Yes, I agree the matching task is daunting and very labor intensive. The guy does have an Amplitrex but not unlimited funds.
This is the reality of the supertriodes---by their very nature, with that insanely tight spacing, they are bound to have sample-to-sample variations greater than a more "relaxed" dimensioned tube.
The guy is also an apostle of simplicity, so a more elaborate approach to paralleling in which each tube is adjusted and outputs combined in a different way than simple paralleling would probably meet with scorn. At least the grids could be commoned.
I was struck by the low distortion of individual tubes too. There was a lot of variation in that as well, with the cited 40ppm the best of the lot of three, and the current source (again used just to circumvent the need for dragging in a higher voltage supply in order to use larger plate load resistors, and to determine the actual mu, Ra, and gm) was hardly perfect---an 88.7 ohm emitter R, MPSW92, and HLMP-6000 LED running with ~3mA bias. With large output swings the thermal distortion at low frequencies could be a problem. The contribution to equivalent input noise by the I source should be small as uncorrelated, but could be smaller. But I'm not trying to sell him on sand, it was just for the measurement. He's talking a stepdown transformer to get the requisite gain for a stepup, and for the paralleled 6922 he had to damp a peak in the freq response with a 3.3k R, which with the 2:1 stepdown means about a load of 13.2k at the 6922 plates.
Brad
The knee for 1/f is probably rather higher than 400Hz. The aggregate noise in the 400-22kHz passband nearly doubled when it was 22Hz to 22kHz. I should have done FT plots too, which would show interference stuff poking above the general trend.WoW, that seems Very good to me for tubes ! I know it's using a HPF though, but hiss would be nice & lowish, for a MM anyway. I know this about MC's, but that spec would be great for MM's
So it "appears" that below 400Hz is where the noise increases. Filtering out PS nasties shouldn't be a problem ?
PS noise wasn't too bad, and I filtered with 200 ohms and 1000uF. But hum from local fields at 60Hz, 120Hz, and 180Hz was one thing difficult to ensure elimination. And since there is so much wireless-related crud in the air and borne on mains these days, I am still hedging a bit on the numbers even so.
The nominal noise expected from a triode that does not include the 1/f component is typically that of a resistor of 1/gm and an equivalent temp of the cathode reduced by the space-charge-smoothing constant of 0.644. So for 1000K, square root of 4kT/gm in this case would manage about 1.3nV/sq rt Hz.
What if Tk could be reduced? At some point you lose the space charge layer and in the limit the effect of the grid altogether. It's possible that the S45 is made to withstand large peak currents without stripping away the space charge layer---I think this is also true of the dual 6H30. But the stepup app is unlikely to need to withstand EMPs
Yes, I agree that the tube as is would be nice as a first-stage MM preamp, especially if the microphonics could be suppressed. The maximum gain of mu could be approached and you're mostly out of noise after that. If a passive RIAA is your thing, you could buffer with another CF before driving the network.
Hi,
That does help somewhat. I've always use 12VDC regs for the heaters on my preamp and headamp designs. Valves last longer and are less noisy too.
Maybe worth doing some measurements here. I no longer have access to the gear unfortunately.
D3As aren't that hard to come by, not in Europe anyway. OTOH the prices I've seen for the 6C45 somewhat what me.....
And I agree with Brad, should we decide on the 6C45 (I still have a preference for the EC86 but it has nowhere near the transconductance of the 6C45 yet would be a good valve to use in my own MC Hammer design given some smallish changes), each paralleled set of valves should have only have the grids in common.
Short of using dedicated sand ccs there's the mu-follower which even though not requiring the same valve as the current reg will also add some noise, cathode resistors would do the same.
Not easy all this......
Brad, I so agree with you on driving a passive RIAA network with a CF. It's a compromise worth studying as the addition of an extra valve will add it's own noise.....No free lunch.
Cheers,
That does help somewhat. I've always use 12VDC regs for the heaters on my preamp and headamp designs. Valves last longer and are less noisy too.
Maybe worth doing some measurements here. I no longer have access to the gear unfortunately.
D3As aren't that hard to come by, not in Europe anyway. OTOH the prices I've seen for the 6C45 somewhat what me.....
And I agree with Brad, should we decide on the 6C45 (I still have a preference for the EC86 but it has nowhere near the transconductance of the 6C45 yet would be a good valve to use in my own MC Hammer design given some smallish changes), each paralleled set of valves should have only have the grids in common.
Short of using dedicated sand ccs there's the mu-follower which even though not requiring the same valve as the current reg will also add some noise, cathode resistors would do the same.
Not easy all this......
Brad, I so agree with you on driving a passive RIAA network with a CF. It's a compromise worth studying as the addition of an extra valve will add it's own noise.....No free lunch.
Cheers,
I've not used Russian pencil tubes in a phono pre, but koi farm and others have without serious issues with microphonics. I have used American pencil tubes like the 5744 and 6022 in a simple phono stage and the performance was OK, my biggest problem was the best performance seemed to be at voltages somewhat beyond the maximum rated plate voltage..
When I first moved to the current place and set up my man cave listening space my very first phono pre here was a 5744/6022 based design. I was at the time not too thrilled with the what I had for vinyl playback which comprised a TD-125 MKI with SME 3009 Series II (Unimproved) with a Grado 8MR on it.. Bored me to tears.. An upgrade to the Grado Platinum Reference was a definite step in the wrong direction, a more boring cartridge could not exist imo.
