Sunteți pe pagina 1din 15

As you can see, unlike some others I am morbidly concerend to keep the signal ci rcuit completely free of electrolytic

capacitors (no Black Gates allowed either) simply because I actually tested them blind against a wire bypass and feel them to be completely unacceptable. Key changes compared to the RCA Circuit are: 1) Converted second stage to ECC88/6DJ8/6922 to reduce output impedance, the Cir cuit is fine to drive a 10K load (tested) 2) Used gridleak bias on the second stage and removed the cathode bypass capacit or in the first stage, re-arranged first stage for low anode voltage operation w hich usually gives much lower noise 3) Changed the the RIAA circuit to whta I call "Kondo" or "semi-split" circuit, this has the advantage to allow for completely seperate adjustment of the capaci tors in the circuit 4) Moved the coupling capacitor behind the RIAA so the RIAA capacitors dielectri c is "soaked" (avoiding any issues from "soakage" [which is also known as "diele ctric absorbtion"] on the signal) and the interaction between coupling cap & fir st RIAA cap (and thus on the RIAA EQ) 350V Supply: If you want to use a 350V supply "Just Do It". Increase the anode decoupling resistor in the first stage to 330K from 180K and increase the anode load resistor of the output stage from 12K to 15K or even 18K . This will give slightly more gain and linearity from the output stage and put those extra 100V to good use. All else remains the same. BTW, note the circuit needs a good deal of current, around 25 - 30mA for both ch annels! Originally posted by philipbarrett but the 2nd stage voltage readings at the pl ate are 75VDC (naturally the 15K resistor got roasting hot). The voltage is fine, just what it should be, if your resistor gets hot I think y ou did not use one having sufficient dissipation (12W Wirewound recommended). Tubes Choice: Don't believe everything you read. Sure a good 6SL7 and 6SN7 is a great sounding valve but so are dozens of others regardless of them being octal, noval, pico 7 or whatever socket and bulb size. What they don't tell you is that most small octal twin triodes from this range t end to be far more prone to microphony than their, say, noval counterparts. Oh, and I've heard 6SN7s and 6SL7s from reputed manufacturers that sounded far w orse than any humble nine pinner....Go figure. Not really, the 7025 is a special quality low noise ECC83. Some ECCs I like, others I don't like at all. One particular brand of ECC83s I never had decent recsults with is the folded pl ate RFTs and some older Shuguang ECCs. Depends. In some ways ECC83 & ECC88 have complementary sonic traits, many modern and NOS options exist, it's just basically a nice and easy to get combo, especi ally considering that you need only one of each, so you can buy singles..... The 7025 is listed as pin and electrical equivalent of the ECC83. If I write ECC 83 I really mean any of the following: 12AX7, 12AX7R, CV10319, CV492, 12AX7A, 12AX7S, 12AX7WA, 5721, 5751, 5751WA, 6057 , 6681, 6L13, 7025, 7382, 7494, 7729, B339, CK5751, CV3970, CV4004, CV4017, CV81 56, CV8222, CV8312, E2164, E83CC, ECC803, ECC803S, ECC863, M8137, QB339 And when I write ECC88 I mean any of the following: 6922WA, CCA, CV10320, CV2492, CV2493, CV8065, E88CC, E88CC01, ECC868, 6DJ8, 6N23 P, CV5358, ECC88, 7DJ8, CV10403, PCC88, UCC88, 6N1P, E188CC,7308, CV5231, CV5354 , E288CC, 8223 Using the standard russian 6N23 (aka 6922) and the EHX 12AX7 makes for a nice so unding phonostage, using NOS Mullaed Box Anode CV4004 and Siemens 7308 sounds no tably better of course. The ECC83 will draw around 1mA, the ECC88 will be (depending upon the specific v alve used) around 14mA. With a "bogey" ECC88 & ECC83 each channel will draw arou

nd 15mA (=/-20%). 6N1P at the output (650ma filament)? That will make the output impedance a littl e higher, otherwise you'll be fine. Paralleling input EC83?: The ECC83 is part of the RIAA (with the large value unb ypassed cathode resistor, so any changes there and your RIAA response (Low frequ encies) will be off. If you can re-tune it it would be easy to use two sections of the ECC83 in parallel. If you use 51K load and 91K +B Dropper and seperate ca thode resistors for the ECC83 sections you will probably need around 270K in the place of 215K for the RIAA EQ. CSS on the output? You can easily change the Output stage to CCS, the Input stag e would (again) require the RIAA series resistor to be adjusted, this time downw ards. As the input stage drops a lot of the +B Voltage anyway you could use a lo wer +B decoupling resistor for the first stage and leave it otherwise as is and add a CCS to the output stage. The input stage is pretty linear (at the voltages it is likel to see anyway) as is. PARTS CHOICE: Bypass all electrolytic caps. The problem is - most HV electrolytic cap's have a rather high ESR and ESL (Equvalent Series R / L) which stops them behaving as c apacitors at higher frequencies. So without bypassing any high frequency noise j ust "rides through" most of the filter chain. A basically poor design will never sound really good, no matter how many fancy c omponets you throw at it. Sadly, many "technically correct" designs also fail to work in reality (shame really). However, depite sounding nice with "generic" parts, do consider upgrading the RI AA Cap's and the interstage coupling Cap. Old Stock Silver Mica > 250V Rating (b uy a few extra and test for leakage at high voltage!!! - If neccesary try to "bu rn out" any leakage) tend to be still best, sonically and in 1nF/10nF are not th at hard to find. Resistors, I'm pragmatic, the best is good enough.... ;-) 12K Resistor need HIGH POWER: Yes, I'd suggest an Industrial TO220 encased type with heatsink. The industrial Meggit / Caddock / Vishay-Sfernice TO220 Power res istors do not give up much in sonics to the expensive audiophile Caddocks (thoug h these are another series). All other resistors are fairly uncritical in terms of dissipation. In the context of my personal sonics I'd use a Rohpoint Wirewoun d as Anode resistor for the ECC83 and selected (after a while "baking" at 60 - 7 0 centigrade) Carbon Composite Resistors in all shunt positions, plus good metal films in series positions. Output Coupling Cap to taste, my recommendation as said would be KP-SN Types nea rest value to 2.2uF but no doubt Copperfoil & Teflon Excotica would be great too , if you have them around.... Well, as said, this Phono is very much a minimalist and basic, entry level desig n, there is bigger and better out there, but as a starting point it should do ni cely. m'I right to prefer molded (axial) instead of dipped (radial) micas?All molded t hat I bought in the past were not magnetic. The dipped were magnetic. Also I fin d the values that I need with same basic ratings (pF, tolerance, voltage) but di fferent codes e.g.CM20E102F,CM20F102F and CM20E601E for 1n 1% 500v. Any quidelin es/suggestions ? Yes... CAP LEAKAGE TESTING: Attach a Voltage equivalent to that seen in operation to th e Cap via a 1M Resistor. There should be no voltage measurable across the resist or, under static conditions. If there is measurable voltage, try connecting the Cap to a voltage equal or close to it's rating with plenty of current reserve. T he "fault" will either burn out or develop into a permanent short. This way I ge t about 2 in 3 "leaky" old Mica Cap's to clear to the fault. They may loose a li ttle capacitance that way but are still fine for use as bypass or in coupling. 2.2UF: You don't absolutely have to have 2.2uF. I use that value as I recommend it as minimum coupling cap for the S&B TX-102 which i use as passive linestage (

