...This amp.
I built this as a design project to get into SS a little. Seems to work well enough; since it's on a breadboard and powered by the bench supply, I can't bring it upstairs to any better speakers to really give it a listen.
Output biases at around 150mA after warming up (the diodes are *not* thermally coupled to the outputs). Maximum power output into 8 ohms seems to be 5W.
Tim
P.S. Huh... I got the input cap reversed on the schematic!
I built this as a design project to get into SS a little. Seems to work well enough; since it's on a breadboard and powered by the bench supply, I can't bring it upstairs to any better speakers to really give it a listen.
Output biases at around 150mA after warming up (the diodes are *not* thermally coupled to the outputs). Maximum power output into 8 ohms seems to be 5W.
Tim
P.S. Huh... I got the input cap reversed on the schematic!
Attachments
> really give it a listen
It works. You can't do a lot better for three transistors and no transformers in an 8Ω load.
Gain is very low, input impedance is low-ish, driver stage eats a lot of power, and power output is 1/2 to 1/3rd what that power supply should be able to do with a better amplifier.
The output DC voltage is not clearly defined. You have it set about halfway between rails but that will vary with transistor parameters. It is very possible the output cap could be reverse-biased. Simple single-ended designs like this usually use single-ended power supplies: +34V/Gnd instead of +17V/Gnd/-17V, so you know for-sure which way the output cap is biased.
Re-rig it with +34V/Gnd supply, reduce 56Ω to zero, use a Darlington for the driver transistor, and re-scale your bias-feedback resistors. You can set the output voltage to a multiple of Vbe, which isn't perfect but does work OK. Keep the 10K and change the 33K to 120K, the output will sit at about 1.2V*(10K+120K)/10K= 15.6V, a workable value. The open-loop gain of the driver will be over 100. The input Z will be about 120K/100, pretty low. Audio gain is not well defined. So add a 5K input resistor in series with the 47uFd to set gain to about 120K/5K=24, 12 or 15dB feedback.
The output power is low because of the 2.2 and 330Ω resistors. The 330 refects through Hfe=50 transistors as about a 6 ohm resistor. So the maximum output voltage is about 8/(6+2.2+8) about half of the available voltage swing from idle voltage to positive supply. For more efficiency you need to reduce the 2.2 resistors (which may force a different bias affair) and reduce the 330Ω (but it is already eating too much power for comfort). The classic answer for 8Ω loads is to use Darlington outputs so Hfe rises over 1,000 and the driver does not have to be so powerful.
It works. You can't do a lot better for three transistors and no transformers in an 8Ω load.
Gain is very low, input impedance is low-ish, driver stage eats a lot of power, and power output is 1/2 to 1/3rd what that power supply should be able to do with a better amplifier.
The output DC voltage is not clearly defined. You have it set about halfway between rails but that will vary with transistor parameters. It is very possible the output cap could be reverse-biased. Simple single-ended designs like this usually use single-ended power supplies: +34V/Gnd instead of +17V/Gnd/-17V, so you know for-sure which way the output cap is biased.
Re-rig it with +34V/Gnd supply, reduce 56Ω to zero, use a Darlington for the driver transistor, and re-scale your bias-feedback resistors. You can set the output voltage to a multiple of Vbe, which isn't perfect but does work OK. Keep the 10K and change the 33K to 120K, the output will sit at about 1.2V*(10K+120K)/10K= 15.6V, a workable value. The open-loop gain of the driver will be over 100. The input Z will be about 120K/100, pretty low. Audio gain is not well defined. So add a 5K input resistor in series with the 47uFd to set gain to about 120K/5K=24, 12 or 15dB feedback.
The output power is low because of the 2.2 and 330Ω resistors. The 330 refects through Hfe=50 transistors as about a 6 ohm resistor. So the maximum output voltage is about 8/(6+2.2+8) about half of the available voltage swing from idle voltage to positive supply. For more efficiency you need to reduce the 2.2 resistors (which may force a different bias affair) and reduce the 330Ω (but it is already eating too much power for comfort). The classic answer for 8Ω loads is to use Darlington outputs so Hfe rises over 1,000 and the driver does not have to be so powerful.
Yup, about what I was thinking, thanks for the reply.
- Oh -- the BD677 *is* a darlington. Nonetheless I added a small preamp transistor on the breadboard.. computer has low output.
I have a new drawing, I'll breadboard it today, oughta get a good bit closer to the rails, and work good with a 4 ohm load as well.
Oh- some things I forgot (hey, it was late), the collector load on the 1st transistor (which is unmarked, that's a 2N4401) is actually 2.2k, no DC coupling but I don't feel like it yet; and I was going to draw 150 ohm B-E resistors on the TIP3xC output.
Improvements I see: more power thanks to the darlington output and CCS loaded no-Re driver stage. More gain with the preamp stage, and global NFB as well.
