Nigel, Mission Cyrus had a vaguely similar arrangement but given the thermal delay (an order of time longer than bass compression concerns), I think that would have just as much to do with reducing assembly costs. A finger says it takes seconds to get the peak temperature rise of the output stage to the heatsink near the Vbe multiplier in this example.
FWIW, Self examines thermal distortion, particularly of chip amps where there is some evidence, in a chapter of that title in a few editions of his Handbook. Very easy to see there, with all mechanisms on the same die.
FWIW, Self examines thermal distortion, particularly of chip amps where there is some evidence, in a chapter of that title in a few editions of his Handbook. Very easy to see there, with all mechanisms on the same die.
Last edited:
Hi Ian. I read the same and it's this question that reminded me of the A&R. They realised I liked electronoics and was given a tour of the amp. The thermal sniffer as we called it was just away from the heat sink.
When I visited A&R ( ARCAM ) I had by chance my first ever cheque book with me. On seeing S No 301 built I was asked if I wanted to take it home? It was my first ever cheque at £90.97 ( trade price, £139 I guess - 35 % paid in advance perhaps ). I had to ask John to how to write the right wording. It might have been this time of year 1976 or 77. Also met Chris Evans co founder in the old French's Mill ( still owned by Spillers-French at the time ).
When I visited A&R ( ARCAM ) I had by chance my first ever cheque book with me. On seeing S No 301 built I was asked if I wanted to take it home? It was my first ever cheque at £90.97 ( trade price, £139 I guess - 35 % paid in advance perhaps ). I had to ask John to how to write the right wording. It might have been this time of year 1976 or 77. Also met Chris Evans co founder in the old French's Mill ( still owned by Spillers-French at the time ).
Reading up other forums the Naim NAP 250 is said to have very very poor slewing coupled with measurable instability. All this comes from Spice simulations which I don't use. The result which I find hard to believe was 4 V/uS falling and 2 V/uS rising.
The suggestion is the Naim preamps had ideal bandwidth limiting to help the power amps. Julian Vereker told me very cryptically not to sell other companies preamps with his power amps as that would be to cheat the customer and any problems would be down to me. He never said why.
One solution is very easy. The 220 pF to TR1 shown on the Avondale version can be tuned. I wouldn't recomend smaller values of C unless certain. A CD player might need this more than a radio tuner. One way to do this is to add 47R to the CD player and 10 nF as an example. To be frank that betters Naim as each source should be tuned and not the preamp. One could use a 10 nF and 220R pot to each source and tune by ear. The resistor also is a current limiter and PSU loop cure ( strange hum problems ). I usually use 2.2 uS as a strating point. 10 nF NPO.
Here is a little specualtion I had about tuning. The idea to get the I to V function of TR1 to TR4 working better. Here I just about get transconductance. It allows removal in mimutes. The Naim situtaion is not ideal as the 1K feeds 300 R ( Z in at best ). The gain of TR4 needs to be about 400 if so. Even then not ideal.
A Darlington could be used. If so use the standard Naim or Avondale circuit. This is a beautiful device ( FT very high for a device like this ). The VAS would then be a VAS as Z in might be 3000 ohms. No emitter resistor needed. As Darlingtons feature current gain the 47 pF should be fine ( no worse than standard transistors I seem to find ). The device itself is about 15 pF which is a nice surprise, some are unusable due to being very high. You might be able to reduce the 47 pF a bit due to that. The VAS drive current would be about 1.2mA on the standard Naim circuit as set by 1K TR1 and Darlington Vbbe od 1.1V( 1 mA is the limit ). The resistor could be changed to 2K and 43K for TR2. Or the tail current doubled ( add 620 R to TR3 ). If so try 68pF tto TR4.
ZTX705 - DIODES INC. - Bipolar (BJT) Single Transistor, Darlington, PNP, -120 V, 160 MHz, 1 W, -1 A, 3000 | CPC UK
The suggestion is the Naim preamps had ideal bandwidth limiting to help the power amps. Julian Vereker told me very cryptically not to sell other companies preamps with his power amps as that would be to cheat the customer and any problems would be down to me. He never said why.
One solution is very easy. The 220 pF to TR1 shown on the Avondale version can be tuned. I wouldn't recomend smaller values of C unless certain. A CD player might need this more than a radio tuner. One way to do this is to add 47R to the CD player and 10 nF as an example. To be frank that betters Naim as each source should be tuned and not the preamp. One could use a 10 nF and 220R pot to each source and tune by ear. The resistor also is a current limiter and PSU loop cure ( strange hum problems ). I usually use 2.2 uS as a strating point. 10 nF NPO.
