advertisement


Starfish revival

I am updating the PCB to a 4layer design (with added combined footprints for films and tants plus the possibiliy to thermally couple the relevant transistors) with hopefully better layout and have a question regarding a possible GND-plane: I plan to have the signal tracks on the bottom layer, the inner layers carrying the signal- and power-GND tracks, and the top layer is almost empty and could be used as a copper shield under the components - if that would be a good idea I don't know since I am a hobby amateur. So I ask for your input on this! Please have a look at attached pictures:

8Zg03Yh.jpeg


The copper plane on the top layer is only connected to PWR_GND at one point (the one in the middle of the PCB). The PWR_GND tracks from there will be routed seperatedly:

H5jNCS8.jpeg


pink = power_GND
yellow = signal_GND

Please let me know if you think using the top copper layer as a shield is a good idea!
 
Naim philosophy is to have minimal grounds, and F-A shielding - they use mains cable for pre-power inter-connect!!

I think a ground plane shield is good idea, and the Maclaren 02 pre-amps all had them back in the early 1980s, as do things like Krells.

But you need to get input here about what people want - it would be a very un-Naim like design choice in a pre-amp to use a ground plane.
 
  • Like
Reactions: jpk
I also upgraded my 7x regulator boards to feed power to the preamp boards: the 317/337s are upgraded to LT1963A/LT3015, and I made another board with LT1763/LT1964 and 317/337 as pre regulators:

l1Xv6QB.jpeg
 
I finally managed to assemble the preamp:

B8SgpRs.jpeg


The knob I made on my lathe with a mother of pearl inlay :cool:

lxSKldD.jpeg


The PSU boards are CRC, each rail feeds a 317/337 regulator, from there I have separate LT1763/LT1964 regulators for each preamp supply. The blue transformer is for the digital and relay stuff, digital-GND is connected to chassis (as is mains PE). Signal-GND comes in from the DIN/RCA sockets, goes through the input selector relay board to the star-GNDs on the preamp boards, from there to the PSU boards. Analog and digital GND are not tied together inside the amp but can share grounds through mains PE if the source has signal-GND connected to PE. I tried a couple of different GND wirings, but hum at full volume (and also some noise) doesn't go away even with shorted inputs. Also the relay attenuator clicks a bit at every step even with no signal. I measured DC at the input pin (after the output cap on the 729 and without signal) and switched MSB: the difference was something like 0.015mV. I played around with the relay timing as described here but didn't manage to get rid of the clicks. I guess it's EMF from the relays: with added diodes across the coils it's much better but still disturbing. Disconnecting the blue transformer and powering the digital circuits from a bench PSU didn't change anything. Also powering the preamp board from the bench PSU didn't change anything. The preamp sounds really great, but with no signal I can hear white noise which increases with volume, and at full volume there is quite some hum. But as soon as music plays it's not disturbing at all...
 
Ingenious, a lot of work, congratulations!

Re: the hum: it could be those little axial inductors in the CLc output filters are actually picking-up the field from the transformers. Have you tried simpy shorting each inductor with a wire bridge on the underside of the regulator PCB..? If that drops hum, you have (an) answer. If not - we'll have to think a bit harder!
 
Last edited:
  • Like
Reactions: jpk
Actually - there's another possibility, one that addresses the noise issue as much as 'hum'. I need to draw-out the equivalent schematic of your wiring pictures above, and think about it.

It's about the whole routing of the 0v returns which you show in white wires - both for the reg boards, and between the reg boards and your starfish pcb. I think there might be a small adjustment needed to the wiring.

It would help if you could share the reg board layout with me for the '0v' layout - via PM if you wish, to keep it confidential
(which I would respect - post such in open forum & it'll end up on internet sales sites sooner or later... : / )
 
Jeff Young said:
Hi Jun,

I'm still digesting some of this, but I have a couple of observations.

I can't see any value in separate regulators for the negative rail. It supplies 5 CCSs (3 double-BJT types and 2 resistive types), so there's heaps of PSRR built in there.

The positive rail is another story.

On dropping the 729/TA buffer, I think we want the filtering, but probably don't need the buffering (who runs a tape loop these days?). I've not yet wrapped my head around whether or not the Air Guitariste is providing the filtering without the buffering, or is dropping both.

The current source mod for the gain stage is replacing a tight-feedback (high impedance) constant current source with a lower impedance but voltage-controlled current source. I haven't read up on it yet, but that's going to throw away some PSRR, no? (You could recover that by feeding the LED with a CCS -- an E-452 between R9 and the LED would do nicely.) But either way it's starting to become a different preamp, and if I'm not re-creating an icon then I'd rather just use my HPA-1 clone. ;)

Doing the 729 CCSs would appear to have even less value.

Anyone know what the story with C7/C8 is? Is that just because a 120pF capacitor is hard to find?

Cheers,
Jeff.
Click to expand...
Naim used 100p and 22p in parallel
 
martin clark said:
It's about the whole routing of the 0v returns which you show in white wires - both for the reg boards, and between the reg boards and your starfish pcb. I think there might be a small adjustment needed to the wiring.

It would help if you could share the reg board layout with me for the '0v' layout - via PM if you wish, to keep it confidential
(which I would respect - post such in open forum & it'll end up on internet sales sites sooner or later... : / )
Click to expand...

