See GBP remarks in post 17 of this thread and have a look at the several LME49811+STD03 designs posted in the chip amps forum. Among other things the 49811 is production rather than obsolete. (TomChr's 49811 build in particular is a good starting point, though a TL071 or TL031 is more attractive as a DC servo than the LF411.) See also the stability criteria for composite amplifiers in the appendix C of Linear AN-47 and note the lack of local feedback around the LM143. With modern audio op amps the BUF143, LME49600, and LMH6321 are pretty much the only output devices which yield stability with only global feedback to the controlling op amp that are available in thermally sufficient packages. None of them are fast enough to avoid response peaking and none have the current capability to be attractive for a 100W build.
Feedback impedances on the order of 10k are typically sufficient to avoid amplifier THD limiting on resistor heating, though it's wise to run an analysis using the thermal resistance and temperature coefficient of the candidate parts to confirm. A few hundred k and up is usually undesirable due to higher noise and larger output errors.
Feedback impedances on the order of 10k are typically sufficient to avoid amplifier THD limiting on resistor heating, though it's wise to run an analysis using the thermal resistance and temperature coefficient of the candidate parts to confirm. A few hundred k and up is usually undesirable due to higher noise and larger output errors.
"No Vbe multiplier? "
Not needed.
You can make diodes out of the same transistors used for Q1, Q2, and glue them to Q1, Q2 (CFP needs to have the bias track the drivers, not the outputs). Or do the same with Q5, Q6.
I have run a basic circuit (no Q5, Q6, replace with a resistor, selected for about 3mA~4mA), and class B bias on Q3, Q4 (0.34V), with no emitter resistors, and it sounded fine.
I used ±30V with MJE3055/2955 driven with MPSA06/56 for drivers. Power bandwidth was 230Khz, with the feedback caps from the outputs back to the drivers at 10nF in parallel with 150Ω. The resistors to ground were 51Ω. 100nF plus 10Ω on the output, no inductor.
Not needed.
You can make diodes out of the same transistors used for Q1, Q2, and glue them to Q1, Q2 (CFP needs to have the bias track the drivers, not the outputs). Or do the same with Q5, Q6.
I have run a basic circuit (no Q5, Q6, replace with a resistor, selected for about 3mA~4mA), and class B bias on Q3, Q4 (0.34V), with no emitter resistors, and it sounded fine.
I used ±30V with MJE3055/2955 driven with MPSA06/56 for drivers. Power bandwidth was 230Khz, with the feedback caps from the outputs back to the drivers at 10nF in parallel with 150Ω. The resistors to ground were 51Ω. 100nF plus 10Ω on the output, no inductor.
Hi mr Ingenieus!
I have power supply Ucc=+/-37V with about 7A. I have the same speakers with impedance of 4ohm and frequency characteristics from the 40 .. 18000Hz to +/-2dB!Give your recommend an audio amplifier class AB (schematics or project) with a pair of output transistors in TO3 package (such as MJ15003/4 or MJ4502/ 802) that I want to use it for home use.
Thanks for the help!
That power supply is perfect for the circuit in post 9. I can adapt it to work with other transistors.
If you want to use a TO3 package, it must mean that you have a heatsink for it. I can't think of many other reasons to use them.
PS: after taking a good hard look at the circuit, I noticed the diodes. Shame on me. Kinda old school to do it that way though.
Last edited:
"I can't think of many other reasons to use them."
How about 250W of SOA de-rated to 200°C junction temperature?
An amplifier driven hard may be running a 100°C junction temperature. With 150°C plastic outputs you only have a Delta T of 50°C to work with. With 200°C metal you have a Delta T of 100°C to work with. What this means is that you need 2x as many 25°C plastic watts to really be equal to the 25°C metal watts. The common plastic outputs are 150W, the ON Semiconductor metal are 250W, so you really need three plastic to be equal to one metal.
How about 250W of SOA de-rated to 200°C junction temperature?
