The Eline is very good at dissipation compared to To92
And is more responsive to an attached heatsink. The two flat faces conduct heat into the sink through two thin layers of insulation.
A To92 might only use one flat face. The curved surface will rarely match the shape of the sink and the insulation is very much thicker.
The MJE has a construction inteneded for good dissipatio, but the transient capabilities may not be as good as Zetex Eline.
Look at some of the Zetex 753 disspation numbers.
It is specified as a continuous 2A, 1W device and it has a 45W capability for one shot 1ms pulse.
Last edited:
BTW. There was a theory in the 1980's that negative feedback causes distortion to arrive in an amplifer at the higher harmonics. I was able to test that theory and find it false ( valve amps make it easier to test ). However it might as well do exactly that in terms of what we hear. If the only distortion that remains in an amplifer is the higher order stuff we seem to hear it even in at - 70 dB or better. Thus less feedback might sound less distorted. As said the stability margins dictate this. The MOS FET's I show simplify things. At least the current amp is super fast. If the higher harmonic is in musically correct proportions although high it seems less harsh or sterile. Some would call it masking. Call it what you like it isn't nice if " unmasked ".
Dave and others. Do you have a voltage for the TR3 emitter resistor ( 620R ) and what transistor type used? Is it 0.6V as assumed?
Hi Andrew, I'm aware of the capabilities of the ZTX65X/75X series but in referring to the package's dissipation I'm not talking about total rated power dissipation but about heat dissipation via the package's surfaces and anticipated lead lengths - only a fraction of what could be expected of comparable larger die semis - say for example MJE172/182. Sorry if that was not a clear distinction
The E-line package is minuscule with little cooling effect compared to even the TO126 package and will operate at a considerably higher package/case temperature. Fine, if the odd burn and curse doesn't deter you but I'm inclined to just fit a more appropriately sized device when high operating temperature is a concern.
Naim also used ZTX652/752 parts as drivers in NAP90/ Nait amplifiers but they weren't the only manufacturers to use them in small-medium power amplifiers. Interesting though, that Douglas Self for one, specified conventional small signal TO92 VAS types and MJE340/350 drivers in his higher powered blameless model designs. They're obviously a lot more capable than the majority of DIY designers here seem to think.
Could not the ZTX be a defacto heat source influencing the Vbe transistor environment - depending on the proximity of the two.
This is my speculation about the Avondale circuit. There is little point having the 22K if TR1 collector sits at 500 mV. The small TR4 emitter resistance will shift the balance ( 10 R ). The 22K should sit at about 8.5 V I suspect ( not 10.5 V ). The 10R should make TR4 easier to drive as a bonus. Loop gain reduces by a factor of 4.3 if so ( VAS gain ). This will make the amp more stable and with modern devices remove the need for phase shift network. The TR1/2 emitter resistors at 0R will help loop gain.
The 100R to TR7/8 will speed up the output switch off faster. Fast transistors are only fast if made fast.
The 2K inverting input is to show what I mentioned before.
That could also depend on whether you consider clones too. Most follow the NAP layout quite closely but the popular, simplified H140 kit from "HifiDIY" is nothing like it. This is probably a good time to eyeball some examples of the parts and boards in question.
An externally hosted image should be here but it was not working when we last tested it.
Here's the link again to Acoustica.org where there are a number of NAP interior pics: Modifying Naim Audio power amplifiers. You can see the layout easiest from the NAP 140 pic, where the input stage and VAS transistors are almost in line from one side the board to the other, about 25 mm from the input end. TR4 is at one side and TR6 on the opposite side - at about 75mm centres. TR5 is 5mm off-centre but still too distant to be directly affected. Below is a good quality Chinese clone board that uses TO92L VAS transistors.
The clone linked at post #1295 is very close to NAP140 with original semis and you can make out the ZTX transistor locations on the larger image posted there.
Last edited:
Thanks Ian,
I had been looking at a Nait 2 layout, where the Vas and Vbe are 1 cm apart which while interesting is not relevant here. I agree with your comments about the post #1295 clone being the more authentic of the two illustrated.
The pictures at Acoustic.org remind me I had an old Naim publicity brochure from the Olive product series. The centrefold was of the internal workings of an NAP 250 from which the circuit diagram could be worked out since the copper trace side is visible in the photo.
From this I see the quite a lot of copper in the trace areas around the Vas transistor where the leads are connected - this will serve to improve heat dissipation of the ZTX transistor which suggests Naim have ticked this box in sofaras Avondale Audio is concerned.
I had been looking at a Nait 2 layout, where the Vas and Vbe are 1 cm apart which while interesting is not relevant here. I agree with your comments about the post #1295 clone being the more authentic of the two illustrated.
