advertisement


Avondale NCC200 mod

Hi,
I just ran a simulation, and set one resistor to 68ohms, and the other to 1kohm. The DC operating point simulation for the resistors shows the 68ohm resistor has 3.1mA, and the 1kohm resistor has 0.21uA.

This is not with the LTP input stage used in isolation, but connected to a VAS and output stage with DC offset servo.

Can you run a simulation and confirm ?.

Thanks and regards,
Shadders.

Measurement trumps simulation!

Are you taking into account that for the amp to work properly the VAS transistor "regulates" the voltage across the input transistor's collector resistor at ~0.6V.
Simulating at LTP in isolation will not work.

Here are the measurements from both channels of my NCC200 at quiescent with +/-40V supplies.
Voltages measured across resistors:
1K 0.500V 0.501V (= 0.5mA i.e. ~1/2 of the tail current)
22K 10.75V 10.66V (= 0.48mA)
620R 0.632V 0.628V (= ~1mA tail current)
68R 0.716 0.705V (~10.3mA VAS current)
220R in + rail 2.52V 2.49V
220R in - rail 2.86V 2.82V

If you change the 22K resistor to 1K you will still get about 0.5mA through it!
 
Measurement trumps simulation!

Are you taking into account that for the amp to work properly the VAS transistor "regulates" the voltage across the input transistor's collector resistor at ~0.6V.
Simulating at LTP in isolation will not work.

Here are the measurements from both channels of my NCC200 at quiescent with +/-40V supplies.
Voltages measured across resistors:
1K 0.500V 0.501V (= 0.5mA i.e. ~1/2 of the tail current)
22K 10.75V 10.66V (= 0.48mA)
620R 0.632V 0.628V (= ~1mA tail current)
68R 0.716 0.705V (~10.3mA VAS current)
220R in + rail 2.52V 2.49V
220R in - rail 2.86V 2.82V

If you change the 22K resistor to 1K you will still get about 0.5mA through it!
Hi,
I simulated with the entire amplifier - VAS, output stage, DC servo etc. The amplifier i built, the simulation closely matches the values measured, where i have used a transistor based current mirror. I have degenerated the emitters - so this may have a bearing - but not to such an extent as to why simulation varies so much from the measured values. I will have to have a think about this and ask on another forum.

Regards,
Shadders.
 
Measurement trumps simulation!

Are you taking into account that for the amp to work properly the VAS transistor "regulates" the voltage across the input transistor's collector resistor at ~0.6V.
Simulating at LTP in isolation will not work.
Hi,
I was using my amplifier design which had a cascade VAS (2 transistors) with 1kohm resistor in the emitter of the first VAS transistor, which modifies the circuit behaviour - trial and error to set the current, without going into equations.

I recreated the NCC200 as per the diagram on page 2 of this thread - and i obtain 0.61mA for the LTP input transistor collector current, and 0.39mA for the other collector. I see now what is happening - as you have said, the Vbe of the VAS sets the current through the input transistor collector, and the remaining current from the current source has to go through the other transistor.

In reading Doug Self book, it indicates that any mismatch in the currents cause distortion. I had assumed that using resistor based current mirror that varying the feedback transistor collector resistance would change the the current - it doesn't as long as the voltage drop is within specific bounds. For the simple VAS such as the single transistor with no emitter resistor, you can fine tune the current mirror to be equal by modifying the input transistor collector resistor.

Do you think you are fortunate to obtain a 0.02mA difference from a 0.5mA current, which is 4% difference ???

Did you modify the resistor values such as using a trim pot ?

I can see why using a transistor based current mirror is advantageous, as it ensures that the currents are very closely matched without any need for trimming, or reliance upon Vbe.

I also now realise that you are not modifying the currents by modifying the resistor (22k), but modifying the base width of the feedback transistor as you indicated.

Regards,
Shadders.
 
Thanks very much for the tecnichal explanation.
If i inderstand modifying the 22K resistor modifying the base width of the feedback (which is very sensitive)

And what do you recommend at the end ? and for the value?
 
Thanks very much for the tecnichal explanation.
If i inderstand modifying the 22K resistor modifying the base width of the feedback (which is very sensitive)

And what do you recommend at the end ? and for the value?
Hi,
The idea behind an amplifier is that it should be a wire with gain, so no distortion.

The use of a resistor based current mirror using standard components will introduce distortion. Changing the value of the relevant resistor makes the distortion greater. This is for the circuit provided on page 2.

I think the Avondale circuit i have seen previously has resistors in the emitters of the input stage, and uses same value resistors in the collectors of the input stage, which reduces the distortion.

In both circuits, changing the resistor will modify the distortion, so this is a way to implement the sound you prefer - selecting the value which sounds best for yourself.

Regards,
Shadders.
 
Hi, I see now what is happening - as you have said, the Vbe of the VAS sets the current through the input transistor collector, and the remaining current from the current source has to go through the other transistor.

