rongon said:
Close but needs some adjustment. If I plug your values into the Hagerman calculator (reversing the order to reflect your order), I get a value of around 42k for the Total series resistance required. From this you have to subtract the impedance of the preceding stage, and account for the following grid leak resistor, to get the resistor value.
So one step at a time. First stage:
Plate resistance of 7788 is about 3K (adjust accordingly, if this is not the correct value at your operating point). Your cathode resistor is unbypassed, so you need to multiply 50R times u (let's say about 50), and add that to the plate resistance. So (2.5k + 3k)//9k is 3.4k. This is in series with your 47k. Now you have to parallel the following grid leak resistor so 50.4k//470k = 45.5k That is your total effective series resistance.
Input capacitance of the following 5687 is about 4pf, plus 4pf times u(or 18). Let's say 80pf total. This is effectively in parallel with your RIAA cap.
Therefore your total RIAA cap for this stage is 1nf + 620pf + 80pf, or 1.7nf total. Plugging this value into Ca for the Hagerman calculator gives a required series resistance of 42.24k.
You can tweak the series resistor or the cap. If you stick with 47k, you need 1.58nf total. Subtracting 80pf, means a RIAA cap of 1.5nf.
I'll leave it to you to do the second stage.
Sheldon
edit: corrected for unbypassed 7788 cathode resistor, and included following stage grid leak resistor.
Why does "using a second section of the 5687 as a third stage amplifier (negate) all the calculations of the 318 + 3180us section"? Wouldn't it be possible to include the effect of the 5687's Miller capacitance in the calculations? That's also unavoidable in a two-stage design, right? No matter what, you're going to see an active stage after the 318uS + 3180uS filter in a two-stage design...
And why do both Arthur and Thorsten Loesch ignore all this? Or did they include that in their calculations?
I'm following the topology of the Arthur Loesch RIAA (three amp stages), also built and schematic published by Thorsten Loesch (no relation). This is a three stage design with the RIAA filters split between the first/second and second/third stages. Splitting the passive filters into two parts is supposed to be a good thing, isn't it?
Arthur Loesch's hi-gain MC version uses a 6GK5 in the second stage. The MM/hi-output MC version uses a 6C4 in the second stage. Both use a 417A in the first stage and half of a 5687 in the third stage (common-cathode, not cathode follower).
Thorsten's version uses a 5842 in the first stage, 6GK5 in the second and 1/2 5687 in final stage. His MC version has a 2SK147 and 5842 in cascode(?). MM/hi-output MC uses 5842 in common cathode (no FET).
Thorsten Loesch's version:
http://www.fortunecity.com/rivendell/xentar/1179/projects/toccata/Toccata.html
I'm hoping to use a triode-wired 7788 (common cathode) as the first stage and a whole 5687 (or 7119) per side for this version, because I have lots of those tubes. I don't need more gain than that because I'll be using a Denon DL-110 with output = 2.2mV (even though the factory spec says 1.6mV, it is in reality a little above 2mV). I have an active line amp after that.
The only reason for not doing a two-stage design is because I heard one of these Loesch preamps years ago, and it was possibly the best hi-fi experience I've ever had. Even with its gross errors, my current version captures some of that sonic magic. All I'm trying to do is get the RIAA values close to correct...
Trade-offs, trade-offs... Two or three amp stages? One complex or two simpler EQ stages? Which is better?
OK, this is what I could come up with quickly, using the GlassWare calc...
rp of 5687 is about 2500 Ohms, but unbypassed with 150R Rk gives about 3450 Ohm rp.
11.1k plate resistor and 499k following grid leak resistor.
C = .062uF
Rseries = 47.5k
Rparallel = 5.1k
How's that look to you?
Lunch is over, so back to work I must go...
EDIT: Forgot about the Cmiller of following stage - Subtracting about 80pF from .062uF = inconsequential?
rp of 5687 is about 2500 Ohms, but unbypassed with 150R Rk gives about 3450 Ohm rp.
11.1k plate resistor and 499k following grid leak resistor.
C = .062uF
Rseries = 47.5k
Rparallel = 5.1k
How's that look to you?
Lunch is over, so back to work I must go...
EDIT: Forgot about the Cmiller of following stage - Subtracting about 80pF from .062uF = inconsequential?
Attachments
Designing an accurate RIAA stage with valves is one of the more difficult things to do. As I said a cathode follower has very high input impedance and very low input capacitance and you can couple the 318+3180 stage to it directly (no coupling capacitor) and it will not affect the action of that stage of the filter. If you use a third stage with gain you have to use that coupling capacitor, the input impedance is much lower and the amplified input capacitance is much higher so that it will interfere with the previous stage and throw it out.
