Hi, I need a low power, power amp for my system (2way with active crossover). I had toyed with the idea of a Pass F5 or a JLH, but they seem over the top since I only need 1 Watt, and I need 4 channels. I came up with this as an alternative. There's no power amp as such, just a class A, unity gain current buffer for the pre-amp. The feedback can be removed from the pre-amp and put on the end of the buffer to get low distortion, or left where it is and have no feedback.
I think of it as a JLH with the middle transistor upside-down, and imagine (no sim software or test equipment at the moment), that the top transistor works as an emitter follower with an active load formed by the bottom transistor (one of the definitions of the JLH).
Is this likely to work?
Thanks,
Brian.
I think of it as a JLH with the middle transistor upside-down, and imagine (no sim software or test equipment at the moment), that the top transistor works as an emitter follower with an active load formed by the bottom transistor (one of the definitions of the JLH).
Is this likely to work?
Thanks,
Brian.
Attachments
It's just a little bit more complicated
The idea is valid, the challenge (as Messrs Hood and Pass have pointed out) is getting a current source with the bandwidth (slew rate) you need.
If you really want to get simple then some PA iron working as an inductor in series (like capacitors in parallel) with some large air core inductors should do well.
Then you "just" need to sort out the bias settings.
Frankly a single FET (a la Nelson Pass) with a large inductor and an "ultra path" DC blocking cap would be the easiest solution. Back to a single device outside your op-amp. If brave, the capacitor is optional with split rails (if you can use the op-amp as a DC servo.
The idea is valid, the challenge (as Messrs Hood and Pass have pointed out) is getting a current source with the bandwidth (slew rate) you need.
If you really want to get simple then some PA iron working as an inductor in series (like capacitors in parallel) with some large air core inductors should do well.
Then you "just" need to sort out the bias settings.
Frankly a single FET (a la Nelson Pass) with a large inductor and an "ultra path" DC blocking cap would be the easiest solution. Back to a single device outside your op-amp. If brave, the capacitor is optional with split rails (if you can use the op-amp as a DC servo.
Hi thoglette, thanks for your reply and suggestions. I can see how an output FET would work open loop, but to work closed loop, would it have to be a source follower?
I think my original idea would be more linear assuming I can sort the current source out as you say. The simple answer to that is to bootstrap it as JLH did, but I would probably try a normal two transistor circuit first. I think I need to get a PC the I can simulate this.
Best regards,
Brian
I think my original idea would be more linear assuming I can sort the current source out as you say. The simple answer to that is to bootstrap it as JLH did, but I would probably try a normal two transistor circuit first. I think I need to get a PC the I can simulate this.
Best regards,
Brian
Yes, otherwise you'd have to feedback to the positive input.
Should work well, just a little more complicated. 1W into 8ohms is 350mA, if you can find a high beta bipolar you may be able to avoid the voltage amp stage
Member
Joined 2009
Paid Member
Why reinvent the wheel - the JLH can be made to work at lower power and you can use smaller transistors for the output. THe basic JLH is simple enough and is a proven design. Scaling it down shouldn't be an issue - in fact I think it's been done before for use as a headphone amplifier.
Hi Bigun, mainly because I don't need the gain or the extra complexity and I expect that the distortion would be lower this way, not that distortion was ever really as issue for the JLH... My idea would also get rid of any DC offset adjustment because the opamp in the preamp would handle that for me.
Best Regards,
Brian
Hi Guys
The JLH structure is riddled with THD and bias problems. Best to avoid it unless you really like rising THD with level. Scaling the circuit up or down simply moves the cusp of muddiness to higher or lower power and SPL levels.
The easiest way to convert the opamp preamp into a power amp is to add a complimentary follower stage - either as CFPs or EFs. Make sure there is a small resistance between the opamp output and EF input, with a local feedback cap around the opamp for stability.
Biasing is achieved using a standard bias regulator (Vbe multiplier base spreader). This can be fed by actual current sources or bootstrapped resistors.
No one has mentioned that the power supply in the preamp will not support the current needed to drive a speaker even to 1W per channel. That level requires about 350mA per channel into 8R, at which point the output stage is dissipating around 5W peak assuming +/-15V rails. So, good heat sinking is required, as well.
TO220 F-paks like the 2SC4793 - 2SA1837 pair allow currents up to about 1A and do not need mica insulators - just bolt them to a heat sink. Parallel pairs would help with peak output power up to 1A5 loads. Larger packages (TO-3P, TO-246) provide lower thermal resistance and safer operation at typical listening levels. These do not need paralleling to handle the current, although there are benefits to paralleling as explained in my books, and others, as well as on various threads here.