Spot on!

Do you think you are fortunate to obtain a 0.02mA difference from a 0.5mA current, which is 4% difference ???

Did you modify the resistor values such as using a trim pot ?

Naim are a lot cleverer than some people credit them for.
Because the VAS runs hot, its Vbe is quite low. Actually, if you look at my measurements it must be 500mV (or 501mV in the other channel).

PS In an NCC it would be better to measure the current balance of the LTP using the emitter degen resistors - this will include the base currents.
Even Doug Self forgot to include the effect of the VAS base current in his first series of articles on amplifier distortion, later on I think he corrected this.
 
Naim are a lot cleverer than some people credit them for.
Because the VAS runs hot, its Vbe is quite low. Actually, if you look at my measurements it must be 500mV (or 501mV in the other channel).
Hi,
I think that the circuit as per page 2, and the unmodified NCC200 is designed to introduce distortion, to give the amplifier a characteristic sound. The distortion is predominantly second order which is more pleasing to the ear.

Using the diagram on page 2 of this thread : I simulated the amplifier once i was able to obtain the spice models, except for the input transistors (BC239C), hence used BC546B. The modification of the LTP resistor causes distortion as expected, but it is quite severe. The distortion rises from 0.024% to 0.29% (1kHz into 8ohms 68watts RMS) when the 22kohm resistor is implemented. This is a 10x increase. At 20kHz, at 6volts peak output into 8ohms, you can see the oscillation on the waveform.

Using the NCC200 with emitter degeneration resistors - which i think is the correct circuit, the distortion is 1.3% at 20kHz, 2.5watts into 8ohms. At 1kHz same power the THD is 0.01%. The distortion at 20kHz is unusual - the waveform is distorted, and does not seem to be harmonically related. The diagram used is from 2017. This was with 1kohm for both LTP collectors. Despite this, again the distortion is still 2nd order, and 10x the third order distortion.

Have you measured the distortion of the amplifiers at 20kHz ?. I realise there is not much information in music at this frequency - at 10kHz the THD for 2.5watts RMS into 8ohms is 0.33%, and the same kink is lessened, but still there.

Regards,
Shadders.
 
I only measured distortion at 1KHz and 1W (approx) for my NAP200 clone:

NAP200-C-Left-at-2-860-V.jpg


Theoretically the TR2 collector resistor is not needed and can be shorted. But this is what happens to the distortion in that case:
NAP200-C-Left-at-2-858-V-with-R-shorted.jpg


I always meant to measure it with equal collector resistors, but I don't have this amp any more.
 
I only measured distortion at 1KHz and 1W (approx) for my NAP200 clone:



Theoretically the TR2 collector resistor is not needed and can be shorted. But this is what happens to the distortion in that case:


I always meant to measure it with equal collector resistors, but I don't have this amp any more.
Hi,
Thanks for that. The simulation for 1kHz shows better performance, approximately half the distortion. What does seem odd is that the distortion is predominantly 3rd order for the measured, and 2nd order for simulated.

Were the above for unequal resistors ? I just simulated with 22k and nearly same distortion at 0.008%, and still 2nd order dominates.

Regards,
Shadders.
 
Hi,
Thanks for that. The simulation for 1kHz shows better performance, approximately half the distortion. What does seem odd is that the distortion is predominantly 3rd order for the measured, and 2nd order for simulated.

Were the above for unequal resistors ? I just simulated with 22k and nearly same distortion at 0.008%, and still 2nd order dominates.

Regards,
Shadders.

My first graph is with TR1 collector resistor = 270R and TR2 collector resistor = 5K6.
The NAP200 uses a tail current of approx 4mA.

The 2nd graph is with TR1 collector resistor = 270R and TR2 collector resistor = 0R.

I have never quite figured out what effect varying TR2 has on the technical performance of the amp. Alegedly it increases 2HD, but I would like to see the measurements. I did measure/estimate that the phase margin is a bit better with lower resistor values in TR collector, which I suppose is logical because it's operated with higher open loop gain and hence will be a bit slower. I can see how this would increase distortion at higher frequencies.
However, the audible effect is very apparent - seemingly out of all proportion to the measurements, which is most intriguing!!
 
My first graph is with TR1 collector resistor = 270R and TR2 collector resistor = 5K6.
The NAP200 uses a tail current of approx 4mA.
Hhi,
In using the NCC200 amplifier schematic, changing the current source to 4mA, and 270ohm, 5.6kohm resistors for the LTP collectors, 1watt (RMS) into 8ohms, @1kHz, has a THD=0.0095% which is just more than half the measured value. Again, this is second order being 4x more than the 3rd order component.

At 20kHz, THD=0.426% (1watt RMS, 8ohms load etc), but the 3rd order dominates - the increase in tail current has increased the slew rate, and the kink in the 20kHz signal seems to have disappeared.