According to Morgan you have to juggle five variables to get the right result, which means that your results for R1, R2 and C will have to change.
According to Morgan you have to juggle five variables to get the right result, which means that your results for R1, R2 and C will have to change.
Oh, I see. That hasn't entered into any of the equations so far?
I could easily change this circuit to 7788-triode > 6DJ8 DC-coupled to 6DJ8-CF.
The mu of 7788-triode is about 40 to 50, 6DJ8 is about 30. That should be enough gain, right?
I could even run them with fixed bias from batteries, yielding as much gain as possible.
--
Personally I used an EC88 (not ECC88 ) which has a mu of 68 for the second stage and an EC86 for the third stage cathode follower, mainly because I had some. Morgan used all ECC88s (6DJ8s) in his first example but that offends against the Morrisonian commandment : thou shalt not use the same kind of valve twice in a row.
If I were you I would sit on my hands until Morgan's book arrives : he does have designs for three active stage preamps in it.
If I were you I would sit on my hands until Morgan's book arrives : he does have designs for three active stage preamps in it.
rongon said:
Looks fine. Yes. The Cmiller is below the error threshold here.
barretter said:
The third stage miller is calculated exactly the same as for the second stage. The coupling cap has nothing to do with it. In this example the Cmiller is much lower than the RIAA cap (62nF vs 0.08nF), so can be safely ignored.
However, the interstage coupling cap will have a slight effect on RIAA curve, as it's reactance at 50Hz is about 15kOhm. It forms a HP filter with the grid leak resistor at about 1.4HZ. You have two in series in the circuit, with one in the previous stage too - sort of a built in rumble filter. I don't think the effect will be significant, but you can easily try a bigger cap for the second stage and see if it sounds any different.
rongon said:
Since your RIAA compensation eats up about 20dB of gain at 1kHz, you need about 60dB to start with. A 40x30 = 1200x, which is about 60dB. That's fine, but you will not realize the full u of either tube unless you are using current sources as plate loads, and fixed bias or bypassed cathodes.
Sheldon
edit: Here's my version of a two stage amp into a follower. I used an Aikido follower to take advantage of some noise cancellation (since changed the first stage bypassed cathode resistor to an LED). The stuff after the follower is part of the noise cancellation. I get about 33dB for the first stage and about 26 for the second. http://www.diyaudio.com/forums/attachment.php?s=&postid=1067656&stamp=1164742822
I've read the RDH4 and RC-19 manual sections on CF design and based what I've done on that.
Seems you need to plot a good loadline, which usually means getting the Ep, Ec and Rk load to yield the desired Ip to get enough current going through the tube to keep it in a linear area of operation all through the likely voltage swing. Also make sure the CF can't be overloaded by the previous stage(s) at any point. Just the usual common sense stuff.
I've used a constant current source in the cathode of a CF and noticed some improvement over the usual load resistor. Used paralleled 1N5309 or 1N5314, that kind of thing. I once tried LM317 as a CCS. Worked well. I'd do that again. I'm sure there are lots of posts on that around here, if I search...
That said, I've always noticed a slight "thickening" or "slowness" of sound from a CF. It's somehow different than a regular common-cathode gain stage. Am I imagining this or do others notice this too?
Or is there something I'm missing in the proper design of a CF? (That's likely, I'm sure...)
--
Seems you need to plot a good loadline, which usually means getting the Ep, Ec and Rk load to yield the desired Ip to get enough current going through the tube to keep it in a linear area of operation all through the likely voltage swing. Also make sure the CF can't be overloaded by the previous stage(s) at any point. Just the usual common sense stuff.
I've used a constant current source in the cathode of a CF and noticed some improvement over the usual load resistor. Used paralleled 1N5309 or 1N5314, that kind of thing. I once tried LM317 as a CCS. Worked well. I'd do that again. I'm sure there are lots of posts on that around here, if I search...
That said, I've always noticed a slight "thickening" or "slowness" of sound from a CF. It's somehow different than a regular common-cathode gain stage. Am I imagining this or do others notice this too?
Or is there something I'm missing in the proper design of a CF? (That's likely, I'm sure...)
--
My goodness! That's a lot of supporting circuitry. I wonder if I could pull that off... I suppose I'll have to try that bipolar transistor in the cathode trick, next time I make a CF. Is this leaps-and-bounds better than using an LM317T to ground? Is LM317 too noisy? Or were you reaching for the stars with this, just because...?
I'm not reaching for the limits of what's do-able at this point. I just want to make something really basic, competently. My layout skills leave much to be desired. I'm working on it...
