Make that a 22k collector resistor. Looking at transistor Ic/Ib specifications. The Faichild BC239 illustrates the relationship is very linear at low Ib values. I will look at that later when I get over a summer sickness.
That's interesting. Here is an amplifier that has asked the same questions in the DIY link below ( Perfect ). The current mirror of 2 x 2N5401 have no emitter swamping resistors. I like that as it maintains a bias towards single transitor errors. It also sits as low as it can to the rail should it matter ( It does if like the Naim there is no VAS emitter uplift resistor, that resistor has no real value as supposed local feedback. Mostly previous stage matching, a good trade off sometimes, try 10 R to 47 R ). The MOS FET output stage is a well known combination that works well with the Exicon varriants. The JFet input as shown might be a simple substiute for the Naim. Note how the feedback lower arm needs no capacitor. This allows a highly desirable 10R resistor. The author has sensibly set the feedback resistors low enough to offer advantages whilst not robbing power. The output devices have what looks like an exponential slope transfer curve and not the on-off of the Naim power transistors. MOS FETs are much disliked by Mr D Self as having no exact turn on point. Lets be very clear nor do bipolar transistors. They are worse because it is take it or leave it. If a MOS FET is biased to a certain point we the listener distort the music more than the gain doubling of the not perfect matching of MOS N/P FET's curve etc. At about 100 mA bias it works well. MOS FET's of this type can be set by ear and nothing else ( and a hand on the heatsink ). If someone likes the amp to work more into class A multiple output devices will help. To me this suggests the speakers are too quiet. Have a look at the circuit as it points the way should anyone be interested. Personally I like the old Hitachi amp as it can do miracles. Many mistake it's simplicity for being wrong. They haven't tested one at 100 kHz full power and found no real distortion. The graphs I show are at 5 watts. It might surprise people to know 5 watts for this amp is it's weak point, far better at 70 watts or 0.05 watts. The analyser is to 16 Mhz so is not showing a filtering effect at 47 kHz. Notice 20 kHz is worse although 2 nd harmonic which at - 60 dB we can ignor ( VAS matching compromise ). The single VAS shows the simplicity of the Hitachi design can not be made better by conversion to a single VAS. Doubtless the other design shown restores the virtues. The Hitachi has far fewer parts than it. The oscillator is working very well for what has a theoretical - 60 dB distortion. It is a design by RA penfold using one NE5534 and one RA53 thermistor and 680R feedback resistor. The caps 10 nF NPO and that's all. I suspect it is a NE5532 with a switchable single pole active filter in addition if I remember. This gives an active buffer with great simplicity. It is an integrator. These graphs were for me and not for showing here. They show a trend I was interested in. The only difference is single or twin VAS. A twin VAS needs an analyser to set it up as it's only demerit. Here is has 12K 6n8 compensation in the VAS No1 that isn't connected to the output devices. The double VAS is symetrical in source and sink ( Between 20 to 35 V/uS depending on desings seen, MOS FET amps do not show slewing problems as the on-off nature of bipolar is not present. Slewing is seen as crossover distortion up a wave when in it's mildest form ) . The single VAS was made as good as it could be. It was used as a subwoffer amp and was tweaked to be good to 22 kHz. It was simple enough to be built on tag strip. It used a bootstrap VAS CCS in the better version. Even though a MOSFET is only about 78% the voltage gain of a perfect current amp it is enough to drive a bootstrap. Do not believe what you read, even 50 % makes a big difference. I have an amplifer I built with only 50 % as that was the free lunch the circuit could give. When valves that would need 1000 V by the purists route. It lowered distortion by a factor of 5 and sounded much better. Positive feedback is sometimes the best.
http://www.diyaudio.com/forums/solid-state/186650-jfet-input-mosfet-vas-lateral-output-perfect.html
I forgot to include the 20 kHz tests and seems I have lost them. What I gave instead was the start of VAS matching when 10 kHz. Nice to think when not so good it's only second harmonic.
