It's interesting. When building preamps with RIAA to me bandwidth really matters. We can't hear that bandwidth in theory. It must be wave shape or leading edge. Class D sounds brighter than it's bandwidth. I specualte the small 400 kHz ( typical ) output is changing the wave shape. Perhaps slight ringing is taken to be brighter. Or the choke is giving an output kick. It's said slightly unstable amplifiers sound better. That seems to say the tests are not realistic. The ringing of the output choke often taken to be a fault. 10R 3 watt to damp it can work.
What Carver said effectively is some designers have better ears, the very thing hi fi engineers avoid talking about. The Elephant in the room that no one would question if a musical instrument. It's like dancing, you can or you can't. I can't.
What Carver said effectively is some designers have better ears, the very thing hi fi engineers avoid talking about. The Elephant in the room that no one would question if a musical instrument. It's like dancing, you can or you can't. I can't.
Member
Joined 2009
Paid Member
Speakers are weird. Whereas an amplifier might be 1dB down at 20 kHz and be considered poor a speaker might be +/- 3 db and considered excellent. That might be further compromised by low damping factor amplifiers that make the response related to the speaker impedance curve and then the way placement changes things. +/- 6 db for a speaker and 3 % THD might be as good as it gets. Poor old JLH often at 0.03% THD is said to be not bad.
Often what I would get 10/10 for at college would get 4/10 in hi fi. What would get 9/10 in hi fi ( 10 never happens even if perfect ) might only get 4/10 at college because there are extra circuits that in subtle ways make it worse. I just designed something like that. I dare say I will do it the wrong way. I am now working out how to make wrong extra circuits mostly harmless. 1% capacitors will help. What college would ask is which thing matters most of a long list? Usually what hi fi has absolutely no interest in is very important. When did you ever see a current waveform for distortion?
Often what I would get 10/10 for at college would get 4/10 in hi fi. What would get 9/10 in hi fi ( 10 never happens even if perfect ) might only get 4/10 at college because there are extra circuits that in subtle ways make it worse. I just designed something like that. I dare say I will do it the wrong way. I am now working out how to make wrong extra circuits mostly harmless. 1% capacitors will help. What college would ask is which thing matters most of a long list? Usually what hi fi has absolutely no interest in is very important. When did you ever see a current waveform for distortion?
Member
Joined 2009
Paid Member
college ? well I don't remember Leeds University having anything to say about my hi-if. But I do remember the Dorm, every Sunday each person wheels out their speakers to just outside their room door, tunes in to the BBC Top 40, settles down in the common area outside the room doors with a hot pizza and cranks the volume to off-the-scale. I don't think THD was that much relevant.
At college it was well known I would use any excuse to discuss hi fi. The guy who taught us was mad keen on TV so it was easy for him adapt a subject to suit either. It might surprise people to know although in the 1970's mostly we still did valves. I also did electrical engineering, it wasn't my first choice. Now it earns my money. 80% of electrical engineering is of use in electronics. Valves or transistors didn't really matter much. I still have no problem thinking that way. The biasing is slightly and usefully different. JFET's and valves make two terminal constant current devices.
I wish we had done CMOS logic. It was just starting to be discussed. Some still is used. A CMOS in audio thread would be interesing. Eg. How to get the CD4066 to work for us rather than against us like in a shunt circuit and how to hard switch it off or bootstrap modulate it's gate to reduce distortion and the 74HC4066 versions.
I wish we had done CMOS logic. It was just starting to be discussed. Some still is used. A CMOS in audio thread would be interesing. Eg. How to get the CD4066 to work for us rather than against us like in a shunt circuit and how to hard switch it off or bootstrap modulate it's gate to reduce distortion and the 74HC4066 versions.
But so do lm317, 337, 338, 350, 7805, 7905 etc and dep. mosfets etc.
Lm317 being usually cheaper than the rest and adequate in 90% of all cases.
Quad use the 4066 (using a + & - 7.5v supply) in their 34 preamp. Lots of them and I shuddered to think what they sounded like until I bought one (& a 405-2 amp) and was very pleasantly surprised.
I've never used a standard MOS FET that way. It becomes a diode? I like the LD1084/85 as they do all of the functions of the LM317 with the same noise, lower loss, higher current when required. I suspect a LM317 could just about work in a JLH. TNT audio tell more.
