two 9V batteries, I was thinking of using a single LiIon cell for the powersupply and adding converters to make +-9V. This can be done with a micro-controller that is also usefull to add a few other features, like:

(see sketch)

* DC output protection / switch on delay.

* input signal detection, to switch off the amp when there is no signal on the inputs for, say, 20 minutes or so, since I often forget to switch off and end with empty batteries ;(

* Battery monitor with automatic shut-off when batteries are empty.

* Battery charger circuitry to be able to charge with 5V (USB) or 12V (car)

* indication LEDs for status, battery full, charging, on/off etc.

All should fit in an enclosure of about 60 x 120 x 30 mm, be easy to assemble and use not too exotic/expensive parts.

Curious to know your suggestions, do's or dont's !

So let me know what you think.

Thanks,

Tom

 

[sgrossklass]

Re: A), I think the one for driving headphones was the OPA1622.

Re: B), I would refine the buffer a bit. Vbe multiplier transistor instead of diode biasing, move opamp output up, leave out upper bias resistor (free opamp class A bias), add series resistor (47-100R) between opamp output and buffer, add (optional) feedback capacitor (xx pF NP0) between opamp out and -in. With the 4R7 emitter resistors shown, optimum output stage AB bias is about 2.5-5 mA.

For the buffer transistors you could use NXP's BCX56/53 (or the lower-voltage and presumably higher-beta 55/52 or 54/51), the surface-mount version of the now-discontinued BC639/640 or Philips' old BD139/140 (R.I.P.). A much faster transistor than your average jellybean 1.5 amp BD139/140, read: suitable for lower idle current. Also available in SOT-223 (BCPxx) for slightly higher dissipation. You may be able to find types with lower beta droop and highe Early voltage though, maybe from Toshiba or something.

In case of a DC fault, it may be more advisable to turn off the power (and flash the LEDs) rather than trying to interrupt the output - both solid-state and regular relays tend to use power, unless you're using a bistable, and even then relays always are potential trouble spots.

I'm not sure what you mean with the microprocessor that does +/-9 V, I guess a microprocessor controlled converter? It's basically doable, switching converters usually need a good bit of attention to EMI though, so keep that in mind and do not skimp on filtering components and layout.

For your input signal detection, you may be able to use a very low-power (FET input) opamp as a buffer to drive a peak detector with a Schottky (1N60P-ish?).

 

[Tom Tech]

filling in the dots...

First of all, thanks for your extensive comments and suggestions! Highly appreciated.

Re: A), I think the one for driving headphones was the OPA1622.

Yes, of course I have been looking at that one with eager eyes too, but although I have a fair amount of SMD experience, I'm a bit hesitant soldering this microscopic thingy...
The course I'm taking with this design also gives me plenty of room for experimenting with different opamps (many available in SOIC8).
In my previous design I noticed big differences in 'musicality' between opamps...

Re: B), I would refine the buffer a bit. Vbe multiplier transistor instead of diode biasing, move opamp output up, leave out upper bias resistor (free opamp class A bias), add series resistor (47-100R) between opamp output and buffer, add (optional) feedback capacitor (xx pF NP0) between opamp out and -in. With the 4R7 emitter resistors shown, optimum output stage AB bias is about 2.5-5 mA.

I have drawn the changes as I understand from your remarks (see new schematic). I have a few questions though:

* What is the purpose of the resistor (R5 56ohm) in the output of the opamp? Just to limit opamp output current?

* the feedback C (C6 56pF) is meant as an 'speed-up' measure for the opamp?

* I have taken E2 to be 47uF would that be enough, roll-off is mainly through E2-R7? (that would be 0,7Hz)

For the buffer transistors you could use NXP's BCX56/53 (or the lower-voltage and presumably higher-beta 55/52 or 54/51), the surface-mount version of the now-discontinued BC639/640 or Philips' old BD139/140 (R.I.P.). A much faster transistor than your average jellybean 1.5 amp BD139/140, read: suitable for lower idle current. Also available in SOT-223 (BCPxx) for slightly higher dissipation. You may be able to find types with lower beta droop and highe Early voltage though, maybe from Toshiba or something.

I have been looking around and indeed found the BCX54/51; gain 100-250 in the -16 class, transistion freq 100-180MHz and 6pF collector capacitance.
Another one I found, maybe even better, is the 2SCR293/2SAR293, minimal gain 270, trans. freq. 320MHz, 7pF col. cap. also SOT 89 and cheap ;-)

In case of a DC fault, it may be more advisable to turn off the power (and flash the LEDs) rather than trying to interrupt the output - both solid-state and regular relays tend to use power, unless you're using a bistable, and even then relays always are potential trouble spots.

Just thinking; of course cut-off the powersupply is effective, but maybe I can also use a crowbar across the supply output (capacitors!) (or even audio outputs!?), to be quick reacting... That might be faster and in the end a less troublesome and a simpler solution!

I'm not sure what you mean with the microprocessor that does +/-9 V, I guess a microprocessor controlled converter? It's basically doable, switching converters usually need a good bit of attention to EMI though, so keep that in mind and do not skimp on filtering components and layout.

I have made several buck and boost inverters for these type of voltages and currents with a PIC processor. A PIC running on 8MHz can generate 100kHz PWM or maybe a bit more. That is still easy to filter, I only wonder if I need post regulation to improve noise numbers, I'll have to check. Many available SMPS chips are relatively expensive and often (again) have un-doable footprints for DIY soldering equipment ;-(
Making the inverters this way it is also easy to do soft-start and soft switch off.

For your input signal detection, you may be able to use a very low-power (FET input) opamp as a buffer to drive a peak detector with a Schottky (1N60P-ish?).