The pictures at Acoustic.org remind me I had an old Naim publicity brochure from the Olive product series. The centrefold was of the internal workings of an NAP 250 from which the circuit diagram could be worked out since the copper trace side is visible in the photo.
From this I see the quite a lot of copper in the trace areas around the Vas transistor where the leads are connected - this will serve to improve heat dissipation of the ZTX transistor which suggests Naim have ticked this box in sofaras Avondale Audio is concerned.
The heat output of the VAS is based on worst case 400 mW which is just inside the limits of most T092 devices. T0126 is easilly 1 watt before any clip on heat sink added. Whilst the VAS will be part of a thermal story the output stage will be dominant. The VAS + Vbe multiplier will reach stasis after a few minutes which is true with or without VAS heat. FET outputs simplify things because they are not critical on bias. Looking again if we wrote a formula Stasis = ( T1+t2 )K t2 being VAS we might find (T1)K is actually slower to reach stasis. In fact the VAS may be usefully masking any transient changes. Doubtful, but possible. Holding between thumb and finger a T092 at 400 mW usually is about 60C which seems less if not holiding tight. The heat can travel via tracks, that is a more serrious question.
Here's a thought for the VAS. They look genuine. If so they will measure > 250 gain. Fakes most likely are 2N2504. 2SB716 is a T092L device at about 0.8 watt. They are centre collector. Often German suppliers are better.
B716 2SB716 2SD756 D756 C2856 2SC2856 A1191 2SA1191(5pcs/lot) | Shop66 | Cheap Online Shop | China
Last edited:
2SA1208 looks a very good risk. It is in the T092L case so needs no heatsink. The Japanese transistors seldom are at the lower end of the gain range they quote. Even so these beat most devices regardless. Cob is excellent. These transistors were designed for high speed audio use. It will run 600 mW at 60 C so that looks about right.
CCTV Camera Systems and DVRs from Cricklewood Electronics
http://www.promelec.ru/pdf/2sc2910.pdf
CCTV Camera Systems and DVRs from Cricklewood Electronics
http://www.promelec.ru/pdf/2sc2910.pdf
Last edited:
Thanks Nigel.
Although the 2sc1208 - bce lead configuration - would need reorienting the leads to suit the cbe footprint of a 140 or clone board the holes would not need drilling out as for a TO126 type.
Since the 2sa1208 has a Cob of 2 pF compared to the Zetex 30 pF, the 47 pF compensation capacitor value would need review as Cob is amplified in proportion to device hFE. The 2sa1208 comes in three hFE classifications.
In these respects the Naim hFE baseline for the ZTX device would need to be known - the lengths they go to with device selection can be seen in an old video - https://www.youtube.com/watch?v=xPfdERaAdNc .
If the Naim ZTX hFE can be established, the compensation capacitor could be increased in value or the Miller capacitance increased by selecting a 2sa1208 device with enough extra hFE over the Naim ZTX or a combination of the two options to align with the nearest standard capacitor value - provided the total of new Miller capacitance and the Compensation capacitor is not increased in the process.
This could be a finicky business however if anyone has instability problems with a clone they could take a punt by replacing a lowish hFE ZTX with a marginally higher 2sa1208 one.
The MJE340 has been touted as a suitable TO126 Vas replacement. Cordell classifies the MJE340 (C suffix) as a Vas/Driver transistor.
His Spice models are at http://www.cordellaudio.com/book/Cordell-Models.txt.
I put heatshrink on one leg usually and that seems OK when ( TO92 JAPAN, T0126 BC639/640) ECB to EBC or whatever.
My instict is Naim painted themselves into a corner. It's what is called " Believe your own BS ". I am the worst in the world for BYOBS, so I am an expert.
What I think I have proved is that the Naim design can only be about is long tail pair balance. The rather sterile sound which would be OK for 1979 is totally avoidable if some principles are observed.
Preserve loop gain at all cost.
Best loose loop gain at TR4 emitter if anywhere. It allows the 1K at TR1 collector to be adjusted and allows easier drive as a bonus. TR4 emitter resistor should be tuned by ear. 4R7 to 16R should be about right, 0R if lucky.
Some sort of oscilloscope should be bought to test TR4 VAS cap. 10 pF may well be OK and 27 pF going too far! Assuming the other issuees are dealt with.
DC offest if below 100 mV is not a problem. I suspect it never is above 50 mV.
I suspect any transistor of hFE > 100 will be fine. Higher the better. > 250 becomes interesting. It will not directly change loop gain. It makes the current flow with more ease. Make a 10 mA hFE tester. That might be interesting.
Get rid of phase shifting networks. It's bad engineering. It's bad enough Cob and Ft are thrown in.
