It was mainly a different (better) layout which made the difference I think. Component values were unchanged from the earlier version and decoupling capacitors were of similar quality etc. I should mention that where the 15003 had been fine the 21194 exhibited a low level hum but still sounded very nice indeed. The hum disappeared the moment I put the 15003 back in. After the later rebuild which used 15003 initially it was the memory of the slightly superior sound with the 21194 that made me try it again, this time with success. If you do have to stick with the 15003 you won't be missing much as the difference is pretty marginal. Changing the input transistor can yield a huge difference, almost as big as between 2N3055 and MJ15003. I like the 2SA970 very much and haven't found anything nicer to my ears. Changing between this and a BC560 for example completely changes the sound of the amplifier. One is relaxing and open, and airy in the treble, the other fast and very detailed but somewhat fatiguing. Very different flavours!
Tim
Tim
Last edited:
On page 496 in post number 4955 I posted a long-tail pair circuit with uneven collector resistors. Between that time and now I have tried to simplify the circuit and to try some more common transistor types.
My aim was to come up with a version that used roughly half the power of the 1996 Class A amplifier in which the output devices run at 44W each and
that would not be a great deal hungry than the 1969 Class A.
At half 44A times 4 the transformer can be reduced to half that for the 1996 amplifier.
The result is attached - some observations on the final circuit are in order.
My simulations show that there is a snag with single transistor input circuits in which the FFT components decay with some of the odd order ones being greater in magnitude than the preceding even order ones.
I think this is a question of the input transistor not being able to correct for anomalies elsewhere in the circuit - since it possible to make the correction with an LTP input.
The bootstrap collector load approach used in the original 1969 seems to have a reputation as being better than a constant current source.
The approach is based on positive feedback from the output to the phase-split transistor collector this then is re-cycled to the output and thus through the nfb loop.
This may not matter if the odd harmonics over the even ones at some point in the FFT pattern depending on where they stick out - I just don't like the thought.
Another observation is that phase and gain stability margins can vary when transistor substitutions are made especially the phase split transistor. In the simulation attached to post 4955 I used a 2N5210 transistor which gave the best THD results but when it came to stability testing I had to incorporate a base stopper resistor and a phase correction network in the base to negative rail.
This gave rise to a kink in the Tian plot close to the critical phase frequency where the decay gradient increased from 6dB/Octave to 12dB/Octave.
With the method of compensation using a single capacitor much rides on the right value which may not be as per that simulated. Ideally that is on a portion of the plot nowhere near any elbow.
It is advisable to use a transistor with not too much gain for the phase split section which means avoiding such as 2N5210 and similar devices in this application (high gain might be useful in yours).
Happily humble transistors like 2N3904 and BC550 will work OK and the complimentary 2N3906 and BC560 partners will also work together.
I have used Bob Cordell's .models and included the link to his website so people can do their own comparisons between these and others in his .model selection.
Those I have chosen given the greatest stability margins however changing these is advisable if the voltage rails are increased beyond limits of the device working ranges.
Increasing the current is best done by varying the value of R14 - increasing this increases the standing current and generally improves THD.
The circuit works better without odd value collector resistors in the LTP.
This circuit operates in much the same way as the 1996 Class A which can deliver peak current of close to 4A into 8R.
The present simulation can deliver 2.2A into 8R.
Last edited:
Just to say. the 8K2/22K is to switch off the second transistor without switching off the DC servo action. As Naim found when 1K 22K the DC offset is not too bad. When viewed the harmonic stucture is as was said before Monotonic with a down slope. It will plateau at or near Ft. The Naim idea will have a slightly raised 3rd harmonic which gives the Naim's that fast sound. From what people say a tad of spice in the JLH would be OK. Someone might simulate it.
This will work if R4=8.2k;R5=1k and 22k is inserted between Q2 collector and the negative supply rail.
I have not done a simulation test for 19kHz and 20kHz combined IMD.
There is a slight improvement in THD at full power at 20kHz and the effect in the FFT screen is to cause odd order harmonics to be dominant.
Only the low order ones reach above -60 dB so it is a matter perhaps of which ones you prefer.
I have attached one amended simulation - feel free to do likewise with any of the others.
In a class A design I tend to take 10 kHz as where I become more relaxed about distortion. The cause is different. All the same if it comes at no cost elsewhere make it better.
