One thing I am trying to do is "future proof" a Class AB design with reasonable performance because many transistors become hard to source over time as commercial designs continue to advance. 2N5551/2N5401 have many alternate and generic sources and have high enough VCE for 200 to 300 W class amps with 60 to 70 volt +/- supply voltages.
However, they are probably not the best choice for many of the parts of the amp, and perhaps another part such as BC546C/BC556C might be a better choice and also easy to source. For example, the long tailed pair input which best has a low noise figure might have a low enough VCE so that a BC546C could be used. For constant current sources that have a low VCE, there might be changes in Hfe that cause nonlinearity that causes distortion for which BC546C/BC556C is a better choice as well. There are likely other issues I am not considering. How serious are these problems if 2N5551/2N5401 is used in these roles?
Also, other TO-126 parts seem to be becoming rarer such as MJE340/MJE350. However, TO-220 parts seem to be available, in particular MJE15032/MJE15033 or TIP47C/MJE5731, with MJE10532/MJE15033 having at least one generic available. So while it might be more slightly expensive, using MJE15032/MJE15033 in place of MJE340/MJE350 might help future proof a design. Increased gain-bandwidth might require extra compensation, and increased base and collector capacitance might be an issue here as well.
As for power NPN/PNP combinations there are fortunately still many choices. However, in the interest of future proofing, reducing the number of types of devices required and making the devices have many generic sources helps to ensure availability. Do the substitutions I am suggesting make sense and would not significantly compromise an amplifier?
However, they are probably not the best choice for many of the parts of the amp, and perhaps another part such as BC546C/BC556C might be a better choice and also easy to source. For example, the long tailed pair input which best has a low noise figure might have a low enough VCE so that a BC546C could be used. For constant current sources that have a low VCE, there might be changes in Hfe that cause nonlinearity that causes distortion for which BC546C/BC556C is a better choice as well. There are likely other issues I am not considering. How serious are these problems if 2N5551/2N5401 is used in these roles?
Also, other TO-126 parts seem to be becoming rarer such as MJE340/MJE350. However, TO-220 parts seem to be available, in particular MJE15032/MJE15033 or TIP47C/MJE5731, with MJE10532/MJE15033 having at least one generic available. So while it might be more slightly expensive, using MJE15032/MJE15033 in place of MJE340/MJE350 might help future proof a design. Increased gain-bandwidth might require extra compensation, and increased base and collector capacitance might be an issue here as well.
As for power NPN/PNP combinations there are fortunately still many choices. However, in the interest of future proofing, reducing the number of types of devices required and making the devices have many generic sources helps to ensure availability. Do the substitutions I am suggesting make sense and would not significantly compromise an amplifier?
I just buy qty=500 pieces of whatever parts make it into the final design. That's future proof enough for me, there's no way I'm going to go through all of those in my lifetime, building more copies of the AB amp.
So far it has worked very well. I'm sitting on a stash of 490 pieces of 2SC5171 and 490 more of 2SA1930. 490 pieces of Sanyo 2SC3503 / 2SA1381, 490 pieces of PN4250A, 490 pieces of MPS6571, 490 pieces of BF862, and on and on.
So far it has worked very well. I'm sitting on a stash of 490 pieces of 2SC5171 and 490 more of 2SA1930. 490 pieces of Sanyo 2SC3503 / 2SA1381, 490 pieces of PN4250A, 490 pieces of MPS6571, 490 pieces of BF862, and on and on.
I am trying to make a design for DIY so they will be sourced in the future by those who have not already accumulated any parts. So I would like to use devices that are likely to be available from generic manufacturers in the future if possible.
No getting away from SMD parts im afraid. SOT-223 is not hard to solder and can replace TO-126 type devices quite readily
eg look at BCP53/56. Theres a few sources of these and they seem roughly equivalent to BD139/140 - you could easily use those for VAS duty
Perhaps the OP is assuming that we will still be able to source some genuine, specified, through-hole semis and components in the future, generic or otherwise. These are already an obsolete class of component for new designs and most remaining types are slowly but surely being whittled away as major clients of the manufacturers shift to SMT with chip-based audio processing and class D power with mosfet devices.
The problem for DIYs and repairers globally, is there wouldn't be enough demand to keep through-hole production lines operating in the future since they require orders for millions of parts to make profitable runs. Already though, we are experiencing a serious problem with existing generic substitutes becoming too unreliable or variable to consider as anything more than junk parts. Fakes don't help the situation either and even the generic types can be fakes when they don't meet their own spec.
