advertisement


Possible B4 Buffer Improvements

john.luckins

pfm Member
I just thought I would share my ideas and findings on this great little circuit. In its standard form the B4 has replaced the Starfish in my system. I think as I age I’m tiring of the Naim preamp circuit sound and moving from the snappy and up front presentation to something a little more “pipe and slippers”. I’m very happy with the B4, fed by a CDS3 and into NAP135 clones. To me it is a clear upgrade from my old NAC 52 as well but better than a passive pre.

Now I’ve read a great deal about how good the B4 PSU rejection is and concur with the figures given. However I can easily hear the changes made when I go from a Toroid to an R core transformer driving my B4, so I know there must be interaction between the supply and signal here that matters and isn’t insignificant to the sound (for me).

Very specifically I have found that I much prefer the B4 when hooked up to a large (432VA) R Core Transformer compared to a very nice 500VA screened Canterbury Windings Toroid. I struggled to admit this as I had bought the Toroid specifically to use with my final B4. It is really nicely built by Terry but the R core is somehow better sounding, particularly in the bass.

Separating supplies
I have now got a completely separate supply for the constant current sources D2 and D4 that drive the LED’s and the cascode/bootstrap. I got an improvement by using a Cap multiplier (PFM gyrator) off a separate transformer and CLC filter rectifier.

Cascoding the constant current diodes
After a bit of googling I found a useful paper by Walter Jung on Constant Current sources – here

http://waltjung.org/PDFs/Sources_101_P1.pdf

Figure 9A shows how a constant current diode can be given improved rejection at low frequencies by being cascoded by just one jfet. I did this on D2 and D4 using 2N5486 devices and could hear an improvement to the sound. Two devices per channel can give the B4 a 10 to 20dB improvement in PSU rejection in the audio range. I’d recommend you give it a try, its any easy and reversible change and mines in for keeps. I got my 2N5486’s from ebay.

http://s869.photobucket.com/albums/ab257/John_Luckins/?action=view&current=CCSCascode.jpg

I’m considering cascoding the other two CC diodes (D1 and D3) to see if the effect is also audible here, but I will need to add extra LED’s into the bootstrap/cascode to give enough voltage for it to work. This makes me a bit wary as the extra LEDs will slightly degrade things elsewhere – but hopefully not by too much.

Input/output caps
After a lot of experimentation I’m using a 100kohm input resistor and 0.47uf input capacitor. It is a Russian Teflon. Big, but only about £6 and a really obvious improvement over polystyrene and polypropelenes. On the output I managed to track down a couple of 3.3uF Teflon/Oil caps. All these caps have excellent clarity and balance. The difference they make is not marginal.

Shunt regulators
To finish off I’m now building separate Shunt regulator supplies to replace the LM317/337 regs and to feed D2 and D4. That means 8 separate regulators on a separate board. I’m using the latest version of the Salas Shunt regulator circuit form DIY audio forum. One extreme option would be to feed each of these regulators from separate transformers or windings. I’m considering rewinding some old R core transformers myself to do just this at low cost.

Thereafter a good case, a remote controlled Lightspeed Attenuator, careful layout and maybe screening of the transformers will see me to the end of my search for the best preamp I can put together.

Of course I'm interested in what other people have done to their B4, chances are I've missed a trick or three. Don't be shy.

John
 
Nice work John, piccies of everything please.

When I come to collect those giga-amp boards i'll be dropping off my B4 if you don't mind :)

Stefan
 
Where did you get the R Core from? I've only seen some on ebay (shipped from China) and from a company in France.
 
I am slightly surprised that cascoding D2 and D4 was worthwhile. I looked at this in simulations, and it looked like the dominant effect limiting the PSRR was the Early effect in the boostrap transistors.

I will look at that again.

The comparison between R-core and toroid does not surprise me. R-cores have much better isolation; the common-mode noise current injected into the earth is smaller. Toroids have a lower output impedance, but that hardly matters for this circuit. Note that the PSRR is not relevant to the issue of noise injection into the earth, which I suspect is the dominant difference between the types of transformers.
 
The piccies will be embarrassing at the moment Stefan. No case yet and flying leads all over the shop, but if you like that sort of thing I can oblige shortly. Perhaps I'll stick to close-ups!