I saw about a almost like new 1971 Stingray being put on the rolling road. 145 BHP and 169 BHP when new ( not much for 350 cu inch ). An electronic ignition fitted with no oth mods giving 195 BHP. That's the sort of thing we want for the JLH.To my eyes nothing like that seems to work when the JLH. Even the bootstrap is great because it has no crossover distortion added. I added a very similar bootstrap to a valve design. It was less than an active stage. It was almost impossible to say from graphs which was which. That one was a 15 Vrms bootsrap fed to a 26 Vrms output. The distortion improved by about a factor of 3.
I saw about a almost like new 1971 Stingray being put on the rolling road. 145 BHP and 169 BHP when new ( not much for 350 cu inch ). An electronic ignition fitted with no oth mods giving 195 BHP. That's the sort of thing we want for the JLH.To my eyes nothing like that seems to work when the JLH. Even the bootstrap is great because it has no crossover distortion added. I added a very similar bootstrap to a valve design. It was less than an active stage. It was almost impossible to say from graphs which was which. That one was a 15 Vrms bootsrap fed to a 26 Vrms output. The distortion improved by about a factor of 3.
Hi!
TimA.
Thanks to you I think I understand the character of the transistors used in this circuit, really your clarifying your intervention! Apart from that in recent days I tried to read the past posts of this forum, and among other things I read that Geoff advised against the use of mj21194 in this circuit, I am still convinced that you have framed the right alchemy of the BJT and I would like to try replacing the input, BC560, with 2sa970. The real problem now and I can not find these semiconductors, my usual suppliers say obsolete is not available and in the free market I think you can only find things fake .... is a big problem, I would also try with substitutes but reliable, but I do not know which of these can be, keeping in mind that today this type has been replaced by SMS elements, I realize that their supply is not so easy.
However, if someone could give me suggestions about it I would be grateful.
Greetings to all.
Mleod
TimA.
Thanks to you I think I understand the character of the transistors used in this circuit, really your clarifying your intervention! Apart from that in recent days I tried to read the past posts of this forum, and among other things I read that Geoff advised against the use of mj21194 in this circuit, I am still convinced that you have framed the right alchemy of the BJT and I would like to try replacing the input, BC560, with 2sa970. The real problem now and I can not find these semiconductors, my usual suppliers say obsolete is not available and in the free market I think you can only find things fake .... is a big problem, I would also try with substitutes but reliable, but I do not know which of these can be, keeping in mind that today this type has been replaced by SMS elements, I realize that their supply is not so easy.
However, if someone could give me suggestions about it I would be grateful.
Greetings to all.
Mleod
2SA1312 seems to be current, and datasheet seems almost identical to 2SA970, except the newer device is SMD. I've often thought a small general purpose adapter board is needed for SMD to leaded replacement, though I have soldered wire wrapping wire to SMD devices ... not recommended, really, but can work(with a wrap or two first).
Seems these devices are built with the newer linear gain technology. BC560 is old style, and in the JLH gain linearity is important for low distortion.
Seems these devices are built with the newer linear gain technology. BC560 is old style, and in the JLH gain linearity is important for low distortion.
The usual drop-in replacements for the obsolete 2SA970/2SC2240 are On Semi's KSA992/KSC1845. These are the commonly used, low noise, small-signal transistors featured in all the serious designs presented here on DiyAudio. They are currently in production and available from major distributors like Farnell, Digikey or Mouser.
Member
Joined 2009
Paid Member
Greetings to everyone!
TimA / Thank you for the suppliers suggested to me, but in your opinion, for the use in question, the 2sa970 or the 2sa970BL would be more suitable, I ask myself on the sound level if there can not be differences and what kind.
However, thanks also to the other forumers who suggested the additional BJTs to replace the 2sa970 {2SA1312, KSA992 / KSC18425, 2SA844}, but also in this case there is the same problem expressed above .... what will be the sonic character of these transistors compared to the cold bc560c.
I could have your impressions / experiences about it.
Thank you and greetings!
Mleod
TimA / Thank you for the suppliers suggested to me, but in your opinion, for the use in question, the 2sa970 or the 2sa970BL would be more suitable, I ask myself on the sound level if there can not be differences and what kind.
However, thanks also to the other forumers who suggested the additional BJTs to replace the 2sa970 {2SA1312, KSA992 / KSC18425, 2SA844}, but also in this case there is the same problem expressed above .... what will be the sonic character of these transistors compared to the cold bc560c.