An amplifier driven hard may be running a 100°C junction temperature. With 150°C plastic outputs you only have a Delta T of 50°C to work with. With 200°C metal you have a Delta T of 100°C to work with. What this means is that you need 2x as many 25°C plastic watts to really be equal to the 25°C metal watts. The common plastic outputs are 150W, the ON Semiconductor metal are 250W, so you really need three plastic to be equal to one metal.
"Kinda old school to do it that way though."
Works well though.
I was doing a commercial design to be built by unskilled labor on the Pacific Rim. I wanted to have no adjustments needed, no thermal contact issues (relating to bias), and be able to substitute virtually any small signal transistor pairs. The output pairs were mounted directly on the heatsinks, no insulators, no shoulder washers. It was deliberately overcompensated.
"I am also trying to figure out the feedback network on the output. "
I was running a voltage gain of 3x, dropping at HF due to the 10nF cap.
Works well though.
I was doing a commercial design to be built by unskilled labor on the Pacific Rim. I wanted to have no adjustments needed, no thermal contact issues (relating to bias), and be able to substitute virtually any small signal transistor pairs. The output pairs were mounted directly on the heatsinks, no insulators, no shoulder washers. It was deliberately overcompensated.
"I am also trying to figure out the feedback network on the output. "
I was running a voltage gain of 3x, dropping at HF due to the 10nF cap.
Here is new version with TO3 output devices. The fly in the ointment is the $10 OPA445 high voltage opamp, but it does allow the use of an emitter follower output stage.
The alternative $1 NE5532 opamp needs a complimentary feedback pair to work. This is not necessarily a bad thing, but is just a tad less stable. With careful design it would be OK.
The alternative $1 NE5532 opamp needs a complimentary feedback pair to work. This is not necessarily a bad thing, but is just a tad less stable. With careful design it would be OK.
Attachments
Careful. This is not a regular CFP because the output stage has gain. As a result the full current of the output transistors does not act on the emitter resistors of the 'drivers' so unlike CFP the ability to track output transistor temperature through the drivers is not total, rather (with the given resistors in the feedback network) about 1/4.
In these kinds of outputs it pays to have minimum gain that will get you full swing with what the OPamp can provide, as it improves thermal tracking. How well it will work at higher standing currents (fortunately CFP usually needs low standing current) and with particular cooling arrangements, is not always predictable, so a VBE multiplier, or fitting drivers and outputs on the same heatsink may indeed be necessary for thermal stability.
Have a look at this implementation of the same basic idea.
http://www.diyaudio.com/forums/solid-state/242014-anyone-recognise-these.html
http://www.diyaudio.com/forums/solid-state/242014-anyone-recognise-these.html
You could also concider one of the Quad 405 mods listed here
— Why not buy Keith Snook a BEER —
on Keith Snooks excellent web site
the Mod3 and 4 offer the lowest componant count, as the output stage is unbiased there are no thermal issues
Stuart
— Why not buy Keith Snook a BEER —
on Keith Snooks excellent web site
the Mod3 and 4 offer the lowest componant count, as the output stage is unbiased there are no thermal issues
Stuart
Hi Ingenieus!
I can not find the IC1=OPA445. For Power supply the IC1 = NE5532 or TL071 can I use this schematics? (see schematics)
Thanks for your help!
P.S.
Hi all!
How to transfer PCB for this amplifier in Protel98 here on this post?
Attachments
Last edited:
While it will work, the output swing of the ampwill be limited to to +15V. To make full use of the+37V from the power supply, you need an output stage with voltage gain, i.e. a complimentary feedback pair with local feedback as already shown.
The emitter follower output stage in my previous posts will only work with a high voltage opamp. There are others available (see above) but if you can't get one of them, the only remaining alternative is a circuit with discrete components.
