As we say in New Zealand when faced with a highly questionable statement "yeah right".
Hello,
just discovered this therad and read it fully in 2 days. As i'm very intrigued by Naim legend and their specific sound i want to give it a try etching some boards and using sone Onsemi samples that i have on hand.
So : what do you think about using some MJL4281 for power output and MJE15034/15035 as drivers?
I remember some highly regarded transistor for VAS i used for Fc100 project finalised some time ago: 2SA1360/2SC3423. What is your opinion about using them in this project ?
Thanks!
just discovered this therad and read it fully in 2 days. As i'm very intrigued by Naim legend and their specific sound i want to give it a try etching some boards and using sone Onsemi samples that i have on hand.
So : what do you think about using some MJL4281 for power output and MJE15034/15035 as drivers?
I remember some highly regarded transistor for VAS i used for Fc100 project finalised some time ago: 2SA1360/2SC3423. What is your opinion about using them in this project ?
Thanks!
Hi, you may be making a mistake in substituting modern, high-quality parts for what was a design from around 45 years ago. It's true that Naim upgraded the output transistors to Sanken LAPTs and their own coded NA007 types over the years but other part specifications remained much as they always were. You can see some variations here: Index of /t/naim/poweramp_pix
It's great to experiment with old designs but you really will be just repeating what has already been tried and reported in several forums. The Forum used by many Naim freaks is here:
Help needed replacing caps in Naim NAP 120. With Pics! - pink fish media
MJLX281 series parts are quite similar to 2SC5200 which are still supplied in H140 kits. Other clone kits kits use Sanken 2SC3585 but none of these substantially improve sound quality. MJE150XX drivers will also work but with little benefit. Otherwise, might I urge you to use the original ZTX653/753 parts from Diodes inc. as the voltage amplifier and current source. These do make a difference, as said many times in the thread.
The other small signal types may be covered by BC550C/560C low noise types for the input pair and 550B/560B elsewhere. There is no point going through a list of rare, obsolete and unnecessary components on the assumption that these will improve anything. Use what works first and play around later if you wish
Last edited:
I agree, this is a good way to go about analyzing the circuit - see post 1426.
Theory says the current in each half of an LTP will be the same if the tail supplies the right level of current. These measurements support this supposition. In the early Nait amplifiers the resistor in the constant current source was 560R so achieving balance was a matter of changing one resistor.
Some other observations.
The relationship between transistor collector current and base current deteriorates as the latter increases so the objective is to keep base current low. If a curve is representative of the relationship smaller segments of a curve approximate more closely to a straight line.
0.5 mA through TR1 collector resistor of 1k drops 0.5 of a volt under no signal conditions. The base current will be 0.5mA divided by the current gain of TR1 which will be miniscule.
The situation with TR2 appears a little different the 22 k collector resistor drops around 11 volts for the same 0.5 mA of current. The base current for TR2 will be similarly small as in the case of TR1.
TR2 output is not committed to drive any circuit element however the structure represents an inverting amplifier having very high voltage gain.
There will be parasitic capacitance between the collector and base junction and feedback via this capacitor will have the effect of improving TR2 linearity while reducing the gain with increasing frequency which is also a stability requirement for the main negative feedback loop.
The Miller capacitor between collector and base of the voltage amplification transistor ensuring stability is 39pF low enough not to cause any charging problems due to the current limit of the LTP current source. There is a buffered lead capacitor in the main loop feedback path. Somehow these two obvious measures don't seem enough.
You forget this was someone I knew quite well. I thought Julian was very knowledgable until the 30 feet statement. Having recieve the letter from Mr Mornington West it's clear he was the brains. It might be like the VW Beetle, Dr P designed it whilst someone famous had the vision. Music is far more worthwhile. For my sins I have a VW.
The 22K (very slightly) hinders loop stability by adding another pole, thereby moving away from the dominant pole ideal.