Given that true listening room SPLs that are tolerable - let alone listenable for hours - are much lower than most people suppose, being able to get a watt or two per speaker is sufficient and does not require high supply voltages. This applies easily to pop music and to movie and TV sound. If you listen to classical music - especially exclusively - then much higher peak power is beneficial as the dynamic range is higher on average - 20db typical.
The above circuit will essentially look like many headphone drivers, based on the classic opamp with discrete current booster.
There is a certain amount of complexity that proper design requires. It is overlooked by everyone that the historic designs of JLH and others, particularly UK designers, was predicated on keeping cost low as BJTs were "expensive" at the time. This frugality lead to many assumptions about circuits being spouted and repeated to the point of being believed as unthwartable. In the proper context, those old rules fall to pieces and those old circuits are best forgotten, except for being steeping stones to truly great performance.
Have fun
Kevin O'Connor
The JLH structure is riddled with THD and bias problems. Best to avoid it unless you really like rising THD with level. Scaling the circuit up or down simply moves the cusp of muddiness to higher or lower power and SPL levels.
The easiest way to convert the opamp preamp into a power amp is to add a complimentary follower stage - either as CFPs or EFs. Make sure there is a small resistance between the opamp output and EF input, with a local feedback cap around the opamp for stability.
Biasing is achieved using a standard bias regulator (Vbe multiplier base spreader). This can be fed by actual current sources or bootstrapped resistors.
No one has mentioned that the power supply in the preamp will not support the current needed to drive a speaker even to 1W per channel. That level requires about 350mA per channel into 8R, at which point the output stage is dissipating around 5W peak assuming +/-15V rails. So, good heat sinking is required, as well.
TO220 F-paks like the 2SC4793 - 2SA1837 pair allow currents up to about 1A and do not need mica insulators - just bolt them to a heat sink. Parallel pairs would help with peak output power up to 1A5 loads. Larger packages (TO-3P, TO-246) provide lower thermal resistance and safer operation at typical listening levels. These do not need paralleling to handle the current, although there are benefits to paralleling as explained in my books, and others, as well as on various threads here.
Given that true listening room SPLs that are tolerable - let alone listenable for hours - are much lower than most people suppose, being able to get a watt or two per speaker is sufficient and does not require high supply voltages. This applies easily to pop music and to movie and TV sound. If you listen to classical music - especially exclusively - then much higher peak power is beneficial as the dynamic range is higher on average - 20db typical.
The above circuit will essentially look like many headphone drivers, based on the classic opamp with discrete current booster.
There is a certain amount of complexity that proper design requires. It is overlooked by everyone that the historic designs of JLH and others, particularly UK designers, was predicated on keeping cost low as BJTs were "expensive" at the time. This frugality lead to many assumptions about circuits being spouted and repeated to the point of being believed as unthwartable. In the proper context, those old rules fall to pieces and those old circuits are best forgotten, except for being steeping stones to truly great performance.
Have fun
Kevin O'Connor
Last edited:
Hi Kevin, thanks for the conprehensive reply. I'm not sure what benefit a follower stage would offer over what I have posted since my idea is a follower anyway, could you please clarify a little. I have no PC or simulation tools at the moment, so I cannot check different topologies.
I shouldn't really of mentioned JLH because my middle transistor is "up side down", and is an emitter follower.
Best Regards,
Brian
I shouldn't really of mentioned JLH because my middle transistor is "up side down", and is an emitter follower.
Best Regards,
Brian
Hi Guys
Brian, just rewire that middle BJT as a bias circuit. Use a typical EF with NPN on V+ and PNP V-. Make the circuit additions I described and you have a complete circuit.
Also, add a 1k to ground at the amplifier output. A DC leak path there helps to stabilise offset and should be a part of every direct coupled output.
With typical 2SA/2SC output devices, beta is about 100. An output current of 1A therefore needs 10mA of drive current to the BJTs. Best to add a driver stage to the output stage to relieve current demands (and thus reduce THD) of the opamp. If the amp is strictly for 1W or less, then you don't need the drivers. However, best to err on the side of reliability and use MJL1302/3281 outputs and 2SA1837/2SC4793 drivers.
Have fun
Kevin O'Connor
Brian, just rewire that middle BJT as a bias circuit. Use a typical EF with NPN on V+ and PNP V-. Make the circuit additions I described and you have a complete circuit.
Also, add a 1k to ground at the amplifier output. A DC leak path there helps to stabilise offset and should be a part of every direct coupled output.