Below is how the single VAS improved ( nothing like it's origine, Hitachi ). The 22 uF VAS CCS was by using the analyser. When the distortion reached a certain level I didn't add more. It was very low at < 1 khz. The 22 K could be 16 K here if you like. It should work fine. No real need as the FET's tend to offer the second harmonic anyway.
The input bootstrap is to allow it to piggy back the Quad 33 without loading it down. It worked fine. As the bootstrap feedback is below the critical level it is OK to do this. A scope is required ( At a gain of 0.97 bootstraps become troublesome, complimentary feedback pairs for example, The 220R I add is not really required as the voltage gain is about 78%). 16R VAS emitter resitor is to force the VAS to be a VAS and less of a TIS, with this transistor 0R should be OK or even preferable. 220R is a nice choice for the FET grid stoppers. The grid oscillates at about 5 MHz ( Be very careful if you look for this, the FET's might go up in flames ). The resistor as much as anything is working as an inductor. I would speculate a high grade resistor of the foil type to be less good here. The reason a simple VAS can drive a MOS FET is effectively the 500 pF to 1 nF gate capacitance is bootstrapped to the source and not 0V. Thus the VAS doesn't really see it. What it will see is the seldom quoted CgD. This amplifier although not intended to be a Naim beater possibly is. It is not on a PCB so needs no kit. It did have a split Cdom with a 100 R to the ouput devices. It's the example in D Self book except the 2K2 is 100 R and not taken to the voltage rail or 0 V. I was able to get it to measure very nicely. MPSA 42 wasn't tried. 220R bias giving 128 mA standing current is a bit high.
The speaker is a 15 inch driver on an open baffle. This amp might have slightly more hiss than is true today. That's simply a matter of some resitor changes. For my needs this one is fine. The 33 uF Non Polar is what I have a bag of and the 22 uF 100 V polyester also. Non Polar caps tend to outperform Audiophile grade polarised caps on distortion tests. The trick is to keep NP caps below 0.4V peak to get the best. This amplifier gets to about 15 V rms before that happens. High gain is useful with that.
Below is how the single VAS improved ( nothing like it's origine, Hitachi ). The 22 uF VAS CCS was by using the analyser. When the distortion reached a certain level I didn't add more. It was very low at < 1 khz. The 22 K could be 16 K here if you like. It should work fine. No real need as the FET's tend to offer the second harmonic anyway.
The input bootstrap is to allow it to piggy back the Quad 33 without loading it down. It worked fine. As the bootstrap feedback is below the critical level it is OK to do this. A scope is required ( At a gain of 0.97 bootstraps become troublesome, complimentary feedback pairs for example, The 220R I add is not really required as the voltage gain is about 78%). 16R VAS emitter resitor is to force the VAS to be a VAS and less of a TIS, with this transistor 0R should be OK or even preferable. 220R is a nice choice for the FET grid stoppers. The grid oscillates at about 5 MHz ( Be very careful if you look for this, the FET's might go up in flames ). The resistor as much as anything is working as an inductor. I would speculate a high grade resistor of the foil type to be less good here. The reason a simple VAS can drive a MOS FET is effectively the 500 pF to 1 nF gate capacitance is bootstrapped to the source and not 0V. Thus the VAS doesn't really see it. What it will see is the seldom quoted CgD. This amplifier although not intended to be a Naim beater possibly is. It is not on a PCB so needs no kit. It did have a split Cdom with a 100 R to the ouput devices. It's the example in D Self book except the 2K2 is 100 R and not taken to the voltage rail or 0 V. I was able to get it to measure very nicely. MPSA 42 wasn't tried. 220R bias giving 128 mA standing current is a bit high.
The speaker is a 15 inch driver on an open baffle. This amp might have slightly more hiss than is true today. That's simply a matter of some resitor changes. For my needs this one is fine. The 33 uF Non Polar is what I have a bag of and the 22 uF 100 V polyester also. Non Polar caps tend to outperform Audiophile grade polarised caps on distortion tests. The trick is to keep NP caps below 0.4V peak to get the best. This amplifier gets to about 15 V rms before that happens. High gain is useful with that.
Here it is simplified. OB speakers can need weird solutions and this was a specualtion mostly.
