A recent discussion over in the Blowtorch thread was debating the problems of opamps driving feedback networks (and external loads), loads that might be considered to low in value. Should this situation occur, an increase in distortion is inevitable... but is it audible. Where are the limits ?
It seemed a fun topic to try a little practical test. Shown below is a simple line level buffer using an LM4562, probably one of the more highly regarded modern devices, particularly considering its relatively low cost. R6 is the vital component. For the test it was either in place, or not. The LM4562 had no problem driving 100 ohm visually on a scope at typical line levels of -/+ 3 volts. The supply was a super clean 12 volt VRLA battery.
And now its over to you 😀 Can you decide which of these files has the 100 ohm load added. The tracks are short (around 46 seconds).
Project X
Project Y
And thats it 🙂 It might be more fun if you pm me your choice and then we can see how everyone voted.
It seemed a fun topic to try a little practical test. Shown below is a simple line level buffer using an LM4562, probably one of the more highly regarded modern devices, particularly considering its relatively low cost. R6 is the vital component. For the test it was either in place, or not. The LM4562 had no problem driving 100 ohm visually on a scope at typical line levels of -/+ 3 volts. The supply was a super clean 12 volt VRLA battery.
And now its over to you 😀 Can you decide which of these files has the 100 ohm load added. The tracks are short (around 46 seconds).
Project X
Project Y
And thats it 🙂 It might be more fun if you pm me your choice and then we can see how everyone voted.
This thing might easily be better than many ADC. What did you use ?
Out of curiosity: what's the point of r3-r4 ? Why not a simple link from output to -in ?
Out of curiosity: what's the point of r3-r4 ? Why not a simple link from output to -in ?
The ADC was just the inbuilt soundcard on my Dell Vostro PC.
R3 and R4 shave a little off the open loop gain (at the expense of noise). Even though classed as unity gain stable, its a technique which is reputed to bring many opamps away from the edge of instability. I'm not entirely convinced of its efficacy though (as a general thing).
This thread shows the technique on an inverting stage although the values I used at the time were far from optimal.
http://www.diyaudio.com/forums/analog-line-level/196461-different-opamp-compensation-technique.html
R3 and R4 shave a little off the open loop gain (at the expense of noise). Even though classed as unity gain stable, its a technique which is reputed to bring many opamps away from the edge of instability. I'm not entirely convinced of its efficacy though (as a general thing).
This thread shows the technique on an inverting stage although the values I used at the time were far from optimal.
http://www.diyaudio.com/forums/analog-line-level/196461-different-opamp-compensation-technique.html
I conducted similar tests years ago with a BTech BT928 (Philips NE5532 based, 47 ohms out, modified for gain of -4.7), and distortion at decidedly elevated levels proved measurable but benign. Like -70 dB 2nd harmonic with nominal 32 ohms.
Driving +/-3 V peak into 100 ohms is not a truly major challenge for a semi-powerful opamp like an LM4562 or NE5532. OPA213x may be more interesting, its load immunity at higher freqs is not that great. Or more wimpy parts like TL072, some 4558, MC33078 or LM833 (at somewhat reduced levels).
BTW, using a zener in the reference divider is no doubt luxurious, but I bet a plain resistor divider would work just as well, given how lightly Vref is loaded...
Driving +/-3 V peak into 100 ohms is not a truly major challenge for a semi-powerful opamp like an LM4562 or NE5532. OPA213x may be more interesting, its load immunity at higher freqs is not that great. Or more wimpy parts like TL072, some 4558, MC33078 or LM833 (at somewhat reduced levels).
BTW, using a zener in the reference divider is no doubt luxurious, but I bet a plain resistor divider would work just as well, given how lightly Vref is loaded...
Yes, I scoped the output and it had no difficulty with 2 volts RMS (-/+ 3 volts or so). What I did wonder was if some form of compression could be occurring on transients.
The listening tests and comments are proving interesting though 🙂
The listening tests and comments are proving interesting though 🙂
I was in need of a cheap and cheerful headphone driver for a disco.
So I used an op-amp with a series 100R resistor into 8 ohm headphones.
It worked very well, just loud enough to hear.
So I used an op-amp with a series 100R resistor into 8 ohm headphones.
It worked very well, just loud enough to hear.
For those willing to dig a little deeper, here is the test loaded in LTspice (save ProjectsX & Y in the same directory as the .asc).
I have placed the cursors over two typical "events": one at 16s is pretty obvious, and easy to correlate to the signal contents.
The other at 13s is more puzzling: the result is a LF/DC shift having no obvious cause (at first sight); certainly worth investigating
I have placed the cursors over two typical "events": one at 16s is pretty obvious, and easy to correlate to the signal contents.
The other at 13s is more puzzling: the result is a LF/DC shift having no obvious cause (at first sight); certainly worth investigating
Attachments
Looks like it doesn't like the space in the filename, try surrounding it with double or single quotes or removing the space from the name of the file itself.
Yes, exactly: save under name ProjectX, not Project X.
Note that it may work if you change the filename in the source as Project X, but I didn't even try, because the syntax in LTspice is generally very basic and archaic (capitals ignored, etc) and I tend to opt for the "first time right" option, even when there is a possibility (or not) for something more evolved or civilized.
Thanks guys. I should have played around more along those lines. Copying and renaming the files worked. Changing the "wavefile=ProjectX.wav" on the sim to match my file names didn't.
