1. Things you need to know about the new ‘Conversations’ PM system:

    a) DO NOT REPLY TO THE NOTIFICATION EMAIL! I get them, not the intended recipient. I get a lot of them and I do not want them! It is just a notification, log into the site and reply from there.

    b) To delete old conversations use the ‘Leave conversation’ option. This is just delete by another name.
    Dismiss Notice

Starfish Improvements (IMHO)

Discussion in 'd.i.y.' started by john.luckins, Dec 18, 2010.

  1. john.luckins

    john.luckins pfm Member

    Three days ago I replaced all of the regulators in my BOM 1.2 Starfish with Cap (Vbe) Multipliers. This was simple and took me two hours to do. It involved removing each of the regulators along with its two associated voltage setting resistors (the 121 Ohm and 931 Ohm ones) and fitting a single transistor in the regulators place and a resistor between its collector and base. The tricky bit was calculating the correct value of resistor for the transistor and its load current so the filtering was low enough in frequency and the voltage across the transistor was in the right ballpark.

    I've done some (hand)written instructions and drawings here for anyone who would like to have a go.



    These transistors are not direct pin for pin replacements for the regulators and two of the leads have to be interchanged - a little careful bending and shaping makes this possible. I added the new resistors to the underside of the board and soldered one end directly to the pad of the collector of the transistor, the other going to the recently vacated adjust pin resistor's pad.

    If you do give it a try you'll need:

    8 X BC550C
    6 X BC560C
    12 X 47 kOhm resistors
    2 X 470 kOhm resistors

    I haven't perfected the resistor values but I think you can get away with 56 and 560k or even higher if you don't have the others handy. There may be other suitable transistors. These have two important features, high Hfe (circa 500) and Ft (200 MHz) at 10 mA collector current.

    The Results?Well I judge this to be a significant improvement. Even when I just changed the +ve and -ve regulators on the first stage of the buffer the sound was simply more musically coherent and enjoyable. Percussion was more natural in its decay and each disk seemed to have more of its own character. I'd recommend doing just these four regulators first to see if you like the results.

    Eventually I tore myself away from listening and replaced all the remaining regulators in one go. Two days of listening since and I am not going back to the regulators. This change is fundamental. It is an improvement not in HiFi terms - you don't get more quantity of anything but you get depth and realism, even Prat, dare I say it. It is a subtle improvement in Hi Fi terms but very significant musically. I've listened to some of my most difficult disks and found some tracks listenable that I never had before. I find myself listening to whole albums right through when before I would skip the end of tracks or jump to another disk. I even managed to listen to Pink Floyd "Animals" all the way through for the first time ever on CD and enjoy it. I've also used the repeat function on my CDS 3 for the first time in a couple of "have got to hear that again" moments.

    Anyway I can only praise Jiim for having the foresight to create such a versatile board. It would be great to know if anyone else gives it a try and agrees with my verdict.

    I have two more improvements to follow in this thread. One is to the input to the Traco, the other a basic component improvement after some comparisons, so watch this space.

  2. Fatmarley

    Fatmarley pfm Member

    I replaced all the regs in mine with Teddy regs and thought it was worthwhile. I would have liked to have done a direct comparison with standard regs, but never got round to it.
  3. john.luckins

    john.luckins pfm Member


    Very similar, perhaps even better.

    Where did you get the volts from ? Or did you run the Starfish at slightly less than the +/-12v?

    I figure the Linear Regulators are more trouble than they are worth when being ratttled by incoming 300kHz pulses and when the DC is already regulated. They have a negative input impedance and have to have both their inputs and outputs snubbed by a capacitor with high ESR (the Tant). The load is sensitive to 300kHz too as the 47uF across the circuit is pretty useless at that frequency. In my view that is why the Cap multiplier is so effective.

    I'm considering trying a second cap multiplier further upstream, just after the filters, or perhaps even before them, i.e. immediately after the Traco?? Its either going to be diminishing returns or a noticable improvement.

  4. john.luckins

    john.luckins pfm Member

    P.S Matt

    I see you were selling an R core transformer. I'm cursing because I missed it.

  5. Fatmarley

    Fatmarley pfm Member

    Hi John,

    I didn't run mine off a Traco, so I had plenty of voltage headroom with my linear supply. I would have liked to have built one with Traco's to compare, but the old hag made me sell it all.

    I've only just sold the R-core. I thought people would snatch my hand off, when I put it up for sale, but I got very little interest...
  6. mr_phil

    mr_phil If it isn't broken, try harder

    I would love to get rid off the 1MHz noise that permeates everything and is on the speaker terminals at 20mV. Is there an alternative to the Traco? Would taking it and the LC filters off the board and into a metal box (with lots of non electrolytic decoupling) be worth a try?
  7. Jo Sharp

    Jo Sharp Pulls on doors marked push

    1. you can go for the option of fitting 2 Tracos which will give you plenty of headroom.

