Cathode follower implementation on phono pre

I'm in the final stages of my phono pre build, but I am stuck with a final design question. The original preamp design was meant to interface with >50K input impedances. I was advised by the designer to tag on a cathode follower so that I could interface with my studio gear (10-15Kohms). Basically, I took an 6SN7 CF from one of his other pres and stuck it on the output of this design.

My question is: Should I leave the output cap and resistor form the original amp design, or tie the 6SN7's directly to the cathode on the 6DJ8? The original cap/resistor are hand drawn on the schematic below.

Thanks,
Chris

Perhaps none of the options will work.

That cathode follower is DC coupled, so it needs the correct DC voltage on the grids (not gnd potential).

Best regards,

Mikkel C. Simonsen

Ugggh. I see that now. So definitely the coupling cap is out. The schematic that I lifted the cathode follower from had 120V on the cathode. This design has 102V at 185 B+. I am running 240V B+ , so I will probably have even more than 120V there. What would an ideal range be for the cathode? My design skills are zero, so I am your mercy :grin: .

[quote author="Emperor-TK"]What would an ideal range be for the cathode?[/quote]
Whatever the circuit was designed for :grin:

But you could just try it without the cap, and see what happens. The other DC voltages in the circuit should not change, and you should have a reasonable amount of current in the cathode followers (probably 2-5mA). The 150n output caps may be too small.

Best regards,

Mikkel C. Simonsen

Take out those scribbled parts. The 6DJ8 plate sits at 102V, which is an appropriate bias for the 6SN7 grid.

C4 C5 MUST be around 2uFd (at 400V!) to feed 10K loads well. R16 R17 should probably be reduced to ~33K to control turn-on thump.

Feed 6SN7 heaters from a separate winding than the one making 6VDC for the input stages. The 6SN7 CF can surely be heated with AC, if you keep the wiring tight and twisted and away from the input stage. Let it float, don't ground it.

I'd actually be inclined to skip the CF and replace 6DJ8 pins567 with a fatter triode like 6BX7(?) with a smaller plate resistor. But Steve took the 2KHz rolloff in that stage. We'd have to re-calculate that (or move it to the 6DJ8 pins123 unit plate and re-calculate). Since 6BX7 has lower Mu, gain is less, which may be an issue.

OK, how about an unpopular idea. You could use a 300V transistor for a follower. Skips the heater issue. Maybe an MJE 340. Only $0.30 at mouser. Keep the cap and use a resistor divider from b+ to base to ground. Maybe lower R14/15 and put them on a heatsink.

A self-biased, AC-coupled cathode follower can be "tacked on" (as it were) easily. Budget about 10mA of plate current for each CF, so 20mA in your case. Actual current draw may be less.



EDIT: I just saw PRR's reply. I agree with his point about reducing the output tie-down resistors to avoid thumps when using a larger coupling cap. I edited the schematic just now and reduced the output tie-down to 100K.

> You could use a 300V transistor for a follower

Sure, and it could be DC-coupled to the previous plate. Do toss a few K in the collector lead so it does not try to pass infinite current on a transient with capacitive load.

But if we are going to use a transistor, very fine all-tranny preamps exist. Some hard, some soft. I assumed this was a religious decision to use tubes (to drive a chip/DAC DAW?).

> Budget about 10mA of plate current for each CF

With given values of ~300V B+ and 47K, drain can't be more than 6.25mA even if the tube shorts. With 1K cathode resistor, and Mu=20, more like 300V/((1K*20)+47K)= 4.5mA.

What do we need? Unlike digi-media, the recorded level on a phonodisc is not standardized, and can vary 20dB from one disk to another. My feeling is that the preamp should have modest gain, followed by a gain control, and a lineamp with gain of 10. A Classic Hi-Fi preamp. This one looks to have a gain of 200 at 1KC. Back in the 1970s I had LPs that would give >4V peaks at that gain, and I've seen recent "12 inch singles" with much bigger wiggles. Such a disk and preamp might beat 8V peak, which is the overload level of many ProTools-class digytal inputs. Taking that as our peak level, we need 8V/10K= 0.8mA peak signal. In a dumb cathode follower, we want more than that. Since distortion is reduced by ~Mu, say >2mA DC idle current.

Holy smokes, thanks everyone for the advice. :thumb: Let me just make a couple of comments on the comments:

Feed 6SN7 heaters from a separate winding than the one making 6VDC for the input stages. The 6SN7 CF can surely be heated with AC, if you keep the wiring tight and twisted and away from the input stage. Let it float, don't ground it.

Click to expand...

I can't exactly do that without changing too much of what I have already built. I went with a G9 style power supply, mostly because I had an extra PCB for the regulators lying around. It wound up being more trouble than it was worth, because when I added the CF stage, I had to add another off-board regulator to handle the heater current. I am heating the 12AY7's with the 78S12 and the 6DJ8's and 6SN7 with an LM317 (trimmed to 6.3VDC). Both are fed by 21VDC . Both are referenced to ground.

