Hi everyone,
This is my first real audio project, Ive done a lot of research and the design below is what ive come up with. With that in mind a big thanks to tangent for his website, ive leart so much and it has been a lot of fun.
Getting to the point it is an op amp based discrete buffered amp, the outputs are two sets of 2N2222A & 2N2907A transistors controlled by an AD8066 and the v ground is another 2 sets of 2N2222A & 2N2907A controlled by one side of a TL072. the other side of the TL072 is used to set the V ground in the AD8066 (Signals)
amp | Flickr - Photo Sharing!
EDIT: i thougt adding a scematic might help, its 4 am. i need to go to bed.
EDIT: corrected some mistakes on schematic
http://www.flickr.com/photos/wessellemmer/6943227533/in/photostream/http://www.flickr.com/photos/wessellemmer/6796513634/in/photostream
4x 2N2222 - one per output channel 2 for Vground output
4x 2n2907 - one per output channel 2 for Vground output
4x 20k 1% - 2 in feedback loops to 8066 2 to vground signal
2x 10K 1% - 2 to vground signal from input
2x 100K 1% - for 50 50 split to TL072
2x 2uf poly - input caps
1x AD8066 - to control output stage
1x TL072 - to control V grounds
2x 270uf 25v - stabalise the v ground out
8x 1ohm 1% - share load from transistors i guess
Ive built the amp on a breadboard for now and there are a couple of issues i havent been able to sort out and cannot explain.
1. There is a minor issue of crosstalk, its very soft and not distorted so it doesnt really bother me but i cant explain why. ive tried adding another TL072 to seperate the signal V ground so each side of the 8066 has its own undistorted reference but this had the same effect. why am i getting crostalk?
2. Most of my concern lies with the way ive implemented the buffering circuits. i have to have 2 caps in serie over the V ground (out) else i get a crunchy distortion(to the right of board). two small 0.1 uf caps have little bass response and sort of flubbers. where as two larger caps say 270uf or even 2200 uf removes the crunchy distortion and it sounds good. but i start to get pops and clicks at higher volumes. from what i can tell i should have no need for caps here as the circuit should be able to handle the load by itself.
3. another clue to the above problem maybe, when i place a cap over the V+ and V- the pop and click type distortion is worse and at lower volumes?
I am driving a set of ie6(18ohm), HD205's and HD212's the 212's are really soft and distorts easily. im guessing they are 60+ ohm
Any feedback, comments, suggestions, explanations etc would really be appreciated.
This is my first real audio project, Ive done a lot of research and the design below is what ive come up with. With that in mind a big thanks to tangent for his website, ive leart so much and it has been a lot of fun.
Getting to the point it is an op amp based discrete buffered amp, the outputs are two sets of 2N2222A & 2N2907A transistors controlled by an AD8066 and the v ground is another 2 sets of 2N2222A & 2N2907A controlled by one side of a TL072. the other side of the TL072 is used to set the V ground in the AD8066 (Signals)
amp | Flickr - Photo Sharing!
EDIT: i thougt adding a scematic might help, its 4 am. i need to go to bed.
EDIT: corrected some mistakes on schematic
http://www.flickr.com/photos/wessellemmer/6943227533/in/photostream/http://www.flickr.com/photos/wessellemmer/6796513634/in/photostream
4x 2N2222 - one per output channel 2 for Vground output
4x 2n2907 - one per output channel 2 for Vground output
4x 20k 1% - 2 in feedback loops to 8066 2 to vground signal
2x 10K 1% - 2 to vground signal from input
2x 100K 1% - for 50 50 split to TL072
2x 2uf poly - input caps
1x AD8066 - to control output stage
1x TL072 - to control V grounds
2x 270uf 25v - stabalise the v ground out
8x 1ohm 1% - share load from transistors i guess
Ive built the amp on a breadboard for now and there are a couple of issues i havent been able to sort out and cannot explain.
1. There is a minor issue of crosstalk, its very soft and not distorted so it doesnt really bother me but i cant explain why. ive tried adding another TL072 to seperate the signal V ground so each side of the 8066 has its own undistorted reference but this had the same effect. why am i getting crostalk?