    2. This is what I did to allow me to fit TRs after the Tracos (one +TR,one -TR) which then feed the 317/337regs.

    I'm very interested in your findings on replacing the linear regs though - I've wanted to try this for a while but have not yet built enough TRs to give it a go.

    Keep us informed of your other mods please.

  8. Bemused

    Bemused What's this all about then?

    Very interesting thread, thanks guys. In a slightly different dirrection I replaced the linier regs on my gain stages with the reg boosters, a little gyrator / cap multiplier circuit to sweep away some junk before the regulator takes over. I found an improvment but it was subtle and not profound. The HF mush is still present. I am not looking to remove the boosters, infact I need to find time to fit more boosters to the filter stage. I would sum up the sq gain as a little finesse not that the fish is lacking at all.
    Other minor bodges to the other wise standard fish are a Hackercap dedicated supply feeding STR and a little play with with Wimas as input / output coupling. I did loose a touch of voltage with the boosters, about 0.3 volt from memory. I need to also find time to try out the improved differential mode filter.

  9. john.luckins

    john.luckins pfm Member

    I agree with you Mr Phil on taking the Traco and the filters off board. Its certainly worth a try. By replacing the Regulators with cap multipliers I am halving the current consumption of the Starfish and this will allow for more severe filtering. It also means the current pulses in and out of the Traco are much reduced. The only danger in going off board is the need to keep the power & ground paths short. A small screened box sited just above or below the main board would do the trick. It could contain the Traco and filters just like Mr Tibbs's original board did. Simple tunable notch filters to take out the 300 kHz and perhaps its higher frequency spurious oscillations from the Traco would be ideal. Another option is a very carefully constructed Cap multiplier right at the output of the Traco. At these frequencies notch filters can be made whilst by adding only an Ohm or two to the supply line resistance.

    I found that I was picking up 5-20 mV of 300 kHz just about everywhere I probed. After reading up on Switrch Mode supplies I can see that this is a combination of common and differential mode noise and is pretty well guaranteed when probing with an unbalanced scope probe. It isn't necessarily indicative of the true quantity of switching noise getting through, just the method of measurement and radiation, although some may be getting through. At 300 kHz plus there are different earth potentials all over the place due to lead and trace inductance.

    What really concerned me was putting a (just) critically damped LC filter right up against the negative input impedance input of the Linear Regulators. The input capacitor and its ESR have to be just right to avoid it going into oscillation at every switching pulse. The cap multipliers don't have enough feedback to have a negative input impedance so they are not going to suffer from this.

    The input to the Traco is also negative impedance i.e it draws less current the when you provide it with a higher voltage as its power input needs are constant. The length of lead feeding it and the capacitance (value and ESR) across its input terminals is very important to its stability. But that is my next "improvement" on offer here - later.

    Jo - Do the TR's have Linear Regulators on them or are they Cap multipliers with no significant feedback? If they have Linear Regs then are they providing anything useful here?? After the change I've just done I'm keeping Linear Regs away from the circuit now.

    I wish I had the time to try all of this...

  10. john.luckins

    john.luckins pfm Member


    Your solution may be better as long as it removes the potential for oscillation at the input of the regulators. It does use more volts and current though.

    I've done some playing with coupling caps too and the differential filter deserves special attention.

  11. Bemused

    Bemused What's this all about then?

    If my attempt is better then its just a little luck, I am not so clever with this stuff. It just ocured to me that little booster circuits would be a simple-ish implication and not requiring the excesive drop out of teddyregs etc could keep the original Traco setup. Cheaper than TR / STR too :)
    Also easly reversed if things get pear shaped.

    I really need to find time to finish this off and find out how it performs.

    I must say your approach has intrigued me since you first mentioned it some months back.

  12. Arfer

    Arfer pfm Member


    I have not built the circuit you are talking about in this thread, but I have had experience of DCDC converters from work.

    I found that once the rabbit is out of the hat there is no getting it back in. You refer to 300KHz, is this when viewed on a wide bandwidth scope and probe i.e. >200MHz. It is probably a spike at a repetition rate of 300KHz, as opposed to a sawtooth etc (which it may appear to be on a low bandwidth scope), and that spike probably contains frequencies in the 20MHz ish region, thats why the datasheets for DCDC converters only specify the noise up to about 10 to 20MHz because after that it increases, and its down to how the DCDC is connected on the board. This noise will get everywhere as common mode noise (I think its common mode, either way its a sod to get rid of). This is due to the large instantaneous currents involved in switching regulator circuits.