The world is a much better place with Schicks, SPUs and TD-124s. lol
The world is a much better place with Schicks, SPUs and TD-124s. lol
You haven't missed anything as far as I can tell from limited experience.. Some people love Grado cartridges and I've tried, but just don't. The SPU (any SPU) is so different in character and is in line with what I like so the Grado never really stood a chance.
The phono stage was probably pretty mediocre; no way to know its potential given the arm/cart/TT combo it was used with, it was an ill-fated experiment that was quickly replaced with the first of my very strange designs; triode connected D3A and 5741A with passive EQ and tube based gyrator loads. The gyrator based unit has moved on to a friend who uses it with LL1941 SUTs and an Ortofon Windfeld LOMC. (Sheu table, high mass Schroeder arm)
The phono stage was probably pretty mediocre; no way to know its potential given the arm/cart/TT combo it was used with, it was an ill-fated experiment that was quickly replaced with the first of my very strange designs; triode connected D3A and 5741A with passive EQ and tube based gyrator loads. The gyrator based unit has moved on to a friend who uses it with LL1941 SUTs and an Ortofon Windfeld LOMC. (Sheu table, high mass Schroeder arm)
I'm happy to hear that someone did tube gyrators. I was about to, for something or other, but didn't get very far before the target moved a lot.
gyrators
A true gyrator transforms an impedance at one port to a reciprocal impedance at the other, hence a C becomes an L.
Approximations to the ideal gyrator can be made from ~unity-gain buffers like cathode followers. If you look at the Sallen-Key highpass unity-gain filter, its input impedance looks like a lossy inductor in series with the filter input capacitor. That's the form that has been used probably the most, with emitter followers or op amp unity-gain buffers, but there are others with various advantages and disadvantages.
When dealing with tubes we're already not too concerned about bulky, usually, so good inductors may be the better choice in many applications, especially as they are easy to float. The gyrator versions get complicated further if the L is needed as a series element. This led to FDNRs, frequency dependent negative resistors, those in turn made from GICs, generalized immitance converters. GICs are usually realized with two op amps.
Not sure what Kevinkr did exactly.Hi,
To be totally honest I don't even have a clue what a tubed gyrator would look like.
The idea is to replace an inductor with an active circuit as the anode load, is that correct?
Cheers,
A true gyrator transforms an impedance at one port to a reciprocal impedance at the other, hence a C becomes an L.
Approximations to the ideal gyrator can be made from ~unity-gain buffers like cathode followers. If you look at the Sallen-Key highpass unity-gain filter, its input impedance looks like a lossy inductor in series with the filter input capacitor. That's the form that has been used probably the most, with emitter followers or op amp unity-gain buffers, but there are others with various advantages and disadvantages.
When dealing with tubes we're already not too concerned about bulky, usually, so good inductors may be the better choice in many applications, especially as they are easy to float. The gyrator versions get complicated further if the L is needed as a series element. This led to FDNRs, frequency dependent negative resistors, those in turn made from GICs, generalized immitance converters. GICs are usually realized with two op amps.
Hi Frank,
Why wouldn't you use 6j32p / Ef86 / 6267 as the input stage?
6c45p is a high-gm triode, with tiny distance between cathode and grid, thus microphonics and large grid (noise) current (as Brad explained). High-gm triodes have more higher harmonics.
If you would like triode no matter what you may have a look on 6c17k phono stage from Eric Barbour
Why wouldn't you use 6j32p / Ef86 / 6267 as the input stage?
6c45p is a high-gm triode, with tiny distance between cathode and grid, thus microphonics and large grid (noise) current (as Brad explained). High-gm triodes have more higher harmonics.
If you would like triode no matter what you may have a look on 6c17k phono stage from Eric Barbour
6c17K is a new one on me, and an odd package. would appear to need and interesting mount, or globs of solder.
Found this link wavebourn posted a few years back. http://91.201.116.112/acustica/6c17k.zip
Found the Barbour article as well. Interesting, but no actual measurements of noise.
You learn something new every day
Found this link wavebourn posted a few years back. http://91.201.116.112/acustica/6c17k.zip
Found the Barbour article as well. Interesting, but no actual measurements of noise.
You learn something new every day
Yes the grid current is substantial, enough so that with Vk = 0 I can't drive it into net forward conduction even with Vg ~ +100mV!Hi Frank,
Why wouldn't you use 6j32p / Ef86 / 6267 as the input stage?
6c45p is a high-gm triode, with tiny distance between cathode and grid, thus microphonics and large grid (noise) current (as Brad explained). High-gm triodes have more higher harmonics.
If you would like triode no matter what you may have a look on 6c17k phono stage from Eric Barbour
The friend is concerned about connecting such a stage (or even worse, with two or more parallel devices) to his friend's $5k MC. I said take out an insurance policy, but the premium will probably be about... $5k.
The biggest worry for me is that the guy would say, even barring a malfunction, that "It never sounded the same after that"
But of the things that could go wrong: a parasitic of great amplitude (years ago I blew up C45's before I determined a safe gate stopper component, although I was running them at high current), or however unlikely, a piece of something shorting plate to grid, in which perhaps the MC coil would protect the tube grid