I built myself a headphone amplifier for my bass. It's mono using one half of a single-supply dual opamp (TLC272A). I'm using a very simple circuit from the data sheet (AC-coupled non-inverting) and so far it's working ok. I'm not expecting high fidelity out of it.
My question: can I simply bridge the inputs and outputs of the two amps to get twice the output power? I somehow feel that's the wrong approach, is it and why? If it's not ok, how much of the circuit do I have to duplicate? I basically have a voltage divider to offset the input to 4.5V (9V battery),
10:1 resistors for feedback and caps everywhere to decouple input, output and the feedback divider.
My several headphones are in the 25-60 ohm range. Not sure I understand your comment, though. I'm a relative noob, as you can probably tell. I thought that by running the two amps in parallel I'd effectively get twice the "available" current at a given voltage. I went with a 10:1 amplifier based on the schematics I was looking at and it turns out to be a good ratio for my bass, which has a pretty low level output voltage. Now I'm wondering if I can use a second parallel amp to get more output power for the lower impedance headphones.
If you google on "headphone-amp output level" you'll find a related thread in rec.audio.pro, May 2004. In that thread, an article by Douglas Self was mentioned,
In that article's section on "driving heavy loads" he discusses paralleling opamps to achieve better current capacity. However, it's not a great technique. As Self's article makes clear, most opamps get pretty lousy with loads below 600 ohms; to get down to driving the 32 or 16 ohms typical for many modern phones and earbuds, you'd need to parallel quite a few opamps. For instance, I own a commercial headphone amp that uses four parallel opamps to drive each output channel. I have measured the distortion of this rig and it climbs steeply up with low (30 ohm) loads.
Your TLC272 is a poor choice for this application; if you look on its datasheet you'll see that, characteristic of CMOS opamps, it is specified only down to 10k loads (so you're off by a factor of 500 or so), and it is not stable with capacitive loads (the typical capacitance of a long-ish headphone cord is enough to set it oscillating). Its current drive capability is very weak - although it's specified for a max current of 30mA, the amount of current it can drive while still having any sort of gain, bandwidth, and distortion spec is more like a tenth of that. You would be far, far better off with something like an OPA2134 (or even a TL072), though IMHO these still do not really have enough current drive, even with two sections paralleled.
Rather than paralleling opamps, you might consider the common approach of adding a discrete class-B buffer to the output. Self gives an example circuit in his article. One caution: although Self says that he did not need any compensation to avoid oscillation, in my own experience I've found it necessary to put a small (47pF or so) capacitor in parallel with the feedback resistor.
Thanks for the info and link, guess I'll have some reading to do this weekend.
I read some articles by Chu Moy and he used the OPA too. We have a lot of different opamps in our bin stock here but no OPA so I picked one with a decent output current but I didn't understand the relationship to distortion and gain. My plan was to learn about opamps with the stuff on hand and then order some parts that are more appropriate. I'm actually surprised that the sound is pretty good to my ears given that I've committed other sins like using cheap caps for decoupling.
This sounds like a good idea - I might go for that.
I assume the TLC272 is a 'rail to rail' amplifier. That means it should be able to swing +/- 4.5V at the output. Into a 25 ohm load the load current would be
+/- 180mA but the chip will current limit long before that.
You'll therefore get more output by operating 2 sections in 'parallel'. You don't need more volts ( as you'd get by bridging ).
Configure the second half of the op-amp as a voltage follower connected to the first half's output and connect the outputs together via a couple of low value Rs ( say 10 ohms ) in series with each output pin ( for current sharing ).
I would recommend that a considerable amount of skepticism be applied when reading articles by anyone who writes about audio without the benefit of either a distortion analyzer or randomized double-blind testing.
On the other hand, it is also true that some of the flaws introduced by running an opamp into a too-low impedance are of the sort that is not always easily heard by untrained ears. Hearing different sorts of distortion is a trained skill; one has to learn what to listen for, and in fact some kinds of distortion are often perceived as positive (as being more pleasant than the undistorted sound) by naive listeners. For instance, a little bit of low-order distortion of low frequencies can make bass sound more "rich" and "full". Clipping (what happens when the volume exceeds the voltage that the opamp can supply, thus turning sine waves into sorta-square waves) is easily audible, and ugly-sounding, when it affects more than a few percent of the wave; things like intermodulation distortion and slew-rate limiting are less easy to hear unless you have appropriate source material and know what to listen for.