With typical 2SA/2SC output devices, beta is about 100. An output current of 1A therefore needs 10mA of drive current to the BJTs. Best to add a driver stage to the output stage to relieve current demands (and thus reduce THD) of the opamp. If the amp is strictly for 1W or less, then you don't need the drivers. However, best to err on the side of reliability and use MJL1302/3281 outputs and 2SA1837/2SC4793 drivers.
Have fun
Kevin O'Connor
Hi Kevin, I understand what you mean now, thank you for that. Will this give lower distortion than the original circuit I posted? Sorry, but I still don't have a PC to simulate these things on (hoping to resolve that in a couple of weeks).
Regards,
Brian.
Regards,
Brian.
Follow the pre-amp with a diamond buffer. This circuit is simple and easy to thermally compensate the output stage. Use J-fets with the gate connected to the source for the current sources. Easy peasy.😉
Have you considered one of the Power Op-Amp family.
http://docs-europe.electrocomponents.com/webdocs/06e8/0900766b806e88be.pdf
The LM386 only needs 4 components.
http://docs-europe.electrocomponents.com/webdocs/06e8/0900766b806e88be.pdf
The LM386 only needs 4 components.
Hi CBS240, a diamond buffer was my second choice, but without any simulation tools I wasn't sure if it would be better or worse than my suggestion. I'll be getting a PC in a couple of weeks, so I can check then and I'll do some reading in the meantime. Thanks for the suggestion.
KatieandDad, I could use a power op amp, but I couldn't find one that is unity gain stable, or that had distortion figures to even get close to my old class A amplifier (<0.001% THD simulated). Unless you can find one of course.... Thanks for the suggestion anyway.
Brian.
KatieandDad, I could use a power op amp, but I couldn't find one that is unity gain stable, or that had distortion figures to even get close to my old class A amplifier (<0.001% THD simulated). Unless you can find one of course.... Thanks for the suggestion anyway.
Brian.
Hi Guys
DO NOT use a diamond buffer for the output. You are trying to make a power amp here, albeit a low power one.
The main attraction of the diamond buffer is its complementary-ness and single-node input, neither of which are of benefit here. The diamond has an illusion of simplicity with its four devices, but requires dual current sources to function at all. These can be dual bootstraps if you do not wish to add further active devices.
The standard circuit that I described above allows easy adjustment of the idle current, where the diamond does not. The described circuit will function far better than anything based on the JLH10. Sometimes conventionality is superior.
The composite amplifier using an opamp front-end with a discrete power buffer works far better than any chip amp or any power opamp. It is well-known that any chip driving a heavy load will have thermal issues where the output stage heat will upset the THD performance of the front-end, all being on the same die. The fascination with chip amps certainly pleases manufacturers of those items, but you will never get the performance of discrete, or even of partial discrete. In any such hybrid circuit, the discrete portion should be the power stage.
Have fun
Kevin O'Connor
DO NOT use a diamond buffer for the output. You are trying to make a power amp here, albeit a low power one.
The main attraction of the diamond buffer is its complementary-ness and single-node input, neither of which are of benefit here. The diamond has an illusion of simplicity with its four devices, but requires dual current sources to function at all. These can be dual bootstraps if you do not wish to add further active devices.
The standard circuit that I described above allows easy adjustment of the idle current, where the diamond does not. The described circuit will function far better than anything based on the JLH10. Sometimes conventionality is superior.
The composite amplifier using an opamp front-end with a discrete power buffer works far better than any chip amp or any power opamp. It is well-known that any chip driving a heavy load will have thermal issues where the output stage heat will upset the THD performance of the front-end, all being on the same die. The fascination with chip amps certainly pleases manufacturers of those items, but you will never get the performance of discrete, or even of partial discrete. In any such hybrid circuit, the discrete portion should be the power stage.
Have fun
Kevin O'Connor
Last edited:
OK, I have a PC now, and have started simulations, but I'm not sure I'm doing it right. I have done the JLH and the diamond buffer; the JLH being considerably higher distortion, but the diamond buffer seems incredibly low. I might have a bulk stock of ZTX devices 😉
Please could someone have a look at this and tell me if I've got it wrong. Distortions down at -150dB for 100mV, and -90dB for 1V seem very low for open loop.
Thanks,
Brian
PS. I will have to figure out how to attach things now :-(
Please could someone have a look at this and tell me if I've got it wrong. Distortions down at -150dB for 100mV, and -90dB for 1V seem very low for open loop.
Thanks,
Brian
PS. I will have to figure out how to attach things now :-(
Last edited:
- Status
- Not open for further replies.