2. Most of my concern lies with the way ive implemented the buffering circuits. i have to have 2 caps in serie over the V ground (out) else i get a crunchy distortion(to the right of board). two small 0.1 uf caps have little bass response and sort of flubbers. where as two larger caps say 270uf or even 2200 uf removes the crunchy distortion and it sounds good. but i start to get pops and clicks at higher volumes. from what i can tell i should have no need for caps here as the circuit should be able to handle the load by itself.
3. another clue to the above problem maybe, when i place a cap over the V+ and V- the pop and click type distortion is worse and at lower volumes?
I am driving a set of ie6(18ohm), HD205's and HD212's the 212's are really soft and distorts easily. im guessing they are 60+ ohm
Any feedback, comments, suggestions, explanations etc would really be appreciated.
Last edited:
Thanks jcx, yes in the schematic the PNP's where upside down, drew that at 3 in the morning so i guess i wasnt paying attention 
scem | Flickr - Photo Sharing!
ill go check out LTspice
scem | Flickr - Photo Sharing!
ill go check out LTspice
Let's hope the reversed IN+ and IN- inputs on U1a are just another result of late night schematic drawing, else you'd be in trouble...
I'm not a fan of basic class B buffer stages, they are pretty lousy in terms of high-freq distortion (real bad crossover distortion at typical headphone levels). At least add a 100..220 ohm resistor from OP out to output if nothing else.
I'm not a fan of basic class B buffer stages, they are pretty lousy in terms of high-freq distortion (real bad crossover distortion at typical headphone levels). At least add a 100..220 ohm resistor from OP out to output if nothing else.
Last edited:
After some more simulation, I can only conclude that even 1 or 2 milliamps of quiescent current for the buffer are well worth it. In a class B buffer using BC337/327s driving 32 ohms at 10 kHz, I couldn't really get a lot below 1% THD at mid-high levels (a few hundred mV), with an ugly spectrum at that. Using simple diode (well, B-E junction) bias and 10k resistors plus a bypass cap drops that to little more than 0.01%, and at halved current it still is 0.03%.
Using a class B buffer stage for a headphone amp really isn't worth it; you might as well be using an opamp "barefoot" with better results. Even AB buffers at sub-1mA quiescent current levels are likely to be of limited use, since that already is what opamps are running their output stages at.
Using a class B buffer stage for a headphone amp really isn't worth it; you might as well be using an opamp "barefoot" with better results. Even AB buffers at sub-1mA quiescent current levels are likely to be of limited use, since that already is what opamps are running their output stages at.
So ive been playing with LTspice program and barely got any sleep. Please dont laugh but ive come to the realisation that im trying to build a mini class B amp and that i have one as a v ground buffer too. Running some tests on the program showed strong spikes in vground until i added a filter cap on the feedback loop of the vgrounds opamp. I did that on my test board and tada click and pop distortions at high volume went away. also i could now add caps over +9 and -9 and no popclicks. really is helping me a great deal to see what is going on, and yet still so much to learn. ive also seen this method of buffering is kind of OK for use as ground but for signal its not going to help at all.
i have ordered 3 LME49600 buffers wich i intend to use in a carryaround amp. But hey, im enjoy playing with the discrete stuff so i was wondering how to do a class A buffer if at all?
Thanks sgrossklass, it seems you went to quite some effort to help. honestly i didnt follow you too well, but im getting there.
i have ordered 3 LME49600 buffers wich i intend to use in a carryaround amp. But hey, im enjoy playing with the discrete stuff so i was wondering how to do a class A buffer if at all?
Thanks sgrossklass, it seems you went to quite some effort to help. honestly i didnt follow you too well, but im getting there.
I've been doing a lot of simulation lately anyway, so it was little more than a refresh on buffer stages.
A class A buffer (able to transition into AB if higher currents are required) can be as simple as in the Eaton amp. (I'd add a 10-22µ bootstrap cap across the LED though, and I'd rather use 4k7s instead of 2k2s, besides you could leave the top one out for opamp class A biasing.) Aside from current draw and heatsinking requirements, they are near ideal for giving a good, well-implemented opamp the ability to effortlessly drive headphones. The currents required easily allow the use of TO-126 medium power devices (typically BD139/140 or the like) without much need to slow down the opamp, if any, and the typically large emitter degeneration resistors (3.3+, often 10..15 ohms) give good temperature stability.
