Yes I notice a lot of circuits use that Darlington. It doesn't seem to be.made by any name brands anymore. Sure there must be a similar fit.
With Darlingtons it works it or it doesn't work. To be honest a 1 kHz ability gets most of the job done. If you use a 0R33 10uF output filter any energy required at > 5 kHz will come from the 10 uF. Panasonic FC being good types. When not doing much it's a RF filter. Be generous with the capacitor to taste, 1000uF perhaps? T0247 ( TIP ) is a nice case size. LM723 can be trouble. I often think LD1084 is sold as a 5 amp device when T0220 is really for 1.5 amps. It's still a great device and should do 2 amps and likely 3.
If you want to build a better Darlington that will be fun and should be easy.
I was told in 1973 that a zener Darlington PSU was history. It would be nice to show it isn't . Peter Platt ex RAF said he taught it as until then it was common. You might see how 1969 was exactly that time. Peter taught me Kirchoffs current laws which has been a living for me. A ground plane is how it works in the real world and is called a Gausian solution. Gause says you may not know everything going on in a town and you don't need to. An electronic ring road makes the problem go away. Simple PSU's can be Kirchoff solutions. How would you know? You should get within 10 dB of the data sheet fo attenuation. LM723 is a Gausian solution and needs an oscilloscope. It can go unstable. It is remarkably good for Aerospace type things.
If I know the house voltage we can do a realistic test. I have a feeling you might have 236VAC +/- 4 V. If an older system perhaps a little more.
If you want to build a better Darlington that will be fun and should be easy.
I was told in 1973 that a zener Darlington PSU was history. It would be nice to show it isn't . Peter Platt ex RAF said he taught it as until then it was common. You might see how 1969 was exactly that time. Peter taught me Kirchoffs current laws which has been a living for me. A ground plane is how it works in the real world and is called a Gausian solution. Gause says you may not know everything going on in a town and you don't need to. An electronic ring road makes the problem go away. Simple PSU's can be Kirchoff solutions. How would you know? You should get within 10 dB of the data sheet fo attenuation. LM723 is a Gausian solution and needs an oscilloscope. It can go unstable. It is remarkably good for Aerospace type things.
If I know the house voltage we can do a realistic test. I have a feeling you might have 236VAC +/- 4 V. If an older system perhaps a little more.
I think a regulator circuit requiring multiple devices is getting a bit too complex and messy for the simple job of powering a simple, low power amplifier. By complex, I mean too much to fine tune or troubleshoot when all doesn't go to plan. You also need at least 2 of them for a dual rail, stereo pair. As Nigel points out, you need good instruments to troubleshhoot supplies and these don't come cheap, nor does some background in linear supplies to know what to do when power circuits don't do what the designer intended.
For an idea of what a relatively basic LM723 application with a single reg. kit looks like, see - link to a currently available kit but the 14 pin DIL chip dates back to 1976.
For an idea of what a relatively basic LM723 application with a single reg. kit looks like, see - link to a currently available kit but the 14 pin DIL chip dates back to 1976.
I agree with that. I have ordered the Darlingtons. The avantage of what I will try is it is harder to get it wrong. If people don't have an oscilloscope even the simplest regulator might be working badly.
My Kirchoff idea is sometimes called Star Earthing. People write about it like the words alone are the secret. In reality it seldom is seen in the conceptual form. Rotel managed it quite well on the RA931 a budget amplifer. Guess what, it sounded great.
The problem of using the LM723 as a double positive PSU joined like batteries is it's complex and mildly worrying.
I suspect the Darlingtons will do 5 amps quite well. With a couple of RC filters you could have two sets cheaply. 22 000 uF 0R33 ( x 2 ) 4700 uF ( x 2 ) TIP 142 ( x 2 ) TIP 147 ( x 2 ). The filters not only filter they split the PSU. I think that will sound better. 2 amps current when 0R33 is 0.66V loss. It's also a place to measure idle current. Myself, 2 amps is a bit more than you need. If 1.5 amps used one set should be OK. It is almost impossible to get a non feedback Darlington to work badly.
