In many ways the Baxandall improved quasi comp is quite cute but this is no time for a baby show.
For now, I am curious about your Class A driver stage since this could explain a few things particularly if you have taken the driver emitters via resistors to the opposite voltage rails.
The issue with the quasi OPS is that it's like the EF1. With a complementary OPS you can use EFII:
Distortion In Power Amplifiers
Sorry I mis-described EFII as having class A drivers.
Distortion In Power Amplifiers
Sorry I mis-described EFII as having class A drivers.
Last edited:
Having never tried this I am not sure how close to the Vbe TR5 it would get. It is just a constant current source using a transistor rather than diodes. On paper the better solution. If you build a replica and put it near the driver or output you could see how well it tracks the TR5. 220R is just about enough in theory. Some would make the reference transistor same as the TR7. I have had the opposite problem when building current sources. They react to heat. That seems a good strating point. Some amplifiers were made this way. What this lacks is the rather cute over or under compensating trick that the 27R to collector of TR5 brings. Or rather doesn't bring. Having built the offered solutions to this problem I have to say most designers live on Fantasy Island. No way do any exactly track, D Self included. The heat sink and box ruin that illussion. The Quad 303 side steps it. Anywhere between 5 and 10 mA standing current works. That is easy to ensure. The Quad is what Naim should have used. Crimson did and Crimson was a gaint killing amp. A work of genius.
If the TR5 rubber diode has one silly advantage it is the setting resistors need a little current which fracionally speeds the T7/8 shut down. I have mostly drawn this to show a more ideal TR7/8 shut down ( 100R + 220R + 100 R ).
Dave I agree about class A stages not needing to be ultra fast. I have said it so many times myself. Here is a thing to ponder. Low Ft transistors often have poor linearity and the reverse is often true. BD139/140 bought with care at between 100 and 190 MHz ( Philips of old ). MJE 340/350 said to be 6 MHz. BD139/140 should have slightly better gain. BC639/640 often have more data, said to be the same device.
I identified what I feel to be a weakness in the Avondale. There should be a >220 uF capacitor near TR9/10. If 2200uF high grade it might be a vast improvement and reduce the need for high grade in the main PSU. The 52/35 V split is to show how the BD139/140 could be used. 1K8 and 100uF will be nice and quiet. 52 V will make someting like the NAP250 to happen. The output stage will be working like a series regulator. The 3V ripple that might happen can be a useful >10V above the signal. I think the weakness was done with good intentions. Les saw the 100 uF as a way to quieten the TR1 to TR6 section. What is left out is stabilising the TR7 to TR10 because of the 220R. Any inductance in the cables or PSU caps needs to be nullified. What I think a German guy forcefully put forward is making PSU low grade ( normal types ) and the TR7 to 10 decouplers high grade beats all high grade. Then the defect of the low grade capacitor becomes a quality of excellence ( CLC filter ). Douglas Self has the decouplers taken to the PSU Star earth.
Nigel - I has some KSA/KSCs on order with Mouser. I will give one a whirl in place of the MJE350.
Interesting about the decoupling. My 002 amp has 390uF Pana FA local decouplers on the front end and also on the OPS. Maybe that is another reason why I prefer the sound to the NCC? The layout is also sort-of-star grounded for these caps.
Interesting about the decoupling. My 002 amp has 390uF Pana FA local decouplers on the front end and also on the OPS. Maybe that is another reason why I prefer the sound to the NCC? The layout is also sort-of-star grounded for these caps.
I built an EF2 in 1980 - a modified version of an RCA application note. I had picked up a RCA handbook (which I still have) at the time. In that decade I built some magazine projects and modified output stages to this format.
One of these was relatively low powered about 15 watts which allowed me to use 30 MHz 50 watt TO220 output transistors that Motorola had introduced.
My benchmark at the time had been a DIY Class A amplifier. I could not pick any difference between the sound of either.
I try to have an open mind about the compromises that have to be made in designing amplifiers and I do not write off quasi complementary amplifiers.
Relative to complementary stages there are differences in doping layers of NPN and PNP types. This will lead to a degree of HF performance disparity between the two, albeit this is less with modern high fT types.
This makes such less of an issue for many but not all designers - certainly not Julian Vereker.
In their earlier years Naim used to limit the bandwidth of their amplifiers with an RC input network to 40 kHz. They were using standard production BDY56 10 MHz power devices - good for the day but pedestrian nowadays.
The current Naim amplifier range varies in bandwidth up to 100 kHz using devices made to their specifications - unlike some I do not regard the claim as hype. In normal manufacture high fT devices are more fragile so the requirement is for a high speed high power device, would require a larger chip.