and recommend).Look here: <http://www.angela.com/catalog/capacitors/Angela_SBE_S CR.html>" target=_blankhttp://www.angela.com/catalog/capac...tml <http://www.ang ela.com/catalog/capacitors/Angela_SBE_SCR.html</a> ELECTROLYTIC CAPS: I tend to use generic Nippon Chemi Con (VX) series axial and/ or radial capacitors. They are very cheap and also used by Sakuma San. They work out fine if you don't want to stretch for motor run MKP's. I would be carefull with bypassing. If you must, 1uF of fairly high esr capacitors seems indicated. That means MKT over MKP etc.... PARTS CHOICE COMMENTS: I don't see anything that is prejudicial to performance. But one or two notes.Modern Holco's have no longer got any advantages over other RC55 standard resistors, sonically but are still sold at a premium.The gridleak resistor and output resistor (10M & 1M respectively) do not require particulary high quality resistors.Critical for sound are the RIAA Resistors, the cartridge load, the first stage cathode resistor and the Anode loads, in that sequence of impact.Try making the gridstopper carbon composition, that is more important th an the value.The +B Dropper resistors are relatively uncritical.I would recommen d for all uncritical positions generic metal films of sufficient power rating fr om Mouser.Mills on the Anode load of the 2nd stage is fine. For the first stage I think a premium grade low noise (Vishay? Caddock?) Resistor is indicated, the same might also be a good choice for the EQ. The 12W are okay and actually suggested for the 12K. The second stage is run "lo w & hot" to minimise anode impedance. For the anode/cathode resistor Dale (1W) RCN55 and similar should be O.K?: Not really. The second stage anode resistor dissipates around 3 Watt. A few notes. The Gridstopper is the 100R, NOT the 10M. The 10M is the gridleak r esistor and the circuit is gridleak biased.Power rating, all resistors are fine as 0.5W excepting the Anode Load and +B dropper resistor for the Output Stage. T hey needs to be 3W & 9W minimum, the Mills is probably overkill for +B Dropper a nd you could try one of the various TO-220 cased power film resistors (Caddock, Meggit, Vishay come to mind). PARTS LIST COMMENTS: Looks good, except there is absolututely no need for the 3W/180K, it can be 1/2" . The ECC83 runs at 0.6mA. The 100K Resistor dissipates 0.036W and the 180K 0.06 5W. I'd probably use standard (Mouser CCF55 Range) metal Films for all but ECC83 cathode, cartridge load, gridstopper & RIAA and of course the big resistors for the output stage. Mrs Annan should like the change in cost. Zout: Around 2500 Ohm, dependiong upon the ECC88 type valve used. Well, the outp ut impedance is around 3K, not exactly low, but with 1nF cable capacitance (whic h is well past what I'd recommend - stick to low capacitance cables) the -3db po int is still 50KHz. I'd leave the cathode follower off if good sound is desired. Paralleling triodes will lower the ouput impedance?: Yes, but I doubt the necces ity of lowering much more than it is already. Separation L+R: Hmmm. It is hard to make a LP Pickup with > 30db channel separat ion. Sharing valves between channels will couple a little at high frequencies, b ut a lot less (at least 20db++ less) than any Pickup. I'd not loose much sleep. On the plus side, the halves within one valve envelope are usually assebled by t he same person from the same tray of parts and hence chances are good for very c lose match, sonically and elecitically. Originally posted by fscarpa58 the ECC88 has a 0 Volts bias. is it intentional? Actually, it has what is normally called Gridleak biasing. It is not ideal for A mplifier driver stages, but with stages working at a fairly low level this works fine. The advantage of very small value coupling cap's and no cathode RC combo have to o much a lure for me to resist. Sonically it works great, as evidenced also by K urt Strains Phonostage (who converted his second stage to gridleak bias on my su

ggestion and gary Piumm's Phonostage who in turn took the gridleak bias from Kur t. Reading them old valve books has many cool ideas rarely seen in modern commercia l gear. BTW, the first stage has an unbypassed Cathode resistor to stabilise the gain an d output impedance (which impacts on the RIAA EQ accuracy) and also helps to red uce the classic "ECC83" sound a little. For implementation, this was meant to be fed 250V from a simple, solid state reg ulated supply, All cap's in the main audio Circuit MKP (PSU) and Silver Mica, al l resistors good quality metal film, PTFE Valve sockets should really be conside red mandatory for a Phonostage. prefer Mica types, I will agree that this may be a matter of taste. quote: Originally posted by GAK Any suggestion for the signal resistors (215K , 48K7 , 100R) ? I use standard precision metal film, nonmagnetic leads, sadly magnetic endcaps. quote: Originally posted by GAK What is PTFE valve sockets?Those with the can that cov ers the valve? No, those where the inserts are PTFE. They tend to be however also those that ta ke covers, often an apropriate choice for low level circuits. quote: Originally posted by GAK In the future I'm planning to use an MC cartridge.Step -up with xformers or active? I like really good transformers better than solid state stepup and never got val ve based stuff to work quietly enough. I had the original S&B TX-103 (quite expe nsive) designed to conform with my personal sonic prejudices. The Phonostage as shown will work well with a TX-103 as MC Stepup and a TX-102 as volume control ( without the +6db stepup recommended). Overall gain for MC will be up 68db, usual ly enough for all but extremely low output (< 0.2mV) MC's.... 2.2uF Cap Choice: My choice would Audyn/Mundorf/Angela KP-SN (they are all very similar). I'm not much of fan of the Hovelands or Jensen PIO'. You can always ex periemnt later if you want a leaner, more clinical or warmer more cuddly sound Power Supply quote: Originally posted by GAK Any comments-suggestions for the following PSU? Looks on one hand a little overkill (a stereo chassis will draw around 30mA only ) and at the same time not anywhere near quiet enough in terms of ripple rejecti on. I would recommend a simple LM317 type adjustable regulator with a voltage limiti ng series transistor. This tends to work well enough, the original design was ai med to be used with that. Use a snubbered (Buddah style) solid state bridge made from soft switching diodes with 240V secondary. The original idea was to use a pair 115V+115V : 9V+9V Transformers back to back, use the 9V+9V with a dual schottky diode for the heater supply and use the seco nd transformers nominal primary as the HT source. Added RC or LC filtering for t he raw voltage feeding the HT Stepup transformer is easily implemented. A simple RCRC Filter after the bridge and an LM317 plus voltage boost transistor give a quiet supply. The RC decoupling per stage makes sure that the Signal cur rent AC loops are closed through the local film capacitors and not the regulator . Tube Supply?: Nope, i found very little difference. The LM317 solution is simple r, cheaper and actually quieter. No Reg IC and LC Filter?: This is what I am doing with my current "in system" ph onostage. But as uncompromising an approach as I took there may not be for every one. The point with the ECC83/ECC88 Phonostage is to provide a good basic qualit

y phonos that is easy and fairly inexpensive to realise (it was originally aimed at a commercial application). Anyone who wants a truely outstanding phonostage has to work harder than that, b ut to many it will likely serve well. Think of it as the "Valve - El Cheapo" to give you an idea where it fits in in the whole scheme of things. It is not "the latest & greatest", far from it. But it is "pretty good". Constant voltage is no t really an issue. I prefer a good quality cap to a gas regulator (tried). Pleas e note that the LM317 is PRE-Regulator followed by an RC filter. For those paran oid enough - you can add another 1k/22uF section to the output stage and split t he input stages 180k Resistor into 2pcs of 91k and use a pair of 10uF Cap's. Wit h the original circuit, the LM317 regulator output (AC - be it noise or reaction to the AC current demand of the output stage) is suppressed by around 23db @ 10 0Hz and 40db @ 1KHz. Doubling the RC circuits doubles that attenuation. Also not e that the RC circuit drastically reduces the current modulation on the LM317 in the first place. At 20Hz only 1/4 of the current flowing in the output stage fe eds back as modulation of the LM317 output, at 100Hz it is only 0.072 times the current in the output stage. Thus the single RC circuit after the regulator insu lates the Signal circuit from any signal interactions with the regulator by arou nd 45db @ 100Hz and much more than that above 100Hz, compared to feeding the cir cuit DIRECTLY from the regulator. So the solid state regulator only becomes audi ble at very low frequencies. If you feel 45db down @ 100Hz is not enough, doubli ng the RC circuits up gets you > 90db reduction of signal related "nasties" from the "bad & evil" solid state regulator at 100Hz.... As always, it is less what you but how.... MOSFET Regulator?: It's not a regulator (1) and it seems to be somewhat testy. I have build quite a few of these and never once did I have one oscillating on me , blowing up etc... It does often seem to happen to others though, hence I prefe r not to recommend it. Again, the sonic differences are small after you interspe rse a sufficiently large timeconstant RC Circuit after the regulator. So a recom mendation based on "get it working easily....". DUAL SUPPLIES?:Nope, one regulated and LOW NOISE supply providing 250V with a li ttle over 30mA current Capacity is fine. An ECL82 on it's own should manage fine , if you want valves. Just keep the ripple low. The seperate RC decoupling circu its to each stage move the sound critical bit's into the decoupling Cap's and st rongly reduce the sonic effect of the PSU. If you retain the RC decoupling as drawn in my schematic one regulator for both channels suffices. Do not delete these RC circuits unless you are very experienc ed in wringing good sound from circuits with loads of loop feedback and invariab ly poor transient response (regulator). NO RC FILTER, JUST CAP?: Then the regulator will dominate the sound due to it's transient behaviour. I find most regulators by far too audible (in a negative se nse of the word) to place them into signal current loops. LARGE CAP?: Why? Because a lot does a lot of good? Like a lot of paint in one co at, instead of many thin coats and like a ton of glue instead of clean, flat sur faces and only the slightest bit of glue? Remember, always use the smallest valu e capacitor you can get away with, the smaller the value the smaller the physica l size and the less usually the various evils plaguing them. Leave th 10uF Cap i n the first stage alone. If you must (like having suitable cap's lying about) us e larger values place them in the anode circuit of the ECC88, up to 100uF make s ense, though a pair of 1k Resistors and of 22uF capacitors will be much more eff ective. I would not recommend this phono as "cost no object" project. It's desig n was emphatically "high cost IS an object".... My "cost no object" design is a notably better but tends to push the parts cost well past the $ 1,000 line. If c ost is really no object build that. Otherwise place your "parts money" wisey, wh ere it counts sonically and use the right parts too (quality over quantity), bas ically the polar oppsite approach of "cost no object".... If you want to regulate I would recommend STRONGLY using current sourced shunt r egulators, one per Stage and you will find a basic TL431 & Cascode FET plus LM31 7 & Cascode FET current source infinitly superior to even the best Valve Series regulators. However, I feel that such extreme effort is not required and is bett