The input impedance of an LM386 is too low for it to be directly driven by a bass (the OP's signal source). He would need an input stage of some sort; his existing opamp stage would probably be appropriate for this.
connecting outputs directly together is generally a bad idea.
ok try this.
connect the non inverting inputs together and use separate feedback sections for the non-inverting inputs connect both outputs to 4.7 ohm resistors and the other end of both resistors to the headphones,
drive the headphones in series. (drive the left channel via a capacitor, ground the right leave the common terminal unconnected)
Not necessarily. The LM386 has an input impedance of 50K and a voltge gain of 200. Keeping in mind that the guitar can put out as much as 1Vpp (at least my one does), there's pleanty of room for a 500K to 1meg series input resistor.
I built a headphone amp for my guitar using an LM386, and I find it has all the volume I could ever want, with both my bass and six string. But you do need to avoid those inefficient 99c store headphones.
I checked out RadioShack's catalog, since there's a store next to my work and I did pick up a LM386 on Friday. Haven't gotten a chance to play with it yet but it will be one of my next experiments. My bass has active electronics so it's probably going to be ok.
My current amp works at low volumes but starts distorting rather severely when I turn up the output of the bass OR the volume control on the headphones (these are cheapo 40ohm headphones with volume pot built in.)
Thanks for the tips - I'm definitely getting a lot of distortion when I turn up the bass or turn down the impedance of the phones (turn the volume pot in the headphones to max). The distortion is very unpleasant, it sounds like a bass played through a cheap fuzz box.
I've tried the amp with my old Sennheiser headphones which have 600ohms and I also get distortion when I turn up the bass. The volume level seems about the same as with the 40ohm phones which is a bit confusing to me. Maybe I'm already current limited with the higher impedance.
So far I haven't had the scope and the bass in the same room (one is at work, one at home) so I can't really tell what's going on but headphone impedance isn't the only thing.
By the way, I'm a total beginner on the bass so I'm sure I won't have "appropriate source material" for quite a while yet.
Speaking of caps - how does the size of the decoupling caps come into play. I know that the capacity affects the frequency response but I don't quite understand how it affects distortion. When I had very small decoupling caps in my signal path I got a clean signal on the scope but when I hooked up the phones the amplitude dropped to near zero and I hear a faint, highly distorted, high-frequency signal. When I placed a much larger cap in the path I got a much cleaner signal.
I understand that low frequencies pass through the large cap but I don't quite understand whether the small cap introduces distortion by clipping. I was, probably incorrectly, assuming that the lower frequencies are attenuated but if I think about it in terms of charge I imagine that with high amplitudes the cap being charged up quickly and then saying "hey, I'm full, I can't get the remaining 50% of your signal". That would result in the kind of fuzz I'm hearing.
Although caps do introduce distortion, it's much subtler than what you're hearing.
Don't try thinking of the caps' effect in time domain ("charging and discharging"), it'll just confuse you. Think of it in frequency domain: the cap is a resistor, whose resistance is different for low- and high-frequency signals. The resistance is Z = 1/(2 * pi * f * C). So if you combine that with the resistance of the headphones (which is reasonably constant for all frequencies in the audio range), you'll see that you get two things: first, a voltage divider which passes more or less signal to the 'phones depending on frequency, so a small capacitance means the 'phones see less signal; second, the total resistance that the opamp sees depends on frequency as well, so a small capacitance means the opamp isn't loaded as heavily.
As a quick rule of thumb: the capacitance you want, for coupling between stages (or to a load), is C = 1/(2 * pi * R * f) where R is the load resistance and f is the lowest frequency you want to pass. So for 16 ohm phones, to pass signals of 40Hz and above, C = 250uF. Notice that it has nothing to do with how much power is involved: it's the same for a milliwatt or a thousand watts. (In truth, there are some issues there; this is just a first approximation.)