I got 3 X R cores from Hifi tuning in Germany about 5 years ago before the stopped doing them. The company moved on to doing turntables - shame. They are 27-0-27v 432VA ones based on a Chinese core. Two are running the output stage of my power amps and sound very nice there, the other jumps between preamp and Headphone amp duties. The R cores I might rewind are the ones from China on Ebay. I've heard rumour that there are some Japanese ones around and companies in India claim to be able to make them. The chinese ones I have are screened and I've found them to be pretty good. The voltages aren't always as you would want them, hence the rewinding.

I've never tried Selectronic R Cores though they have a good range and I was thinking of 2 X 500VA for future power amps. I understand that the core is less prone to saturation with an R Core compared with a Toroid and this may explain the sonic differences. I'm concerned because I've just ordered 2 X 1200VA Toroids from Canterbury windings and I'm hoping they can match or better my lower power R Cores in a revised HackerNAP! I really hope they do.
 
To be honest PD I wasn't expecting to hear any improvement from the jfet cascode on D2 and D4 either. I thought any effect would be way above the audio range and due mainly to the lower capacitance of the resulting 2 terminal device. With the right jFETs it is any easy change to make and doesn't screw up the boards looks too much.

I've read a paper today that concurs with your observation that the Early effect dominates and suggests output transistors with a much higher Early voltage. It too relies on simulations to come to this conclusion (nothing wrong with that -probably more reliable than my ears). I would assert that the cascode can do no harm as it increases the impedance of the CCS and hence improves the PSRR even if this isn't the dominant factor. I'm keen to see what others find (hear) when they put such a cascode on. Jungs graph on the link is quite convincing to me and I prefer the cascode to be there.

I was wondering PD, is there any way to reduce the common mode noise cureent injection when using a Toroid. Does balancing the incoming power to earth help perhaps or do we have to use filters before the transformer?
 
Les is the man to speak to about transformers.

You can add a shield winding to a toroid design, and that will help; this is a single layer of wire, usually fairly fine, between primary and secondary, which is earthed. It greatly reduces the capacitive coupling between primary and secondary.

Balancing the supply might not make that much difference; the trouble is that the primary will be several layers, so the layer nearest the secondary is probably at one end of the winding, and I would expect this to dominate the electrostatic coupling.
 
Slectronic sources trafos from India:shilchargroup.com <[email protected]>
I'm happy with them though I don't have the theoretical background as you fellows. I use them in pre. PSU's and intend to use them for the output of my HackerNAp.
Nevertheless I thought one only needed 12-0-12 40va or so for the B4. Am I wrong?
 
12-0-12 40VA would be fine. I don't know ifJohn compared his big transformers with more standard size ones, so can't say anything about the benefits of an oversized one. 15-0-15 would also work. I suspect you could go as small as 10VA without anything too bad happening.

As a generalisation, a smaller transformer should have less interwinding capacitance, so may actually inject less noise current into the earth; this is a tradeoff against the supply impedance.
 
Size doesn't help here, 432VA was all I had hanging around and as PD says it may bring with it a few problems that a smaller one wouldn't and is not good VFM.

I'm really saving the big one for a headphone amp and even there its a bit OTT.

I wondered where Selectronic got their cores from. I'm going to try some (smaller ones) for the B4. Do they come with an earthable screen Harthold?
 
My 300va R trafos from Selectronic come with the "jaune" wire to connect the shield to earth. Goagle for Selectrònic France
Cheers
 
Another type to look at are split bobbin designs; they have no overlap between windings so very good isolation; the down side is that the leakage inductance is a bit higher, so they will give a higher source impedance. For a pre-amp supply, this is a good tradeoff.

I think Les used to do a split bobbin type.
 
I am using STRs (4) with no onboard regs, Mundorf Supremes on input, no output cap (hackernap has input cap) and lightspeed attenuator and think it is bloody marvellous. I use a toroid for what it is worth
 
Thanks PD, I have a couple of 100VA Split bobbin EI cores that I have used to wind inductors in the past. I'll try winding my own 15-0-15 before going for some small R Cores.

I'm glad someone else likes the lightspeed attenuator. I've read some bad reports on its linearity (distortion of signal through nonlinear resistance) and I'm not sure how much this will detract from the advantages of having no moving contacts. I have an RC unit with a 100kohm dual motorised pot on board to drive the lightspeed when it arrives.