I suggest you study NwAvGuy's blog, Douglas Self's various articles (and books) and Samuel Groner's opamp measurements.
Here's a little schematic to get you started. I think the BD models were from a BJT library in the LTspice group's files section that I merged into cmp\standard.bjt, as was the UniversalOpamp2 symbol. (You can also obtain BD139/140 models from Bob Cordell's website as an extra to his power amp book.)
A class A buffer (able to transition into AB if higher currents are required) can be as simple as in the Eaton amp. (I'd add a 10-22µ bootstrap cap across the LED though, and I'd rather use 4k7s instead of 2k2s, besides you could leave the top one out for opamp class A biasing.) Aside from current draw and heatsinking requirements, they are near ideal for giving a good, well-implemented opamp the ability to effortlessly drive headphones. The currents required easily allow the use of TO-126 medium power devices (typically BD139/140 or the like) without much need to slow down the opamp, if any, and the typically large emitter degeneration resistors (3.3+, often 10..15 ohms) give good temperature stability.
I suggest you study NwAvGuy's blog, Douglas Self's various articles (and books) and Samuel Groner's opamp measurements.
Here's a little schematic to get you started. I think the BD models were from a BJT library in the LTspice group's files section that I merged into cmp\standard.bjt, as was the UniversalOpamp2 symbol. (You can also obtain BD139/140 models from Bob Cordell's website as an extra to his power amp book.)
Thanks for tolerating another "my first amp" thread. I can now see i have made quite a few mistakes and those just happen to be where i have made assumptions, funny how that is.
I got my LME49600's YAY, built a little test circuit with 3 of them, one for buffering v-ground and the other as output stages... and for the first time in my life i was smiling
ill continue with my discrete endeavor, and someday soon i could have a face off
here is a list of some more assumptions that Ive made, please correct me if needed.
true or false? Its better to have a faster transistor i chose 2n2222 and 2n2907 because they are 300MHz and 200MHz and i chose ug06d diodes for the same reason as they are "ultra fast recovery"
the 2n2222 and 2n2907 can handle a max of 0.6 amps, is this not enough? i thought one could get 200ma easily. "why use TO-126 medium power devices"
true false? input caps should be as large as possible and shouldn't be electrolytic. Then resistors that bring the incoming signal up should be as small as possible, this is to minimize a high pass filter effect. i got 10uf and 200 ohms, is there some flaw here?
is the AD8066 a good op-amp? I like the sound of it but thats compared to a TL072. will something more expensive be much better.
is there any benefit in using separate single channel op amps
is there any benefit to have separate op amps to create V-Ground references for each channel, in other words having 2 op amps per channel?
Maybe i should stop here for now
I got my LME49600's YAY, built a little test circuit with 3 of them, one for buffering v-ground and the other as output stages... and for the first time in my life i was smiling
ill continue with my discrete endeavor, and someday soon i could have a face off
here is a list of some more assumptions that Ive made, please correct me if needed.
true or false? Its better to have a faster transistor i chose 2n2222 and 2n2907 because they are 300MHz and 200MHz and i chose ug06d diodes for the same reason as they are "ultra fast recovery"
the 2n2222 and 2n2907 can handle a max of 0.6 amps, is this not enough? i thought one could get 200ma easily. "why use TO-126 medium power devices"
true false? input caps should be as large as possible and shouldn't be electrolytic. Then resistors that bring the incoming signal up should be as small as possible, this is to minimize a high pass filter effect. i got 10uf and 200 ohms, is there some flaw here?
is the AD8066 a good op-amp? I like the sound of it but thats compared to a TL072. will something more expensive be much better.
is there any benefit in using separate single channel op amps
is there any benefit to have separate op amps to create V-Ground references for each channel, in other words having 2 op amps per channel?
Maybe i should stop here for now
Last edited:
OK, Im guessing this is actually Bandwith. The question then is pretty much the same. is it better to have more?