    The best way to minimise this is at the source by using ground planes and good grounding techniques at the Traco, preferably it will have a metal can that can be grounded too. You will probably also need common and diff mode chokes etc, with v low esr caps. Basically you need to treat the DCDC a bit like an RF circuit. Do you have a diagram of your Traco part of the circuit, and maybe some pics of your layout, maybe people could then make some suggestions.

    Even if you did build the Traco into a separate box you would still need to clean the noise up at the source to start with, if you didnt then your interconnecting cable will act as a nice aerial to spread the noise all around your system where it can get into the feedback loops of your amplifiers and make thinks sound hard and harsh. All these
    problems are the reasons why not many hifi manufacturers use switching regulators, and those that do, such as Chord employ experts in the subject to make them work properly.


  13. Bemused

    Bemused What's this all about then?

    A further interesting post thanks Ian.
    My head is starting to swim a little.

    I remember jacking the Traco supply voltage up when when I built my fish a year or more ago. Back then I was just trying to distribute heat evenly between my STR feeding the Traco and the Traco as I had a large supply overhead. I was not aware of the noise back then and it may be worth a revisit and seeing if switching noise atenuates a little with higher supply voltages / lower currents. Also my system is now so much more revealing since adding a pair of Carls Hackernaps, back then I still had a NAP150.

    Ian, I must say the Starfish preamplifier we are all talking about is not at all "broken", its just that a few guys are trying a little guilding so to speak.

    My understanding is the schematics are not normaly posted as Jiim their designer wishes to restrict them to builders only but here is a picture of a fish.

  14. mr_phil

    mr_phil If it isn't broken, try harder

    I came to the Starfish rather late so may have missed earlier discussions, please indulge me. Firstly there is no fault, but maybe there is potential for improvement. What is the reason for the Traco? Is there another way of doing it? I ask because the high frequency rubbish we all observe can't improve the sound (can it?) and it will be 'better' if gone.

    My limited understanding is that the Traco is there to enable the 'fish to float from earth but if I don't tie the transformer to earth then it floats anyway doesn't it? Arfer believes once you have the noise it is very difficult to remove, seems very likely to me and others belive the quantity of noise is very dependent on layout and that a screened Traco box is hard to implement. So with a Traco we are stuck with noise - probably

    Back to my 2nd question, is there a way of doing the power without a Traco? I am willing to experiment within reason if anyone has any good ideas

    Thanks, Phil
  15. Bemused

    Bemused What's this all about then?

    Well I guese the simplest is to pull the differential mode chokes and just stuff a conventional two rail supply in keeping the Nert, maybe a Hicap type of thing, taking the fish back to its Naim roots.
    You could use a three rail supply and remove Nert.

    I understand the Traco more or less in the same manner as you, ticky to get a gound loop though the Traco.
  16. Jiiim

    Jiiim bodger extraordinaire

    This is one of the build options mentioned in the build manual - It's quite possible to run the fish from a conventional linear supply. You MUST be carefull with earth potentials though - directly plugging in something like a hicap would be unwise. You may also end up feeling slightly disappointed.

    The use of a traco has many technical and sonic advantages - but it is also priceless seeing the look on someones face when you say the preamp they've just listened to is powered by a crappy wall wart.

  17. Bemused

    Bemused What's this all about then?

    Hi Jim

    I quite agree and have no wish to tear out the Traco, I just mentioned that its quite easy / quick to go linier if Phil wishes to experiment.

    But the HF noise makes us all wonder if the fish could get up a step if this issue could be addressed.
    Maybe we all like the fish because of the HF noise :)

    Would you perhaps be giving us a release of the next bom / snert / diff filter as a christmas present :D
  18. mr_phil

    mr_phil If it isn't broken, try harder

    Thanks Bemused, you have expressed my feelings very well - addressing the HF noise issue is an opportunity to enhance an already great preamp and I am willing to experiment if pointed in the right direction
  19. simonbrown

    simonbrown The Dogfather

    Somewhere in here you will find my report on the effect of adding additional LC filter stages. Jiim has mentioned an upgrade in progress around inlet filter/NERT, which I will certainly be looking forward to.
  20. Arfer

    Arfer pfm Member


    Could somebody post a pic of the underside of the board, I assume its double sided but no more than just the two layers.

    I found the secret circuit, amazing what google can do :)

    Didnt see too much high frequency on there (only had time for a quick glance though), electrolytics are not going to do much up in the tens of megs. Might be able to do something with some bits of copper tape to provide low impedance paths for a proper HF filter, maybe even stick the traco on a bit of veroboard with a copper foil plane stuck on the opposite side of the board to the tracks, then connecct this to the PCB.



Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice