More to the point, do you know John Luckins. There is a SIM of the Naim circuit in among his other treasures.
It's important to have the other point of view on this. I remember my brother explaining to me that PP valve amps are not as well liked as SE valve designs because the distortion cancelling makes for an unmusical presentation. My ears agree with that. Despite SE amps having higher distortion they sound much less distorted if an optimum design ( most are not). There is a piece of very bad thinking that has persisted for decades. " The ears seems to be able to ignor second harmonic distortion ". That's actually is very wrong. " The ear seems to like second harmonic distortion ". That is even worse. What the ear actually likes is the second harmonic to be in the correct proportion. 1% THD with ideal distortion sounds clear and neither warm nor cold. In subtle ways ( very ) an amplifier of 0.1% or 0.01% of the same type will sound slightly better assuming class A type behaviour or close. The other distortion mechanism is relative to ideal curve distortion. The 1% THD amp could have 0% relative distortion in curve shape. Below is a graph of Quad 303. It gets close. Before anyone says, yes the whole system needs to do the same. Most other pieces of hi fi do except CD players ( ??? ). They never seem to work at HF, mine does almost . The alternative is 0% distortion. Anything else should be wrong if so ( I suspect that's as near as anyone got to the truth on this ). Naim were not stupid. It's not MSG, it's preserve the flavours already there.
How this works is like TV. As long as the ways colours interact looks correct the eye will see black when not there. Get the colour white wrong and nothing looks right even if the percentage is small. This was even worse when CRT as the colour was a dynamic quality ( as amplifiers ) for the reason that they used SE amps, BF720/721 are typical TV devices that live on and are excellent . When vision everyone is an expert. When sound people are not so sure. To be clear. Some music will sound great with some extra distortion. The majority of music will not.
A very strange thing. My friend has an Audiolab amp and Quad 303 ( 34 pre ). Mostly I find the Audiolab a bit closed in when Harbeth speakers. A 2D sound with no zest. The Quad always sounds OK, albeit slightly soft. I took him some BBC LS3/5A clones with a Audax tweeter to try. The crossover is more like Linn Kans. For the first time ever the sound was everywhere in the room on the Audiolab, the sort of sound I like. For the second time in my life I could like this amplifier. I came to the understanding that the Audiolab is a great amplifer that has one minor defect. It doesn't like speakers. I think I know why. The speaker will enter the feedback loop. It's not rocket science and has been said many different ways before. It is totally logical. It's not tonal if wondering. It is space and fizzy treble and wanting to watch TV more than listen to music when it's in use. Below is the 303.
How this works is like TV. As long as the ways colours interact looks correct the eye will see black when not there. Get the colour white wrong and nothing looks right even if the percentage is small. This was even worse when CRT as the colour was a dynamic quality ( as amplifiers ) for the reason that they used SE amps, BF720/721 are typical TV devices that live on and are excellent . When vision everyone is an expert. When sound people are not so sure. To be clear. Some music will sound great with some extra distortion. The majority of music will not.
A very strange thing. My friend has an Audiolab amp and Quad 303 ( 34 pre ). Mostly I find the Audiolab a bit closed in when Harbeth speakers. A 2D sound with no zest. The Quad always sounds OK, albeit slightly soft. I took him some BBC LS3/5A clones with a Audax tweeter to try. The crossover is more like Linn Kans. For the first time ever the sound was everywhere in the room on the Audiolab, the sort of sound I like. For the second time in my life I could like this amplifier. I came to the understanding that the Audiolab is a great amplifer that has one minor defect. It doesn't like speakers. I think I know why. The speaker will enter the feedback loop. It's not rocket science and has been said many different ways before. It is totally logical. It's not tonal if wondering. It is space and fizzy treble and wanting to watch TV more than listen to music when it's in use. Below is the 303.
This is a mild mod to an amp called RH34. One I rejected as I could by using an input pentode drastically reduce distortion ( 0.2% ) whilst keeping an ideal curve ( I had a bucket of pentodes to be sure any would work, some were only 50 % emmision. I only needed 20 % ). The experts without ever trying my design said it was a one off. No,it was tried about 20 ways. The difference good to bad 1 to 1.5 % THD full power when 5 % is not unusual ( and totally wrong when so ).