Anyway the business, along with Baxandall Improved Quasi Complementary output stage still lives on despite a change of ownership and the passing of Julian Vereker. The Bentley car company has adopted Naim in car sound systems.
While the availability of high power high fT has helped progress with complementary output amplifiers, the same can apply with quasi-comp stages. The really high fT ones from way back used to be NPN who knows what the capability of the current Naim types might be.
Yes you are right my VAS is NPN = MJE340.
The special Naim transistors were Sanken die packaged (and selected?) by Semelab.
Naim uses complementary OPS in some of their later amps and the big ones use FETs.
The special Naim transistors were Sanken die packaged (and selected?) by Semelab.
Naim uses complementary OPS in some of their later amps and the big ones use FETs.
Hello, as already showed you my first NAP clone was made using McBride layout and i intend to redesign my own board and i have a question regarding grounding design as i read Naim amps are very fussy regarding groung and i've seen different aproaches:
1. local grounding for every channel connected to power supply star gnd. Speaker has own return to star gnd
2.more extremist grounding: one localgnd for frontend goes to power star gnd, feedback cap goes to power main star gnd. Zobel goes to power main star gnd and naturally speaker goes to power main star gnd.
So in the second case almost everything is connected to power main star gnd and not so much small signal local gnd.
If i look on the Avondale's NCC200 implementation seemts to be my first case described and also the one that i'm using right now in my V 1.0 🙂
Do you have any recommendation for the best approach on grounding ? I will be using a single power supply for both channels.
Cheers!
Adrian
1. local grounding for every channel connected to power supply star gnd. Speaker has own return to star gnd
2.more extremist grounding: one localgnd for frontend goes to power star gnd, feedback cap goes to power main star gnd. Zobel goes to power main star gnd and naturally speaker goes to power main star gnd.
So in the second case almost everything is connected to power main star gnd and not so much small signal local gnd.
If i look on the Avondale's NCC200 implementation seemts to be my first case described and also the one that i'm using right now in my V 1.0 🙂
Do you have any recommendation for the best approach on grounding ? I will be using a single power supply for both channels.
Cheers!
Adrian
Hi. How about this kit:
Black Box ?clone Naim NAP200 Amplifier Kit DIY Power Amp Kit 75W 75W L1511 29 | eBay
Is it good as well, anyone has experience with it?
Thanks
SB
Black Box ?clone Naim NAP200 Amplifier Kit DIY Power Amp Kit 75W 75W L1511 29 | eBay
Is it good as well, anyone has experience with it?
Thanks
SB
Good luck with your Vas update I assume you have a stash of suitable caps for C.dom. Any ideas when Naim switched to complementary and FET's.
This is how a FET version might look. TR6 68R can be made 100R. That might sound softer/nicer. Double FET's can be one set or three. When three it approximates the bipolar loss. One set should be fine for most people. No need for 1K + 22K to TR1 + 2 as FET's are rich in second harmonic.
Last edited:
This is how a FET version might look. TR6 68R can be made 100R. That might sound softer/nicer. Double FET's can be one set or three. When three it approximates the bipolar loss. One set should be fine for most people. No need for 1K + 22K to TR1 + 2 as FET's are rich in second harmonic.
Hi
Very nice, did you simulated the circuit or these is a option to someone may do that.🙂
Greetings
I think they use complementary OPSs on some amps but not all. Maybe SuperNAIT.
The Statement would appear to use (loads of) FETs based on the observation that the TO247 centre pin seems to be the output. and some of the comments made in an article/interview about it. IIRC they run the OPS without feedback.
Please don't quote me on any of the above. In general I only have a passing interest in Naim and I have never owned any of their products (except a rather boring demo CD🙂). FWIW I like Naim amps but the rest of their kit is nothing special.
Hi Nigel,
9-10 mA of current may be not enough for charging the gates of the FETs, especially when you have two in parallel. I thought about that too some time ago, and gave up on the idea. For higher current you'll need different VAS transistor. That's why my design used BJT drivers after the VAS.
Lateral MOSFETS work great by the way. They have generally higher output impedance, but in exchange you get rid of the emitter(source) resistors and you forget about gm doubling, thermal problems and so on. Their Vgs is about 0.8V for some 150mA of quiescent current. So, not very difficult to bias either.
My version sounded different, but not worse than the clone. It's more expensive though. The only proven not fake Lateral FETs that are available nowadays are from Semelab. Not cheap 🙁
Hi,
This is the first time I see this kit. It looks as good as any. My guess is that having the power supply capacitors closer to the amps, and boards that copy 100% the originals is good thing. Still, nobody can tell what is the quality of the boards, what is the copper thickness etc. As usual, semis can be safely considered fake.
If you buy one let us know what you think.