I think Mark's plan is the best for the near (say 10 years) future. Beginning with small signal types, buy up large stocks of the best low noise, general purpose audio transistors you can find at a bulk price. Use them freely and sell off small quantities to friends at a nominal profit to lighten the burden. I'd go for the favourites here: KSC1845 and KSA992 BJTs. You can use a lot of these in discrete audio designs up to 120V - very handy!
The problem for DIYs and repairers globally, is there wouldn't be enough demand to keep through-hole production lines operating in the future since they require orders for millions of parts to make profitable runs. Already though, we are experiencing a serious problem with existing generic substitutes becoming too unreliable or variable to consider as anything more than junk parts. Fakes don't help the situation either and even the generic types can be fakes when they don't meet their own spec.
I think Mark's plan is the best for the near (say 10 years) future. Beginning with small signal types, buy up large stocks of the best low noise, general purpose audio transistors you can find at a bulk price. Use them freely and sell off small quantities to friends at a nominal profit to lighten the burden. I'd go for the favourites here: KSC1845 and KSA992 BJTs. You can use a lot of these in discrete audio designs up to 120V - very handy!
Last edited:
One thing I have noticed on Shenzhen parts suppliers such as lcsc.com is that parts like the 2N5551 and 2N5401 have generic version from Chinese companies like Semtech, Changjiang semiconductor, Unisonic Tech, and other manufacturers. For example, look at:
Transistors (NPN/PNP) | Transistors | LCSC.com
Transistors (NPN/PNP) | Transistors | LCSC.com
Transistors (NPN/PNP) | Transistors | LCSC.com
Transistors (NPN/PNP) | Transistors | LCSC.com
and that just in general there are more generic TO-220 parts that are still available than TO-126 parts. So I would think a design that uses TO-220 parts with generally sufficiently high VCE, Hfe, and transition frequency could be used as long as the compensation was set correctly.
So I think if there are readily available alternatives available it may be from these suppliers who may be willing to source these parts if they are phased out by the original manufacturers.
Transistors (NPN/PNP) | Transistors | LCSC.com
Transistors (NPN/PNP) | Transistors | LCSC.com
Transistors (NPN/PNP) | Transistors | LCSC.com
Transistors (NPN/PNP) | Transistors | LCSC.com
and that just in general there are more generic TO-220 parts that are still available than TO-126 parts. So I would think a design that uses TO-220 parts with generally sufficiently high VCE, Hfe, and transition frequency could be used as long as the compensation was set correctly.
So I think if there are readily available alternatives available it may be from these suppliers who may be willing to source these parts if they are phased out by the original manufacturers.
Manufacturers like Unisonic, Wing Shing, ISC and Mospec are well known to long time DIYs. Some of them already have distributors of good quality semis, including useful TO126 audio types, in several countries. With marketing companies like LCSC, you would be dealing directly with a local agency, not a selling platform like Ebay that is set up like a mass retailers' online operation. Be prepared for some extra requirements if you choose to buy that way.
It's not clear either, if import duties and taxes applicable to your country will be added to your bill at the time of purchase or if you will need to negotiate them with each transaction by supplying specific, authorised information on how they need to be applied either before dispatch or on arrival at your local distribution centre. It could get very sticky, slow and expensive even for small quantities in the latter case.
Why not raise an order first and inform us how long it took, any special procedures, difficulties etc. before suggesting others do business that way?
It's not clear either, if import duties and taxes applicable to your country will be added to your bill at the time of purchase or if you will need to negotiate them with each transaction by supplying specific, authorised information on how they need to be applied either before dispatch or on arrival at your local distribution centre. It could get very sticky, slow and expensive even for small quantities in the latter case.
Why not raise an order first and inform us how long it took, any special procedures, difficulties etc. before suggesting others do business that way?
I buy from lcsc.com all the time to the USA and usually get my order within 2 weeks. I usually buy less than 500 USD worth of parts because I am just ordering for my own hobbyist use, not to set up a manufacturing supply chain.
That's good to know. So what taxes are applicable and where does it happen? In this country, a goods and services tax now applies to all private and commercial transactions both local and import, with no minimum threshold.
