advertisement


The unlikely star-candidate - Linn LK-100

yairf

Trade: Teddy Pardo Audio
Hi all
The legend has it that once upon a time in the previous century, Ivor Tiefenbrun, met with his then best mate, Julian Vereker, and handed him a document containing a list of drawbacks that Linn engineers found on Naim's amp design. The list included:
- No front-end regulation (resulting in audible noise-floor).
- Asymmetrical input stage (resulting in increased distortion).
- Inherent tendency to oscillate (resulting in blown tweeters).
- the preamps are based on mechanical switches and knobs, which tend to lose contact after some years.
As the legend has it, Julian kicked his friend out of the door, suggesting him to focus his efforts on designing speakers and turntables.
They say that at this point in time (1985), Ivor decided to introduce amplifiers under the Linn badge.

The older-generation (non-chip) Linn amps are pretty similar to the Naim design, with some modifications as per the above list.

I have seen very little views on Linn amps on PFM, most of them were not extremely warmhearted. It seems that Linn amps are always describes as "Lean", "Dry" and so forth.

A while ago, I gave a close look at the LK-100. It's a very nice piece of equipment, typically available on ebay for 250 UKP - almost half the price of its counterpart - the NAP-140, and with some small and simple changes, it can be made equal, or better than the 140.
For a start, the LK-100 has a very good build quality, aided by a nice trafo and good choice of components. In addition, its extremely easy to dismantle (good news for bodgers), and there is no limit to the mods you can apply to it.

See the schematic below:
ooRexQ8.jpg


Note the following changes. Start with those I consider to be mandatory:

1. C201 - the input stage low-pass filter, set to 2.2nF - about 8-10 times the normal values we come across ! no wonder the thing has no highs !!
Change it to 390pF, even 470pF will be fine to restore high frequencies.
2. Feedback loop: the feedback resistors are extremely small - while Naim has 27K--1K resistors, the LK-100 has 3.9K -- 150R, which is actually a good thing. However, the feedback cap is only 122uF - no wonder the LK-100 has no bass !
I actually prefer the lower-valued resistors as-per Linn, however, the feedback cap MUST be replaced to something in the range of 1000uF - use a good 'lytic bypassed with a decent Poly cap and you should be fine.

In addition:
1. The main caps are OK, and you can leave them, I usually replace them with Mundorf M-Lytic - the height of the chassis limits cap choice to 10,000uF, which is just fine.
2. The front-end regulation is implemented with a bog-standard zener-transistor arrangement. You can always replace it with your best choice of regs.

The result - the LK-100 has now restored its lost "oomph", and it really kicks butt! It can deliver plenty of drama to demanding speakers without introducing its own drama - it is extremely quiet and smooth.

Given its ridiculous selling price, I think it's really a star candidate !

Cheers, Yair
 
Its a truly state of the ark design, you would be better of scrapping the whole lot and putting in a Doug Self blameless type design in. 250UKP is quite a lot of money for a box and traffo. There does not appear to be any method of setting the bias current, I suspect it is a pure class B design and is never really going to cut the mustard compared to a more modern design.
 
I liked the careful choice of words. In any case, bias is set by R209 / R210 - it's a bog standard Class A/B design, just without a pot that will CHANGE it. I have seen this in many other amps (look at Cyrus 1/2 as an example).
Cheers, Yair
 
R209 and R210, being equal, will have a voltage of 2Vbe drops across them. To any current in the out put stage you need 4 Vbe drops, so it looks like pure class B.
 
That would be true in the case of Darlington output stage, however the LK-100 output is using the Sziklai arrangement, so you only need 2 x VBE.
 
Ah, I did not see the 0.47 in parallel with the diodes D210 and D211 still not guaranteed to be AB though.

Edit. If you want to improve it get rid of the three parallel driver transistors and replace it with one http://www.profusionplc.com/pro/gex/pcatdtl0?ipartno=2SB649AL-T60-K

and its complement http://www.profusionplc.com/pro/gex/pcatdtl0?ipartno=2SD669AL-T60-K

replace Q208 with a 2SD669 mount this and the driver transistors on common heatsink, not the same as the output transistors, make R209 variable, you can then set the output current properly and it will be stable.
 
1K & 2n2 gives a perfectly reasonable -3dB point of around 70KHz (depending on the driving impedance).
At the bottom end -3db at ~9Hz should be OK as long as there are no other poles near this frequency.

I'm not a fan of this o/p stage, difficult to stabilize and not the best sounding. Looks like they had some fun with it (C206, 7 and R217).
 
1K & 2n2 gives a perfectly reasonable -3dB point of around 70KHz (depending on the driving impedance).
At the bottom end -3db at ~9Hz should be OK as long as there are no other poles near this frequency.

I'm not a fan of this o/p stage, difficult to stabilize and not the best sounding. Looks like they had some fun with it (C206, 7 and R217).