er reserverd for more extreme Phono Stage designs. After all, despite a lot of f ancy components the actual design is quite basic and cheap to realise. oreover, regulators are in most cases based around looped feedback amplifiers, which by d efinition implies inherent instability, so layout and implementation can be crit ical. If you don't get it right you might spend days or weeks chasing humps and pumping that are reallyvhf oscillations of the regulated PSU which you cannot se e on the usual 20MHz 'scopes.... The "sound" of regulatorsmis another topic agai n.... LCLC FILTER?: Most likely it will make things worse. LC circuits imply resonance s, these can be difficult to handle. The noise is rejected very much by the RC c hain, adding more filtering will not improve anything. If you build a fully LC f iltered chain you can reduce the number of secytions and you need less raw HT, b ut you are now dealing with highly resonant circuits having their resonances in the same region as record warp etc. I'd recommend sticking to RC for the time be ing, less to go wrong. NOTES: I think I've got the idea.Splitting the resistor the characteristics off the filter remain (cutoff frequency, ripple factor,..) remains the same ? ANSWER : In essence yes, but you get added common mode noise filtering. In the case of LC filter with L for both lines ( + and -) do the 2 Ls summup and the filter calculations are based on the sum ? ANSWER: In principle - yes. You can add a neat trick, make the choke one choke with two sections wound as a "rea lly bad" (high leakage inductance) 1:1 transformer. Gives compact but effective filtering not only for differential mode 100/120Hz noise but also for common mod e noise, especially if both sections in turn are wound for a minimal shunt capac itance (very high resonance). Having "lowest ripple" and impedance as the only target of a PSU is achievable i n many ways. ANSWER: Yup. As so many other things. You can do a lot of things in different ways. It often pays to have a good hard how it's done in truely legen dary commercial gear. Gives good ideas what is easy, reliable and simple, withou t being too expensive.... I (as GAK and many others I believe) am confused when the PSU time comes. Are th ere any basic rules for it? ANSWER: Yes, unless you are spectucaluray good in "t uning gear" by ear and have endless amounts of time always a RC section as final section, this removes most of the Regulator/LC filter ringing sound. In poweram p's that of course becomes difficult. For example - which should be the roll off frequency of a filter block ? Depends... - which should be the resonance frequency and Q factor for a LC filter ? Low to very low on both. Implies large chokes and large cap's - when does a decoupling filter really decouple ? (I'm sure it is not like throw in as much capacitance as you can and you've done it): ANSWER use the term "decoupling" in any case where we want to isolate several br anches of a (+B) circuit from being modulated by the signal current in one branc h as much as possible. This implies first that the "decoupling" RC/LC circuit mu st reject audio band signals with a pretty substantial rate. If you look at the suggested circuit from my Phono, we have a 100uF Cap or a regulator as the final output of the PSU (before we enter what is really the signal circuit). With 22u F/1k you have a -3db point at 7Hz and a -20db point at 70Hz. This means any nois e from the PSU (or reaction from the psu to a signal related current change) is rejected by > 20db @ 100Hz. This means at 100Hz our signal circuit for the outpu t stage is DE-COUPLED from the PSU by over 20db. BTW, for the first stage the "d ecoupling is more like > 60db @ 100Hz. Considering the characteristics of this filter techically adequte for the specif ic design, the decision to alter the values eg. 47uF/500R (nearly identical char acteristics) has more to do with sonics or affects seriously the whole design ? ANSWER: The 1k/22uF is actually based on "the lowest value cap I can get away wi th. Increasing capacitor values means using larger capacitors. This is not alway s a good idea.

