Well the current values are good enough, if we can assume that rail voltage will average out at +-15V.
For a 4AA, 1mW (95db) is .5V, and I hope you don't listen that loud... Distortion at this voltage will be very low. THD is almost purely dependent on Ie(Q7 and Ie(Q6). The rest of the circuit produces relatively no distortion. So in fact, if you wanted to hear the "sound" of the circuit, you would probably decrease the rails to decrease the impedance seen by Q6/7. There are other ways to do this without sacrificing output voltage swing.
How about you pick a decibel value for the second harmonic, H2, at your own listening volume? Working from that, I can give you something most certain to show the circuit's own colors.
- keantoken
For a 4AA, 1mW (95db) is .5V, and I hope you don't listen that loud... Distortion at this voltage will be very low. THD is almost purely dependent on Ie(Q7 and Ie(Q6). The rest of the circuit produces relatively no distortion. So in fact, if you wanted to hear the "sound" of the circuit, you would probably decrease the rails to decrease the impedance seen by Q6/7. There are other ways to do this without sacrificing output voltage swing.
How about you pick a decibel value for the second harmonic, H2, at your own listening volume? Working from that, I can give you something most certain to show the circuit's own colors.
- keantoken
Last edited:
It will probably work well, I never thought about this option (silly me, I went through all that trouble with BJT's! Sim time.).
It looks to me like M1 will take more current than M2, because of the current through R7. To help balance this, I suggest to the same value resistor as R7 across the G-D of M1, and use R1 to adjust bias. Depending on J1's impedance, the imbalance between M1 and M2 would probably be greater than in a BJT stage because of the high impedance MOSFETs.
I would suggest you increase R4 to at least 1k for about 4mA. I would not use over 10mA here.
The MOSFET output will have greater DC impedance but less HF impedance. I don't think this will negatively affect stability.
My efforts on offset so far were meant to keep the current through Q6 and Q7 even. Your method will work fine I believe; I don't know what good it is getting precise current matching when real devices vary a lot. But if you're using, say, matched pairs, odd harmonics could be practically noexistant.
- keantoken
It looks to me like M1 will take more current than M2, because of the current through R7. To help balance this, I suggest to the same value resistor as R7 across the G-D of M1, and use R1 to adjust bias. Depending on J1's impedance, the imbalance between M1 and M2 would probably be greater than in a BJT stage because of the high impedance MOSFETs.
I would suggest you increase R4 to at least 1k for about 4mA. I would not use over 10mA here.
The MOSFET output will have greater DC impedance but less HF impedance. I don't think this will negatively affect stability.
My efforts on offset so far were meant to keep the current through Q6 and Q7 even. Your method will work fine I believe; I don't know what good it is getting precise current matching when real devices vary a lot. But if you're using, say, matched pairs, odd harmonics could be practically noexistant.
- keantoken
Last edited:
Hi keantoken
Nice circuit!
Here's an alternative output stage that may (or may not) come in handy.
It's similar to what you showed a few days ago, but with an LED instead of a PNP bjt.
Not really relevant here, but it's a nice circuit for high-voltage amps where suitable PNP devices are hard to find - something to bear in mind in case there's an electrostatic headphone in your future.
Anyway, as shown it should idle at about 100mA (+10 for the bias)
Regards - Godfrey
Nice circuit!
Here's an alternative output stage that may (or may not) come in handy.
It's similar to what you showed a few days ago, but with an LED instead of a PNP bjt.
Not really relevant here, but it's a nice circuit for high-voltage amps where suitable PNP devices are hard to find - something to bear in mind in case there's an electrostatic headphone in your future.
Anyway, as shown it should idle at about 100mA (+10 for the bias)
Regards - Godfrey
Attachments
Well I suppose I'm a little dull in the MOSFET realm, I'm not sure why I'd go beyond what you've already done (But that's just because I don't know much).
A poweramp is certainly possible for the NTP. What I have done here could easily be modified for a first-watt type amp for very sensitive speakers (MOSFETs would be real convenient here). I've tried several poweramps in the simulator; the NTP is certainly not lacking, although it's stranger compared to the LTP. I'd like to show my designs before someone beats me to it, but it would just be simulations, and no real prototypes soon.
Your schematic should work just as well (and probably better) than my BJT version.
I've attached the schematic with the suggestions I described for the MOSFET version. Here I've added Q1. This increases PSRR, as well as acting as offset control. My main reason for doing this however is that by decreasing the impedance seen by Q6 and Q7's emitters, 2nd harmonics are increased. This helps put them in the audible range. At 1V output into any load, 2nd harmonics will probably be audible, but headphones are sensitive enough that mine only need around 50mV for normal listening. So to actually hear the 2nd harmonics I'm sure they need to be increased. (no big deal though, even without them it probably sounds great.)