This is exactly why Kirchoff and Gause were taught in electronics. Kirchoff very badly stated says the sum of the current entering a nodal point is equal to the sum of the current exiting that node. Where Kirchoff goes wrong is when things get complex. Gause stepped in and said one can say the current paths could be seen to be like entering a town with a ring road. We can then use Kirchoff to solve the problem as what is going on in the town is not our concern. Thus a LM317 is a Guassian 2 terminal device. It is said a LM317 is mostly a uA741 and pass transistor. A NE5534 and 3055 would be better and slightly faster. Negative feedback looks to make a Gaussian solution in a way. LM317 can be made into amplifiers.
This is exactly why Kirchoff and Gause were taught in electronics. Kirchoff very badly stated says the sum of the current entering a nodal point is equal to the sum of the current exiting that node. Where Kirchoff goes wrong is when things get complex. Gause stepped in and said one can say the current paths could be seen to be like entering a town with a ring road. We can then use Kirchoff to solve the problem as what is going on in the town is not our concern. Thus a LM317 is a Guassian 2 terminal device. It is said a LM317 is mostly a uA741 and pass transistor. A NE5534 and 3055 would be better and slightly faster. Negative feedback looks to make a Gaussian solution in a way. LM317 can be made into amplifiers.
Douglas Self shows if the load is 47K or above the distortion can be very low. Not as low as he would like. He shows an inverting stage of 22K + CMOS and 22K to the virtual earth point ( OUT to - input, + to 0V often ). Here the distortion is very low as the PN junction is 50 ohms or lower + 22Kohms. 4066 is better than 4016. The 34 is very nice. The MC stage is sort of the same idea as a CMOS switch in bipolar. That looks even more unlikely to work. Renardson do a very JLH type design .
MC Phono Preamp
Passing a signal thru a 317 ... I don't know that would be the "no-go" 10% use-case for me. But i know there were designs with 317's in there as ccs or even as an amplifying element ... so what do I know.
As a ccs in a psu for example, I wouldn't even blink to drop a 317 there... Cheaper, more predictable, less fuss than a jfet... just one use-case f.e.
As a ccs in a psu for example, I wouldn't even blink to drop a 317 there... Cheaper, more predictable, less fuss than a jfet... just one use-case f.e.
In valve circuits a LM317 works very well. If bootstraped by a capacitor it can be said they almost are not there. Ironically the valve can do this itself. I found a BD139 with 3 diode bias equalled a LM317 for linearity. Often valve people will only use MOS FET's if not valves. This is crazy as bipolar devices are super linear as current amplifiers. Never overlook MJE350/340 in valve circuits, 6 MHz is OK.
Greetings to everyone!
I have deepened my experiences of replacing BJT output. I completely replaced, on the two channels, the mj15003 with the most modern mj21194. The result was confirmed disappointing, the sound became unkempt, with a clear predominance of high tones and a decrease in bass, a decrease in the harmony and warmth that characterized the previous transistors, even if the notes seem clearer and more analytical, l - listening is fatiguing, whereas before the music never tired, now after a little it pushes to constantly change the song. Definitely a step of the shrimp .... disappointing in a word. I am surprised that all this happens without making any changes to the circuit, is the first time I do such experiments on transistors, I had previously encountered substantial differences with the tubes, but I did not expect a similar effect with the BJT. Someone would know how to explain the reason for such results. Thank you. Greetings to everyone!
Mleod
I have deepened my experiences of replacing BJT output. I completely replaced, on the two channels, the mj15003 with the most modern mj21194. The result was confirmed disappointing, the sound became unkempt, with a clear predominance of high tones and a decrease in bass, a decrease in the harmony and warmth that characterized the previous transistors, even if the notes seem clearer and more analytical, l - listening is fatiguing, whereas before the music never tired, now after a little it pushes to constantly change the song. Definitely a step of the shrimp .... disappointing in a word. I am surprised that all this happens without making any changes to the circuit, is the first time I do such experiments on transistors, I had previously encountered substantial differences with the tubes, but I did not expect a similar effect with the BJT. Someone would know how to explain the reason for such results. Thank you. Greetings to everyone!
Mleod
The design will be unstable if you use "modern" transistors without adjusting the feedback loop.
Read through the earlier parts of the thread: the answers are all there.
many thank's for this really good idea !!
I had not returned to the beginning of the thread for 4 or 5 years and it really made me feel good.
by the way and to synthetize, you have to know the most essential essentials of the jlh69 to get the quintessence and that's probably why I always come back to the simplest version of this schematic with well selected components.
If I get time I will finish the idea of a Naim JLH hybrid for discussion. The problem with the JLH 69 is you need a very advanced understanding of electronics to understand what you see and what you hear. A nice book by Douglas Self or Ben Duncan won't help you much.
Self in a few words says don't bother. His dislike seems to suggest the JLH and others are what my brother called a Dogs breakfast. That is it's made by experiment and can not be simply explained as to what the thinking that made it was.