If you read back in the thread, the general consensus is that the 22K is there to add 2HD to the signal. The reason for this is to mask the higher order distortions.
IMO this is readily apparent on auditioning - listen to the decay of piano, bell/glockenspiel sounds and female vocals like Alison Kraus. The "purity" of the notes is dirtied a tiny bit. OTOH brass and electric guitar etc. sound really vibrant. The 22K resistor is effectively the "fruitiness" control.
If you mean that the Miller compensation capacitance seems low, the Cob of the original voltage amplifier and CS transistors is quite high at 30 pF. IME, If low capacitance transistors such as supplied in the kits presently available are substituted, the miller cap needs to be upped to 60pF or more to get it sounding right - or at least to restore that familiar Naim-like sound.
I haven't simulated this due to suspect models and doubts over the Chinese substitute parts marked as 2SB647/2SD667 with Hitachi look-alike trademarks. Obviously not genuine and Ccb is just 3pF but I don't know much more.
it seems to me that 30pF fixed cap + 30pF of Cob (variable cap) does not equal 60pF fixed cap + 3pF of Cob
even adopting 57pF for the fixed still cannot make the 3pF of variable Cob perform as a 30pF Cob.
The Naim amplifier performance is dominated by the choice of loads and currents and devices such that the output is not a scaled copy of the input.
If one wants to build a clone of a Naim, then all the devices and all the currents and all the loads should be duplicated as closely as possible.
even adopting 57pF for the fixed still cannot make the 3pF of variable Cob perform as a 30pF Cob.
The Naim amplifier performance is dominated by the choice of loads and currents and devices such that the output is not a scaled copy of the input.
If one wants to build a clone of a Naim, then all the devices and all the currents and all the loads should be duplicated as closely as possible.
Douglas Self likes Cdom to be large compared with Cob when the VAS( TIS ). His arguement is long and very interesting. If I read him correctly he sees Cob as a low grade capacitor with a bad reference point. The external Cdom will swamp the internal ( good ). NPO/COG capacitors are excellent for this job. I often use SMD types. After years of silly ideas I now use my finger nail to hold them down when soldering. If it's too hot for me it's too hot for them. I only use 60/40 these days as solder. As I can buy it why not?
Hi Andrew. Cob capacitance specifications do not necessarily equate to that in the fixed Miller capacitor. That fixed cap is a standard 39 pF value actually - see schematics posted several times in the thread. Whilst this reduces the apparent difference, the estimation of transistor Cob is carried out at a fixed 10VDC bias which we don't have in either transistor here. This capacitance arises much as in a varicap diode so the voltage dependence is significant.
However, I take the point you make about duplicating as much of the circuit componentry and operating conditions as possible if we hope to make reasonable clones. I think that few people now posting here to discuss their builds, have set out just to make clones though. Most want either a cheap basic amplifier kit to kick off something of their own invention or attempt to improve on the outdated NAP design.
I have been lucky enough to own an original 140 for many years and can retrieve it from being on loan when closer comparisons are needed. As it turns out, it's not too difficult to get very close results if you keep the case, PCB and cooling arrangement (that's the 3mm aluminium case) about the same, as there isn't a lot of difference otherwise, so long as key component types are more or less the same.
Of course it works with a separate heatsink too but the bias regulator/ VBE multiplier or whatever sits in the middle of the board but this may take forever to respond - seeming stable but drifting very slowly, particularly when mounted low in a ventilated case as DIY builds often are.
However, I take the point you make about duplicating as much of the circuit componentry and operating conditions as possible if we hope to make reasonable clones. I think that few people now posting here to discuss their builds, have set out just to make clones though. Most want either a cheap basic amplifier kit to kick off something of their own invention or attempt to improve on the outdated NAP design.