My Kirchoff idea is sometimes called Star Earthing. People write about it like the words alone are the secret. In reality it seldom is seen in the conceptual form. Rotel managed it quite well on the RA931 a budget amplifer. Guess what, it sounded great.
The problem of using the LM723 as a double positive PSU joined like batteries is it's complex and mildly worrying.
I suspect the Darlingtons will do 5 amps quite well. With a couple of RC filters you could have two sets cheaply. 22 000 uF 0R33 ( x 2 ) 4700 uF ( x 2 ) TIP 142 ( x 2 ) TIP 147 ( x 2 ). The filters not only filter they split the PSU. I think that will sound better. 2 amps current when 0R33 is 0.66V loss. It's also a place to measure idle current. Myself, 2 amps is a bit more than you need. If 1.5 amps used one set should be OK. It is almost impossible to get a non feedback Darlington to work badly.
JLH Class-A Amplifier
I like the idea of simple. I am thinking of copying the version linked above.
What is a reasonable target for psu ripple is an amp like this? <1mV?
Any ideas of approx ripple I would get from the LM338? If I understand correctly, assuming something like a 2.5V2ripple after the rectifier, attenuation of around 80db would acheive this. At least as a starting point. I wish I had access to a scope...
I like the idea of simple. I am thinking of copying the version linked above.
What is a reasonable target for psu ripple is an amp like this? <1mV?
Any ideas of approx ripple I would get from the LM338? If I understand correctly, assuming something like a 2.5V2ripple after the rectifier, attenuation of around 80db would acheive this. At least as a starting point. I wish I had access to a scope...
Last edited:
Your link shows a rather massive and likely expensive assembly. In the proximity of hot heatsinks, wood is not a good idea for the housing or structural assembly either. If you shop around the aluminium sellers such as awning, window and caravan suppliers, you can often pick up affordable short lengths of aluminium in useful thicker sections like channels, angle and flat bar forms and these can be used to make the whole box and even the heatsinks when fastened with csk screws or even pop rivets if you're in a hurry to begin. I have found some nicely profiled and great looking sections over the years I've been tinkering with DIY builds. I've had to wait months for the right sections to come along though.
Regarding the reg. noise attentuation, there is also the noise of the reg. itself to consider and LDO type regs like LM338 aren't as quiet as some standard types. https://e2e.ti.com/blogs_/b/powerhouse/archive/2017/06/14/ldo-basics-noise-part-1
Regarding the reg. noise attentuation, there is also the noise of the reg. itself to consider and LDO type regs like LM338 aren't as quiet as some standard types. https://e2e.ti.com/blogs_/b/powerhouse/archive/2017/06/14/ldo-basics-noise-part-1
Sorry I should have clarified. I am not looking to copy the chassis design, only the specific amp and power supply circuit. I too have collected a supply of aluminium sheet and heavy gauge angle.
The internal noise issue I am trying to understand. It was my reason for looking at the LM723.
The tnt audio article:
Simple Voltage Regulators Part 1: Noise - [English]
Indicates that with proper adjustment bypass cap, noise is at around -80dB across the whole audio frquency range (they also indicate that in this case no output cap is required). Wouldn't this include the internal noise of the reg? Wouldn't this be considered relatively low noise at the voltages we are talking about?
The article was referring to LM317, I am assuming (perhaps wrongly) that the LM338 would behave the same.
The internal noise issue I am trying to understand. It was my reason for looking at the LM723.
The tnt audio article:
Simple Voltage Regulators Part 1: Noise - [English]
Indicates that with proper adjustment bypass cap, noise is at around -80dB across the whole audio frquency range (they also indicate that in this case no output cap is required). Wouldn't this include the internal noise of the reg? Wouldn't this be considered relatively low noise at the voltages we are talking about?
The article was referring to LM317, I am assuming (perhaps wrongly) that the LM338 would behave the same.
Noise can also be internal grounding, interconnection grounding and stability related. Maybe you could compare the output to the supply, considering the PSRR.