The idea of the Naim NAP140-JLH69 hybrid is to make 0V as the centre output. One then has the choice as in a LM1875/ TD2040 to be AC or DC coupled. Already we get to a point many would prefer. Following from this the setting of the standing current is far more predictable. That is because regardless of what you do zero is zero. The amplifier has a simpler DC servo loop as a result.
The next advantage. The AC feedback loop can be anything you want with the hybrid. This means a nicer quality of capacitor can be used. One can feed the signal in here if you prefer. This is voltage shunt feedback type. In mathamatical terms it's the prefered type in that gain is -G= A/B. This has a small advantage over G= 1 + ( A/B ) of the standard input. -G means it inverts which is no big deal. In fact it can simplify other things. 1 + ( A/B ) is not ideal. Often a tip down filter is added to RIAA stages to offset it's mathamatical compromise ( 2 uS ). MC stages suit inverting very well. One of the more facinating things with the inverting input is near enough nothing can be seen at the base of that transistor and yet the collector has output. Although I know this it is close to magic.
Naim type input means it almost can minic the JLH single transistor input. Why I didn't do this is I have no time to devellop the idea. Some of it will be wrong which I already know. Which compromise to use is a mystery. The first problem is can I use 8K2 and 180 K as the long tail pair collectors ( 8K2 as the JLH ). 15/180K = 80uA ( circa ).That's only i max. My guess is it would be 40 uA if the collectors sits at half voltage or whatever. The NPN driver ( before outputs ) sits at 1.23V/ 8200= 150 uA. If we guess the tail to be 190 uA at perhaps 15V 22K + 56 K tail resistors with 100 uF to the split. When I write this down it looks OK. There is a small chance it could be better than the JLH beacuse it will adapt better.
https://www.sanfoundry.com/wp-conte...mcqs-voltage-shunt-feedback-amplifier-q14.png
Self in a few words says don't bother. His dislike seems to suggest the JLH and others are what my brother called a Dogs breakfast. That is it's made by experiment and can not be simply explained as to what the thinking that made it was.
The idea of the Naim NAP140-JLH69 hybrid is to make 0V as the centre output. One then has the choice as in a LM1875/ TD2040 to be AC or DC coupled. Already we get to a point many would prefer. Following from this the setting of the standing current is far more predictable. That is because regardless of what you do zero is zero. The amplifier has a simpler DC servo loop as a result.
The next advantage. The AC feedback loop can be anything you want with the hybrid. This means a nicer quality of capacitor can be used. One can feed the signal in here if you prefer. This is voltage shunt feedback type. In mathamatical terms it's the prefered type in that gain is -G= A/B. This has a small advantage over G= 1 + ( A/B ) of the standard input. -G means it inverts which is no big deal. In fact it can simplify other things. 1 + ( A/B ) is not ideal. Often a tip down filter is added to RIAA stages to offset it's mathamatical compromise ( 2 uS ). MC stages suit inverting very well. One of the more facinating things with the inverting input is near enough nothing can be seen at the base of that transistor and yet the collector has output. Although I know this it is close to magic.
Naim type input means it almost can minic the JLH single transistor input. Why I didn't do this is I have no time to devellop the idea. Some of it will be wrong which I already know. Which compromise to use is a mystery. The first problem is can I use 8K2 and 180 K as the long tail pair collectors ( 8K2 as the JLH ). 15/180K = 80uA ( circa ).That's only i max. My guess is it would be 40 uA if the collectors sits at half voltage or whatever. The NPN driver ( before outputs ) sits at 1.23V/ 8200= 150 uA. If we guess the tail to be 190 uA at perhaps 15V 22K + 56 K tail resistors with 100 uF to the split. When I write this down it looks OK. There is a small chance it could be better than the JLH beacuse it will adapt better.
https://www.sanfoundry.com/wp-conte...mcqs-voltage-shunt-feedback-amplifier-q14.png
Hi!
thanks for the answers, I'll go to reread the past tread.
However I would like to ask TimA, if you remember, what were the practical interventions that have led to rectify the situation ... lay out typology, component replacement, change of the values of negative feed back ... this to understand if it is worth to continue the implementation or pragmatically take a step back and eventually look at another project.
Thank you. Greetings to all.
Mleod
thanks for the answers, I'll go to reread the past tread.
However I would like to ask TimA, if you remember, what were the practical interventions that have led to rectify the situation ... lay out typology, component replacement, change of the values of negative feed back ... this to understand if it is worth to continue the implementation or pragmatically take a step back and eventually look at another project.
Thank you. Greetings to all.
Mleod