--
I'm not reaching for the limits of what's do-able at this point. I just want to make something really basic, competently. My layout skills leave much to be desired. I'm working on it...
--
Hey SY,
Thanks for the info on the CF. I'll definitely try your idea for bipolar transistor in the cathode of the CF. I'd really like to be able to make a CF I like the sound of...
Following Sheldon's lead,
In the meantime, I couldn't resist playing with RC values some more, so here's what I've got today:
The plate resistors for the first and second stages are 15k Mills WW-NI paralleled with 1W metal films to arrive at the desired values. They're working, so I'd like to leave them alone.
I looked at the layout (point-to-point wired) in my phono pre and I don't want to dig in and change those 100k series resistors. My clumsy layout put them soldered in under other parts. As painful as it is, I've got to keep them and find a workaround. I can get to the other RIAA parts quite easily, though.
Fortunately, I have a decent collection of polystyrene C's and metal film R's.
I used Sheldon's calculations as a blueprint and the GlassWare calc to do a double-check.
I decided to look in the parts drawers to see if I could parallel parts to come out with the correct values (or very close). I think I can...
So here's the latest schematic, showing these paralleled Rs and Cs.
edit: Forgot to mention...
5687
rp = 2500
u = 16
grid C = ~4pF
7788 triode
rp = 3450
u = 40 to 50 (no data, so have to guess)
Thanks for the info on the CF. I'll definitely try your idea for bipolar transistor in the cathode of the CF. I'd really like to be able to make a CF I like the sound of...
Following Sheldon's lead,
In the meantime, I couldn't resist playing with RC values some more, so here's what I've got today:
The plate resistors for the first and second stages are 15k Mills WW-NI paralleled with 1W metal films to arrive at the desired values. They're working, so I'd like to leave them alone.
I looked at the layout (point-to-point wired) in my phono pre and I don't want to dig in and change those 100k series resistors. My clumsy layout put them soldered in under other parts. As painful as it is, I've got to keep them and find a workaround. I can get to the other RIAA parts quite easily, though.
Fortunately, I have a decent collection of polystyrene C's and metal film R's.
I used Sheldon's calculations as a blueprint and the GlassWare calc to do a double-check.
I decided to look in the parts drawers to see if I could parallel parts to come out with the correct values (or very close). I think I can...
So here's the latest schematic, showing these paralleled Rs and Cs.
edit: Forgot to mention...
5687
rp = 2500
u = 16
grid C = ~4pF
7788 triode
rp = 3450
u = 40 to 50 (no data, so have to guess)
Attachments
Funny, that didn't change much. It brought the total series resistance down by roughly 2k Ohms.
rp = 1k
Rp = 8.9k
u = 40
Rk = 60
Unbypassed Rk * mu
60 * 40 = 2400
Rk*mu + rp
2400 + 1000 = 3400
Now, 3400 Ohms in parallel with 8900 Ohms (Rp)
3400 * 8900 = 30,260,000
3400 + 8900 = 12,300
30,260,000 / 12,300 = 2460
That's the 7788-triode Zout if its internal plate resistance is ~1k.
Now, Zout is added to the Rseries:
2460 + 100,000 = 102,460
Then that in parallel with the 470k grid leak resistor of following stage...
102,460 * 470,000 = 48,156,200,000
102,460 + 470,000 = 572,460
82,142 Ohms is total series resistance
Plugging values into GlassWare RIAA calculator, it gives C1 = 890pF (so 820pF plus following 5687's Cmiller of about 75pF should do it) and R1 should be 100k.
If that and the second stage look OK, then I think I'm close enough for the time being.
How's that look? Did I make any mistakes?
--
rp = 1k
Rp = 8.9k
u = 40
Rk = 60
Unbypassed Rk * mu
60 * 40 = 2400
Rk*mu + rp
2400 + 1000 = 3400
Now, 3400 Ohms in parallel with 8900 Ohms (Rp)
3400 * 8900 = 30,260,000
3400 + 8900 = 12,300
30,260,000 / 12,300 = 2460
That's the 7788-triode Zout if its internal plate resistance is ~1k.
Now, Zout is added to the Rseries:
2460 + 100,000 = 102,460
Then that in parallel with the 470k grid leak resistor of following stage...
102,460 * 470,000 = 48,156,200,000
102,460 + 470,000 = 572,460
82,142 Ohms is total series resistance
Plugging values into GlassWare RIAA calculator, it gives C1 = 890pF (so 820pF plus following 5687's Cmiller of about 75pF should do it) and R1 should be 100k.
If that and the second stage look OK, then I think I'm close enough for the time being.
How's that look? Did I make any mistakes?
--
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.