Hi djk,
Referring to the circuit in your post #22, I would like to have your opinion on
the following subjects:
1. Is it beneficial to feed the CCS's comprised of Q5, Q6 and their associated
components from the 15 V regulated supplies instead of the full supply
voltages? Afterall, the voltage swing at their collectors is in line with the
output of the opamp.
2. Is there any downside to dispense with the CCS's altogether and replace
them with bootstrapped or plain R's?
3. Is there an ideal value of R from the output of the opamp to the midpoint
of the bias chain? 100 ohms?
Thanks in advance
Selim
Referring to the circuit in your post #22, I would like to have your opinion on
the following subjects:
1. Is it beneficial to feed the CCS's comprised of Q5, Q6 and their associated
components from the 15 V regulated supplies instead of the full supply
voltages? Afterall, the voltage swing at their collectors is in line with the
output of the opamp.
2. Is there any downside to dispense with the CCS's altogether and replace
them with bootstrapped or plain R's?
3. Is there an ideal value of R from the output of the opamp to the midpoint
of the bias chain? 100 ohms?
Thanks in advance
Selim
"1. Is it beneficial to feed the CCS's comprised of Q5, Q6 and their associated
components from the 15 V regulated supplies instead of the full supply
voltages? Afterall, the voltage swing at their collectors is in line with the
output of the opamp."
I fed the CCS from the ±15V supplies.
"2. Is there any downside to dispense with the CCS's altogether and replace
them with bootstrapped or plain R's?"
I chose to use plain R's fed from the ±15V rails. Using an active CCS will improve measurements by a factor of 10x (20dB), but it does not sound any better. One could argue that the R's from the higher voltage from the output stage would result in a more constant current, but there is the power supply ripple as well. I was more interested in getting some bias in the diodes under small signal conditions, nothing more.
"3. Is there an ideal value of R from the output of the opamp to the midpoint
of the bias chain? 100 ohms?"
IIRC, I used 51Ω. A different opamp may work better with a different value.
I tried a few opamps, TL072, LF353. The TL072 sounded better than the LF353, but when driven into clipping it latched to the negative rail and sounded horrible (typical bi-FET behavior). The 5532 sounded better at any volume level, and did not misbehave even hard into clipping. We actually destroyed a transformer from driving into clipping for an extended period of time.
components from the 15 V regulated supplies instead of the full supply
voltages? Afterall, the voltage swing at their collectors is in line with the
output of the opamp."
I fed the CCS from the ±15V supplies.
"2. Is there any downside to dispense with the CCS's altogether and replace
them with bootstrapped or plain R's?"
I chose to use plain R's fed from the ±15V rails. Using an active CCS will improve measurements by a factor of 10x (20dB), but it does not sound any better. One could argue that the R's from the higher voltage from the output stage would result in a more constant current, but there is the power supply ripple as well. I was more interested in getting some bias in the diodes under small signal conditions, nothing more.
"3. Is there an ideal value of R from the output of the opamp to the midpoint
of the bias chain? 100 ohms?"
IIRC, I used 51Ω. A different opamp may work better with a different value.
I tried a few opamps, TL072, LF353. The TL072 sounded better than the LF353, but when driven into clipping it latched to the negative rail and sounded horrible (typical bi-FET behavior). The 5532 sounded better at any volume level, and did not misbehave even hard into clipping. We actually destroyed a transformer from driving into clipping for an extended period of time.
Last edited:
Here is a low voltage opamp driving a CFP output with voltage gain. It works (in a simulation at least) but performance is not the very best. It gives 70W into 8ohm. THD at 1kHz is a very nice 0.0068% but this shoots up to 0.27% at 20kHz. Also, setting the bias is still very tricky. I used the OPA134 (around $3 in my part of the world) because I couldn't get the NE5532 to work properly.
Those output transistors are very reluctant to switch off, as shown by the output currents and the distortion waveform.
Those output transistors are very reluctant to switch off, as shown by the output currents and the distortion waveform.
Attachments
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.