Thank you so much for your kind help! I have no problems sharing my layout (those guys will find enough other designs to copy), so here is the PSU board:

8Y90kaj.png


The 0V tracks are highlighted. Here is the 7x regulator board:

5qbVOZw.png


The GND tracks for the positive and negative regulators are completely separate but could easily be joined. The voltage trimmer footprints are populated with fixed resistors and zeners for the 317/337s. The inductors can be linked out as you suggested...

One question: I have the input and output caps not as in the original starfish. I measured the voltage across them and have drawn all 4 of them the other way round. Here is the original starfish:

L58C8wQ.png


The positive legs point inwards - but in my preamp the negative legs point inwards. Except my input caps are populated with SMR films as recommended on your website. Could the polarity be related to the clicking in the relay attenuator? I understood the slightest DC would be audible...
 
The polarity of C2 & C112 shown above is correct, due to the base biasing resistors. The circuit in post #1 is very different.
 
My latest schematic is in #120. The one in #1 is guess-work before I saw the starfish schematic. Some values are different, so I still wonder about cap polarity...
 
Thanks for pcb layouts and schematic - i think the reason for noise and hum is the way the 0v side of things is wired up; you have both raw PSU side and quiet-side 0v sharing a common impedance (long white wire, wide-open 'loop'). This means the decoupling caps on the output side of the regulator boards are not doing very much bypassing at all, and their filtering effect is entirely-defeated. I am slightly surprised you are getting good audio at all, and not odd sounds, a bit of 'motorboaring and similar!

It also means those caps on the regulator board are not doing their job - which is to decouple the various stages on the starfish board, at all - they are referenced to the raw PSU 0v, when they should be as close to each section they serve and referenced to the star ground on the starfish pcb. Similar, for the Zener diodes you have used to set the output voltage. The '0v' end of those need to be as close as possible, electrically, to the pre-amp board '0 star. As do, the '0v' ends of all related downstream bypass capacitors. Yet your 'regulator' board has no direct 0v links on the output side..!

So: while this is bad news - the best way to improve matters with your PCBs as built:
  1. Remove all the the wiring from reservoir caps 0V output to the reg board.
  2. On the reg board, tie the two '0v' input-side traces together - use a short length of thick bare wire to join the 0v ends of the pair of zeners.
  3. Use one wire from the centre of this link, back to the 0v output Reservoir cap PCB. Twist this with your red and blue supply wires, so you have just one, 3-wire twisted cable - from the reservoir pcb to each reg board (one 3-wire connection per channel.)
  4. Use a second white wire from the bare wire 0v 'link' centrepoint, to the Starfish 0v centrepoint. Makes this wire as thick and as short as possible! Mounting the reg board above the starfish board might help you facilitate this.
It's still far from perfect, but with the steps above changed:
  • then the decoupling caps , and the LM3x7's own regulator reference (0v end of the zeners) are more-correctly arranged with respect to the task of each part;
  • the supply ( and decoupling path) loop areas are better-separated; and very much smaller, both of which is critical
...and it should work very, very much-more as-intended : )

HTH.
 
Last edited:
jpk said:
My latest schematic is in #120. The one in #1 is guess-work before I saw the starfish schematic. Some values are different, so I still wonder about cap polarity...
Click to expand...
In the circuit above, R103 & R104 set the right-hand side of the input capacitor (and the base of Q102) to somewhere between the positive supply voltage and 0V. Let’s assume it’s at 12V. The other side of the capacitor is close to 0V via R101. Therefore, that DC (no signal) voltage across the capacitor dictates the direction you have used above and it is correct.

Looking at the circuit in post #120, the base of Q2 will be somewhere between +12V and -12V. Therefore, close to 0V, but it could be slightly negative, which seems to be what you’ve measured, or it could be slightly positive. It’s inconsequential, so don’t worry! The magnitude of the voltage across the capacitor will certainly be less than 12.
 
Thanks a lot! So I will link out the inductors and feed the 0V line through the regulator board as suggested by Martin - and report here...
 
If you try what is in my longer post #132 just above (text just updated for clarity) - then you can leave the inductors in, just as they are.

There are far bigger issues at play - which is all to-do with the present wiring architecture.
That is the whole source of your hum & noise issues.
 
Last edited:
  • Like
Reactions: jpk
Ok! So what is my mis-thinking about the 0v layout, and how should the GND tracks on the 7x regulator board be routed if I would re-design the PCB? I thought the starfish is all about star grounding... Now it seems I need some kind of series grounding. I would love to understand / learn! Do I need to route the clean side 0v lines of each of the 7 regulators out to the edge of the board where the out-connectors are and tie them together there and run one wire from there to the preamp's star-GND? Wouldn't that result in multiple GND-loops on the 7x regulator PCB?
 
I'm away from my own webspace access at the mo - when next in range, I'll scribble a few diagrams & post, to show what is going on : )

The take- away is: the physical geometry of circuit layout really matters. The map (the schematic) is not the territory (real world side effects; the mountains, chasms, and difficult terrain) More .. soonish.
 
  • Like
Reactions: jpk
martin clark said:
The take- away is: the physical geometry of circuit layout really matters. The map (the schematic) is not the territory (real world side effects; the mountains, chasms, and difficult terrain) More .. soonish.
Click to expand...
Or (same facts, different viewpoint) the physical layout can add extra components that aren't on the original schematic.

Tom Christiansen (Neurochrome) does quite a lot of modelling of layout effects, by adding these pseudo-components onto his simulations.
 
Yep - it's long been essential, to understand such things, to get the best results.
Also a solid reason why surface-mount was invented; and subsequently took over the world
 
You must log in or register to reply here.


advertisement


Back
Top