Here is a little specualtion I had about tuning. The idea to get the I to V function of TR1 to TR4 working better. Here I just about get transconductance. It allows removal in mimutes. The Naim situtaion is not ideal as the 1K feeds 300 R ( Z in at best ). The gain of TR4 needs to be about 400 if so. Even then not ideal.
A Darlington could be used. If so use the standard Naim or Avondale circuit. This is a beautiful device ( FT very high for a device like this ). The VAS would then be a VAS as Z in might be 3000 ohms. No emitter resistor needed. As Darlingtons feature current gain the 47 pF should be fine ( no worse than standard transistors I seem to find ). The device itself is about 15 pF which is a nice surprise, some are unusable due to being very high. You might be able to reduce the 47 pF a bit due to that. The VAS drive current would be about 1.2mA on the standard Naim circuit as set by 1K TR1 and Darlington Vbbe od 1.1V( 1 mA is the limit ). The resistor could be changed to 2K and 43K for TR2. Or the tail current doubled ( add 620 R to TR3 ). If so try 68pF tto TR4.
ZTX705 - DIODES INC. - Bipolar (BJT) Single Transistor, Darlington, PNP, -120 V, 160 MHz, 1 W, -1 A, 3000 | CPC UK
Here is how a Darlington VAS might look. As the emitter of the Darlington section one drives section two base it seems a better driving system. That is tons of current to do the job. ZTX705 claimes minimum 3000 gain. The collector of TR2 sits at a different voltage now. I don't think that would matter. The Balanced version is in the top left corner using 2K4 to TR1 and 0R to TR2 collector. Note ZTX705 sits at 1.2V and not the 0.7 of the ZTX753.
When looking for nice VAS devices I regreted Naim using a PNP. ZTX694B NPN looks the most ideal device for a single VAS type. Above is a specualtion which is almost identical something I built. The feedback to VAS can be removed ( 220R dotted ) and the Bootstraped input that I did need in my version can be removed ( 100 uF dotted ).
The output to VAS feedback does tidy things in terms of HF distortion. Don't try it with the Naim as I doubt it would ba happy.
The input is a double BC556 in surface mount. 2x 100 R emitter resistors very likely would be OK as an improvement. 2 x 27R might be optimum. I show 0R.
The 100uF bootstrap current source could be 22uF. I usually find the amplifeir is OK up to 1 KHz without any bootstrap. 22uF 100V polyester would do well.
The bias will give about 1.4V which is just enough ( 100mA ). Should it fail to 220R the outputs should be OK.
I have not used semiconductor current sources just to say they are not 100 % must have components. Also when not using a PCB it gives a greater chance of sucess.
I don't intend anyone to make this. However like cooking sometimes the very best ingrediants and a simple recipe works best. My version was on tags, not even veroboard.
I wouldn't use the ZTX694B myself as I have some 2SD756 of similar gain. They were made for this exact use.
The 220R gate resistors are about right, 100R is more common. They must be on the device.
The gain of the amp is 1+ ( 33/0.62 ) = 54 like a Quad 405. This is a drawing I did perhaps 10 years ago updated. I don't have an exact FET drawing.
I suspect it would beat the Naim to many ears.
Just to say. The Avondale could be swaped to this as a ZTX694B VAS and BC556 TO92 inputs. The 1N4004 diodes in the power rails removed and every transistor swapped in polarity type ( BC546 becomes BC556 and 2N5551 becomes 2N5401 etc). 1N4148's reversed and caps also swapped + for -. Very risky if you are not sure. It could be done and I suspect it would be better. The output would stay as it is and wires cross the PCB where the 1N4004 were to the opposite sides. The stablity should be OK as the rail capacitors are that side of the PCB. Be sure to keep the ouput rails unchanged. I don't exspect anyone to do this. Just to say it wouldn't be impossible.
My 002 amp has BC556 LTP and MJE350 VAS (2SA1360 current source for VAS). It also has a fully comp o/p stage MJE15030/1 and Sanken 2SC2922/2SA1216.
I am very happy with the sound. Must admit is seems to be better with no emitter degen!
Of course it's really not a NAP clone any more. I do prefer the sound with the 22K TR2 resistor - sounds bright and "solid state" with it shorted.