Last I read, in many US states the government no longer has a minimum threshold on import taxes applied to electronic products including transistors either. I could be wrong there but it appears that you would be required to pay on all GSP transactions up front, if the seller is set up to remit the tax automatically via, for example, Ebay's GSP (Global Shipping Program). 'Not sure how that works with independent sellers like LCSC but I imagine it would have to be managed differently and likely by you, before local delivery.
Last I read, in many US states the government no longer has a minimum threshold on import taxes applied to electronic products including transistors either. I could be wrong there but it appears that you would be required to pay on all GSP transactions up front, if the seller is set up to remit the tax automatically via, for example, Ebay's GSP (Global Shipping Program). 'Not sure how that works with independent sellers like LCSC but I imagine it would have to be managed differently and likely by you, before local delivery.
With due respect, I think you are overthinking this.
Even more:
Ok, maybe so, wish you the best ... but I´d wait for sales, say, during the first year (in which all parts will be as available as today) to have some hard numbers and estimate future sales.
Even more:
You seem to be certain your design will be hugely popular and in high demand for a long time. 🙂
Ok, maybe so, wish you the best ... but I´d wait for sales, say, during the first year (in which all parts will be as available as today) to have some hard numbers and estimate future sales.
I am more designing for open source, and I provide all of the gerbers/BOMs (and source code where applicable) for all of my projects to on my github
profdc9 * GitHub
including this project which is currently in progress:
GitHub - profdc9/PowerAmpAudio: Power Amplifier based on Michael Chua's C300 amplifier
I have projects already there for a open source Vector Network Analyzer I designed, a HF (up to 30 MHz) RF amplifier for amateur radio use, an automatic tuning unit for amateur radio use, solid state Tesla coil control circuits, a Nixie Clock, a theremin, a constant-current constant voltage linear power supply, among others. The designs are all in Kicad and can be downloaded and modified using this software as needed. I also use Qucs SPICE simulator to model them. They are intended to be working examples that can be download, studied, and modified as needed for educational purposes.
To that end, I was thinking of using 2N5551/2N5401 transistors for small-signal transistors, a pair of TO-220 type high voltage transistors like MJE15032/MJE15033 for the VAS (there seems to be other TO-220 types available as well that can fit this purpose), and then a pair of power NPN/PNP like MJL3281/MJL1302 for the power output stage. Try to keep it as simple as possible so that there is minimal risk of part obsolescence while maintaining good performance.
profdc9 * GitHub
including this project which is currently in progress:
GitHub - profdc9/PowerAmpAudio: Power Amplifier based on Michael Chua's C300 amplifier
I have projects already there for a open source Vector Network Analyzer I designed, a HF (up to 30 MHz) RF amplifier for amateur radio use, an automatic tuning unit for amateur radio use, solid state Tesla coil control circuits, a Nixie Clock, a theremin, a constant-current constant voltage linear power supply, among others. The designs are all in Kicad and can be downloaded and modified using this software as needed. I also use Qucs SPICE simulator to model them. They are intended to be working examples that can be download, studied, and modified as needed for educational purposes.
To that end, I was thinking of using 2N5551/2N5401 transistors for small-signal transistors, a pair of TO-220 type high voltage transistors like MJE15032/MJE15033 for the VAS (there seems to be other TO-220 types available as well that can fit this purpose), and then a pair of power NPN/PNP like MJL3281/MJL1302 for the power output stage. Try to keep it as simple as possible so that there is minimal risk of part obsolescence while maintaining good performance.
Make the circuit design tolerant of transistor substitution. Most solid state designs are, to some extent. Some more than others, and you won’t get the absolute pinnacle of performance. Use local feedback where possible to reduce dependency on Hfe, choose operating points carefully, design so you can swamp out high output capacitances in power devices. When the ‘ideal’ transistors disappear, something close or even generic can be substituted.
BC546C/556C aren't specified as low noise either. I don't have the datasheets handy, but I suspect their noise figures being not too different from the 2N5401/5551's. The low noise ones in that group are BC550C/560C - with lower VCE ratings, though.
Best regards!
Another, completely different, approach is: radically re-design the amplifier, so that it will function correctly using a wide variety of different part types. No matter which parts are available in Y2055, they will work in your amplifier.
You'd have to assume your BJTs have low beta, low fT, high Cbc, low ICEmax, and low VCEmax.