I assumed this was a religious decision to use tubes (to drive a chip/DAC DAW?).

Click to expand...

It was based on Kiira?s recommendation. She had built a few phono pres and liked the sound of this one best. Steve Bench also seems to think it?s the best of his 5 designs that he offered up. I also have fun building tube stuff. I will be interfacing it with both analog and digital inputs, depending.

Such a disk and preamp might beat 8V peak, which is the overload level of many ProTools-class digital inputs. Taking that as our peak level, we need 8V/10K= 0.8mA peak signal. In a dumb cathode follower, we want more than that. Since distortion is reduced by ~Mu, say >2mA DC idle current

Click to expand...

I think I understand this about 50%. So you are saying that if we tune the CF to output 8V peak, that would give 0.8mA idle current through the CF? The CF really wants to have 2mA though to reduce distortion. If we did give 2mA idle, then we would be getting 20V peak at the output (which of course would smash anything I was feeding it into).

A self-biased, AC-coupled cathode follower can be "tacked on" (as it were) easily.

Click to expand...

Excellent! Dave, I can do this. Knowing that my B+ will be about 240VDC and my grid bias is about 120-130V, what kind of Vout should I expect with this design? Will it still work under these conditions?

Final question (for now): Is there anywhere that I can throw a stepped attenuator into to adjust gain? Before CF, after CF?

Update:
I actually have the unit together and working. Voltages are pretty much what I expected for the higher B+. I don't have the CF hooked up yet, but rather ran off the original design. The output is low and a little choked sounding, with some low end missing, but I suppose that is the impedance mismatch. But overall the sound is clear and well defined. There is also a 60Hz hum I need to suss out. Other than that, I can't believe that it is actually working the first time that I plugged it in. :shock: :grin: I'm kind of proud that I got this far, doing my own layout and all. This has been my most ambitious project yet.

> So you are saying that if we tune the CF to output 8V peak, that would give 0.8mA idle current through the CF?

No. I'm saying that IF the disk level, the needle sensitivity, and the preamp gain (not known), conspire to give 8V peak (which I think is just possible), and the load is 10K, then we need 0.8mA of signal current.

We want bias current to be somewhat larger. Using my way, 102V/47K= 2mA idle current, which will do. NYD's way gives 3 to 5mA depending, which will also do (at a skooch less THD). BTW, the original 6DJ8 stage runs 80V/20K or 4mA, so it has enough Power: the problem is that he designed the EQ around that node assuming negligible loading, and 10K is not negligible here. It will also be lower in gain, higher THD than designed. And those 0.15uFd caps with 10K load will be somewhat bass-shy.

> There is also a 60Hz hum I need to suss out.

mmmmmmmmmmmmm...... that power transformer looks awful close to the tubes. In a Dynaco it is in the far opposite corner of a 14-inch chassis from the phono tubes. That aluminum wall is not going to stop magnetic hum. I fear there may be a total re-layout in your future. As a quickcheck, unbolt the transformer and gently flop it over the right side. If the increase of distance and change in angle changes the hum, that's something to ponder. (And if it doesn't change the hum, then you can ignore that and look elsewhere.)

Note that regardless of the particular cathode follower you should still have another floating filament supply---although, you are not too far from the heater-cathode limit (100V quiescent) with the direct-coupled approach, so you might get away with it.

Are you really taking 21V down to 6.3 with a 317? Sounds like it's getting mighty hot---1.33A times (21-6.3) = 19.55W! Imax for the 317 is 1.5A, and Pd max is only 15W. Maybe you have a resistor in series with unreg d.c. that is handling some of that dissipation. And maybe your 21V was measured open-circuit, and sags a lot with load.

PRR: OK thanks, I think I get it now. I listened again. In addition to the anemic bass, the overall EQ doesn't sound exactly right (A/B vs. a CD of the same source). That, plus the low output, all signs of the preamp being loaded too much by 10K. I'll play around with the transformers. The separator is actually steel, so it might be providing some magnetic shielding. I am almost ready to rip the whole PS out and start again (or at least mount it in a separate chasis). It's been a big PITA so far. I'd also do the layout different if I were starting from scratch (I'm learning :green: ). I'd probably bring the tubes inside and closer to the turret board (away from the PS).

Brad, thanks for pointing it out. I had 300mA for the 6SN7 written down. I don't know where I got that number from. I just checked, 600mA is the right value. So are you saying that I need to bias the 6SN7 heater, or just float it from ground? Is this because of the large bias on the grid, or because of plate voltage? Looks like I will need a separate heater transformer. Ugggh.