The split 22/220 pF is not yet tested on this exact amp. It was tested on one nearly identical. The concept is good. I think I tested a 50/50 version of 2 x 33 pF with interesing results. When with a complimentary output stage it offers no improvement with split Cdom ( a very complex set up ). Adding this closed the gap to the point of not being able to see a difference at 50 kHz.
The split 22/220 pF is not yet tested on this exact amp. It was tested on one nearly identical. The concept is good. I think I tested a 50/50 version of 2 x 33 pF with interesing results. When with a complimentary output stage it offers no improvement with split Cdom ( a very complex set up ). Adding this closed the gap to the point of not being able to see a difference at 50 kHz.
Here is the ultimate step too far. In a way it works. What it does show is the 50 kHz clean up is working. It won't take too much imagination to realise this is a spectacular result. Don't bother to make it as the previous amp has more to offer. It is how to make a better class A design of the old school. An OPA604 and Complimentray feedback pair was no better at 50 kHz. This design was improved. It would be wasting everyones time to try to find what file name I put it under ( not sure ). This is just to show split Cdom working. The non polar output cap a precaution, Celestion speakers sometimes had the same.
Nigel,
I have a couple of Technics amplifiers one at a holiday property and one in my garage cum man cave at home. The at the holiday property is a Class H+ 2 channel amplifier receiver an SA-GX170 home entertainment unit I bought second hand 10 years ago. I find I can adapt to this and other amplifiers in my collection without fretting.
For a time I had one of the Class AA bipolar amplifiers in the garage and I swapped this for one that offered a few more facilities including a mm/mc phono option and a remote control.. This is one of the mos Class AA variety. I thought this somewhat preferable to listen to. One can buy these for less money it would take to build. a DIY project.
While this is all interesting it is taking the discussion away from the Naim topic.
I have a couple of Technics amplifiers one at a holiday property and one in my garage cum man cave at home. The at the holiday property is a Class H+ 2 channel amplifier receiver an SA-GX170 home entertainment unit I bought second hand 10 years ago. I find I can adapt to this and other amplifiers in my collection without fretting.
For a time I had one of the Class AA bipolar amplifiers in the garage and I swapped this for one that offered a few more facilities including a mm/mc phono option and a remote control.. This is one of the mos Class AA variety. I thought this somewhat preferable to listen to. One can buy these for less money it would take to build. a DIY project.
While this is all interesting it is taking the discussion away from the Naim topic.
Last edited:
My calculations were based on TR2 having a current gain of 200 based on the Fairchild specification sheet for BC239. I understand these were used in the LTP of Naits 1 and 2.
Using a higher current gain device will reduce the pole frequency since this increases the resistance in and the collector to base capacitance value. A current gain of 300 will reduce the pole frequency to 12kHz and with 400 the pole will reduce further to 9kHz.
There is a Miller capacitance calculator which explains things at http://www.ecircuitcenter.com/Circuits_Audio_Amp/Miller_Effect/Miller_Effect.htm For my personal use I found it easier to make up a spreadsheet and plug in the values into the optional fields.
If any of the kit clones come with high current gain devices for TR1 and TR2 and the sound does not emulate the originals, that should be no mystery.
Last edited:
It is also the Naim topic made larger. Some of the things I show can be tried in the Naim design.
Intersting to try what you suggest. I also will look at some of my older amps. A class H could really benifit. Did you notice my circuit became better at 50 kHz when the loop gain dropped? I inserted a 47R emitter resistor in the VAS of the class A (AB ) design. My explanation is sometimes the VAS overloads the driving stage even though text books say it won't. If a bipolar design I would exspect worse. The faster speed of the MOS FET's makes for a better design. Books like Self seem to not think slew rates are a function of the output speed. I am sure they do, they forget to think it worth saying. Think about it for a moment. If the MOS FET is super fast the driving stage is not working so hard. That conneced with less need to maintain bias as the device is not a switch.
What I hoped to do was get someone to say could the Naim be improved by making it simpler? One way would be to fit a simple tail resistor like I do. Everyone will say it will suffer. Simple truth is it will not as the input stage is current opperated. Next simple truth is it is cheap to fit a current source. Not fitting it is a sound quality issue and not money.