I'll have a look now 🙂
I'll have a look now 🙂
I'm curious: why is everybody trying to adjust the op amp to their requirements by adjusting their results so they suit the op amp, rather than changing the op amp section to adjust it to the user's requirements?
Specifically, if you are researching a specific op amp, that's just fine, but if you have a specfic need, why not simply add a pair of compementary BJTs, such as a say MPSA 06/56 pair, and raise the current capability that way?
If that's not enough, you can use the MPSA transistors as drivers, driving a pair of say BD 139/140 BJTs. There's almost no end to varying this theme.
I have yet to hear an op amp, any op amp by anybody, which did not sound better with a current booster than otherwise. All sorts of distortion either drop or disappear from the menu. As an added benefit, you also get a much improved output voltage capability. While this is hardly needed for headphones, which in most cases blow one's mind by about a 2V output, it does provide for better rendition of transients.
Lastly, I honestly don't see why people fall head over heels to get a good damping factor for their loudspeakers, but treat the headphones as if the were not in effect a loudspeaker load. Add a 100 Ohm resistor and it vecomes the amp's output impedance, so with modern headphones, you get a NEGATIVE damping factor - someone please tell how this does not affect your transfer curve.
Specifically, if you are researching a specific op amp, that's just fine, but if you have a specfic need, why not simply add a pair of compementary BJTs, such as a say MPSA 06/56 pair, and raise the current capability that way?
If that's not enough, you can use the MPSA transistors as drivers, driving a pair of say BD 139/140 BJTs. There's almost no end to varying this theme.
I have yet to hear an op amp, any op amp by anybody, which did not sound better with a current booster than otherwise. All sorts of distortion either drop or disappear from the menu. As an added benefit, you also get a much improved output voltage capability. While this is hardly needed for headphones, which in most cases blow one's mind by about a 2V output, it does provide for better rendition of transients.
Lastly, I honestly don't see why people fall head over heels to get a good damping factor for their loudspeakers, but treat the headphones as if the were not in effect a loudspeaker load. Add a 100 Ohm resistor and it vecomes the amp's output impedance, so with modern headphones, you get a NEGATIVE damping factor - someone please tell how this does not affect your transfer curve.
Last edited:
Mooly, did you have a mobile, wifi, DECT or bluetooth device in the vicinity of your test setup?
Fair enough, Mooly, but I would suggest you try adding a pair of BJTs as a current booster, and then try all the versions. That would give you insight across the board and you could check out the two most obvious aspects.
However, do remember that the results may vary widely depending on the impedance of the headphones. It's hardly the same thing whether they have and impedance of 32 or 600 Ohms. Add to this the already wildly varying headphone efficiency, and you have a chaotic situation. The resulty you obtain will always be locked to the headphones you tried it with and similar, but hardly across the board.
And you have not mentioned the damping factor. Headphones are, afterr all, transducers.
Look, I'm really not trying to complicate your life for the fun of it, I am passing on my own experiences from a few years back, when I was designing my own headphone amps. One was an all BJT affair, the second was a FET input MOSFET output device, and the third was an all tube device. The tube version has an arguably best midrange, but is not strong on bass and is much influenced by the headphone impedance. They sound wonderful with Beyer Dynamic 600 Ohm cans.
But the overall winner is the BJT version. On balance, it doesn't do the mids as well as the tube version, it doesn't do the highs as well as the FET/MOSFET version, but the differences are very small. However, the bass line difference was not so small, its bass lines leave both other versions in the dust. It's a fully complementary BJT amp, which uses 50W output devices (MJE15030/15031) mounted on local heat sinks, and runs in pure class A as it's biased at 85 mA of standing current.
There was a fourth version, using a humble OP 37 op amp, feeding two MPSA driver transistors, driving humble BD 139/140 transistors. Sounded really good. If OP37 is replaced by AD829, it becomes ambitious and aims for the top. Again, bias was set at 85 mA for the output stage.
However, do remember that the results may vary widely depending on the impedance of the headphones. It's hardly the same thing whether they have and impedance of 32 or 600 Ohms. Add to this the already wildly varying headphone efficiency, and you have a chaotic situation. The resulty you obtain will always be locked to the headphones you tried it with and similar, but hardly across the board.
And you have not mentioned the damping factor. Headphones are, afterr all, transducers.
Look, I'm really not trying to complicate your life for the fun of it, I am passing on my own experiences from a few years back, when I was designing my own headphone amps. One was an all BJT affair, the second was a FET input MOSFET output device, and the third was an all tube device. The tube version has an arguably best midrange, but is not strong on bass and is much influenced by the headphone impedance. They sound wonderful with Beyer Dynamic 600 Ohm cans.
But the overall winner is the BJT version. On balance, it doesn't do the mids as well as the tube version, it doesn't do the highs as well as the FET/MOSFET version, but the differences are very small. However, the bass line difference was not so small, its bass lines leave both other versions in the dust. It's a fully complementary BJT amp, which uses 50W output devices (MJE15030/15031) mounted on local heat sinks, and runs in pure class A as it's biased at 85 mA of standing current.
There was a fourth version, using a humble OP 37 op amp, feeding two MPSA driver transistors, driving humble BD 139/140 transistors. Sounded really good. If OP37 is replaced by AD829, it becomes ambitious and aims for the top. Again, bias was set at 85 mA for the output stage.
Last edited:
- Status
- Not open for further replies.