Was looking around a bit and I saw some British guys mainly at pink fish media tweaking their Naim preamps in some older threads. Mostly their phono boards for parts quality only and changing regs. Here is a NA323 schematic reverse engineered from an enthusiast. As Q1 he names 5pcs ZTX384 (BC184) in parallel with individual 15R degeneration and a group's common 15R to ground. Between them feedback returns via a 470R. Those BJTs are not carrying any detailed noise plots for a range of source impedances, Ic, & frequency, in datasheets that I saw. They seem to be stated 4dB NF for 2K RS @ 200uA 5V, that's all.
The circuit in general is two loop feedback amps. NPN, PNP common emitter cascades, buffered by an emitter follower each. Splits Riaa into an HF passive EQ in between them, and LF shapes as active EQ feedback in the second stage. Output buffer Q8 is CCSed by a ring of two. There is a Zener referenced 20V local stabilizer with NPN pass element for B+. By designating R1 4k3, R2 100k, R17 200k, R23 82k, we go to mark 5 newer board. It simulates for 56dB gain and the attached frequency response.
Those PCBs are idiosyncratic, clinging to their 70s heritage. Bandwidth limited on purpose, tipped up in Riaa, and using tantalum polarized couplings. Noise specs of those BJTs are vague but I would roughly expect CCT En 2nVrtHz and strong enough 1/f ascend towards the lows since the currents in the 5 pack are just around 100uA each. 0.53nVrtHz of that we can attribute straight to the sum total degeneration. The feedback resistor can be almost ignored since its contribution is divided by the stage gain (30dB).
When Rs is 10-50R of an MC cart things can go much worse for En especially towards the lows. A-Weighted perception of ear in low SPL helps there, and Riaa's shape helps higher. Still bad. Would expect a dry mid-bass, & bright mid-high sound by its Riaa EQ, PSU, and self noise level-shape. What could we attempt to address if we were tweaking it or cloning it?
Well, 4-8 times better BJTs as for noise in 1st loop, less extra noise contributed from the resistors ladder, bit more gain to play LowMC too, bit more bandwidth, revised flatter Riaa, better PSU, RC decoupling of the pre-pre, MKP coupling caps. Attempted in simulation, looks like working. R3 can be a 10k trimmer to bias different types and to center the DC level at Q7's e. Still, when there is loop feedback employed, and transistor types changed also some patches removed, someone has to prototype the thing to finally say its stable. Can be wanting its pF step network patches back. Q3 (Q7) is running at 190mW on the original, a high constant for a TO-92. To drive the lowered value Fb resistor we need employ TO-126 on mini sink. Will go 600mW. Showing a nice FBET, but any good VAS grade transistor will do. At least a Philips made BD139. Can be made with Q1-Q5 BC550-Q6 560, less en less linear with 2N4401-4403, doing both well with 2SC2240-2SA970, doing best noise with 2SC2547-2SA1085. Japan to the rescue.
Edit: Also a 1mA per input's 5xBJT example CCT (shown with 2N4401s) has been added later. This one moves even further from the original as for bias and coupling caps values also. Better for noise with lower source impedance. If anybody NAIM user ever attempts the tweak ideas let us know.
The circuit in general is two loop feedback amps. NPN, PNP common emitter cascades, buffered by an emitter follower each. Splits Riaa into an HF passive EQ in between them, and LF shapes as active EQ feedback in the second stage. Output buffer Q8 is CCSed by a ring of two. There is a Zener referenced 20V local stabilizer with NPN pass element for B+. By designating R1 4k3, R2 100k, R17 200k, R23 82k, we go to mark 5 newer board. It simulates for 56dB gain and the attached frequency response.
Those PCBs are idiosyncratic, clinging to their 70s heritage. Bandwidth limited on purpose, tipped up in Riaa, and using tantalum polarized couplings. Noise specs of those BJTs are vague but I would roughly expect CCT En 2nVrtHz and strong enough 1/f ascend towards the lows since the currents in the 5 pack are just around 100uA each. 0.53nVrtHz of that we can attribute straight to the sum total degeneration. The feedback resistor can be almost ignored since its contribution is divided by the stage gain (30dB).
When Rs is 10-50R of an MC cart things can go much worse for En especially towards the lows. A-Weighted perception of ear in low SPL helps there, and Riaa's shape helps higher. Still bad. Would expect a dry mid-bass, & bright mid-high sound by its Riaa EQ, PSU, and self noise level-shape. What could we attempt to address if we were tweaking it or cloning it?
Well, 4-8 times better BJTs as for noise in 1st loop, less extra noise contributed from the resistors ladder, bit more gain to play LowMC too, bit more bandwidth, revised flatter Riaa, better PSU, RC decoupling of the pre-pre, MKP coupling caps. Attempted in simulation, looks like working. R3 can be a 10k trimmer to bias different types and to center the DC level at Q7's e. Still, when there is loop feedback employed, and transistor types changed also some patches removed, someone has to prototype the thing to finally say its stable. Can be wanting its pF step network patches back. Q3 (Q7) is running at 190mW on the original, a high constant for a TO-92. To drive the lowered value Fb resistor we need employ TO-126 on mini sink. Will go 600mW. Showing a nice FBET, but any good VAS grade transistor will do. At least a Philips made BD139. Can be made with Q1-Q5 BC550-Q6 560, less en less linear with 2N4401-4403, doing both well with 2SC2240-2SA970, doing best noise with 2SC2547-2SA1085. Japan to the rescue.