If we simplify the overall schematic into a basic negative-feedback amplifier, ignoring the compensation routes, we get the following graphs:
ujjfVeFl.jpg


OTOH, if you look at Naim's standard configuration, you get this:
zPV7Umcl.jpg


Which looks far better, especially at the bottom end.
Cheers, Yair
 
1K & 2n2 gives a perfectly reasonable -3dB point of around 70KHz (depending on the driving impedance).
At the bottom end -3db at ~9Hz should be OK as long as there are no other poles near this frequency.

I'm not a fan of this o/p stage, difficult to stabilize and not the best sounding. Looks like they had some fun with it (C206, 7 and R217).

Yes, I have been playing with a MOSFET current dumping design in LTspice which uses Sziklai bipolar MOSFET combination, a few issues getting it tamed.
 
C206 just looks wrong! And C202 vs C203-R207 is a peculiar little wrinkle; layout-related?
Also no cap across the Vbe bias spreader - I wonder why.
 
C206 just looks wrong! And C202 vs C203-R207 is a peculiar little wrinkle; layout-related?
Also no cap across the Vbe bias spreader - I wonder why.

Yes, C206 is more likely to induce ringing than anything else, maybe it was bit soft sounding and needed brightening up. Looks like someone designed it for the sake of being different rather than to do a specific task of amplifying a signal correctly. Where is the inductor//8 ohm for the zobel network?, its not like they do any harm and are highly beneficial as far as stability goes.
 
C206 just looks wrong! And C202 vs C203-R207 is a peculiar little wrinkle; layout-related?
Also no cap across the Vbe bias spreader - I wonder why.

I agree about C206 being too large, I havent yet fiddled with it - keep in mind it's there to replace the series coil, so you do need it there.
As for the bias cap, AFAIK its usually there for biasing cross-over considerations, and should have no effect on frequency response, so I guess you can do without it.
Yair
 
Yes, C206 is more likely to induce ringing than anything else, maybe it was bit soft sounding and needed brightening up. Looks like someone designed it for the sake of being different rather than to do a specific task of amplifying a signal correctly. Where is the inductor//8 ohm for the zobel network?, its not like they do any harm and are highly beneficial as far as stability goes.

C206 replaces the output coil, and creates no ringing whatsoever, as long as the speaker impedance at 20KHz is lower than ~ 20 ohms, so should be fine with most speakers. Ribbons OTOH, may suffer loss of energy at high-frequencies.
 
C206 replaces the output coil, and creates no ringing whatsoever, as long as the speaker impedance at 20KHz is lower than ~ 20 ohms, so should be fine with most speakers. Ribbons OTOH, may suffer loss of energy at high-frequencies.

I really do not like the circuit at all, you couldn't use a passive pre-amp with a high impedance pot, its not biased correctly, output stability is questionable, I do not think its any improvement on the Naim. Its just a box to put a decent amp in, just rip the damn thing out and start again!
 
Sorry mate, none of those claims can be justified clear cut. Im actually fairly positive you would speak differently, had this thread mention "Naim" instead of Linn.
I can tell from hands-on and ears-on experience, its a lovely piece of kit (after the mods, of course).
Cheers, Yair
 
Sorry mate, none of those claims can be justified clear cut. Im actually fairly positive you would speak differently, had this thread mention "Naim" instead of Linn.
I can tell from hands-on and ears-on experience, its a lovely piece of kit (after the mods, of course).
Cheers, Yair

I do not think Naim designs are that good either, Doug Selfs series on amp design are a good place to start, the blameless design has a lot of good ideas in it, Bob Cordells book on audio power amplifiers should be on any DIYers reading list as well, well worth checking out if you have not got it.
 
As for the bias cap, AFAIK its usually there for biasing cross-over considerations, and should have no effect on frequency response, so I guess you can do without it.
I don't agree with that ;)

A cap across the Vbe bias spreader cannot ever be too large - it minimises several sources of error in the OP stage bias current: e.g. from temperature, from Vbe efffective impedance vs frequency, from PSU rail noise; and has a wholly-positive effect on driver base charge suck-out also.

And whatever cap one uses - regardless of size - cannot provoke instability here either. Win - win-win. So there's no reason not to use one.
 
C206 just looks wrong! And C202 vs C203-R207 is a peculiar little wrinkle; layout-related?
Also no cap across the Vbe bias spreader - I wonder why.

C202 is the miller compensation cap for the VAS, completely standard method of loop compensation. C203 + R207 are jsut tweaks on this.
The loop looks quite heavily compensated - driven from the high impedance current mirror. This is a high feedback amp. Probably sounds quite "solid state".
 
Heavily compensated was really my point; C203/R207 straps things down at 270Khz, presumably because the miller cap isn't enough on its own - or a larger miller cap causes some other problem. It just looks like a patch for something else.
 


advertisement


Back
Top