Good misakes, learning mistakes. :green:

[quote author="Emperor-TK"]PRR: OK thanks, I think I get it now. I listened again. In addition to the anemic bass, the overall EQ doesn't sound exactly right (A/B vs. a CD of the same source). That, plus the low output, all signs of the preamp being loaded too much by 10K. I'll play around with the transformers. The separator is actually steel, so it might be providing some magnetic shielding. I am almost ready to rip the whole PS out and start again (or at least mount it in a separate chasis). It's been a big PITA so far. I'd also do the layout different if I were starting from scratch (I'm learning :green: ). I'd probably bring the tubes inside and closer to the turret board (away from the PS).

Brad, thanks for pointing it out. I had 300mA for the 6SN7 written down. I don't know where I got that number from. I just checked, 600mA is the right value. So are you saying that I need to bias the 6SN7 heater, or just float it from ground? Is this because of the large bias on the grid, or because of plate voltage? Looks like I will need a separate heater transformer. Ugggh.

Good misakes, learning mistakes. :green:[/quote]

It's all good. You are getting just enough positive results to motivate further work :grin:

The problem is the maximum rating of the voltage between heater and cathode---the grid and cathode are a bunch of volts positive from ground, and your heater supply is ground-referenced. If it floats, or is d.c.-biased around (best a bit more positive than) the cathode voltage then you can swing lots of volts out and be well within the ratings. Since you are probably well out of noise by the followers an a.c. supply is probably adequate if done with care per PRR. You can probably find room for a 6.3VAC ~1A trafo in the box---either a little toroid, or an EI core carefully located and oriented to minimize magnetic radiation into the more sensitive sections.

As far as mag shielding---it's a bitch. The only really effective way more-or-less completely encloses the affected part. The steel will reshape the field, and may help, but may also make matters worse depending on all the details.

Toroids are self-shielding to a point but not near the leadouts and also not when big charging current pulses with higher freq energy are present, pushing the core toward saturation. The distribution of turns needs to be very uniform too.

The great thing about mag fields is the falloff with distance tends in near field situations to fall off with the third power of distance. And then for a given pickup loop somewhere in your sensitive circuit section, tuning the orientation of the mag field source can null the signal. This is unlikely to be perfect for all the loops though.

> the overall EQ doesn't sound exactly right (A/B vs. a CD of the same source).

Unless you have signed notes from both mastering engineers, assume that LPs and CDs should NOT sound the same. At best, a CD may be mastered direct, though I think they are often "tweaked". But phonodisk cutting almost requires heavy massaging.

You check the EQ with an inverse-RIAA network. I wouldn't think about a DIY RIAA without one: too many ways your calculated EQ could be wrong, and hand-testing RIAA curves is maddening. Use an iRIAA. AudioXpress had one, and Hagerman does, at http://www.hagtech.com/iriaa.html Yes, it ain't cheap, but is accurate and will last a lifetime (and Hagerman is not getting rich). Hagerman has full details for DIY, and theory if you want to study. (He also has a very fine little preamp, the Bugle Phono, that would already be working for you, but it doesn't glow.)

Thanks PRR, I am going to get one of those. I knew that vinyl gets its own "special" mastering job. I just forgot to take that into account. I've had a few vinyl records that I recorded come back from mastering that left me scratching my head (to say the least).

Latest update: I hooked up the CF as per NewYorkDave. Output is about right for a -10dB unbalanced input with a few different test records. The clip light on my signal router ocasionally flickers. The LM317 is running quite hot. Hum is still crazy. I think I need to look for other sources besides the xformers. Perhaps I missed something in my grounding scheme. The reason I say this is because both L and R channels hum equally, despite the tubes for the right channel being about 3 inches further from the power supply. On the plus side, the sound stage is incredibly deep and wide, and the detail is great. I think this will be really nice when it's done.

Would ripple on the B+ cause a loud hum (I guess it would, no)? I haven't checked that yet. When I turn off the unit, the volume drops while its running temporaily off the caps, but the hum instantly goes away.

-Chris

Your last sentence tells me that it's either ripple on the B+, or the heater supply coupling into the audio. Have you put a scope on the B+ and looked for ripple?

How about a schematic of your power supply? That would help.

One note about the preamp schematic: the junctions of R1-R4 and R17-R20 should be very well-bypassed. Unless your supply is super-clean, 0.1uF won't be enough. Try much bigger caps for C2 and C10, 1uF and up.

> the junctions of R1-R4 and R17-R20 should be very well-bypassed.

No. The gain from that grid is 100K/Rp(V2). The main source of rail hum is V1a's high effective plate resistance (due to the V2 plate in its cathode) allowing ALL the rail hum to appear at the output R11.