Currently looking around for a low power opamp that can do the job...

Thanks again.

Tom

 

[sgrossklass]

I have drawn the changes as I understand from your remarks (see new schematic). I have a few questions though:

* What is the purpose of the resistor (R5 56ohm) in the output of the opamp? Just to limit opamp output current?

* the feedback C (C6 56pF) is meant as an 'speed-up' measure for the opamp?

These two are a common technique for improving capacitive load driving, especially at lower gains (it would be quite unusual to run a portable headphone amp at 20 dB - if you need that much gain, it may be more advantageous to add a line-level amplification stage in front of the volume control for an additional 6-10 dB, possibly inverting, but make sure to include a unity gain setting or bypass to avoid clipping). (Speaking of low gains, note the increasing importance of common-mode distortion performance at these. The FET-input OPA16x2 should be quite good in that respect.)

The opamp has an output impedance that rises with frequency, behaving like an inductor. When combined with the buffer's input capacitance, this tends to result in peaking which you don't want. Emitter followers really aren't into inductive source impedance especially when faced with capacitive loading, they easily break into oscillation then. The resistor basically dampens the LC, you could say it's a base stopper.

The cap provides local feedback, essentially taking the buffer out of the loop at high frequencies. The feedback lowers output impedance and flattens it out (making it dominantly resistive) over a certain frequency range, while also reducing the opamp's external GBW. So the opamp gets slower, the buffer gets faster, making the whole affair more well-behaved. Typical pole splitting approach.

Of course taking away GBW from global feedback also means less feedback for reducing buffer stage distortion at high frequencies, so one has to be careful not to overdo it, and when on a tight power budget, may have to resort to 2nd-order compensation schemes (TMC etc., the Solid State section should have a fair bit of discussion on these).

* I have taken E2 to be 47uF would that be enough, roll-off is mainly through E2-R7? (that would be 0,7Hz)

It's mostly of interest for the higher frequencies anyway, where loop gain is inevitably reduced. Time constant is formed with the parallel bias network impedance. One might typically use 100-220 µF.

I have made several buck and boost inverters for these type of voltages and currents with a PIC processor. A PIC running on 8MHz can generate 100kHz PWM or maybe a bit more. That is still easy to filter, I only wonder if I need post regulation to improve noise numbers, I'll have to check.

A bit of supply cleanup (maybe using a lil' cap multiplier to save space) can be the smarter option compared to relying on brute force PSRR easily marred by layout issues. Be sure to give the power amp stage a few hundred µF of local buffer capacitance with low parasitic inductance in the layout, that plus a few ohms in series might actually already provide quite adequate filtering for audio-frequency noise as well. (You can get arbitrarily fancy with this, up to dedicated RC filtering for the buffer, main opamp, and potential input amp.)
Depending on what kind of ground return impedance you can achieve, consider using dedicated signal and power ground returns to avoid coupling amplifier switching noise into signal ground via your decoupling capacitance (said series resistors in the supplies tend to help a fair bit with that already).

 

[Tom Tech]

Thanks you for your explanations. It's clear you have a better (theoretical) background and understanding of these things than I do. I studied electronical engineering only for 1 year before dropping out, mostly because I'm more a hands-on learning type of person.

I do have basic insights, some very good 'teachers', plus "The art of electronics" from Horowitz & Hill and many years of experience, so I can follow your explanations reasonable well.

The gain of the amplifier as I build it previously was around 10x, which gave plenty of power for both my Sennheiser PX100 (32ohm) and my old Sennheiser HD430 (600ohm!), so I won't need extra gain. I will reserve space on the board for the feedback C and experiment later.

Concerning the value of E2 in my last schematic, you wrote:

It's mostly of interest for the higher frequencies anyway, where loop gain is inevitably reduced. Time constant is formed with the parallel bias network impedance. One might typically use 100-220 µF.

I thought E2 was there to ensure that both output transistors bases see the same signal and that you would size it for the lower frequencies? Am I mistaken?

As far as the power/microcontroller is concerned: I will make it on a separate board and take good de-coupling measures with LC and/or RC filtering on all lines coming in and out.

Once I have done the improved schematic and board layout, I will post them here as well.

Thanks again!

 

[sonidos]

Hi Tom Tech. I am also going down a similar road so I've been lurking a bit on this thread. Keep up the good work and I look forward to reading up on your progress.

 

[Tom Tech]

Current state of afairs

Hi Tom Tech. I am also going down a similar road so I've been lurking a bit on this thread. Keep up the good work and I look forward to reading up on your progress.

I haven't build anything yet, but so far I have drawn 2 schematics, one for the audio part and one for the power board.

As you can see, I changed the audio part according some remarks I received here. Depending the type of headphones you want to drive, you can possibly skip the power output drive transistors.

For the power board I decided to use a small ST ARM processor. Not that it is necessary to have that type of processing power, but for me it is also a learning project while I'm switching from PIC to ARM processors.

The power board contains the following parts:

* the processor is an STM32L011 in 20 pin SSOP package, the processor drives a fly-back inverter that makes both the pos. and the neg supply voltages and takes care for smooth start-up and close down (soft start) so avoiding power pops and clicks.
Furthermore it monitors the following:
- Battery voltage
- DC detect on audio outputs
- signal detect on inputs
Besides the processor regulates the charging of the battery by either an USB (5Volt) input or from a aux. input voltage between 8 and 16 Volt (think: car battery)
There is one push button to switch on/off and two indicator LEDs to display status, battery health, error etc.

Please note that components values are *NOT* correct at this stage!

It will take some time to finish this project, because I also have other projects in the pipe-line.

Let me know what you think.