I could have your impressions / experiences about it.
Thank you and greetings!
Mleod
The cold sound may be from hf peak that is not mitigated. The original 2n3906 is good.
I said if I had time I would post a Naim JLH hybrid. Like the Underground map this is to guide. It's about where I would start. On the DC coupled version I have also give a 3K resitor to set gain ( choice ). That is a value that should be high enough to drive with a preamp of the modern type. I also put a RF input filter on the DC version.
It's for conversation mainly. I would imagine more tail current to the longtail pair worth trying. I don't like the VAS cap of 0 to 100 pF. Best to have something then debate better ways of doing it later. Transistor are best of easy to get devices. AC circuit is typical of TDA2030. 100 uF BP is bipolar type that have low distortion.
You were probably much better off being in the electrical side.
Most places need electrics.
I studied industrial electronics and found it of little use as a qualification.
It was just a basic level of electronic understanding.
I was lucky in I got a job but just in repair, it did lead on to R&D and some software work. But very few companies wanted electronics engineers at city and guilds level, they mostly wanted degree level people.
We Nigel's have very similar lives!
My JLH concept.
I would play with the 180K for about a month. If I liked it with lets say 47K I would add a bit of current to the tail resistor. Measure the lets say 47K for voltage and by V/R = current, add that to the tail current. The tail hardy moves so within reason can be anything you like. Ask if not sure. I suspect given enough listening something as good as the JLH can be made which can do extra things and be easier to say if it's correct. I gave the AC coupled version as it should sound closest. The 180K is unlikely to be right as the Naim example runs at 40V and if I remember sits at an odd voltage. As longtail pairs are so forgiving one can do virtually anything you like. A sign of the 180K being too large is if DC offset is > 100 mV I suspect it will be <50 mA. To be honest 100 mV is fine. Less than 1/10 of a watt at the speaker.
Then play with capacitor values. If you use 10K input resistor on the DC coupled version and 220R 2K7 the true virtues of 2SA970 will be heard. You can not do all these things on the real JLH as the 2K7 feedback resistor sets the DC conditions. On the LTP version it doesn't so greatly. Talking about transistors should also included capacitor values. Again take time to listen. 1 uF input with 100 uF PB and 3 K should be better than JLH. When people say they hear this and that I am convinced it's not THD. I retained the bootstrap as I really like them. You can even bootstrap the input, the JLH is an amplifier that might do that very well. Again to exploit 2SA970 better. Being class A it should not inject high order distortion. No maths needed as the logical point is at unity gain with the input and in phase. One would use 1K+ 1K and a bootstrap capacitor and limiting resistor. Your preamp even if valve anode follower might drive it. They are the nice ones using 75R TV cable to connect ( 300 mm = 20 pF as Leak Varislope ).
My JLH concept.
I would play with the 180K for about a month. If I liked it with lets say 47K I would add a bit of current to the tail resistor. Measure the lets say 47K for voltage and by V/R = current, add that to the tail current. The tail hardy moves so within reason can be anything you like. Ask if not sure. I suspect given enough listening something as good as the JLH can be made which can do extra things and be easier to say if it's correct. I gave the AC coupled version as it should sound closest. The 180K is unlikely to be right as the Naim example runs at 40V and if I remember sits at an odd voltage. As longtail pairs are so forgiving one can do virtually anything you like. A sign of the 180K being too large is if DC offset is > 100 mV I suspect it will be <50 mA. To be honest 100 mV is fine. Less than 1/10 of a watt at the speaker.
Then play with capacitor values. If you use 10K input resistor on the DC coupled version and 220R 2K7 the true virtues of 2SA970 will be heard. You can not do all these things on the real JLH as the 2K7 feedback resistor sets the DC conditions. On the LTP version it doesn't so greatly. Talking about transistors should also included capacitor values. Again take time to listen. 1 uF input with 100 uF PB and 3 K should be better than JLH. When people say they hear this and that I am convinced it's not THD. I retained the bootstrap as I really like them. You can even bootstrap the input, the JLH is an amplifier that might do that very well. Again to exploit 2SA970 better. Being class A it should not inject high order distortion. No maths needed as the logical point is at unity gain with the input and in phase. One would use 1K+ 1K and a bootstrap capacitor and limiting resistor. Your preamp even if valve anode follower might drive it. They are the nice ones using 75R TV cable to connect ( 300 mm = 20 pF as Leak Varislope ).