The problem with excess bandwidth is that it increases the likelihood of HF instability (oscillation) and you can't always detect it audibly, although it can be subtly degrading the sound quality. Sometimes it's obvious, the active devices self-destruct due to overheating. A headphone amp realistically only needs a slew rate of ~1V/uS.
Without entering into a mathematical treatment, input caps should be dimensioned according to the input (shunt) resistance. The low frequency cutoff occurs at the frequency where the capacitive reactance equals the resistance.
Read the nwavguys blog with reference to virtual grounds. It's a topology that's better avoided.
I've left a few terms hanging here, look up the things you don't understand.
Actually, having higher fT transistors in the buffer is a good thing. You want to be something like a factor of 3 faster than the opamp (GBW / gain) to avoid instability or having to reduce loop gain. (Alternatively, "a lot slower" is another - if not too well-performing - option.)
If, for example, you want to operate an opamp with a 20 MHz GBW at 3x gain, you'll want at least 20 MHz of buffer fT. A BD139/140 pair can do it but they need to be run a little warm.
Note that practical fT depends on current plus transistor tf and Cbc/Cbe. (Formula given here, for example. fT vs. Ic also may be given in the datasheet.) This is why transistors with higher power handling and thus bigger dies tend to be slower. Cheapo oldschool power transistors like TIP31, TIP41 or BD24x tend to reach about 3 MHz only.
Excessive fT may be a problem since emitter followers like to oscillate when fed by an inductive source impedance, and an opamp output is about as inductive as they come (its output impedance steadily rises as its open-loop gain falls, up to the point where it runs out of OLG altogether). Using a feedback cap on the opamp or artifically increasing Ccb (making the transistor slower) can help in such cases.
The biggest problem with any TO-92 case transistor is cooling. That kind of case has a thermal resistance of like 200 K/W in free air and not a lot under 100 on a heatsink. You'd probably be well advised not to run one at more than about 10 mA in a typical headphone buffer circuit, and I wouldn't expect massive amounts of output power.
2N2222s or BC337s aren't a bad compromise for TO-92. Still quite fast and reasonably robust. BC639s already need a good bit more current but are more rugged in return, though in that case they can't really live up to their full potential. Philips used to package the same die in TO-126 as BD139, those weren't bad parts (fairly little beta droop, too). The BD139s you can buy today can handle more current, but their capacitances are about twice as high, and beta droop starts to set in above about 200 mA already. Still quite useful for a headphone buffer though.
If, for example, you want to operate an opamp with a 20 MHz GBW at 3x gain, you'll want at least 20 MHz of buffer fT. A BD139/140 pair can do it but they need to be run a little warm.
Note that practical fT depends on current plus transistor tf and Cbc/Cbe. (Formula given here, for example. fT vs. Ic also may be given in the datasheet.) This is why transistors with higher power handling and thus bigger dies tend to be slower. Cheapo oldschool power transistors like TIP31, TIP41 or BD24x tend to reach about 3 MHz only.
Excessive fT may be a problem since emitter followers like to oscillate when fed by an inductive source impedance, and an opamp output is about as inductive as they come (its output impedance steadily rises as its open-loop gain falls, up to the point where it runs out of OLG altogether). Using a feedback cap on the opamp or artifically increasing Ccb (making the transistor slower) can help in such cases.
The biggest problem with any TO-92 case transistor is cooling. That kind of case has a thermal resistance of like 200 K/W in free air and not a lot under 100 on a heatsink. You'd probably be well advised not to run one at more than about 10 mA in a typical headphone buffer circuit, and I wouldn't expect massive amounts of output power.
2N2222s or BC337s aren't a bad compromise for TO-92. Still quite fast and reasonably robust. BC639s already need a good bit more current but are more rugged in return, though in that case they can't really live up to their full potential. Philips used to package the same die in TO-126 as BD139, those weren't bad parts (fairly little beta droop, too). The BD139s you can buy today can handle more current, but their capacitances are about twice as high, and beta droop starts to set in above about 200 mA already. Still quite useful for a headphone buffer though.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.