Let me describe the circuit a bit. People at DIY Audio insisted to discredit the inventor by saying it had been seen in 1937 ( Fender, Schade, RCA ). Not really. The seemingly unsuitable ECC81 RF triode has a very high Rp. Now the Rp of a triode is usually is very low. If a capacitor is added to the 510 R cathode resistor Rp will tumble. Suprisingly the outcome is not very different. However how the circuit works is this. Some call this circuit anode to anode or plate to plate feedback via the 71K67 ( a bunch of resistors, 68 or 75 K would be OK ). It is not because the anode will not work as a feedback node. The truth is the high anode plate resistance ( Rp ) will not stop the grid of the KT88 receiving feedback. The outcome is not unlike the KT88 becoming a VAS in the Naim amp. The ECC81 could be said to be part of a V to I converter and the KT 88 a I to V converter ( it looks standard and is not, V to I is a bit tenuous although on the scope it looks like it is ). The 390R and capacitor on the KT 88 cathode is the bias.
One pentode valve EF184 ( TV type ) has almost exact ECC81 curves when made up as a fake triode. That was my starting point.
In the end I got the curve shown moved up to 5 watts with a power reserve. I also got 620 mV sensetivity which is not true here.
One thing I did find is that the swapping to a conventionel layout gave near identical curves. This was using ECC 81 with cathode cap or ECC82 and ECC 82 cascode and 12BH7A. There are no free lunches. Alex Kitic will insist that the locked feedback makes in my words a Super triode. He claims the Rp total is about 80% that of the KT88 strapped as a fake triode. If so that might be a good enough reason to do this. The ECC82 cascode was a bit sad. After doing many curves I found it was almost exactly like the ECC81. One would think the sheilding of the valve No 1 to be almost a pentode. Not really. The pentode and triode have opposite curves that can be reliably cancelled. If the pentode can be at very high voltage some magic can happen ( or USE a CCS of PNP 350 V transistor ). The cascodes tend to flash over if the voltages too high. ECC82 will tollerate 100 V on cathode No 2. It's not enough.
The RH84/34/88/300B designs are as simple as valve designs ever gets if wanting to know more. RH34 I think is best. RH807 perhaps caused the Schade label. Mr Schade did use this idea about 1937 to tame the 807. I have seen it used when a 1960's RCA design idea looking much like a Dyncaco that didn't have it ( 7199 input pentode/triode ). It is interesting that the higher order harmonics of the 807 were even then thought unpleasent in 1937. The 2A3 was then and now the worlds better amplifying device. KT88 is not much better than 807. The EL34 to my ears is second only to 2A3 and better than 300B in typical use. A poormans PX25. If I build this amp again it might be with GU50. The ouput transformer was for a 845 here which was OK. I could get the matching right by ratios. That's why I show the coil combinations Sowter suggest. 807 was from a 1935 British Harries valve company design sold to RCA I think. 6L6 being another spin off. The same returned to the UK to become KT66. GU 50 seems to predate the EL34 as LS50. I wonder if Philips used LS50 as the initial design? Philips did orininate pentodes as far as I know circa 1927. Not sure about power pentodes?
I thought I should attempt show how a simple valve amp works. I doubt anyone will be as interested as me. No problem, it is intended as a one off open and closed story. It shows where the differences in ideas are widest. One has to remember the transistor industry just like CD had to convince the world " worst is best ". The only real way valves were inferior was damping factor and price. As one man I knew said " damping, that was the job of speaker at one time ".
My speakers are about 102 dB/watt if wondering.
A few clone kits and designs have already specified those drivers and as far as I know, without problem. They are TO220 size, 8A max. continuous current rated so there is plenty of capacity to drive a larger output stage.
http://www.onsemi.com/pub_link/Collateral/MJE15028-D.PDF
http://www.onsemi.com/pub_link/Collateral/MJE243-D.PDF
I prefer the original TO126 types MJE243/253 there but I guess this a matter of availability for you.
That could well be true, Andrew but the maximum current must still be derated or rather limited so that the nominal 50W package dissipation spec. is met for all types in the series at any likely Vce. I included the original MJE243/253 types for comparison as they are also 4A rated but the smaller package only permits 15W dissipation so there should be few problems here with ~40V rails and any of the 50W types.
SOA is given in the datasheet. That does take account of Ic, but max Ic is only a small portion covering the very low voltage end of the SOA.
And gets smaller when temperature de-rating is taken into account.
Driver SOAR can be analysed in the same manner as Output SOAR.