With the NCC200 tail current at 1mA, the limited slew rate has an effect - and perhaps people like its effect as it gradually increases as the frequency is increased.

i set the quiescent current to 80mA, so if less this this will have an impact on the simulation results.

Regards,
Shadders.
 
by segregating power supply rails on the NCC200 so that the Front end and output stage are fed separately (as in the Voyager) I was able to measure the quiescent current of the 1st stage and output stage separately and it measured 11mA Fe & 38mA OP
Power supply rail voltage for Fe 55V output 53V

I have attached a copy of the NCC 200 Diag FYI

ncc 200 notes by Alan Towell, on Flickr

Alan
 
by segregating power supply rails on the NCC200 so that the Front end and output stage are fed separately (as in the Voyager) I was able to measure the quiescent current of the 1st stage and output stage separately and it measured 11mA Fe & 38mA OP
Power supply rail voltage for Fe 55V output 53V

I have attached a copy of the NCC 200 Diag FYI

Alan
Hi Alan,
Thanks - the figures seem to correlate with the diagram i have and the simulation. 8.5mA is used by the VAS current source (TR6), and the remainder from 11mA is the LTP current source TR3, with other devices using some current (diodes, base currents etc).

The diagram i have is from 2017. What is the voltage drop across the 0.22ohm resistors - i set the output devices quiescent current to 80mA, but from your figures for the output stage, the current seems to be 30mA or less.

I assume here that you connected the front end to measure the output stage ?? Or did you measure the total current and subtract the 11mA ? (so 49mA total current)

Regards,
Shadders.
 
Hi Shadders
The total was 11 + 38 so 49mA as you said
Alan
Hi Alan,
Thanks.

I ran the simulation again, with a 35mA output stage quiescent current, 1watt RMS into 8ohms :

1kHz : THD = 0.009%, predominantly even order distortion
10kHz : THD = 0.78%, predominantly odd order distortion
20kHz : THD = 2.24%, predominantly odd order distortion

The 1mA current source for the LTP is causing the issue for the high distortion for 10kHz and above. In addition, 1watt onto 8ohms at 1kHz with THD=0.009% is not that good either, but since it is even order, makes it sound nicer.

Regards,
Shadders.
 
Hi,
With the collector resistors at 1k and 22k, the distortion changes :

1kHz : THD = 0.009%, predominantly even order distortion
10kHz : THD = 0.82%, predominantly odd order distortion
20kHz : THD = 2.39%, predominantly odd order distortion

Below is the 20kHz 1watt r.m.s. output into 8ohms.

044toXz.png


If the current source is increased to 4mA, the kink is removed. Then again, there is not much material at this frequency, so it will not be a problem.

Regards.
Shadders.
 
Just a thought to add to the mix here. The value of the resistor in the collector of TR2 will have quite dramatic effect on the Miller capacitance of Tr2 and unbalance the input pair at HF. The Miller capacitance is non linear (Cbe varies with Vcb) and if I'm not mistaken will add a zero to the closed loop gain of the amp at a frequency dependant on this non-linear capacitance and the value of TR2's tail resistor. Increasing the resistor in TR2's tail will reduce the amps phase margin and 22k brings the zero down to less than 1MHz. That's deffo gonna sound different to no resistor in the tail, or even 1K. Just how that effects the harmonic distortion with real reactive loads is possibly beyond simulation.

John
 
Just a thought to add to the mix here. The value of the resistor in the collector of TR2 will have quite dramatic effect on the Miller capacitance of Tr2 and unbalance the input pair at HF. The Miller capacitance is non linear (Cbe varies with Vcb) and if I'm not mistaken will add a zero to the closed loop gain of the amp at a frequency dependant on this non-linear capacitance and the value of TR2's tail resistor. Increasing the resistor in TR2's tail will reduce the amps phase margin and 22k brings the zero down to less than 1MHz. That's deffo gonna sound different to no resistor in the tail, or even 1K. Just how that effects the harmonic distortion with real reactive loads is possibly beyond simulation.

John
Hi,
I used Doug Self book, 6th edition page 376 circuit for a simulated speaker load and the amplifier functions better with the complex load than a resistive load. Possibly because it has a greater than 8ohm impedance.

At 20kHz, 1watt r.m.s. into the load THD=0.036%, which is much better than the 8ohm load THD=2.24%. The reactive load causes the 2nd order component of the THD to be 5x than the third.

The above is the NCC200, 1mA tail current, and 1k and 22k resistors.

Regards,
Shadders.
 
Just a thought to add to the mix here. The value of the resistor in the collector of TR2 will have quite dramatic effect on the Miller capacitance of Tr2 and unbalance the input pair at HF. The Miller capacitance is non linear (Cbe varies with Vcb)

John

Agreed.
We must also remember that Tr1 collector resistor drives the VAS - which also has its nonlinearities.
What is not clear to me is if the 22K resistor's effect is to cancel some of this, or to mask the higiher order distortions by increasing the 2nd HD.
 


advertisement


Back
Top