I have been lucky enough to own an original 140 for many years and can retrieve it from being on loan when closer comparisons are needed. As it turns out, it's not too difficult to get very close results if you keep the case, PCB and cooling arrangement (that's the 3mm aluminium case) about the same, as there isn't a lot of difference otherwise, so long as key component types are more or less the same.
Of course it works with a separate heatsink too but the bias regulator/ VBE multiplier or whatever sits in the middle of the board but this may take forever to respond - seeming stable but drifting very slowly, particularly when mounted low in a ventilated case as DIY builds often are.
I think only the Self books get close to explaining the Naim design as a bi-product of his own philisophy. Some here seem close to going that way so I believe it is their best route to follow. As far as I know Self thinks Cob is a vital consideration. He goes a very complicated route to saying make it large, then make it smaller if you can ( I doubt he would agree ). Very wise advice. What he seems unwilling to discuss is what some call TIS. He says if we look at the input voltage waveform to the VAS it will seem very distorted. He rightly says it can be regarded as a measuring mistake. However I have always had doubts about that. When very high gain transistors are used as the VAS we reduce that ghost ( use a cascode, again see Self, interesting to debate the Tr1 collector z out and Early effect. Could be worse than TIS effects ). I suspect all the slewing nonsense is not understanding transconductance is not an easy state of being. Self is not a slewing nonsense person before anyone tells me. Apart from that, be it a duck or a goose it looks much the same to me.
Here is a point worth making. I designed an amp for a friend. She is thinking of improving it by using NPN and PNP symetrical circuits. She has a transistor curve tracer so as to spot fakes ( it costs about £100 she says, no idea they existed ). She has 1000 fake 2SA970 she says. I told her to very carefully log her curves to be sure she gets the same results each time. In my view the double VAS Hitachi design will do all of that and is understandable from Naim and Self. Matching NPN / PNP stages may only work well when the NPN and PNP do not match ( what I suspect ). Therefore a complicated way of doing a simple thing. Although the Hitachi is very simple it ticks many boxes. Even it's resistor to the long tail pair input is no disaster. The Naim current source is not better when the real problems start > 20 khz. The voltage clamping of the input pair is to 2% without a current mirror and no fancy parts. The drive is symetrical in source and sink ( 35V / uS in my version ). It needs a small tweek in the dead side of the VAS to get good > 20 kHz balance. Self suggests it is a poor piece of engineering. The distortion curves say very differently. I have seen the circuit adapted to bipolar transistors. Goldmund and H-H used versions with MOSFET's.
Julian Vereker asked me if I could get MOSFET's in about 1978. He seemed to think at the time they would be better. I think they are and Julian by then had lost Mr West to help him. If he had persisted I feel he would have used them. I am sure the Naim clones would be fine with MOS FET's . Many parts could be junked. You would be on your own if you do. Make Cdom large at first. Bias is usually about 100 mA ( about 1.2 V if Exicon FET's gate to gate ). The bias is a resistor as the FET's are not on off devices. They will not thermally run away. I built some of Self's ideal bias devices. They seem to drift as badly as usual designs. I don't doubt he got them to work. I just doubt that I can. Self talks of Gm doubling. If it is at 1 watt output I doubt it matters and will sound much like class A below 1 watt ( or better ). I would call Self as critical underbiasing. It seems near impossible to fix. I measured a Hypex class D amp after I spent a week designing a way to do it. Below 1 watt was impressive.
Here is a point worth making. I designed an amp for a friend. She is thinking of improving it by using NPN and PNP symetrical circuits. She has a transistor curve tracer so as to spot fakes ( it costs about £100 she says, no idea they existed ). She has 1000 fake 2SA970 she says. I told her to very carefully log her curves to be sure she gets the same results each time. In my view the double VAS Hitachi design will do all of that and is understandable from Naim and Self. Matching NPN / PNP stages may only work well when the NPN and PNP do not match ( what I suspect ). Therefore a complicated way of doing a simple thing. Although the Hitachi is very simple it ticks many boxes. Even it's resistor to the long tail pair input is no disaster. The Naim current source is not better when the real problems start > 20 khz. The voltage clamping of the input pair is to 2% without a current mirror and no fancy parts. The drive is symetrical in source and sink ( 35V / uS in my version ). It needs a small tweek in the dead side of the VAS to get good > 20 kHz balance. Self suggests it is a poor piece of engineering. The distortion curves say very differently. I have seen the circuit adapted to bipolar transistors. Goldmund and H-H used versions with MOSFET's.