For example with the all passive RCRCRCRCRC supply the PSU impedance becomes ver y high: No, the impedance is that of a 100uF Capacitor. Is the 1:10 perfectionistic and trading this rule (when you must) for smaller ca pacitors is sonically preferable? The 1:10 is for transformer loaded stages. You cannot simply take that forward to Resistor Loaded ones I can undestand that bigger capacitors are bad but taking in consideration your article http://www.fortunecity.com/rivendel...fiertheory.html <http://www.fortun ecity.com/rivendell/xentar/1179/theory/seamptheory/SEAmplifiertheory.html> resis tance also has a limit. SOLID STATE HV POWER SUPPLY, see schematic The +250V output voltage present a 107db 100hz PSSR. What did you think about th is regulation and do we need other capa (what amount) after regulation? ANSWER: I am not sure if the transistors have sufficient voltage ratings. Also, circuits like this one tend to be instable without load capacitance. Past that I suspect a "boosted LM317" circuit would probably provide superior objective per formance, but by all means build as you like. Be aware that most new designs do not quite work the way simulations predict and need debugging. I just look MJE340 datasheet , it's given up to 300V. So I needto donwscale the voltage before MJE340 since my trafo has 230V secondary. Why not just substitute a transistor whose voltage rating is suitable, the circu it is not all that critical: Such as the TIP50 <http://www.onsemi.com/pub/Collat eral/TIP47-D.PDF>? MC First Stage (See Schematic) --------------------------------------------------------------------------------------------------The basic topology is that of the second step-up amplifier by "Maison de L'Audio phile" of Paris/France. The first 100R is the cartridge load, which must be adapted for the cartridge yo u use. Using a paralleled 1nF cap also makes the circuit immune against RF noise . The 47R resistor in the source is not necessary, if the JFET has an IDSS(on) o f 2mAmps or less. As the signals are very sensitive, the voltage must be very clean, a 100,000 uF electrolytic cap for blocking power supply noise, computer grade favourably, is advisable. (note - we used 2 X 2,200uF Elna Starget and a 1k Decoupling resistor .) Suitable JFETS are 2SK170, 2SK240 (double JFET), 2SK97 (double JFET), sorry only these japanese types are suitable, the European BC264 and BF254 and so on do no t work here (I tried !). All resistors must be big metall resistors, e.g. Beysch lag 1Watt, or tantalum resistors. No carbon resistors allowed here, tubes addict s. The output cap is preferably polystyrene or polypropylene or paper-in-oil, as yo u like it, as they are different in sonic balance. If the load is lower than 100 K you need to increase the capacitor proportionatly. Can be supplied from 12V DC but not 6V.... I would recommend making a solidly regulated and low noise 15V Heater supply and tapping of via resistors the 12V for the ECC83 and the 6V for the ECC88. Note t he statement of 6V, not 6.3V, underheating Valves slightly lengthens life and so und better. Then tap of the 15V via a RCRCRC circuit using 100R/1,000uF...2,200u F per filter cell. If you use a passive RCRCRCRCRC filter chain with 1k/100uF cells you will have 3 0mA X 5k voltage drop in the RC filter chain, namely 150V. Add to this the 250V +B and you need 400V. If you use a 6X4 or EZ80 rectifier valve and a 10uF reserv oir capacitor you will be needing a 320V-0-320V Transformer with a DC rating > 3 0mA or an AC rating of at least 50mA or thereabouts. RCRC Resistor Type: Uncritical, even "white coffin" Units should be fine, 2W Met al Films will work okay as well. RCRC Cap Type: The parts quality is really a fundamental question. If you want a n "all out" design, I can suggest something considerably different, which will b

e a LOT better. The PSU and signal circuit discussed are aimed at providing a re asonable performance, to be easy to make work and to be inexpensive. There is no harm in using MKP's but I would restrict myself to the Input Cap's and the fina l Capacitors in the Phonochassis, if good performance is desired, without going overboard. I have attached the suggested PSU schematic for a simple and cheap PSU that is d ead quiet, uses a valve rectifier and is entierly passive and hence should be pr etty safe and reliable to assemble. The Electrolytic Capacitors can be quite ine xpensive types but should be bypassed, the 10uF reservoir capacitor however shou ld be a decent MKP type, to shortcircuit as much HF noise as possible. I use both a FET based and ECC88 MC step-ups. All with tantalum resistors and MI T RTX output caps and NiMH batteries and always thought both sounded great in co mparison to transformers. Recently, i have been using a cheap Lundahl which sounded ok with my Kontra b. A fter about a 100hr of break in its sound changed dramatically, up to the stage o f being clearly preferable to the active step-ups. TUBE PSU It may prove difficult to make valvee regulated supplies sufficiently quiet, oth er than that, no issue. quote: The conclusion is to use four separate PSUs?One for each stage?? I wouldn't consider it overkill for a phono preamp... quote: Why not to use one power xformer,rectification,p-filter and then split in two se parate regulated supplies (tube or SS).One for each channel with an RC filter af ter each section.This for the PSU chassis. After the series reg I'd just put some reservoir caps, no further filtering shou ld be required. quote: Frank: It's also an overkill to use your PSU with EL86,right?Or not? Well, compared to the circuit with the 6AS7G, no, I wouldn't consider it overkil l. If you need a regulated 250VDC B+ a ECL805/85 and a 85A2 would do per stage ( de pending on current requirements perhaps one per channel may do ) and should be b etter and maybe also cheaper than just one with a 6AS7G which is only needed if the reg needs to work for (a) stage(s) drawing much more current. If the ECL82 i s considerably cheaper than the ECL85/805 than that would be fine too... I shoul d have a schematic for one of those in my drawer somewhere... To my mind there's little point in adding a series or shunt reg and than share amongst stages/chan nels in a preamp. The least you should do is have one per channel to avoid inter channel IMD. I'm not frank, BUT, your input LC filter has very low levels of capacitance and inductance, meaning comparably little filtering. Combine that with the regulator circuit shown (which fails in several areas to r ealise the maximally possible PSRR), which will have a lot worse ripple rejectio n than a LM317 you may very well find that the Circuit has too much noise. To put this simple, the Output Stage of the circuit of my suggested Phonostage ( which dominates the PSRR of the Circuit) is around is around 40db. If we assume a 2.5mV Phono cartridge and thus 250mV output from the Phonostage we would like a S/N ratio for the 100Hz PSU component AT LEAST 80db (1:10000) below this. In o ther words, the supply noise at 100Hz needs to be 40db below the expected signal level as the circuit with it's RC filter already rejects the PSU Noise by 40db. In plain english, the noise applied to the Phonostage needs yto be 40db below 25 0mV or no more than 2.5mV. A typhical Valve based regulator does not display all that much 100Hz noise rejection, simply because it tends to have a rather low o pen loop gain. The Series Pass Valve is a simple cathode follower. It's grid will receive an AC voltage that is a fraction of the noise on the valves anode (unless the gridvol tage is additionally filtered) determined by the value of the Anode resistor of