One place I never see bipolar current sources is valve amps. This is stupidity on steroids and is an anti bipolar bias. Often the upper voltage limit is reached so fair enough. Even so P type FET's at high voltage are rare. A bipolar current source with a larger voltage loss than is typical is the most linear thing we have ( use 3 diodes rather than 2 is usually enough to avoid Early effect ) . Mostly a resistor. If you double the resistor of the Naim VAS CCS and add one more diode it would be slightly more linear for the loss of 0.7V. You could try a capacitor bootstrap CCS. Two LED as the CCS referece really works well. When you have 330 V to play with it's fine.
Intersting to try what you suggest. I also will look at some of my older amps. A class H could really benifit. Did you notice my circuit became better at 50 kHz when the loop gain dropped? I inserted a 47R emitter resistor in the VAS of the class A (AB ) design. My explanation is sometimes the VAS overloads the driving stage even though text books say it won't. If a bipolar design I would exspect worse. The faster speed of the MOS FET's makes for a better design. Books like Self seem to not think slew rates are a function of the output speed. I am sure they do, they forget to think it worth saying. Think about it for a moment. If the MOS FET is super fast the driving stage is not working so hard. That conneced with less need to maintain bias as the device is not a switch.
What I hoped to do was get someone to say could the Naim be improved by making it simpler? One way would be to fit a simple tail resistor like I do. Everyone will say it will suffer. Simple truth is it will not as the input stage is current opperated. Next simple truth is it is cheap to fit a current source. Not fitting it is a sound quality issue and not money.
One place I never see bipolar current sources is valve amps. This is stupidity on steroids and is an anti bipolar bias. Often the upper voltage limit is reached so fair enough. Even so P type FET's at high voltage are rare. A bipolar current source with a larger voltage loss than is typical is the most linear thing we have ( use 3 diodes rather than 2 is usually enough to avoid Early effect ) . Mostly a resistor. If you double the resistor of the Naim VAS CCS and add one more diode it would be slightly more linear for the loss of 0.7V. You could try a capacitor bootstrap CCS. Two LED as the CCS referece really works well. When you have 330 V to play with it's fine.
Here is the last useful design and graphs to add. Something as far as I know not mentioned by most. A complimentary feedback pair VAS. The massive advantage is it behaves like a super high gain single device with some advantages. This means it is highly unliky to load down the driving stage. The Cdom keeps it stable without any protest. I think It did sound better than a Darlington, more open . Notice also a very nice fast output stage. The arrow diode is a ready made diode type CCS JFET to the VAS output. Notice also that the FET amp slightly beat this. The output stage was not to blame and is better than the FET's if driven by an op amp like OPA604. With care the FET's can be made to work.
In my haste to fit it where the single transistor VAS went a slight error. Note the very high Cob BD136. This was a tentetive first step so I used the 1960's solution.
Hello Nigel,
I am still settling in from a trip up country to attend a family funeral and have been feeling fatigued so I confined my posts to the subject of the Naim clones. Having looked at the points you raised you have aroused my interest and I have book marked the page as a reference for a future project amplifier.
As a rule I have given up on building complex circuits. I have a Kenwood KA3500 amplifier (simple) awaiting repair in my garage. I am retired but never seem to have enough time to finish my projects - a consequence of having to maintain two properties and keeping my wife entertained - she is partially sighted.
The Kenwood amplifier has a fault at turn on emitting a loud hum before settling to normal. It does sound surprisingly good. It has a resistor load in the tail as per your approach - I have seen this before in Hugh Deans early designs. In a valve/tube amplifier the tail resistor value will be very high in relation to solid state amplifiers which might increase the LTP tail impedance with a CCS.
Returning to the Naim circuit and the subject of Miller capacitance, I chose the typical current gain in the Fairchild BC239 datasheet.
This leaves the question unanswered as to whether Naim used typical current gain values in their amplifiers or whether these were measured and selected for specific values.
The 22k collector load for TR2 is used in all the circuits I have seen. The early Nait amplifiers used 5 megahertz output devices. Using higher fT devices would allow a wider gain bandwidth product for the up market versions. We don't know whether this affected any current gain selections.