I have posted modified versions of Bensen's spreadsheet for BJTs, MosFETs and for BJT Drivers.
The 50W capability combined with the high corner where second breakdown takes effect means the mje15034/35 pair make enormous amplifier drivers.
And gets smaller when temperature de-rating is taken into account.
Driver SOAR can be analysed in the same manner as Output SOAR.
I have posted modified versions of Bensen's spreadsheet for BJTs, MosFETs and for BJT Drivers.
The 50W capability combined with the high corner where second breakdown takes effect means the mje15034/35 pair make enormous amplifier drivers.
Last edited:
I had thought transistor models for SIM might map transistor behaviour more precisely than the range of unit values given in datasheets. There has been no interest in that. Like you I want to find out in more detail what is going on with the LTP tweak.
Anyway the datasheets for BC546 and BC239 give Base-emitter (Vbe) values of 0.55 volts minimum (0.62 volts typical BC239) and 0.7 volts maximum where Ic =2 mA and Vce =5 volts.
What the Vbe values are at the values of 0.499 mA and 0.49 mA as per your measurements for TR1 and TR2 is uncertain. The respective Vbe's have to be measured. With TR3 the figure might be 0.62 volts.
It would find it helpful to know the Vbe values in these three cases.
I took some Linn Kan speaker clones to a friends house. We used a NAP 250 and a NAD 3020. Having used the NAD as a preamp I can say it is a low distortion design. The NAP 250 shows no sign of second harmonic distortion as a sound colour. In fact it is ( as always ) as dry as a bone and rather correct in how it sounds. The NAD uses a single transistor input stage. I have no graphs so have to guess it is what it should be. That is an amp with a pentode/transistor curve distortion albeit low in quantity. The NAD is very nice and not far behind the NAP 250 considering the price difference. Where they differ is the warmth. The NAD is rose tinted and lumpy, not to he extent of saying it's wrong. Reading people here they assume the NAP second harmonic is raised to the point where it colours the sound. No, the other way. Raising the second harmonic so as NOT to colour the sound by loosing the second harmonic. If the second harmonic is correct the fourth also. From memory the NAP 250 fourth is slightly wrong in proportion.
My advice is try larger than 22K if you like. DC offset < 100 mV. ( A Rotel RA931 I repaired was 75 mV on both left and right ). There is a theory a bit of DC offset sounds better. It makes sense if you think about it. Inwards even more so ( - 100 mV ).
My advice is try larger than 22K if you like. DC offset < 100 mV. ( A Rotel RA931 I repaired was 75 mV on both left and right ). There is a theory a bit of DC offset sounds better. It makes sense if you think about it. Inwards even more so ( - 100 mV ).
Here is a good read. As he says, Nait 2 is more or less the same generic circuit. As he also says the Naim circuit is a low distortion design. I had better data than this. Sadly lost.
Naim, NAPs, Naits, distortion and bias-current settings
Naim, NAPs, Naits, distortion and bias-current settings
Shorting the 22K makes little difference to the dc offset in my amps. There's a limit to how much it can be increased due to reduction in the Vce of TR2.
I have my 002CCN up and running and initial impressions are that it sounds very similar to the NCC200 (using the same power supply). Both of them sound a touch "thickened" compared to my Leach amp. As mentioned before the 002 is "upside down" compared to the NCC and has a complementary o/p stage. The pcb layout (originally for a DX Blame) is quite poor. My current thinking is that the two Naim tweaks (22k TR2 collector resistor and the 1K impairing the VAS current source) do seem to be responsible for the "Naim sound". Further experimentation should provide more clues.
I have my 002CCN up and running and initial impressions are that it sounds very similar to the NCC200 (using the same power supply). Both of them sound a touch "thickened" compared to my Leach amp. As mentioned before the 002 is "upside down" compared to the NCC and has a complementary o/p stage. The pcb layout (originally for a DX Blame) is quite poor. My current thinking is that the two Naim tweaks (22k TR2 collector resistor and the 1K impairing the VAS current source) do seem to be responsible for the "Naim sound". Further experimentation should provide more clues.
Yesterday I posted a valve amp design to show ideal graphs. At 1 watt although in the 1% THD range the input and output waves look very similar. Surprisingly the middle wave in this group is not greatly higher in distortion. This should make you think. Is OK for the eye OK for the ear? The truth is that is exactly right and was stated as fact in the latter 1950's. When stated the Quad 57 speaker was available. It's distortion is that of an amplifer of good quality. It was suggested that 1 % THD of the type I showed is hi fi. To give a safety factor for less favourable distortion curves 0.1% became the standard. Naim is typically 20 db better than that. And yet people will say " could do better ". That's bonkers.