Godfrey:
Thanks. Convenient power light, and I think it could also act as an indicator if the output stage burns out somehow.
- keantoken
Attachments
Last edited:
Okay, here's the deal. I haven't done more experiments yet, since the outputs blew out after I made a stupid mistake.
I have a huge trafo sitting around, chopped out of a surge protector with battery backup function. I'm pretty sure I can use it to make a mean PSU for a class A headamp, if I can figure out how it works. So far I think it's an 8-1 ratio, which gives about 15V output, but other tests suggest a 5-1 ratio, not sure. In any case I think all it's good for is a class A headphone amp, due to the low output voltage and bulkiness.
Of course this will necessitate turning it into a single-rail design, which I'm not totally comfortable with. If I'm that worried, I can make a virtual ground with an opamp. Odd as it may sound, if I go through with this it will be the first project I've ever completed.
As for the possibility of MOSFET outputs... They seem to work better than BJT, but with low rails output swing may be limited in the listening range. So for my next build I think I'll have to go with BJT. Also, my latest schematic may have issues with thermal runaway, I'm just not sure (I don't know how to tell). A safer and no more compromising approach would be something along the lines of the my schematic in post 74 or Godfrey's in post 89, with outputs replaced as MOSFETs. This arrangement will have a negative bias tempco and should be totally immune to thermal runaway.
- keantoken
I have a huge trafo sitting around, chopped out of a surge protector with battery backup function. I'm pretty sure I can use it to make a mean PSU for a class A headamp, if I can figure out how it works. So far I think it's an 8-1 ratio, which gives about 15V output, but other tests suggest a 5-1 ratio, not sure. In any case I think all it's good for is a class A headphone amp, due to the low output voltage and bulkiness.
Of course this will necessitate turning it into a single-rail design, which I'm not totally comfortable with. If I'm that worried, I can make a virtual ground with an opamp. Odd as it may sound, if I go through with this it will be the first project I've ever completed.
As for the possibility of MOSFET outputs... They seem to work better than BJT, but with low rails output swing may be limited in the listening range. So for my next build I think I'll have to go with BJT. Also, my latest schematic may have issues with thermal runaway, I'm just not sure (I don't know how to tell). A safer and no more compromising approach would be something along the lines of the my schematic in post 74 or Godfrey's in post 89, with outputs replaced as MOSFETs. This arrangement will have a negative bias tempco and should be totally immune to thermal runaway.
- keantoken
I certainly will continue this project, and I've developed a way to increase 2nd harmonic to what should be fairly audible levels. I can't say how often there will be updates.
I've been looking at that transformer, and I'm not sure it's a good idea to plug it in. In most mains transformers the impedance flattens out before it reaches 60Hz, but I'm afraid it may have too little inductance and will just waste large amounts of current. It came from a surge backup box, so for all I know it may have been running at 10KHz instead of 60Hz. Good news is, I took apart a karaoke machine and the trafo from that has 16, 12 and 20V outputs, 1A, 1.5A, and 100mA respectively. It's not such a hulk, but It's foolproof. Also, due to the 20V rails, I could run the frontend at higher voltage to compensate for MOSFET swing. This gives me more slack on what output devices I can use with a clear conscience.
Currently I don't get to the bench much because 1/3 of my life is consumed with school, another 1/3 consumed in sleep. 1/3 of my life I can call my own, and that could easily be filled with homework. Other things are taking precedence now in the time I have to myself.
Now imagine when I finally get a job.
- keantoken
I've been looking at that transformer, and I'm not sure it's a good idea to plug it in. In most mains transformers the impedance flattens out before it reaches 60Hz, but I'm afraid it may have too little inductance and will just waste large amounts of current. It came from a surge backup box, so for all I know it may have been running at 10KHz instead of 60Hz. Good news is, I took apart a karaoke machine and the trafo from that has 16, 12 and 20V outputs, 1A, 1.5A, and 100mA respectively. It's not such a hulk, but It's foolproof. Also, due to the 20V rails, I could run the frontend at higher voltage to compensate for MOSFET swing. This gives me more slack on what output devices I can use with a clear conscience.
Currently I don't get to the bench much because 1/3 of my life is consumed with school, another 1/3 consumed in sleep. 1/3 of my life I can call my own, and that could easily be filled with homework. Other things are taking precedence now in the time I have to myself.
Now imagine when I finally get a job.