Julian Vereker asked me if I could get MOSFET's in about 1978. He seemed to think at the time they would be better. I think they are and Julian by then had lost Mr West to help him. If he had persisted I feel he would have used them. I am sure the Naim clones would be fine with MOS FET's . Many parts could be junked. You would be on your own if you do. Make Cdom large at first. Bias is usually about 100 mA ( about 1.2 V if Exicon FET's gate to gate ). The bias is a resistor as the FET's are not on off devices. They will not thermally run away. I built some of Self's ideal bias devices. They seem to drift as badly as usual designs. I don't doubt he got them to work. I just doubt that I can. Self talks of Gm doubling. If it is at 1 watt output I doubt it matters and will sound much like class A below 1 watt ( or better ). I would call Self as critical underbiasing. It seems near impossible to fix. I measured a Hypex class D amp after I spent a week designing a way to do it. Below 1 watt was impressive.
re my above post #1471:
For anyone interested in some of the more unusual parameters of transistors mentioned here and elsewhere on the forum, this paper makes a brief and easy to follow primer on testing some parameters including Cob, that may be required to understand how they operate. Sure, use a simulator and learn very little but do learn something helpful to at least know how things work and where to look when they don't: http://www.camsemi.com/pdfs-technical/CamSemi_AN-2576_HV_BJT_production_inspection.pdf
For anyone interested in some of the more unusual parameters of transistors mentioned here and elsewhere on the forum, this paper makes a brief and easy to follow primer on testing some parameters including Cob, that may be required to understand how they operate. Sure, use a simulator and learn very little but do learn something helpful to at least know how things work and where to look when they don't: http://www.camsemi.com/pdfs-technical/CamSemi_AN-2576_HV_BJT_production_inspection.pdf
Last edited:
Post 1426 again for the circuit details.
I had read that and wanted to see a deeper level of discussion about the causes that the general consensus rests on.
With regard to your observations - considered separately each half of an LTP is a single ended amplifier - distortions therein will be even order. If seen in equal proportion on both sides of a "balanced" LTP these will cancel each other through push-pull action of the stage and be absent in the output.
I don't see the Naim variation on this theme as obviating the cancellation of even order distortions necessarily. I don't see second harmonic as something that is deliberately added into a subtraction process.
How this changes the equations and leaves the (observed) second harmonics more prominent after the LTP distortion reduction process has not been elaborated out and the possibility of other mechanisms being involved given much thought.
As I see it, if TR2 can be made into a more linear amplifier than TR1 the corrective focus would be on correcting errors due to the power stages and the load and less on second harmonics due to TR2. This would allow a more direct comparative process of the input and output signals and do a better corrective job.
My thinking about TR2 is the collector to base capacitance will be considerably enlarged due to the increased voltage gain due to the 22k collector load - the Miller effect.
The reduction in gain at high frequency could well act as a goal keeper to feedback loop in a position where it assists gain reduction at high frequency - increasing the phase margin and filter out high frequencies that might inter-modulate in the audible frequency range.
In football terms the voltage amplifier is more like a midfield player having an interception job with bigger and slower players (power stages) unable to defend the goal mouth.
The questions here are whether or not the TR2 "synthesized" capacitor can be adequately charged and discharged through the feedback loop and what other effects may arise out of this. There is also the buffered lead capacitor in the feedback path. I have not thought too much about that yet as I will be away from home for several days without internet access.