TNT's is a nice, easily understood explanation of regs for low power applications like the preamps discussed. For an amplifier PS though, were looking at operation approaching max. regulator ratings which brings in a different set of concerns.
I won't try to describe and compare the operation and noise of regulators as well as TI engineers because there are stacks of articles, tech forums, blogs etc. on linear and switchmode regulators at their site, following the link I posted, if you need one. The 317 and 338 are conventional regs with about 1.5V dropout but 338 is much larger, has better regulation, more current and slightly less residual noise. Read what you can there at the TI site, before taking decisions on what products you want to work with versus availability, price, form factor and any necessary support circuitry etc. It's most likely that when the heat is on, some compromises will be necessary anyway.
Note that the 138-338 series is obsolete now so you may want to look elsewhere rather than pay inflated prices for NOS or fake semis. Their noise specs aren't really all that different to 317-317K but their current ratings were certainly attractive.
On the other hand, using regulation to reduce noise to inaudible levels with your ear hard up to the tweeter is beyond necessary but if you still want it somewhere near that, you might consider both a more sophisticated amplifier and power supply designs. Followers of Jean Hiraga's "le monstre" designs, which have even worse PSRR, even use battery supplies in an effort to cut noise levels to something acceptable. However, I don't recommend anyone use wet or gel cells indoors, just for safety issues (acid vapour emission, hydrogen emission, corrosive electrolyte spillages and creep.) I trust that I don't need to explain why those qualities are considered dangerous.
My approach - build the amplifier first, power it with a simple filtered DC supply and measure what you have, listen to estimate whether noise is really much of a problem before building some complex regulation into the DC supply. At least you'll know what actual noise is being considered and how effective the remedies are.
I won't try to describe and compare the operation and noise of regulators as well as TI engineers because there are stacks of articles, tech forums, blogs etc. on linear and switchmode regulators at their site, following the link I posted, if you need one. The 317 and 338 are conventional regs with about 1.5V dropout but 338 is much larger, has better regulation, more current and slightly less residual noise. Read what you can there at the TI site, before taking decisions on what products you want to work with versus availability, price, form factor and any necessary support circuitry etc. It's most likely that when the heat is on, some compromises will be necessary anyway.
Note that the 138-338 series is obsolete now so you may want to look elsewhere rather than pay inflated prices for NOS or fake semis. Their noise specs aren't really all that different to 317-317K but their current ratings were certainly attractive.
On the other hand, using regulation to reduce noise to inaudible levels with your ear hard up to the tweeter is beyond necessary but if you still want it somewhere near that, you might consider both a more sophisticated amplifier and power supply designs. Followers of Jean Hiraga's "le monstre" designs, which have even worse PSRR, even use battery supplies in an effort to cut noise levels to something acceptable. However, I don't recommend anyone use wet or gel cells indoors, just for safety issues (acid vapour emission, hydrogen emission, corrosive electrolyte spillages and creep.) I trust that I don't need to explain why those qualities are considered dangerous.
My approach - build the amplifier first, power it with a simple filtered DC supply and measure what you have, listen to estimate whether noise is really much of a problem before building some complex regulation into the DC supply. At least you'll know what actual noise is being considered and how effective the remedies are.
Absolutely correct. Mild stability problems change the sound. A circuit unlikely to have that is a good starting point.
1 mV ripple noise is not a bad assumption. I was hoping for 100 uV. If I succeed I will draw the layout using tag strip.
Image you are trying to keep water clean in close proximity to dirty water. That's very much the situation. The water companies use water pressure to make it work. The main thing is to build up a strategy. Politicians think water loss is stupidity on the part of the water engineer. Maybe, but more likely the cheapest option for good health. If it is understood that this cross contamination of a power supply circuit is likely it says simplicity is the better weapon.
1 mV ripple noise is not a bad assumption. I was hoping for 100 uV. If I succeed I will draw the layout using tag strip.