Tim
- Oh -- the BD677 *is* a darlington. Nonetheless I added a small preamp transistor on the breadboard.. computer has low output.
I have a new drawing, I'll breadboard it today, oughta get a good bit closer to the rails, and work good with a 4 ohm load as well.
An externally hosted image should be here but it was not working when we last tested it.
Oh- some things I forgot (hey, it was late), the collector load on the 1st transistor (which is unmarked, that's a 2N4401) is actually 2.2k, no DC coupling but I don't feel like it yet; and I was going to draw 150 ohm B-E resistors on the TIP3xC output.
Improvements I see: more power thanks to the darlington output and CCS loaded no-Re driver stage. More gain with the preamp stage, and global NFB as well.
Tim
Update
I got it working, mostly... here's the current version sitting on the breadboard.
Input has to be referenced to the -17V rail, otherwise ripple completely screws the signal.
Clipping is asymmetrical, occuring first at the positive peak, then a volt or two later on the negative as well. DC output voltage reads -1V or so in this condition, so it's actually a little on the low side. (The reason this happens is - I presume - because of PSU sag, and the driver acting as a Vbe multiplier.) Why does this happen? Shouldn't the CCS be able to go a volt or two higher than is indicated by this?
In any case, I get a good 22Vp-p at clipping into 2 ohms, which is all of 30W...
Tim
I got it working, mostly... here's the current version sitting on the breadboard.
Input has to be referenced to the -17V rail, otherwise ripple completely screws the signal.
Clipping is asymmetrical, occuring first at the positive peak, then a volt or two later on the negative as well. DC output voltage reads -1V or so in this condition, so it's actually a little on the low side. (The reason this happens is - I presume - because of PSU sag, and the driver acting as a Vbe multiplier.) Why does this happen? Shouldn't the CCS be able to go a volt or two higher than is indicated by this?
In any case, I get a good 22Vp-p at clipping into 2 ohms, which is all of 30W...
Tim
Attachments
> wondered about the purpose of the resistors on the darlington...
If you omitted them on the old Germanium transistors, leakage current would cause a melt-down.
With Silicon that is unlikely.
But consider what happens when you go to turn the transisor OFF. (That must happen every cycle in AB mode.)
The first transistor goes off. Its Emitter pin becomes an open-circuit. The output transisor, which has recently been sucking large current, has a large stored charge in its Base. That has to be drained off before the transistor will turn-off. It will leak away as Base current, but as Emitter current drops the Base current drops, so in theory it will take forever to turn off. In practice, it won't turn off entirely in a 10KHz wave, the other side of the push-pull has to feed the load and the slow-turn-off of the "off" transistor, and things get hot.
You can breadboard, even use, a Silicon Darlington without the resistor. But it will work a bit better with them, especially at high audio frequencies.
Don't DC-couple into a speaker unless your bias arrangement absolutely holds the output DC within a few mV of 0V. DC level trim pots will NOT hold the output DC constant enough. If you are not yet ready to play with differential inputs with naturally low DC offsets, get a 2,200µFd cap and put it on the output to protect the speaker. Very-fine amps can be built with output caps. There is some performance justification for no-cap outputs, but the killer reason they vanished in commercial design is that when you must build Stereo, two power supply caps is cheaper than three caps (one power, two output). For mono, or for breadboarding, and especially for LEARNING, don't be ashamed to buck fashion and use an output cap.
If you omitted them on the old Germanium transistors, leakage current would cause a melt-down.
With Silicon that is unlikely.
But consider what happens when you go to turn the transisor OFF. (That must happen every cycle in AB mode.)
The first transistor goes off. Its Emitter pin becomes an open-circuit. The output transisor, which has recently been sucking large current, has a large stored charge in its Base. That has to be drained off before the transistor will turn-off. It will leak away as Base current, but as Emitter current drops the Base current drops, so in theory it will take forever to turn off. In practice, it won't turn off entirely in a 10KHz wave, the other side of the push-pull has to feed the load and the slow-turn-off of the "off" transistor, and things get hot.
You can breadboard, even use, a Silicon Darlington without the resistor. But it will work a bit better with them, especially at high audio frequencies.
Don't DC-couple into a speaker unless your bias arrangement absolutely holds the output DC within a few mV of 0V. DC level trim pots will NOT hold the output DC constant enough. If you are not yet ready to play with differential inputs with naturally low DC offsets, get a 2,200µFd cap and put it on the output to protect the speaker. Very-fine amps can be built with output caps. There is some performance justification for no-cap outputs, but the killer reason they vanished in commercial design is that when you must build Stereo, two power supply caps is cheaper than three caps (one power, two output). For mono, or for breadboarding, and especially for LEARNING, don't be ashamed to buck fashion and use an output cap.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.