Edit: Also a 1mA per input's 5xBJT example CCT (shown with 2N4401s) has been added later. This one moves even further from the original as for bias and coupling caps values also. Better for noise with lower source impedance. If anybody NAIM user ever attempts the tweak ideas let us know.
Attachments
Enjoy yourself. Seems terribly complicated for a preamp device. Lot of very special, slightly obsolete transistors. Do you have a moving coil cartridge to use with this? I believe that is what it said it was designed for. On moving magnet cartridge, using my ears, I get approximately equivalent performance of 12AX7 PAS2 with a ST33078 op amp ($.37 at Newark).
While you're looking, what does NAIM mean? Is it a big european manufacturer of audio gear? I have a schematic from DJoffe of a NAIM power amp, which I have decided I'm not going to copy.
While you're looking, what does NAIM mean? Is it a big european manufacturer of audio gear? I have a schematic from DJoffe of a NAIM power amp, which I have decided I'm not going to copy.
I have arranged it for 60dB gain so it could play a DL103R. 2SC2240/2SA970 can still be found on ebay for $0.30 each. Essential for the first amp. Second stage can have the same or BC550/560C. The first amp will dominate the noise spec. That arrangement can take its noise down to a Pearl II level. 0.7nVrtHz equivalent input noise theoretically. True, the Hitachi or equivalent ROHM ones are more obscure. Under 2 Ohm Rbb' devices are not to be scattered in test concepts if not well stashed up. Can take it down to 0.46nVrtHz though. Naim audio is a legendary British hi-fi company that was going hand in hand with Linn for many years in the past. Had won royal distinction for its exports even. Don't know if there is meaning in Naim word, could have been conjured up to sound like ''name''. They do that in their products too, like Linn does. Making up words with weird spelling that refer to known meanings.
The ST33078 claims
Equivalent Input Noise Voltage (RS = 100Ω, f=1kHz) 4.5 nv/sqrtHz. Of course, my input resistance is at least 20000 ohms on magnetic phono cartridge, so that low impedence is artistic specmanship. It hisses less than a 4558 (OEM)however.
My noise floor in the living room is usually set by the factory across the street and the **** backup alarms on their forklifts loading trailers 70 M from my house. On Sunday mornings when they shut down, I've built a noisemaker for the dynakit ST120 amp. A pair of 200 mm muffin fans salvaged from PC power supplies installed like earmuffs around the ST120 grill. They blow on the 2mm thick finless power transistor heat sinks in case I forget to turn it off before 8 hours.They howl a little, possibly because of the finger guard I've installed in front of them. I'm operating off a 9VDC wall transformer on 12 VDC rated fans. I'm going to hide the amp behind the couch, and think about thermistors and salvage PC fets as a speed control device.
Thanks for explanation of NAIM. As famous as McIntosh or Dynaco, so famous nobody ever explained it before.
Last edited:
There are two things about what noise does. One thing is hiss that many shrug it off depending on the ambient noise level, total system gain (including speaker sens), and personal hierarchy of subjective priorities, the other thing is eating resolution of the very small MC signal. For MM its much easier. We generally must shoot for SNR equal or higher of what our cartridge's nominal output and internal resistance give. A 14R Denon will have 68.5nV Johnson at 30C for the audio bandwidth for instance. With 0.25mV nominal output it will have intrinsic SNR of 71.25dB. To put it in another way, DL-103R's 14R is 0.48nVrtHz. You would need the Hitachi transistor to fully serve that. 40R of the normal 103 model is 0.81nVrt Hz. You will need the Toshiba. 1kOhm example MM cart is 4.07nVrtHz so you will need a 5534 OP-AMP. A DL-160 is 1.63nVrtHz so you could parallel 4 5534.
Made by hand back in the 70's and carried on the same obviously in this range of modules. Curves are good they used to say, but I am no expert in PCB's so to reliably comment on their merit, I am a P2P guy.
Salas, I've modded quite a few sets of 323 cards, and the Prefix.
prefix_open2 | Flickr - Photo Sharing!
Whenever I used Wima caps for input and output coupling on the 323 boards they've always seemed to broaden but soften the bass. More texture than tants but at the expense of some 'snap' to kickdrum skin.
prefix_open2 | Flickr - Photo Sharing!
Whenever I used Wima caps for input and output coupling on the 323 boards they've always seemed to broaden but soften the bass. More texture than tants but at the expense of some 'snap' to kickdrum skin.
I'm sticking with 47uF silmic for the feedback cap, with teflon bypass and 10uf on the input.
I am wondering about the input loading resistance and how that interacts with the bias on the base of the input transistors. Am I wasting my time having a 47k in there or simply upsetting the input bias to the transistors?
I am wondering about the input loading resistance and how that interacts with the bias on the base of the input transistors. Am I wasting my time having a 47k in there or simply upsetting the input bias to the transistors?
Last edited:
interesting play ;
that reminds me of my old - never tried - puzzle ;
not wanting to pee in your pool
-maybe you'll find something interesting :
http://www.diyaudio.rs/blog/6/entry-70-everyday-electronics-apex-preamp-from-olde-times/
that reminds me of my old - never tried - puzzle ;
not wanting to pee in your pool
-maybe you'll find something interesting :http://www.diyaudio.rs/blog/6/entry-70-everyday-electronics-apex-preamp-from-olde-times/
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.