Possibly there would be a feedback loop that excludes the output devices, anyhow this sort of amplifier is more for millionaires. I won't quote you if you don't quote me.
Re your KSA, KSC devices that you mentioned ordering, I looked back to post 1679 where Andrew T mentioned a super version of the MJE340 in the form of KSA 1381 as having a decent datasheet. I wonder if this is your "it" device.
The link given was to Fairchild Semiconductor. http://docs-europe.electrocomponents.com/webdocs/12cf/0900766b812cfc1e.pdf
I have found wanted information in Fairchild datasheets - simply not there with On Semiconductor. BC239 and BC546 refer.
There is also decent information for Sanyo 2SA1208 mentioned by Nigel P.
Looking at KSA 1381 the Figure 1 graph - "Static Characteristic" shows the relationship of Ic against specimen Ib values over a range of Vce.
The relationship is reasonably linear for most types of transistor but this device is exemplary in terms of flatness of the line - Andrew T is on the money.
Figure 6. "Base on Voltage" shows the relationship of Ic to Vbe at low Vbe - this resembling a curve, rather than a line. However the anomaly can be ironed out since Ib - see Fig 1. will be very sensitive to minor movement along the Fig 6. Vbe curve.
The better such movement can be confined to small segments of the curve the more the result will approximate to a straight line.
Yes KSC3505 (and the 2SA1381 complement) are the ones I have ordered.
I do wonder if linearising the amp further than with the ZTXs might be pointless with the 22K fitted. Or perhaps the 22K value can be reduced in order to preserve the balance of harmonic distortions, but at an overall lower level?
I must get my mate's spectrum analyser on the job!
I've also got some old Tosh 2SC3423 parts, not sure how they compare to the Fairchilds (soon to be ON). I am using the 2SA1360 complement as the current source. Surely this matters as much as the actual VAS device???
I do wonder if linearising the amp further than with the ZTXs might be pointless with the 22K fitted. Or perhaps the 22K value can be reduced in order to preserve the balance of harmonic distortions, but at an overall lower level?
I must get my mate's spectrum analyser on the job!
I've also got some old Tosh 2SC3423 parts, not sure how they compare to the Fairchilds (soon to be ON). I am using the 2SA1360 complement as the current source. Surely this matters as much as the actual VAS device???
2 mA is more than enough to drive FET's. Two things people get very wrong. The 500pF to 1nF is not connected to 0V. It is connected to speaker plus. This effectively bootstraps the capacitance and it vanishes. Were this not true the Hitachi amp of circa 1979 would never have worked. It you exspect 100 kHz full power to be impossible then think again. Audio FET's are not the same as IRF number series. So far we are talking CgS ( gate source ).
Error two. CgD ( gate drain ). This is seldom quoted and is important. I have found 5.6 mA to be a reasonable drive current for 4 devices. Certainly 9 mA will be enough. Be care to test the low current version as you might prefer it. What you are doing is calling the VAS current a resistance and CgD as a capacitance to an effective ground ( +/- rails are same as ground in this ). Thus a simple RC filter.
Where my " what if " designs fails is that the Naim and others with single VAS are very poor at driving symetrically. Thus the VAS and TR1+2 need to be enhanced just to get OK driving. Symetrical driving is not possible. In a twin VAS we have advantages, double slew rates and near perfect sourcing and sinking of current. When we enhance the Naim design we get instability and DC offset as our reward.
My 2mA TR1+2 ( + 2 x 24 R ) #1675 gives similar linearity as the Naim ( better ) and slightly better drive. DC offset is about typical of a good design. I got rid of current sources to prove one can if careful. #1675 was a 15 inch bass amp with filter. It was very OK as a general amp.
Error two. CgD ( gate drain ). This is seldom quoted and is important. I have found 5.6 mA to be a reasonable drive current for 4 devices. Certainly 9 mA will be enough. Be care to test the low current version as you might prefer it. What you are doing is calling the VAS current a resistance and CgD as a capacitance to an effective ground ( +/- rails are same as ground in this ). Thus a simple RC filter.
Where my " what if " designs fails is that the Naim and others with single VAS are very poor at driving symetrically. Thus the VAS and TR1+2 need to be enhanced just to get OK driving. Symetrical driving is not possible. In a twin VAS we have advantages, double slew rates and near perfect sourcing and sinking of current. When we enhance the Naim design we get instability and DC offset as our reward.
My 2mA TR1+2 ( + 2 x 24 R ) #1675 gives similar linearity as the Naim ( better ) and slightly better drive. DC offset is about typical of a good design. I got rid of current sources to prove one can if careful. #1675 was a 15 inch bass amp with filter. It was very OK as a general amp.