everyone loves headphone amps!

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

tmbg

Well-known member
Joined
Nov 7, 2004
Messages
438
Location
Atlanta, GA
Hi folks!

Here's my eventual goal:

My portable recording rig consists of a couple of B* ADA-8000 boxes. I use all them wonderful mic pres and line ins of course, but I'm only using one pair of the 16 line outs I have for my reference monitors.

What I'd like to do is build a couple of headphone amps that can take a balanced line from the ADA8K and drive a couple of pairs of headphones. I will probably make a mono version and a stereo version. I did some playing around last night, breadboarding a couple of designs that I googled up, they all work, but not how I'd expect. I'd feed it a sine wave, and the output would be nice and clean, but once I hooked headphones up to the output, it got really horribly distorted.

So, basically I just want a little bit of guidance to make sure I'm on the right track here... the line outs of the B* are gonna be +4dBu balanced, so I should be designing so that maximum input level without the output clipping is roundabout 1.23V, correct? What sort of output level should I be shooting for? +20dBu? +24dBu?

If I'm doing my calculations right, +20dBu is 7.75VRMS, which is about 22V peak to peak, which is a reasonable swing for a +/-15V supply. +24dBu is closer to 35Vpp, which I can't do on +/-15V.

next, I need to know about output impedance. Seems like most designs mention using a 50R resistor in series with the output, is the goal to match Zout to the Zload of the headphones? If I'm using 32R headphones, should I be shooting for a 32R Zout?

How does the load of headphones affect the output of a good amp? If the unloaded output is a nice clean pretty sine wave, should the loaded output be a nice clean pretty sine wave of half the amplitude? My tests on the breadboard last night showed nasty distorted destruction when loaded.

7.75V into 32 ohms is 250mW. I'm using OPA4228's, which have an output drive capacity of around 45mW. Perhaps that is what was causing the distortion? I may have to look into either some higher drive opamps or possibly just a simple emitter follower that can handle that sort of drive and swing.

So, my plan at this point I think will be one op amp for debalancing and buffering, one with 16dB gain, and a 2N4401 or TIP31 or something for output drive.

Now that I'm done rambling, anyone care to comment? (PRR this means you, you naughty headphone guru!)
 
[quote author="tmbg"]So, basically I just want a little bit of guidance to make sure I'm on the right track here... the line outs of the B* are gonna be +4dBu balanced, so I should be designing so that maximum input level without the output clipping is roundabout 1.23V, correct?[/quote]

No; the line outs on the B* are +4dBu at "nominal" level, loosely defined as the level that would have read 0 VU back when we used VU meters, analog tape and the like. Typically this nominal level is set around -20 - -15dBFS, so the input needs to handle 15 - 20 dB of signal above the nominal +4dBu level, or +19 - 24 dBu.

What sort of output level should I be shooting for? +20dBu? +24dBu?

If I'm doing my calculations right, +20dBu is 7.75VRMS, which is about 22V peak to peak, which is a reasonable swing for a +/-15V supply. +24dBu is closer to 35Vpp, which I can't do on +/-15V.

Right. Probably the best strategy is to incorporate a unity-gain balanced input circuit, then a level control, then the actual headphone amp; I'd try for a gain of 10x (20dB) on the latter.

next, I need to know about output impedance. Seems like most designs mention using a 50R resistor in series with the output, is the goal to match Zout to the Zload of the headphones? If I'm using 32R headphones, should I be shooting for a 32R Zout?

Not really; headphone circuits (in common with most current designs) don't work on the principle of impedance matching. The 50R resistor serves two functions. First, it provides a certain amount of isolation for the amplifier from the cruel world outside, to help preserve the amp's stability. Second, it serves as something of an automatic gain adjuster. Low-impedance headphones tend to be quite sensitive, while higher-impedance phones (like the 600-ohm units popular for many years) tend to be less so. The series resistor acts as a voltage divider, causing a good deal of attenuation when used with 8-ohm phones, less with 32-ohm phones, and very little (about 0.7dB) when used with 600-ohm phones. It also kills your damping factor on low-impedance phones, which may or may not make a difference in the sound.

How does the load of headphones affect the output of a good amp? If the unloaded output is a nice clean pretty sine wave, should the loaded output be a nice clean pretty sine wave of half the amplitude? My tests on the breadboard last night showed nasty distorted destruction when loaded.

7.75V into 32 ohms is 250mW. I'm using OPA4228's, which have an output drive capacity of around 45mW. Perhaps that is what was causing the distortion? I may have to look into either some higher drive opamps or possibly just a simple emitter follower that can handle that sort of drive and swing.

Yep; however, that's also probably enough to blast a hole in your ears with typical 32-ohm headphones. Depending on the sensitivity of your favorite cans, you probably can get away with a lot less juice.

So, my plan at this point I think will be one op amp for debalancing and buffering, one with 16dB gain, and a 2N4401 or TIP31 or something for output drive.