the differential Amp valve and it's anode resistance. The cathode will attempt t o follow this voltage but will experience additional pertubations due to the lim ited transconductance of the Valve. Combine that with a fairly high ripple pre-r egulator the result may simply not be good enough. I would move one of the 220uF cap's to the position after the choke (or both) an d limit the capacitive load after the reggulator to what is needed for stability (usually 10 - 20uF). You could even dump the Regulator and choke and use multiple RC circuits (Shindo Style), it sounds good and is very cost effective. Simply use a chain of 5 1k r esistors and 5 100uF Electrolytic Cap's and make sure your raw DC after rectific ation and Reservoir Cap is around 400V. The 5 X 1k/100u Filter chain after a 10u F reservoir cap will produces a very low ripple DC (< 1uV 100Hz noise) quite che aply and effectively. IMO the use of a tube rectifier is not going to bring any sonic or other benefit in this case. You won't hear the difference, I tried this with this circuit several times and never heard anything in favour of the valve rectifier. If you want to at all cost than so be it but as TL has pointed out you'll need s ome extra filtering as the rectifier is limited in the cap it can have straight in front of its output. If it's the slowstart you're looking for, the regulator provides this plus it is olates the PS nicely from the mains, keeps B+ rocksteady and if you use one per stage IMD will be a thing of the past. All in all extra expenses that can and are avoided by the BY133 Shottky rectifie r bridge. The regulator will reject alot of ripple but it is limited by its transconductan ce. Regarding the 0.330F cap, you can use an Audyn MKP here too, they're relatively c heap and sound fine. The regulator as shown has served in a phono stage since 1985 and with a cascade d triode phonostage measured noise is -70dB for phono and -90 dB for line level. Input sensitivity is 2.25mV into 51K (TL is not alone here) with 100pF in //. You can put the caps behind the reg in the preamp chassis if you like but I'd ad d some more MKP caps locally close to the circuit proper. Super Tube Preamp: If you want something that is a little past the design discussed here, look here : http://www.jogis-roehrenbude.de/Les...OCCATA_RIAA.htm <http://www.jogis-roehrenb ude.de/Leserbriefe/TH-Loesch-Line/TOCCATA_RIAA.htm> This Phonostage uses only MKP Capacitors in the PSU, has a heavily overbuild PSU , uses rare and expensive Valves (now E810F in the Input and D3a in the output) as well as expensive Stepup transformers and an LCR RIAA. Those who have build c opies have been very taken with the performance, it IS worth all the money pumpe d into it. STEP UP TRANSFO: I have some experience with Lundahl LL9206 and while not perfect it seems amazin g value for the money. I also have little doubt that the TX103 is much better, e specially in the bass, as it should be at 4 times the price. TX-103 Distributors:Check with the distributors (Bent Audio & DIY HiFisupply). I f you want an idea what others think of it, look here: http://www.high-endaudio.com/RC-Step-ups.html#ClB <http://www.high-endaudio.com/ RC-Step-ups.html> http://www.hagtech.com/trumpet.html#stepup <http://www.hagtech.com/trumpet.html> <http://www.positive-feedback.com/Issue11/hagerman.htm> I use both a FET based and ECC88 MC step-ups. All with tantalum resistors and MI T RTX output caps and NiMH batteries and always thought both sounded great in co

mparison to transformers. Recently, i have been using a cheap Lundahl which sounded ok with my Kontra b. A fter about a 100hr of break in its sound changed dramatically, up to the stage o f being clearly preferable to the active step-ups. The levels with MC cartridges are so low that the transformer will NEVER burn in . There are material issues around the setteling of the magnetic core material a nd the wire. Simply follow the instructions I send to Arthur Salvatore, quoted here from his site: <http://www.high-endaudio.com/RC-Step-ups.html> This worked very well for me, so I recommend it highly. This is what Loesch wrot e, with a little editing: "(The TX-103)... will require a substantial period of "forced burn in" to give i t's best, simply because the magnetic core is huge and will not see much magneti sation with normal MC signals. Please consider connecting a CD-Player to the sec ondary (Output) of the TX-103 and then terminate the input with a low resistance resistor (quality uncritical), I'd say 27 Ohm when connected for 14db gain, 6.8 Ohm when connected for 20db gain and 2.2 Ohm when connected for 26db gain. Leav e with a highly dynamic, wide bandwidth signal CD to play for a week or two. I w ould use music, but I'd expect pink noise to work well too." Personal Note- While on a week's trip, I connected the signal (a tuner on a 24 h our "grundge" station) to the primary (Input) instead, and had great results. Th e rest of the circuit followed Loesch's instructions I recommend the full 9 inch from a CD Player (0dbfs pink noise) for the SECONDAR Y of a MC stepup heavily loaded and about as much level as you can get (somethin g like a 50W RMS Amp!!!!) for something like a 102.... The key is the core mater ial. I hear transformers get better over YEARS. I got fed up with waiting.... Sa me thing for speaker drivers.... On BG Cap's it seems even with "forced burn in" my patience is insufficient though...Yes, JB from S&B was EXTREMELY sceptical. He heard the differences having 102's in his home system on the sky digibox (dig ital satellite TV) during footballing football games and was heard to mutter "I' d never".... Transformers beat ANY other audio component I know, excepting certain speaker dr iver on burn in. Next to them most resistors and capacitors barely change and va lves a little.Must be all that magnetic core material, as holding a srong magnet next to transformer pretty reliably reverses the "burn in" process.... GROUNDING and CONSTRUCTION NOTES: Finally, this time, I have followed the Dead bug style to construct the circuit. That means that I used the ground plane as the 0 reference. I have seen Pete Mi llet (Line & phono) and Gary Pimm (Line) to use this method. I can understand th at in this case currents are diffused on the ground plane in an uncontrollable m anner. On the other hand to implement star grounding and use the ground plane on ly as a shield means longer leads to ground aka more inductance and parasitic ca pacitance. Are there any measured / sonic deferences between the two methods ? ANSWER: I use a mathode that considers each stages current loops and gives each stage a complete star-ground, to which all current carying components are return ed. Thus the first stage "star" has the input signal loop (with the input resist or across the RCA Jack), Cathode resistor, PSU decoupling Cap, RIAA shunt elemen t Grounds and following stage gridleak resistor ground returned to it. The outpu t stage star ground has the Cathode, the PSU decoupling Cap return and the retur n from the output Jack attached to it. Between the two stargrounds a low inducta nce ground bus (0.1mm X 5mm solid silver, multiple twisted silver wires or laque red solid copper - wire or foil) is employed. Makes sense? This way only the PSU return current (hopefully mostly DC) and signal current return from the output stage grid (which should be minimal) traverse the actual "groundplane". The PSU return is connected to the output stage starground ? ANSWER: I would do it like that, as the Bulk of the PSU current flows in that st