Image you are trying to keep water clean in close proximity to dirty water. That's very much the situation. The water companies use water pressure to make it work. The main thing is to build up a strategy. Politicians think water loss is stupidity on the part of the water engineer. Maybe, but more likely the cheapest option for good health. If it is understood that this cross contamination of a power supply circuit is likely it says simplicity is the better weapon.
This is a concept of a simple PSU. As the fixing tab is the collector raw DC can be added there. The power to the 1 K resistor is available from this connection.
There are many improvements that can be made. The obvious one is to add a capacitor to the zener diode. This can be wrapped to the zener diode to maintain the Kirchoff 0V point ( green ).
Ths circuit might be the best sounding despite what people say. No harm in listening to it.
A primitive Darlington using BC327 and BC337 with 2SA1943 and 2SC5200 might be better. A primitive Darlington has no feedforward resistors to speed up the device. Equally these are low capacitance devices so might be OK. Not feeding forward ripple is no bad thing. This might be a big deal.
Zener diodes need tweeking a bit. The 1K also. It is possible for very little money to give the 1K it's own supply. The Creek CAS 4040 did this for it's preamp section. Just a sepparate rectifier and capacitor to the same transformer.
The Sziklai pair is often used as the dropout is reduced. It is slightly less trustworthy.
The primitive form of both is seen in the 1957 design of transistor amplifier. This was the first to be able to set a more ideal bias. H C Lin of Silver Springs Maryland it's designer seems not to be known to the internet! Daddy of Op amps. I used to visit Silver Springs when he was alive. I had no idea! He was honest enough to say he borrowed his friends idea.
1964: The First Widely-Used Analog Integrated Circuit is Introduced | The Silicon Engine | Computer History Museum
George Clifford Sziklai - Wikipedia
It differs from the capacitance multiplier by the included Zener diode. As the voltage rises, it works as a stabilizer. When reduced as a capacitance multiplier.
I dug up a 1965 circuit DIY with a floating quiescent current on Ge.
Class A 10 W. Current without input signal 20% of nominal. Phase splitter and darlingtons PNP GE. The current depends on the output signal.
http://forum.vegalab.ru/attachment.php?attachmentid=349000&d=1560765341
Class A 10 W. Current without input signal 20% of nominal. Phase splitter and darlingtons PNP GE. The current depends on the output signal.
http://forum.vegalab.ru/attachment.php?attachmentid=349000&d=1560765341
Last edited:
An interesting circuit. Germanium usually meant PNP. I notice primitive Darlington's and AC coupling from the voltage amplifier. If anyone needs to fix such an amplifier MOSFETs work well on lower bias. BUZ 900 or 905. In the shown circuit less likely. Leak or similar perhaps.
Usually at this point I would tweak and get a better result, I don't think I will this time. The 120R mostly can be tried. However just as OldDIY said it is a hybrid capacitance multiplier. I was able to go 207 to 253 VAC for 0.4VDC variation. The reasonable 1.2 mV rms noise is OK. Both side were identical and the residual noise is more sine wave than sawtooth. It hardly dropped out at 200 VAC and didn't drop below 18V.
Put fuses in. I would be tempted to make them 5AF. It worked first time. The layout is exactly as diagram which is OK but not as good as it can be. I wouldn't change it myself.
The transistors have internal protection diodes and plastic through the holes. Very helpful. CPC UK supplied them, £1.20 ish. The resistors are dummy amplifiers.
We most of us have a good audio memory. If you have a scrap speaker and 1 uF polyester capacitor you have a poormans oscilloscope for PSUs. The 1uF is conected to the speaker plus to protect it from DC. Even 1 mV is audible. If you get it down to 1/10 mV you will notice. This is just for ripple. If you do it for the 22 000 uF it will be loud. 1 uF is a wise choice and not the -3dB point.
The JLH will probably take the 1 mV of the PSU and make it well below 1/10 mV. JLH didn't bother with very fancy PSUs so I assume it is so.
The JLH will probably take the 1 mV of the PSU and make it well below 1/10 mV. JLH didn't bother with very fancy PSUs so I assume it is so.