This is a "fault" of this amp compared to a classic common cathode input. The conventional way gives 6dB-10dB power supply rejection, this way gives 1dB-2dB PSRR. Not a serious DIY fault because we can afford more than the 10uFd-20uFd that Fisher and folks used in phono power filters. But don't forget that there were usually 3, 4, even 5 stages of R-C decoupling from the rectifier to the phono stage, and it was not uncommon to lose 70+% of the original B+ along the way to the phono.

Lessee... ripple at the first PS cap may be 2% or 6V. Ripple referenced to the input should be a microvolt or less. Ripple at the first stage plate can be ~50X higher, 50uV. 6V/50uV= 100dB ripple reduction!!!

An old commercial rule of thumb is to take 20dB per R-C stage: to get 40dB, it is cheaper to use two 20dB stages than one 40dB stage. That says 5 stages (C)-R-C-R-C-R-C-R-C-R-C. Fat asian caps are cheap, we might go 30dB/stage for 4 stages, or 33dB/stage for 3 stages.

The hacked plan shows a 290V dirty rail and a 185V rail that must be clean and supply 10mA. (290V-185V)/10mA is 100V/10mA is 10K total dropping resistor. Say three 3.3K resistors. To get 34dB in each stage the cap reactance must be 3300/50= 66Ω at 100/120Hz. 24uFd, which really is not bad.

Duncan's PSD, for 60Hz 200VAC, FWB rect, 25uFd-3.3K-25uFd-3.3K-25uFd-3.3K-25uFd and 18K load, says 179V final output and 8.8uV p-p ripple. That's hardly an obscene cost.

Changing to 250uFd-10K-250uFd gives 129uV ripple, which shows the folly of trying to do it all in one stage. A couple 470uFd caps around a 10K resistor might work.

100uFd-5K-100uFd-5K-100uFd gives ~4uV and at today's cap prices might be the sweet-spot. This goes up to 7uV when I tap 10mA (for the CFs) from the first cap. How-ever... that's 1.4V of ripple at the CF plate, and 12AU7 CF ripple rejection won't be more than ~20dB. Even dropping another 1K-100uFd in there gives 15mV at CF plate, over 1mV of ripple at the output jack, just too much.

Yes, power supply cleanliness IS a problem for a hungry lo-PSRR phono preamp.

First off, I want to thank both you guys for taking the time to walk through this with me. I am learning a lot, but even more-so, I am learning more about where I need to learn. I feel like I have the oportunity to be in the heads of a couple of great designers.

So PRR, if I am reading you right, the only viable option that you have discussed is the 3 stage RC filter per Duncan. At this point, I have over 40 hours of build time sunk into the unit. I see no reason to cut costs by trying to eliminate a couple RC stages. I am not afraid of overdesigning the PS at all.

I don't understand this calculation:
>>>To get 34dB in each stage the cap reactance must be 3300/50= 66Ω at 100/120Hz. 24uFd, which really is not bad.

I assume that the 24uF here is a minimum? I also assume that the 66ohm reactances is why the B+ drops down to 179V rather than the 185V calculated for the 10K dropping resistors. Is that 66ohms related to ESR?

With a "price is no option" approach, would this be a better design:

200VAC, FWB rect, 25uFd-(2K-25uFd)x5?

I can also lower the values of the dropping resistors further, because B+ for the 12AY7's and 6DJ8's can be up to 250V according to S Bench's specs. Again, I don't understand the calculation you did to determine ripple reduction, so I don't know how playing with the dropping resistors affects performance per RC stage.

Side note: This is pretty cool. The physics here parallel what we often see in chemical engineering. Distillations, ion exchanges, and heat transfer processes are always more effective using a series of lower performance stages rather than one or two high performance ones.

-Chris

[quote author="Emperor-TK"]I don't understand this calculation:
>>>To get 34dB in each stage the cap reactance must be 3300/50= 66Ω at 100/120Hz. 24uFd, which really is not bad.

I assume that the 24uF here is a minimum? I also assume that the 66ohm reactances is why the B+ drops down to 179V rather than the 185V calculated for the 10K dropping resistors. Is that 66ohms related to ESR?

-Chris[/quote]

The caps are like open circuits at d.c. Three 3.3k in series (9.9k total) will give the same result as one 10k, d.c.-wise.

The calculation is to determine what C has to be for its capacitive reactance to be 50 times less than 3.3k---PRR used 50Hz full-wave-rectified, hence with the lowest frequency energy at 100Hz, as the conservative assumption. A factor of 50 in voltage is about 34dB. 1/50 of 3.3k is 66 ohms, and at 100Hz a 24uF cap has the required |Xc|. If we were splitting hairs we only need 1/49 of 3.3k for 34dB, but this is a damn power filter :grin:

The calculation for multiple stages assumes no loading to speak of on the output, and no loading of one section by another, both of which are good approximations in this example. Also, with this much attentuation per stage you don't need to worry about phase.