age and any signal current impressed will be proportionally larger in this stage as well. What about the PSU side ? Star or Bus connection there? ANSWER: Neither of course. Always consider where the current will flow and where you want it to flow. Think of current/energy as a bucket of water poured out so mewhere. On a perfectly flat and level surface it spreads evenly. Start placing the slightest indentions and the bulk will flow with direction. Place full bullw arks and "shore fortifications" and you will be able to direct a huge flow that would "saturate" your "plain with dug channels" making sure nothing is wasted or available to cause problems elsewhere.Now replace "indention" with "impedance" and "bullwark" with "insulation" and "water" with "current". Clearer?I personall y would split each filter resistor into 2 seperate ones with 1/2 Value and then would use one each (impedance) for both negative and positive line, making the f ilter symmetrical. The signal reference ("ground") appears at the last filter ca p negative terminal.I wish people would get rid of this straw ("ground") they cl utch. Remember, ground isn't.... GROUND STAR?: Apply the usual methode. Each stage receives a local "star" ground where the PSU Filtering capacitors and all other current loops drawing current around this stage are returned. So for example the RIAA "ground" is returned to the first stage star ground.In fact, the best thing to do is to forget anything about "ground" and remember that ground is actually the Signal (return) shared w ith the Powersupply (return). WIRE CHOICE: Bare solid fine silver wire, ideally goldplated in ovesized PTFE Sl eeving, Litzwire from Nebraska Surplus, Enameled copper wire, in order of prefer ence. Any of three are quite cheap. Goldplated 0.2mm Silver Wire from wires.co.u k is around 5 for 10m with the PTFE sleeving costing around the same. DC FILAMENT GROUND: Typhically the best place is the input stage starground. DISCUSSIONS ON DESIGN: quote: Originally posted by Johan Potgieter 1. I must have been a singularly lucky dud e to have been able to get by without very fancy power supplies. My circuits w ere mostly feedback affairs, NFB increases PSRR. Consider a Circuit with very little PSRR and no NFB. quote: Originally posted by Johan Potgieter 2. I would strongly echo someone's thought s early in the discussion on attenuating at sub-audio frequencies. I would strongly advise against. Speakers usually already have well enough LF gr oup delay to cause audible problems, no point excebarating it in a phonostage, e specially not if you can ge a turntable that does not rumble and that does suppo rt the well and makes sure it is quite flat. One may have the occasionally exceedingly warped record that causes problems, in such cases a switchable HPF may be preferable, but it should NEVER be designed into a quality phono stage without bypass option. quote: Originally posted by Johan Potgieter 3. I have found the use of grid leak bias in any but a low gm sharp cut-off tube (say ECC83) a distortion generator, espec ially when fed from a relatively high impedance circuit like a passive RIAA. Adm ittedly the signal level is low here, but at +20 dB peaks (which is posible with some recordings!) this will count. First, no cartridge I ever had tracked +16db ref. 5cm/S @ 300Hz cleanly and the LP standards prescrive =14db re. 5cm/S. Secondly, I found that in this specific circuit the normal signal (0db) on the f inal stage grid is only around 15mV, with the maximum signal being 75mV. The "co ntact bias" system develops around 0.5V static bias (no signal) in this particul

ar circuit. I found a rather substantiative absence of distortion at any normall y expected signal levels and well past. quote: Originally posted by Johan Potgieter When one considers some of the special com ponents mentioned here to get better performance, I personally would not comprom ise by using a grid leak bias with an ECC88. When one considers the absence of observable problems due to grid leak bias and the problems added by using other means of bias (cathode bypass capacitor) it se ems gridleak is not a "compromise", but rather a bonus. quote: Originally posted by Johan Potgieter 4. Then I am not sold on RIAA accuracy nee ding precision components to reach <0.5 dB accuracy etc. Neither am I. quote: Originally posted by Johan Potgieter In fact, I normally over-compensate below 70 Hz to 40 Hz by a few dB, because everything else usually start tapering off t here. It makes for a nice deep-bass effect (but don't overdo it). It will also excebarate any very low frequency noise, which might explain partia lly the need for steep highpass filters. quote: Originally posted by Johan Potgieter Just a few side thoughts, for what they ar e worth. Indeed, just a few thoghts back, fwtaw.And you should really try one of these da ys a non nfb Phono and even gridleak/contact bias. You might be surprised. I hav e done it the other way many times as well. Three short comments: "Consider a Circuit with very little PSRR and no NFB."Why? It is possible to des ign excellent circuits with high PSRR, negative feedback or no. ANSWER: True, bu t it is eaqually easy to make a PSU that is dead quiet simply and cheaply, so wh y bother adding all that complication. "First, no cartridge I ever had tracked +16db ref. 5cm/S @ 300Hz cleanly and the LP standards prescrive =14db re. 5cm/S. "It's the stuff that's not intentional which is the problem. Real-world LPs often have very high level ticks and pops w hich can be dealt with effortlessly- or cause a preamp to choke for some millise conds after, just to make the overload more annoying and noticeable. ANSWER: Yes , this tends to be the problem for NFB Eq'ed stages. The specific design discuss ed has well over 30db overload margin on a 5mV input, which is more than enough for that. IME, this is the single greatest difference among decent phono stages; they can make ticks'n'pops very annoying or let you listen through them effortlessly.And finally, subsonic noise can have more sources than just warps- a LOT of records do not have perfect center holes. ANSWER: The excentric centerhole causes 1.8Hz stuff, as does usually record warp. Even my "fairly flat at 20Hz" phonostage has a good deal of attenuation that far down. Points taken with respect, as were your usual earlier posts. We appear to have had different experiences. As for group-delay effects, acknowl edged; but to my experience there is far more of that and well into the lower au dible band (or not) from loudpseaker design than from a properly designed subson ic filter. One can compare phase effects of various approaches on Spice. ANSWER: I do not know what sort of speakers you are used to, they may explain the diffe rence in experience. In my case you find that I tend to prefer systems with a we ll above average efficiency and high resolution, plus they tend to be impulse co herent or at least nearly so (wideband "fullrange" systems, tannoy coaxials wher e necesary with sealed subs extnding the LF). Conventional so called "High Fidel ity" Speakers (they are all but any fidelity) will give very different results. Yes, a "better" turntable. But it is difficult to convince a customer to pay a 4 -figure amount extra for the same audible gain achieved rumble-wise by just a fe w extra $ worth (or none) of components (but disregarding other possible advanta