This leaves the question unanswered as to whether Naim used typical current gain values in their amplifiers or whether these were measured and selected for specific values.
The 22k collector load for TR2 is used in all the circuits I have seen. The early Nait amplifiers used 5 megahertz output devices. Using higher fT devices would allow a wider gain bandwidth product for the up market versions. We don't know whether this affected any current gain selections.
Best of luck with then funeral. As I get older they are more frequent. Never gets easier.
As you say about valves. They often really need some help, more so than transitor amps. Mine are SE amps so just enjoy the help of a CCS or CCSink. I often fit them just to see what is possible. I have left the amplification device out as it is a bit special. Look how low the distortion is seeing as it is zero loop feedback. The CCSink better than a LM317 people use. The output is driven as a longtail pair with one side grounded. This is also a test bed 211 driver as this design needs 56 V RMS. As you can see the distortion is very low. The output transformer very low cost. Although a PP design it behaves like a SE. The hum was better than I show. No idea how an older graph got in. The LED is an ideal reference. I know this is away from Naim amps. Not really as it is saying how simple CCS improvements can be made.
As you say about valves. They often really need some help, more so than transitor amps. Mine are SE amps so just enjoy the help of a CCS or CCSink. I often fit them just to see what is possible. I have left the amplification device out as it is a bit special. Look how low the distortion is seeing as it is zero loop feedback. The CCSink better than a LM317 people use. The output is driven as a longtail pair with one side grounded. This is also a test bed 211 driver as this design needs 56 V RMS. As you can see the distortion is very low. The output transformer very low cost. Although a PP design it behaves like a SE. The hum was better than I show. No idea how an older graph got in. The LED is an ideal reference. I know this is away from Naim amps. Not really as it is saying how simple CCS improvements can be made.
Ever thought of buying a Nait 2 ? I suspect these amps are not greatly like the originals. Hard to be sure.
Ever since Tom Tom Audio reviewed the Nait range these have been extraordinarily expensive on auction sites. Confession time - the amplifier hooked into the domestic system in my living room is a Nait 5i I have not been tempted to look at the electronics inside the case.
I have had to review my stance about the base current measurement for TR2 and the explanation I gave in a more recent post needs enlargement - in regard to the collector current and the base current relationship.
The starting point for my amended theory is the sequence of events when the amplifier is powered up. Initially the emitter will briefly pass 0.499 milliamps to the collector of TR2 and a smaller current will flow along the feedback resistor.
However the collector junction will develop an instantaneous Miller capacitance in opposite phase which will be fed to the base region - the structure is an inverting amplifier.
The value of the Miller capacitor with TR2 current gain of 200 is 800 odd pf. Current will flow from the collector to the base across this capacitance accordingly - I am arguing this accounts for the .01 mA Ic deficit in TR2 collector load as compared with TR1.
Apart from a small amount of leakage across the electrolytic cap in the decoupling arm of the feedback loop TR2 collector resistor and the feedback resistor stops the charge on the capacitor from going anywhere other between TR2 collector and base.
For whatever reason there have been no replies from contributors other than Nigel who is working on an entirely different theme. I am beginning to feel this is a complete waste of time with everyone else.
Last edited:
Not sure if you comment is directed at me, if so I don't have a lot more to say. The simple fact is that despite Andrew T's theoretical approach and Ian Finch's simulations (post #1492) the collector currents in the LTP are quite well balanced.
The reason for this is that the VBE of the ZTX753 is quite low (cannot remember exactlty but about 0.5V instead of the more usual 0.6V with other trs). These ZTXs must have quite a large chip inside, to achieve their current ratings - that's why the VBE is rather low and the junction capacitances quite high.
In my 005 amp the VAS is an MJE340 and this has a more normal VBE at 10mA of ~0.6V. To get the LTP to balance, with the same tail current, TR1 collector resistor had to be increased to 1K5 (I also have a bit of emitter degen in the VAS in this amp).