The second graph is a very high grade 100 kVA diesel generator on light load. The flat top wave is not how folklaw has it. It is simply how voltage is regulated by core saturation.
The first wave is the 42 watt anode dissipation KT88 trying for 8 watts. What you can see is the curve of the shunt created triode ( very rare in power valves ) unable to hide the pentode curve. Bottoming is happening. The THD is perhaps > 10 %. It is far more than the diesel generator. The difference is the KT88 will still sound like music. The speaker due to it's similar non linear field strength will do the same. If we make the speaker very sensetive the valve amp will be the lessor evil. If the sound of the 50 Hz 100 kVA sine-wave was put through the speaker at 1 watt it would be like a buzz saw.
The simple thing to say is. What you see is what you get. The valve amp I gave as example has a perfect set of matched harmonics.
Here is something. If I was to use Quad 57's with my 8 watt valve desing and NAP 250 on a blind test I doubt anyone could easilly guess which is which. Put the Quad 303 in to be more than the flip of a coin. The big shock would be the valve might on all counts win. SE valve designs can be giant killers. This is because the real sound is at about 1 watt to 5 watts. Below 1 watt the valve is the very purist class A with a DC bias to the transformer to keep in in the most linear range on the BH curve. This produces or retains micro detail. The ESL is not suited to > 15 watts so is happy with 8 watts. It will go surprisingly loud.
The trend is to use a current mirror instead of 1K 22K and lets say 47 R emitter ( source ) resistors in the long tail pair input pair. ( 1935 Blumlein patent. The ECC series valves for the use of in typical circuit ). The funny bit is to use JFet inputs to have an exponential curve in the device. That's a bit mad and yet is state of the art. I looked for someone who converted a NAP 250 to JFet. It would be interesting. The weird thing is these people insist the JFet sounds better although they seem to be trying to make it more linear!!!!! Very well respected people say this. They and Analogue devices all state the same. It is not very good science. I like JFets as they need nearly zero current to work. That means the feedback is easier. I have always wondered how a JFet DC servo works. It must be like a valve that some current flows. If a high grade capacitor as the gate it can't. Having said that it's only JFet op amps I have studied. TL074 make equally good comparators as lets say a LM324 and often as good as real comparators that only can sink currrent. That is a surprise. If a valve I have no trouble as the grid is in the electron flow if such things exist ( one is never sure, good enough for me that they do and even have mass ).
415V AC would indeed be a loud buzz though most speakers, although not for very long
On the subject of distortion I am pretty much in agreement with you. I find it amusing that we are repeating the 1970s Japanese amp situation with all the these ppm designs and lots of transistors straightening out the large signal transfer function. Of course there is also a resurgence at the mo of just how wonderful old Japanese amps are/were!?
TBH sticking JFETs and a cascode in the IPS of a blameless type amp does not make much difference to the sound IME. Replace the VAS transistor with a FET though and things start to get interesting!
After further listening to the 002CCN amp at daytime volume levels (as opposed to previous listening at family-in-bed levels) I like it a lot. Better than the NCC!!
You've mentioned this amplifier a few times now. I haven't heard of it before but with a handle like 002CCN, it seems unlikely to be a commercial product so do you have any details, illustrations or references to design?
[/I]
Probably the same circuit retaining the NPN quasi output, changing the other transistors in the LTP /Vas to opposite polarity and swapping the supply lines. I guess the Vbe transistor would still be NPN.
Anyway there was no reply to my query about vbe of the transistors in the small signal stages but as it turns out I don't need them anyway.
Increasing the value of the TR2 22k collector resistor will cause a greater Miller capacitance to be developed across the base emitter junction of TR2. From my calculations this will result in a reduction in the pole frequency from 18kHz to around 14 kHz. This may seem controversial but I am happy to provide the detail for people to scrutinize for themselves. I am grateful that you provided the information which made this possible.
As to the difference between the collector currents for TR1 and TR2 of 0.499mA and 0.49 mA respectively, the difference of 0.1 mA more or less I would put down to the fact that forcing a voltage across a capacitor causes a current to flow between base and collector. Logically it does not flow into the 22 collector resistor.