Not unreasonable, but do put a level control in between the input section and the gain section. Headphones vary all over the place in sensitivity, and you'll need to adjust.

Peace,
Paul
 
wow, thanks Paul, that's good stuff :)

In the time since I posted that, I've been reading and thinking and scheming and drawing, and basically doing anything I can to make it be closer to the time when I can go home and resume experimenting :)

Ok, so I didn't factor in that the output of the +4dBu lines might be well above +4dBu... I'm so used to dealing with digital that I assumed the 'nominal' level was also the full scale level.

Since these particular lines are on the back end of a D/A converter, there is obviously some 0dBFS value which will be the maximum level that is ever coming out of these things, and it's probably pretty trivial to measure, and I'll do that when I get home.

Assuming that it really is up in the neighborhood of +15-19dBu, that's a LOT of voltage pumping out of that thing! ~20Vpp. Good thing I have a stout +/-15V supply...

I looked up the specs of the actual cans I own, which gives me a good initial target. They're 24 ohm, 500mW max, which translates to 3.46VRMS or 13dBu. So I'll definitely need some form of attenuation...

Ok so my first shot will be a debalancing buffer + attenuator + 20dB op amp gain block + power stage. For the attenuator I'm just gonna use a pot-based voltage divider, and go from there.

Thanks again! :)
 
tmbg, I agree with Paul on the danger to your ears at anythng approaching 250mW! It may be nice to have that as headroom, and there may be some folks out there who are nearly deaf, but....ouch! For these old ears a few mW sounds adequate at least as an average level.

As far as the damping factor issue, I suspect many 'phones are developed assuming some fair amount of series R, so lower output Z for your amp may not necessarily always be a good idea. And as Paul points out the series R does have a benefit in terms of facilitating the use of many different Z phones, as well as isolating reactive loads from the amp.

It is interesting that this R could be realized synthetically if one wanted to have higher efficiency of the overall circuit, although this is hardly likely to be necessary. That is, with a combination of voltage and current feedback the amp could be made to look like a higher R than the physical series resistor, while burning less power in the resistor in the process. It's easiest to do this if one has access to all four wires from the two sides, which is an unusual situation.

If you do roll your own output buffer it would be wise to include it in the overall feeback loop. Beware of making the whole circuit oscillatory though, and if possible make the circuit symmetrical, even if you operate the stage class A (some single-ended aficionados will no doubt disagree here though). I believe there are still some decent unity-gain fast buffers out there that would be suitable, although not as much fun as wiring your own ;-). The BUF04 (Analog Devices, used to be Burr-Brown iirc) is a little shy on output current but would probably be adequate and is likely fast enough not to screw up stability when paired with a decent op amp.


Cheers,

Brad
 
I read somewhere that there's an IEC standard that says any modern headphones should play well with 120R series output resistance.

As for using a high-current op amp for my output buffer, that's a great idea, but I don't have any on hand, and the early stages of this are gonna be 'on-hand' projects definitely :)

At any rate, I just need to scoot home and start playing with it.
 
> They're 24 ohm, 500mW max

What is the peak current needed to reach that level? (Hint: about 10 times higher than a generic op-amp.)

What is the rated sensitivity? 90dBSPL/mW is typical ball-park (it varies WIDELY). What is the dB of 500 times more power? Hint: a LOT! Now, you might need that much for short periods in live recording, but better get the real sensitivity value and do some sanity-checking. 500mW is a hefty little amplifier, and you do not HAVE to supply enough power to melt the cans it might lead to a nicer design.

> a few mW sounds adequate at least as an average level.

Yes, but musical peaks are 10 to 50 times higher power than average (a fact that has been clear since the 1920s, when a 100mW amp was asked to fill a room). So if "a few mW" means "4", then you are looking for 40mW with occasional clipping (good enough for radio) or 200mW to get transient clipping down to "never" (once every week of 24/7 audio). So powers in the 50mW-500mW range are reasonable for high-clarity high-accuracy hearing.

Here is some theoretical meat to chew on:

Headphone Power

Amplifier Capabilities

But the short/quick answer for you is: Loudspeaker amplifier chip, +/-12V or +/-15V rails, 20-30Ω output series resistor.

Working from -10dBV sources, and using high gain to check for noise in the quiet parts, you could need a gain as much as 50, but it better be able to turn way down. For a nominal +4dBu system, gain of 12 is equivalent. If you are not noise-checking, gain about 4 time lower will cover all normal musical listening. (This assumes ~30Ω series resistor, and no wildly hot/dead phones.)

> there's an IEC standard that says any modern headphones should play well with 120R series output resistance.

I believe that is a test standard, a paper test, like the US goverment MPG test. It allows some comparison between models, but isn't how we really drive (cars or cans). AFAIC, most low-Z phones do not care much about damping, but some of the high-Z ones seem flatter with 120Ω or less. In point of fact, a 120Ω will generally not drive low-Z phones as loud as high-Z phones, though the WIDE variations in headphone efficiency mean your results will vary.
 
Back
Top