In recent years there has been much interest in two pole filter compensation around the VAS in amplifiers in the interests of increasing the feedback factor at high frequency without compromising stability margins - a midfield goal keeper perhaps.
With regard to Miller effect, the diffusion capacitance between the collector and base junction of a transistor is inversely proportional to the square root of the voltage applied to the collector so there will be an imbalance between the amounts for TR1 and TR2.
Again these are just observations and I will keep my mind open to comments and other ideas.
Are you observing or speculating?
My amp is an NCC200. The latest version does not use the 47pF lead cap in the feedback path. So it's just straightforward dominant pole compensation.
When I shorted the 22K TR2 collector resistor I observed marginally improved transient response behaviour - not enough difference to indicate that compensation is the reason for the resistor being there.
I suppose I ought to measure the distortion profile with and without the 22K, it would be interesting to quantify its effect.
Not sure about your football analogies!
My amp is an NCC200. The latest version does not use the 47pF lead cap in the feedback path. So it's just straightforward dominant pole compensation.
When I shorted the 22K TR2 collector resistor I observed marginally improved transient response behaviour - not enough difference to indicate that compensation is the reason for the resistor being there.
I suppose I ought to measure the distortion profile with and without the 22K, it would be interesting to quantify its effect.
Not sure about your football analogies!
You could add 1K and a diode to simulate balance ( 1K047 + 1N4148 forward biased ). The diode could be the same type of VAS transistor converted to a diode as in a current mirror if wanting the best chance of the intended result. The beauty of this is fitted and removed in minuntes if placed across the 22K. I would imgine the amp would sound more like something from Douglas Self if so.
The Signal Transfer Company
The Signal Transfer Company
To be very clear. Most simple valve designs and most single input transistor amps produce what Jean Hiraga might say is less distortion even if the THD looks high. That is zero distotion compared with an exponential decay curve. It is supposed a human can not hear 1% THD if this way. As a turntable at best will be that a Quad 303 and ESL will realise the dream. The longtail pair input is a nasty thing. It makes the DC conditions very stable. The trade off is very unfortunate harmonics. It is assumed that if this is reduced to near zero we win back the problem we caused. I suspect this is right. A friend suspects very low noise helps also. He conjecture is transitor noise is white to blue. Pink would be better. If reduced to near zero my friend think the sound becomes warmer.
My simple spultion is Naim found the sound of the excellent " Mr West's" amplifer was a bit dry compared with real music and other amplifiers. My furthur specution is Mr W solderered in some posts and offered Julian some resistors. Juilian chose the 22 K by ear. Julian said ( perhaps to me ) that he tape recoreded things only to find when going home it sounded very different. The tape recorder might have been a Chilton where he was engineer. Julian would have been as far from Mr Self as it is possible to go. Mr West or whoever would have set safe limits. The sound would be Julians choice and no other. The ear sets the exponential decay comparrison, it's that simple. There is nothing stopping someone re-running history. Remember this kit amp is nothing like a real Naim amp so why shouldn't you ? When I say nothing it is closer than lets say a Denon PMA 250 that was Denons take on a UK amp.
My simple spultion is Naim found the sound of the excellent " Mr West's" amplifer was a bit dry compared with real music and other amplifiers. My furthur specution is Mr W solderered in some posts and offered Julian some resistors. Juilian chose the 22 K by ear. Julian said ( perhaps to me ) that he tape recoreded things only to find when going home it sounded very different. The tape recorder might have been a Chilton where he was engineer. Julian would have been as far from Mr Self as it is possible to go. Mr West or whoever would have set safe limits. The sound would be Julians choice and no other. The ear sets the exponential decay comparrison, it's that simple. There is nothing stopping someone re-running history. Remember this kit amp is nothing like a real Naim amp so why shouldn't you ? When I say nothing it is closer than lets say a Denon PMA 250 that was Denons take on a UK amp.