The reason for this is that the VBE of the ZTX753 is quite low (cannot remember exactlty but about 0.5V instead of the more usual 0.6V with other trs). These ZTXs must have quite a large chip inside, to achieve their current ratings - that's why the VBE is rather low and the junction capacitances quite high.
In my 005 amp the VAS is an MJE340 and this has a more normal VBE at 10mA of ~0.6V. To get the LTP to balance, with the same tail current, TR1 collector resistor had to be increased to 1K5 (I also have a bit of emitter degen in the VAS in this amp).
Measure it I think. 6R load would be enough to compare. Think yourself lucky it's not class D. They need special test gear. I read the AP version and had a guess how to do it.
I hope I have pointed out to people that the set up of the Naim amps might actually be more accurate in musical terms than the cheap approach to amplifier design and they cost the same. The last valve amp I showed brakes the mold of how an amp might work. It requires double the voltage swing of usual to work. I even found a way to do it that was possible in 1954. Can you imagine 54 Vrms from 2.5V in and some MOS FET's. I recently did this to build a micro adjustable variac. 4 MOS FET's ( Exicon , not power FET's like IRF ). Bias and capacitor coupling.
The output transformer I used was a Danbury as was sold by Maplin. I0 watts at clipping is not too bad for triode's. I dare say 25 watts if the UL taps used. I hate most UL feedback amps. I like UL without feedback ( as I should ).
From the valve amps a bit of Naim crossover. The cathodes of the EL34 are set to 1.25V loss. This by pure chance is the same as LM317 that people seem to like. It also returns for very different reasons identical distortion. LM317 is mostly a LM741 1960's op amp with pass transistor and 1.25V reference. It has many tens of transistors to do exacly the same job as I show. BD139 is just about OK to do what I show. A 30 MHz 250V T03 device would be better. BD139 might just sound better.
Now the Naim bit. If I drop the voltage loss to 0.7V and use 5R6 perhaps the valve circuit will function as now. However the second harmonic distortion will be noticably higher. How come ? Simple. The transistor is being asked to work right down at the bottom of it's curve. It is no longer the perfect current amplifer. This problem is also found in cascode transistor coupling. Read very old text books if wanting to design better cascodes. Although it won't be much to get excited about the VAS CCS could be tweaked or made as a 1960's bootstrap. If an LED set the resistor higher than you think and then try other values. A LED will be about 1.8V. The JFET CCS I show is more interesting than useful, a valve can do the same. People like them as they need no reference voltage. That's not a good reason.
This is worth a read. Apart from a cable and big resistor you are mostly where you need to be. I suspect someone can offer a test tone CD or download. It doesn't matter that it is perfect. It only matters that it measures things the same everytime. In our company we have an AP test set. My friend has better not from the audio world. I always get results that are mostly the same as the AP. I am not really interested in sub - 70 dB tests. I look for them by using a very small signal into the amp to be sure nothing bad is happening. As bad as a NE5532 is it can be configued as two times gain 10 to see trends ( x 100 ). If weird harmonics show up the design is not nice. My analyser has noise averaging. Silly me didn't use it on the class A 4 device FET amp. I think it would have been even lower distortion than it looked. In real life noise and distortion is, what you see is what you get. When class B more so. If you build a RC filter it often has lower noise than an active circuit ( 3 dB perhaps ). However the noise of the active circuit is a blanket that looks like pink noise very often. The active circuit for this reason alone might sound better. Thus the person becomes a believer in active circuits for a reason they never would guess. It is valve hiss in type. Like MP3 some hiss might be useful. This idea was via Michael Creek. I hope he doesn't mind me telling the story as it brings no shame on him. The original CAS4040 by a design error had no bias. It was real class B ( Class CB ). It was very OK. Michael said the amp had enough hiss to help it switch. The latter CAS4040 was corrected. The original was not a bad amp.
http://www.diyaudio.com/forums/soft...gestions-free-spectrum-analyzer-software.html
http://www.diyaudio.com/forums/soft...gestions-free-spectrum-analyzer-software.html
Last edited:
I have a hand drawn sketch from Julian Vereker of the original NAP amp I''ll dig it out and post on this thread.
Sharif.