STRs look like a good option and slightly more refined than my very basic PFM vbe multipliers.
 
I compared the lightspeed with 2 pairs of good quality fixed resistors - no difference. Can any attenuator be better than that? How about running B4 of batteries?
 
PD did some interesting analysis of the order of distortion from CSd cells a while ago. It is a bigger number than you might expect. I tried them about 12-14yrs ago as a student and didn't like the sound or poor pair-match/tracking of the bits I could get.

Split-bobbin type transformers - yes, my first choice for low-current-draw items, esp since the slightly increased impedance is negligible when compared with CRC supplies and so forth anyway. It even contributes to useful 'smoothing' by reduction of peak charging currents. But even so, when I last bothered to try it, swapping between a few small toroids and the (encapsulated) ' flat' split bobbin type my preamp normally uses... the toroids-driven supply showed on even fairly crude sweeps (soundcard spectrum analysis , o'scope) as rather noisy. And without further (CRC) attention ahead of regs they sounded a bit ... ragged; 'unsubtle' perhaps is a better description. And this at low power levels (2 x 24v, <10VA).

NB split-bobbin encapsulated transformers are also 1) cheap 2) totally silent mechanically 3) often short-circuit/ fire/rodent/proof and UL-rated and 4) easy to mount. Nice, if you don't need more than 20VA, say.
 
Rodent proofing never got a mention in my requirement spec. That was a serious oversight in my neck of the woods. The cat brings in a couple a week and they come here on their own for winter holidays.

I agree with the description of toroids without CRC or CLC before the regs as unsubtle sounding. I've also found them to subdue dynamics (subjectively speaking) particularly in the bass on my B4.

I'm using a Valab 20k stepped attenuator on mine at the moment. It's more clinical and dry sounding than the pot that was in my NAC 62 which I would probably go back to if a lightspeed wasn't on the cards. The Valab doesn't mask very much but can be too revealing of top end issues but this makes it good for prototyping IMHO. I also found the Valab more suited to a B4 than say a Starfish or NAC as you need to use its entire range in the B4.

I have found some problems with microphony of the teflon and polystyrene caps I'm using when near the speakers. To make it easier to work on my B4 it sits on the floor within 6 inches of a PMC transmission line speaker. The teflon 0.47uF Russian input cap appears particularly sensitive to ground borne vibration. Just placing a thin sheet of polystyrene foam underneath it really helps relax the presentation and removes some stridency at the top end. As the big caps (input and output) on my B4 are on flying leads anyway I'm contemplating mounting them either under or above the B4 board on a separate circuit board on leaf springs or some sort of suspension, relying on the flying leads for some damping. I don't think the capless main board will benefit from any suspension. Perhaps this is exteme perfectionism but the possibilities will only nag me if I don't try these things out before or as part of the final build.
 
Military spec caps prone to vibration? I'd suggest you try to measure that effect and look to the quality of your soldering first. Many a man has slightly melted a polystyrene cap with too much heat so that it gives a dodgy connection inside the cap where the leg terminates.

;-)
 
Oh dear, maybe you're right sq225917. That's the first place I should look for a cause. I was careful but perhaps not careful enough. However even the military "cannae change the laws of physics" and quite what the military specs for these russian caps were is unknown to me. They are still wound caps in a metal case. There will still be resonance and movement dependant on the looseness of the winding. I'm thinking that these caps were designed to be clamped securely and possibly tightly to a larger body. Their main beneficial qualities are a high bipolar capacitance and voltage for a low loss angle and leakage along with the ability to self heal.

I've noticed that the metal cases pick up hum very easily and their orientation with respect to the transformer matters. Maybe earthing would help. You can hear the case being tapped when volume is high as well.

When I was in the forces 80's and 90's the russian military were well known for using valve communication equipment in armoured fighting vehicles. It has greater immunity than transistor equipment to the electromagnetic pulse from a nuclear weapon. If you're having to use such extreme size to get the component qualities you want there are going to have to be compromises elsewhere. I may open one up to see how precise its construction is. One thing is for sure, it wasn't designed or specced for audio.
 


advertisement


Back
Top