I am very happy with the sound. Must admit is seems to be better with no emitter degen!
Of course it's really not a NAP clone any more. I do prefer the sound with the 22K TR2 resistor - sounds bright and "solid state" with it shorted.
Hey Dave, could you say in what way(s) does it sound better without the degeneration resistors? I'm curious as I have a Hackernap that I'm contemplating doing some work to.
I can only suggest listening for yourself. It's also worth checking stability as this increases loop gain.
Soundwise IMO it seems to have a touch more body and detail. I also thought I noticed it being easier to follow subtle percussion and rhythms.
Soundwise IMO it seems to have a touch more body and detail. I also thought I noticed it being easier to follow subtle percussion and rhythms.
But probably more important, the ksa is specified much more fully in the datsheet. One can assess it's suitability for the duty.
The 340/350 have almost no useful data to guide a designer.
TR4? It will be PNP if Avondale dia. Could you measure the 1K. I wonder if it is 0.55V? I wouldn't be surprised. If so the Naim might be more subtle than I imagine in both forms ( Avondale also ).
Spot on Andrew.
Spot on Andrew.
My NCC200 measures 0.5V across the 1K (ZTX753 B-E).
My 005CCN measures 0.7V across the 1K5 (MJE340 B-E & 10R emitter resistor i.e. 0.6V + 0.1V).
My 005CCN measures 0.7V across the 1K5 (MJE340 B-E & 10R emitter resistor i.e. 0.6V + 0.1V).
That is as you said before. A big surprise, thanks. If we assume 620R to TR3 = 1mA ( good enough for now ) we might have 500 uA/500 uA NCC200. It would be interesting to know what the 620 R ( TR3 ) sees.
Here is the sub woofer amp idea I tested on tag board ( 15 inch OB driver ). You might ask why this simple? The answer is easy. Any other amp failed on tag board due to oscillation. Look how low the Cdom is and how 100 pF makes it unity gain stable like a good op amp. I am a believer in point to point construction and this amp is just possible. The tiny bit of second harmonic at 22 kHz can be cured using the split Cdom I gave before ( single pole here ). Sorry to say I lost the actual version. It was braver than my guess I gave before with split Cdom to speaker terminal. No smoke or fire! I will be honest and say I am not totally sure why it works. Is it the extended split pole like Self describes or inseting an error signal? Perhaps both ? As this amp was not intended for HF. The 22 kHz is just to know how good it could be ( 50 kHz was fine ). I made a mistake with these graphs. The noise averaging was switched off. My VAS is 2SC2910 at 600 mV ( 720mV - 120mV RE ). 5 watts is where I set the worst distortion point. At 1 kHz it is unmeasurable. Even at 22 kHz the oscillator is starting to distort ( - 73 dB ? ). The FET output bootstrap ( 22 uF to VAS collector 3K6+3k6 ) is said to be unworkable due to lack of voltage gain ( 78 % ). It works very well. As you can see with this amp I have plenty of second harmonic so can use a simple TR2 0R collector. you can just see the third is nicely where it should be. The amp actually reduces in distortion at higher or lower power. If the bias was reduced the bad point would shift down. 5 watts on 100 dB/watt speakers is loud. The amp will give 150 watts 4R safely. If double MOSFET's used that would be 180 watts 4R and 300 into 2R. This VAS will happilly drive two and even three sets of FET's. The 1 nF gate capacitance CgS is not seen by the VAS as it is bootstrapped to the speaker and not 0V as people imagine. The real problem in CgD which is seldom mentioned. Even so this amp will give 100 kHz 100 watts 8R without even thinking about it.