You'd have to assume your MOSFETs have a threshold voltage anywhere in the range (0.4<Vth<4.6), low current, low VDSmax, and high gate charge
You'd have to assume that JFETs are unavailable
You'd have to assume that Zener diodes are only available in the E6 standard values
You'd have to assume that resitors are only available in the E24 standard values
But- this is not impossible! Douglas Self built a power amplifier using 64 parallel IC opamps. Nelson Pass built a power amplifier ("The Beast Of A Thousand JFETs") using 1100 small signal JFETs. Ham Radio operators routinely hold contests to see who can build the most innovative and elegant transmitter+receiver using ONLY 2N2222 transistors.
If they can do it, you can do it.
Pretend you're working in the Soviet bloc, in 1969, with very very few semiconductor devices available, and all of them primitive. Take what you have and make it work! Failure is unacceptable, Comrade.
You can use series stacks and/or cascodes, to circumvent voltage limitations. Nelson Pass's very first power amplifier at Threshold, had an output stage with a stack of 3 bipolar transistors in series, to get around the voltage limits and (more importantly) safe operating area limits. If he can do it, you can do it.
You can use parallel devices, to circumvent current limitations. Ballasting resistors allow you to build output stages with four or more parallel devices. Today, in the year of our lord 2020, Parasound brags that they have eight parallel PNPs and eight more parallel NPNs in the output stage of their power amp. If they can do it, you can do it.
You can circumvent low-fT limitations with clever circuit design and clever feedback topologies. I will leave that as an exercise for the reader; it will be one of the most fun pieces of your project.
Here's a back-of-the-envelope calculation for you:
You can surf around Nelson Pass's website to find photos of his "Beast Of A Thousand JFETs" -- it's built on a single PCB -- to get an idea of the physical size of an amplifier containing 1600 TO-92s.
_
You'd have to assume your BJTs have low beta, low fT, high Cbc, low ICEmax, and low VCEmax.
You'd have to assume your MOSFETs have a threshold voltage anywhere in the range (0.4<Vth<4.6), low current, low VDSmax, and high gate charge
You'd have to assume that JFETs are unavailable
You'd have to assume that Zener diodes are only available in the E6 standard values
You'd have to assume that resitors are only available in the E24 standard values
But- this is not impossible! Douglas Self built a power amplifier using 64 parallel IC opamps. Nelson Pass built a power amplifier ("The Beast Of A Thousand JFETs") using 1100 small signal JFETs. Ham Radio operators routinely hold contests to see who can build the most innovative and elegant transmitter+receiver using ONLY 2N2222 transistors.
If they can do it, you can do it.
Pretend you're working in the Soviet bloc, in 1969, with very very few semiconductor devices available, and all of them primitive. Take what you have and make it work! Failure is unacceptable, Comrade.
You can use series stacks and/or cascodes, to circumvent voltage limitations. Nelson Pass's very first power amplifier at Threshold, had an output stage with a stack of 3 bipolar transistors in series, to get around the voltage limits and (more importantly) safe operating area limits. If he can do it, you can do it.
You can use parallel devices, to circumvent current limitations. Ballasting resistors allow you to build output stages with four or more parallel devices. Today, in the year of our lord 2020, Parasound brags that they have eight parallel PNPs and eight more parallel NPNs in the output stage of their power amp. If they can do it, you can do it.
You can circumvent low-fT limitations with clever circuit design and clever feedback topologies. I will leave that as an exercise for the reader; it will be one of the most fun pieces of your project.
Here's a back-of-the-envelope calculation for you:
- Plastic PN2222 / PN2907 transistors in the TO-92 package, are rated 625mW max.
- Assume you are so conservative that you permit yourself to dissipate 250 mW per device
- Assume the "power efficiency" (power delivered to speakers / power dissipated inside the amplifier chassis) is only 20%
- Assume the power delivered to speakers is 80 watts RMS
- Then the power dissipated in the amp chassis is (80 / 0.2) = 400 watts
- Then you need about (400 / 0.25) = 1600 of the PN2222 / PN2907 transistors to build the amp
- Toss in a couple of ultra quiet Noctua DC fans: done!
You can surf around Nelson Pass's website to find photos of his "Beast Of A Thousand JFETs" -- it's built on a single PCB -- to get an idea of the physical size of an amplifier containing 1600 TO-92s.
_
Last edited:
I also forgot to mention that I am including a power conditioning board with the amplifier so that switching power supplies can be used for the rails which will filter out the switching noise. A power supply like the Meanwell LRS-350-48.