- keantoken
Alright, the new transformer is mounted. Now what's left is to figure out how I'm going to get the circuitry on protoboard so I can fit the heatsinks in. I could mount the outputs on the bottom, but I'm not sure if that will be good enough cooling.
Hey, is it okay to use the heatsink as the star ground? Also, I don't see any problems with using the P55NE06 as the output MOSFETs, but really I know zero about MOSFETs. It's the only part number I have 4 of, and they came from the mentioned battery backup box.
- keantoken
Hey, is it okay to use the heatsink as the star ground? Also, I don't see any problems with using the P55NE06 as the output MOSFETs, but really I know zero about MOSFETs. It's the only part number I have 4 of, and they came from the mentioned battery backup box.
- keantoken
Alright, this is the version I plan to use. R9 and R10 are what determine the amount of H2.
My plan so far...
The transformer doesn't have any center taps for ground, but for headphones I don't think this is terribly important. Bypass each cap with a resistor to ground and we will have a good virtual ground in the audio range. I've decided I'm simply going to run the whole thing off of the 15V rails. Filter caps will be 10mF each, properly decoupled.
With MOSFETs, this could drive sensitive loudspeakers pretty well.
- keantoken
My plan so far...
The transformer doesn't have any center taps for ground, but for headphones I don't think this is terribly important. Bypass each cap with a resistor to ground and we will have a good virtual ground in the audio range. I've decided I'm simply going to run the whole thing off of the 15V rails. Filter caps will be 10mF each, properly decoupled.
With MOSFETs, this could drive sensitive loudspeakers pretty well.
- keantoken
Attachments
I've been playing around in the simulator for a little while and it seems that the second harmonics are produced most strongly when Q7 and Q6 are Vbe matched at the given bias current. Gm for BJT's is fairly constant, so it's just Vbe we have to worry about (we can even use non-complimentary part numbers as long as the Vbe's are close). After Vbe's are matched better H2 generation can then be attained by tweaking either R9 or R10. Generation is pretty good with a match within 10mV.
Matching can probably be done with a cheap multimeter; I don't see why a complex setup would be needed. Matching within 10mV is good enough for me judging by the simulation results.
A spectrum analyzer is not needed. It is possible to match and adjust transistors with an oscilloscope measuring across R3 (in the attached schematic). Even without that, an opamp could be connected across R3 and connected to headphones, where you could adjust until you couldn't hear the fundamental.
Oops, maybe I've spilled the beans on how to add that "H2 sound" to solid state circuitry.
This is the principle behind HeadRush. Most of the H2 generation takes place in the prebuffers Q6 and Q7 in the amp. The Rush Cascode makes this setup convenient (I thought the use of the Rush Cascode/NTP was the feature of this amp, but my offset canceling scheme turned out to be the main feature).
- keantoken
Matching can probably be done with a cheap multimeter; I don't see why a complex setup would be needed. Matching within 10mV is good enough for me judging by the simulation results.
A spectrum analyzer is not needed. It is possible to match and adjust transistors with an oscilloscope measuring across R3 (in the attached schematic). Even without that, an opamp could be connected across R3 and connected to headphones, where you could adjust until you couldn't hear the fundamental.
Oops, maybe I've spilled the beans on how to add that "H2 sound" to solid state circuitry.
This is the principle behind HeadRush. Most of the H2 generation takes place in the prebuffers Q6 and Q7 in the amp. The Rush Cascode makes this setup convenient (I thought the use of the Rush Cascode/NTP was the feature of this amp, but my offset canceling scheme turned out to be the main feature).
- keantoken
Attachments
Well, I've been trying to put together the circuit in post 98, but everything has utterly gone wrong.
Touching my finger to any spot instantly shifts the trace up several mV, even when the circuit is connected to ground. That makes no sense whatsoever. My body should not have a DC charge that won't dissipate after a few seconds into such low loads. Touching something that's grounded does not help at all. It's not AC causing the scope to trigger, since the same thing happens when I turn triggering off. I'm hoping my test bench doesn't violate the laws of physics, since my hobby revolves around physics.
I unplugged the circuit, tested the scope and it's working properly.
I'll try moving to my other breadboard.
- keantoken
Touching my finger to any spot instantly shifts the trace up several mV, even when the circuit is connected to ground. That makes no sense whatsoever. My body should not have a DC charge that won't dissipate after a few seconds into such low loads. Touching something that's grounded does not help at all. It's not AC causing the scope to trigger, since the same thing happens when I turn triggering off. I'm hoping my test bench doesn't violate the laws of physics, since my hobby revolves around physics.
I unplugged the circuit, tested the scope and it's working properly.
I'll try moving to my other breadboard.
- keantoken
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.