Here is the sub woofer amp idea I tested on tag board ( 15 inch OB driver ). You might ask why this simple? The answer is easy. Any other amp failed on tag board due to oscillation. Look how low the Cdom is and how 100 pF makes it unity gain stable like a good op amp. I am a believer in point to point construction and this amp is just possible. The tiny bit of second harmonic at 22 kHz can be cured using the split Cdom I gave before ( single pole here ). Sorry to say I lost the actual version. It was braver than my guess I gave before with split Cdom to speaker terminal. No smoke or fire! I will be honest and say I am not totally sure why it works. Is it the extended split pole like Self describes or inseting an error signal? Perhaps both ? As this amp was not intended for HF. The 22 kHz is just to know how good it could be ( 50 kHz was fine ). I made a mistake with these graphs. The noise averaging was switched off. My VAS is 2SC2910 at 600 mV ( 720mV - 120mV RE ). 5 watts is where I set the worst distortion point. At 1 kHz it is unmeasurable. Even at 22 kHz the oscillator is starting to distort ( - 73 dB ? ). The FET output bootstrap ( 22 uF to VAS collector 3K6+3k6 ) is said to be unworkable due to lack of voltage gain ( 78 % ). It works very well. As you can see with this amp I have plenty of second harmonic so can use a simple TR2 0R collector. you can just see the third is nicely where it should be. The amp actually reduces in distortion at higher or lower power. If the bias was reduced the bad point would shift down. 5 watts on 100 dB/watt speakers is loud. The amp will give 150 watts 4R safely. If double MOSFET's used that would be 180 watts 4R and 300 into 2R. This VAS will happilly drive two and even three sets of FET's. The 1 nF gate capacitance CgS is not seen by the VAS as it is bootstrapped to the speaker and not 0V as people imagine. The real problem in CgD which is seldom mentioned. Even so this amp will give 100 kHz 100 watts 8R without even thinking about it.
Thanks, I've been concerned about throwing away so much feedback with the two 100 ohm resistors. Not only HD, but PSRR too suffer. Getting rid of the degeneration resistors is one way, another would be to reduce the closed loop gain, which I don't currently need. This would keep the positive effect of the 100Rs.
I'm also thinking to re-instate the phase lead network to maintain stability with higher feedback.
I would.
The ZTX753 datasheet shows, to output a collector current of 10 m.A., a VBE of 0.6 volts is required.
To do the tests Ic/Ib has been restricted to 10 which is not the sort of condition a common emitter amplifier should operate under.
The separate graph for Ic/Ib (hFE) gives a current gain of 225 for the same Ic of 10 m.A.
TR1 Ic is 0.5 m.A. and 10 m.A./225 of this is all that is required to drive TR4.
For many transistors the conduction threshold will be in the region you have quoted i.e. around 0.5 V - rising rapidly from there. It appears from your observation that whatever TR4 hFE value has around 0.5V, it is enough.
The ZTX753 is said to be a low saturation device.
At last!!!
It's also worth bearing in mind that the ZTX is running pretty warm (maybe Tj = 100C) at 400mW.
Also, the ZTX is a 2A lump of silicon, the MJE is 0.1A
The simple soulution is make Cdom larger. Below is a way of getting more from the loop gain compromise. Usually making the amplifer lower gain makes it universally worse. The sweet spot seems to be about gain of 30 to 60. Strangely gain of 1 often sounds good. Gain of 5 might have something to offer. It would also make the inverting input a sensible option ( 27K/6K8, gain= -4 if inverting used ). That is to feed in via TR2. 22 uF 100 V polyester could be used.
I don't think a gain of lets say 10 automatically sounds worse. The problem is the amplifer was never concieved that way. An analogy is like holding an engine in too low a gear. If a gain of 10 is desired it means a reworking of stability margins, firstly stable and then best sound. Here is a test you can try. Take an op amp like MC33078/9. Try is at unity gain. It will sound excellent. Now try it at gain of 2 ( 10 K/10 K ) . To my ears it is like it is a lower grade type. Strangely at gain of 50 is comes alive again. Some slight colour can be heard. Otherwise perfect. My theory is the need to be unity gain stability has made the op amp optimum when unity gain ( if made to be that ). In doing that they have taken something away. A supurb low cost op amp if asking. To my way of thinking Cdom is the sound of op amps and power amps. If you are forced to have a gain of 2 when op amps use the inverting option ( 10K/5K1 = -2 ). An inverting preamp and power amp is the textbook best possible senario. The Naim is really a very low grade op amp. As long as that is understood it can be made to work. I would suggest it works in a very very narrow window.
http://www.diyaudio.com/forums/solid-state/177734-questions-about-2-pole-compensation-vas.html
Not quite as dramatic as that. It's actually 0.5A and the issue is dissipation really, which is not the E-line package strong suit, as the approach there is to "tough it out", whatever its rating. Admittedly there are worse limits on slugs like MJE340/350 but nobody likes a 100°C surprise as they fiddle about inside the box.
If you refer to Motorola or On-Semi datasheets, you may find more data than you do with second sources like ST micro, Fairchild etc. who may omit the SOA graph for example. Useful if you want to use them say, for drivers.