I wonder if MJE340/350 are liked becuase like a reliable car old driver they at least get you home? I see no real virtue in them . BD139/140 run at +/-35 V would be my choice if wanting easy to get T0126. BC639/640 are mostly the same devices. The outputs run on whatever you like if 250 V types. If the Avondale change the 220R next to 1N4004 rail resistors if so. 1K looses about 10V. This would reduce the hum and make for softer clipping ( and sound more like a NAP 250 ). One could run +/- 45 without too much trouble. Many times even +/- 57V as long as the VAS is at +/- 35 V ( BD139/140). 57V is about where a 63 V cap is OK. If you did that the TR3 isn't really needed.
I suspect most of you guys don't use scopes? That really is very hard work. Not least many amps are very wrong yet don't kill the speakers. Some say a mildy unstable amp sounds better. The Naim is one by reputation. The reason this might be true is the stability is not music. It is a laid down standard which might not be 100% correct and spoils the music. If you see the progression you will know what you like and why.
I have 4 scopes or perhaps more if I look in all my places I hide things. At work we even have an AP. I never really need that. My little schools scope doing 10 MHz always finds any fault. Even if above 10 MHz there is always sometime showing. My signal generator is up to 5 MHz at 1% THD. I have a home built one for <-70 dB testing. My analyser will give OK results at - 135 dB ( below -125 dB needs observation to use ). I do use it there very often for noise tests. If above - 130 dB I feel I have failed. I exclude hum as it seldom is simple to get right. -90 dB hum is about where I get excited. A recent design - 110 dB at 1.1A, -132 dB reference 12 V, broadband to 16 MHz.
One thing you guys didn't react to was reducing the impeadance between TR7/8. This is not trivial. If Formula One this would be a chance to win. Also the horrid phase shift networks will be gone. The phase shift networks are like not saying bad words when still thinking them. Sometimes that's worse. Best is not to think them or say them. This is about using the scope and very carefully adjusting. The Zobel resistor getting hot means the scope isn't hooked up right. Gets some very cheap speakers you find OK for testing.
I have come to the conclusssion Les of Avondale doesn't really like a standard Naim amp. He likes a Sugden amp. Nothing wrong with that as I also like Sugden amps. A48 is a bargain.
My instict is Naim painted themselves into a corner. It's what is called " Believe your own BS ". I am the worst in the world for BYOBS, so I am an expert.
What I think I have proved is that the Naim design can only be about is long tail pair balance. The rather sterile sound which would be OK for 1979 is totally avoidable if some principles are observed.
Preserve loop gain at all cost.
Best loose loop gain at TR4 emitter if anywhere. It allows the 1K at TR1 collector to be adjusted and allows easier drive as a bonus. TR4 emitter resistor should be tuned by ear. 4R7 to 16R should be about right, 0R if lucky.
Some sort of oscilloscope should be bought to test TR4 VAS cap. 10 pF may well be OK and 27 pF going too far! Assuming the other issuees are dealt with.
DC offest if below 100 mV is not a problem. I suspect it never is above 50 mV.
I suspect any transistor of hFE > 100 will be fine. Higher the better. > 250 becomes interesting. It will not directly change loop gain. It makes the current flow with more ease. Make a 10 mA hFE tester. That might be interesting.
Get rid of phase shifting networks. It's bad engineering. It's bad enough Cob and Ft are thrown in.
I wonder if MJE340/350 are liked becuase like a reliable car old driver they at least get you home? I see no real virtue in them . BD139/140 run at +/-35 V would be my choice if wanting easy to get T0126. BC639/640 are mostly the same devices. The outputs run on whatever you like if 250 V types. If the Avondale change the 220R next to 1N4004 rail resistors if so. 1K looses about 10V. This would reduce the hum and make for softer clipping ( and sound more like a NAP 250 ). One could run +/- 45 without too much trouble. Many times even +/- 57V as long as the VAS is at +/- 35 V ( BD139/140). 57V is about where a 63 V cap is OK. If you did that the TR3 isn't really needed.
I suspect most of you guys don't use scopes? That really is very hard work. Not least many amps are very wrong yet don't kill the speakers. Some say a mildy unstable amp sounds better. The Naim is one by reputation. The reason this might be true is the stability is not music. It is a laid down standard which might not be 100% correct and spoils the music. If you see the progression you will know what you like and why.
I have 4 scopes or perhaps more if I look in all my places I hide things. At work we even have an AP. I never really need that. My little schools scope doing 10 MHz always finds any fault. Even if above 10 MHz there is always sometime showing. My signal generator is up to 5 MHz at 1% THD. I have a home built one for <-70 dB testing. My analyser will give OK results at - 135 dB ( below -125 dB needs observation to use ). I do use it there very often for noise tests. If above - 130 dB I feel I have failed. I exclude hum as it seldom is simple to get right. -90 dB hum is about where I get excited. A recent design - 110 dB at 1.1A, -132 dB reference 12 V, broadband to 16 MHz.