ges of exotic turntables). ANSWER: A "better turntable" is not neccesarily extre mely expensive and the benefits extend considerably past rumble and other VLF no ise. My comments regarding grid-leak bias was general. I have used it on low-impedanc e fed ECC83s and EF86s, but found a measurable difference in performance compare d when used with less sharp cut-off tubes, back in the days when I was priviledg ed to work for a top research organisation where expensive spectrum analysers we re available. (Sadly no longer; retired now and such exotics have become nice-to -have's.) ANSWER: Well, any decent laptop (or desktop) PC with AC97 audio can be easily turned into a pretty decent FFT with > 80db dynamic range and if you bui ld a 1KHz notchfilter plus amplifier you can get at least 20-40db more range. I will not dispute that distortion maybe a little higher with gridleak bias at hig h levels, but at operational levels (which is all that interrests me) distortion in my circuit is low enough (and only very low order) that together with the di stortion from LP & Cartridge we are well in the real of academic differences. There is witness that your circuit performs excellently - so be it; I am glad. A gain our approaches are simply different. E.g. I am not that biased against the use of the right kind of bypass capacitor (sacrilege!); we use electrolytics in power supplies where they are also in the signal circuit. In an excellent series of articles about 6 years ago in Electronics World, Cyril Bateman investigated capacitors over the spectrum; he found that bi-polars peformed very acceptably, with graphs to prove. ANSWER: Well, having done quite extensive testing I do not use Electrolytics anywhere if I can. In the powersupply at the very least the f inal filter section capacitor (which is the one that is actually in the signal c ircuit - earlier stages tend to be insulated by the divider formed by the series impedance vs the final capacitors impedance) in my use is ALLWAYS a film type c apacitor.I agree that one may very well use a bypassed cathode resistor, but 100 uF/35V rubycon metalised polypropylene capacitors are first a major PITA to get hold off at all (you need to get them from shops that only speak japanese) and t hey cost over 50 bucks a throw, plus the rippoff from your credit card company o n the exchange rate, shipping and import taxes and duties. I think the key diffe rences between us are that you view everything in the context of commercial HiFi while I reject such mediocrity all across the board and hence do not concern my self with designs that fit into this context (not strictly true, I actually got against better knowledge and judgement involved again in designing commercial au dio gear, but my commercial stuff is nothing like my DIY Designs, simply because they are meant to sell first and perform second).I will suggest again that if y ou find the time you try building a copy of the "el cheapo", including the sugge sted valve rectified RC filtered supply and try to have a good listen, using a s ystem with proper studio grade monitor speakers (or Quad ESL's at a pinch) and d ecent Valve Amp's (preferably triode, low/no NFB). Please do not bother with com mon HiFi stuff as ancillaries. When designing while using them, you do you invar iably design the mediocricy of the ancililaries into your gear and when audition ing other stuff you fail to notice even quite significant differences. Most comm on "HiFi Speakers" are great equalisers, they make everything sound equally bad by their own severe limitations in fidelity. To me DIY was always about creating what you could not more or less readily buy but wanted anyway (this often invol ves a variety of techniques and features not commonly employed in commercial gea r), not as a way to save a few bucks by copying generic commercial gear (better off working a few weeks heavy overtime and buying the commercial unit me sez). As for "trying a non-nfb phono stage and grid-leak bias one of these days and be ing surprised..." I have tried all and explained my reasons for certain preferen ces. I am saddened by my inability to convey that my experiences did come over d ecades, they were not merely blinkered recent fixations. Also my customers appea red to have been quite satisfied (some were musicians). In that sense not too ma ny surprises for me, but let us allow for our different approaches. I distinctly hope though that I will never cease to find surprises - I cannot imagine a more dull life! (Or, woe of woes - a greater sign of old age.)

Sockets Type The lower the levels the more you get trouble with DA (dielectric absorbtion), e specially in areas where the dielectric is not "prebiased", but also where mille r capacitance etc. is concerned, also PTFE Sockets tend to have a somewhat dampe ning effects on microphonics. Clearly all this is beneficial everywhere, but the lower the signals the greater the benefit. Current production PTFE Sockets cost a mint, old stock military sockets are stil l quite common however. Once of these days I will actually buy some ceramic Octa l & UX-4 sockets, salvage the contacts from them and apply them to homemade sock ets in solid PTFE stock or maybe C37 infused wood (loads of air in wood - surpri singly good dielectric once suitably dry and sealed)....

S-ar putea să vă placă și