One thing you guys didn't react to was reducing the impeadance between TR7/8. This is not trivial. If Formula One this would be a chance to win. Also the horrid phase shift networks will be gone. The phase shift networks are like not saying bad words when still thinking them. Sometimes that's worse. Best is not to think them or say them. This is about using the scope and very carefully adjusting. The Zobel resistor getting hot means the scope isn't hooked up right. Gets some very cheap speakers you find OK for testing.
I have come to the conclusssion Les of Avondale doesn't really like a standard Naim amp. He likes a Sugden amp. Nothing wrong with that as I also like Sugden amps. A48 is a bargain.
I suspect MJE340/350 is as good as most things for a VAS transistor in the real world. I accept its datasheet is pretty minimalist.
The phase correcting (or whatever they are) networks have been ditched in my amp, along with the queer o/p stage.
I would never plug any new amp into my speakers without a few basic measurements and a squint at some waveforms.
The phase correcting (or whatever they are) networks have been ditched in my amp, along with the queer o/p stage.
I would never plug any new amp into my speakers without a few basic measurements and a squint at some waveforms.
It is so slow. In transistor terms it is iron pistons as in the 1920's cars. I have used them myself and found them totally OK. They are as bad a choice as we can make. I am a fan of 2N30555 outputs. I must say that as they are equally awful. I guess the rule should be MJE340/350 or 2N2955/3055, but not both. I would liken MJE340/350 as like entering a race and knowing 5th would be OK, so 5th it is or maybe even 4th. Problems is you could have come 1st and still been conservative. The Sanyo/Toshiba 2SA1208/2SC2910 ( I think they are the numbers ) are 0.9W and 160V plus centre collector. They are drop in devices for MJE types.
Still no one says about the low Z into TR7/8? Sweeping out the minority carriers to switch off as fast as possible? It is reduced from about 1K2 from to 380R in mine. A bit less in truth. Possibly could get down to 120R with a bit of thought and testing. 380R should be coming 1st and being conservative.
Still no one says about the low Z into TR7/8? Sweeping out the minority carriers to switch off as fast as possible? It is reduced from about 1K2 from to 380R in mine. A bit less in truth. Possibly could get down to 120R with a bit of thought and testing. 380R should be coming 1st and being conservative.
My spare parts collection includes MJE340/350, BD139/BD140 and BC639/640 and I have used them in projects for various uses.
There can be instances where a simple approach delivers a resultant sound that is to ones liking and can settle on that accordingly. I think people are not all sensitive to the same minor distortions. Some people can relax to music with a glass of wine, it makes the experience worse for people like me.
I have built a lot of amplifiers in my time in a seemingly endless quest for improvement and this can become an obsession. As interesting as this all was, in the end I wound up with the an update of a simple circuit where my journey had started 40 years ago.
As far as removing stored charges is concerned, this has to do with current as has been discussed by Cordell see http://www.cordellaudio.com/papers/another_view_of_tim_II.pdf
Last edited:
TR7 and 8 run in class A in my o/p stage and the Sanken o/p transistors are fast - that's one benefit of getting rid of the queer quasi.
The VAS is the slowest part of the amp and is slugged even further by Cdom. A well behaved 4th placer is more than adequate. That's what dominant pole compensation is all about.
Why all the fuss about the ZTX running at 100C? It'll last longer than us, especially with minimal thermal cycling. They don't burn yer fingers cos the plastic is a good insulator.
The VAS is the slowest part of the amp and is slugged even further by Cdom. A well behaved 4th placer is more than adequate. That's what dominant pole compensation is all about.
Why all the fuss about the ZTX running at 100C? It'll last longer than us, especially with minimal thermal cycling. They don't burn yer fingers cos the plastic is a good insulator.
The difficulty is music signals vary in level and are asymmetric in form and so if a passage of music demands more power to be delivered by the output stage the power devices will generate more heat which will take some time to be communicated to the Vbe transistor to alter the Iq to a change in equilibrium.
The inside of a Naim amplifier may be a little hot house centred around the power devices, however heat rises causing cooler air to be drafted from colder regions removed from the heat source. One would not want cooler air wafting past the Vbe transistor.
With the Nait amplifier where I got the idea from, the Vas transistor is close to the Vas transistor which is positioned between the former and the main heat source of the power output stages.
This was just an observation I had seen this as a strange twist worth mentioning. I don't think things are always as simple as they might first appear.
The simple advice might be to avoid having ventilation holes in the metal work where cool air can flow past the Vas. Many of the cases on the market have ventilation holes everywhere.