I've been reading up a bit on hybrid cascodes and mu followers lately and I've got an itch to build another phono preamp. I came up with the above (this would be fed by some kind of regulated supply).
Notes:
- - I based my calculations on a 10mA Idss Jfet lower device on the input with a Vgs(off) of -1.1V. The above would be a transconductance of 14 mS for a gain in the first stage of 175x. The gain could definitely be increased with a larger load resistor on the tube, but I wanted to go with a lower overall impedance in the RIAA (because I did a 200k impedance network for a 12AX7 not long ago and feel like trying something different). Different Idss JFETs will probably require tweaking the source resistor.
- - I used a 6DJ8 as the upper device because the high transconductance looks to me like it will lower the load on the JFET, maximizing the benefit of transconductance there and voltage gain "in the tube."
- -The upper part of the voltage divider for the grid reference is bypassed because I'm calculating the gain here to be about -1. In theory I'll be letting through all of the power supply ripple due to the high impedance of the cascode and relatively low impedance of the load. Injecting the ripple into the upper tube will invert the ripple at the output, counteracting the ripple allowed through by the poor PSRR of the cascode topology.
- - The all in one RIAA is pretty straightforward (I have a spreadsheet that calculates the values if anyone is interested).
- - I think a mu follower makes a lot of sense as the output stage here. I'll still get the benefit of voltage gain, but also lower output impedance. I drew up both a tube and a MOSFET mu follower upper device, but I think eliminating the extra heater requirement (and accompanying regulation) is worth letting more SS into the preamp. I'm hoping the lowered PSRR with the MOSFET won't be an issue (I'm planning a quiet supply anyways).
- - Although it will change Miller Effect capacitance (which technically should result in minor tweaks to the .0068uF capacitor), it looks like I can sub various tubes in the mu follower to get different levels of gain. I get a value of .00681 if this capacitance is 50pF or .00671 if the input capacitance to the second stage is 150pF. So .0068uF is within 1.5% either way (and most of these caps are rated to 1%). Based on datasheets, the 250V B+ should allow for plenty of headroom on the 10M45S with the tubes listed.
- - I don't have a MC cart, but I think 60db looks ok with some of the output specs I've seen. 50db is a little high for MM, but a nice, strong 1Vrms signal out of a phono preamp would be nice in my system.
Cascodes and Mu Followers are both new topologies for me. So where am I going wrong? What have I overlooked?
I've been reading up a bit on hybrid cascodes and mu followers lately and I've got an itch to build another phono preamp. I came up with the above (this would be fed by some kind of regulated supply).
Notes:
- - I based my calculations on a 10mA Idss Jfet lower device on the input with a Vgs(off) of -1.1V. The above would be a transconductance of 14 mS for a gain in the first stage of 175x. The gain could definitely be increased with a larger load resistor on the tube, but I wanted to go with a lower overall impedance in the RIAA (because I did a 200k impedance network for a 12AX7 not long ago and feel like trying something different). Different Idss JFETs will probably require tweaking the source resistor.
- - I used a 6DJ8 as the upper device because the high transconductance looks to me like it will lower the load on the JFET, maximizing the benefit of transconductance there and voltage gain "in the tube."
- -The upper part of the voltage divider for the grid reference is bypassed because I'm calculating the gain here to be about -1. In theory I'll be letting through all of the power supply ripple due to the high impedance of the cascode and relatively low impedance of the load. Injecting the ripple into the upper tube will invert the ripple at the output, counteracting the ripple allowed through by the poor PSRR of the cascode topology.
- - The all in one RIAA is pretty straightforward (I have a spreadsheet that calculates the values if anyone is interested).
- - I think a mu follower makes a lot of sense as the output stage here. I'll still get the benefit of voltage gain, but also lower output impedance. I drew up both a tube and a MOSFET mu follower upper device, but I think eliminating the extra heater requirement (and accompanying regulation) is worth letting more SS into the preamp. I'm hoping the lowered PSRR with the MOSFET won't be an issue (I'm planning a quiet supply anyways).
- - Although it will change Miller Effect capacitance (which technically should result in minor tweaks to the .0068uF capacitor), it looks like I can sub various tubes in the mu follower to get different levels of gain. I get a value of .00681 if this capacitance is 50pF or .00671 if the input capacitance to the second stage is 150pF. So .0068uF is within 1.5% either way (and most of these caps are rated to 1%). Based on datasheets, the 250V B+ should allow for plenty of headroom on the 10M45S with the tubes listed.
- - I don't have a MC cart, but I think 60db looks ok with some of the output specs I've seen. 50db is a little high for MM, but a nice, strong 1Vrms signal out of a phono preamp would be nice in my system.
Cascodes and Mu Followers are both new topologies for me. So where am I going wrong? What have I overlooked?
I know it takes 2 more tubes but the tetra phono circuit works wonderfully.
I'm using a 6N1P-12AT7-6N1P-6N1P set up. RIAA vaules were 21.8K instead of 110k, 0.033uf instead of the 0.01x2, and 3.18k into 0.1uf instead of 0.058/15k9.
The benefit is the circuit appears as a resistor to the power supply simplifying the filtering and lower resistances which will add less noise.
Here's a blank schematic of it.
http://www.tubecad.com/2009/10/23/Tetra Schematic Large.png
EDIT: I don't see anything wrong with your circuit though. It should work properly.
EDIT2: My phono cart outputs about 7mv so a gain of 35db is plenty. If you need more you can use a 6N2P or 12AX7 in place of the 12AT7.
Last edited:
A few ohms, no. 20000+ is enough error to introduce a ~ 0.75 db reduction at the 20hz pole.
Is this true? Referring to the upper triode, i am seeing a grounded cathode amplifier with a stiff unbypassed CCS on the cathode. This means whatever signal (ripple) is injected to the grid will be absorbed by the CCS on the cathode to maintain constant 10mA plate current which means no inversion on the plate AND still poor cascode PSRR. Plus, i am not sure how you are getting 125V on the plate with at least 10mA flowing through 12k5+6k from 250V.
I think you will get better chance at increasing PSRR by injecting a portion of the ripple at the lower JFET gate (AC-coupled of course).
Oops, thanks for pointing this out. I'll fix it in my copy!
ballpencil said:
Hi Ballpencil. The Idss of the JFET is 10mA, but I'm calculating the current draw (Id) as 6mA due to the source resistor (so ~75V across the 12k5 resistor). I am ballparking Vgs(off) as -1.1V, so I'm sure I'd have to tweak the JFET operating point a bit one way or the other (or rework it completely if I can't find matched Idss 10mA devices). I do not work with JFETs much though, so I might be off on this.
Looking at 2SK170 datasheets, it appears to me that the Rd is about 12k. So the gain of the upper triode at the grid is:
Mu * Rload / (Rload + Rp + Rd * (Mu +1))
With Mu of 35, Rload of 12k5, Rp of 2500, Rd of 12k, I am getting a gain of -0.98. The Mu and Rp numbers may not exactly reflect real-world devices, but I would think that using the grid to inject inverted ripple to the anode will help the poor inherent PSRR of cascodes even if they're slightly off. Injecting it at the JFET would result in a huge amplification of the ripple and over-correct for the ripple at the anode/load junction, wouldn't it?
I think you're right. I was trying to put the grid about 8V above zero so that the JFET would see 10-12V Vds. But I forgot when I added the 6k RC to drop some volts that I had to rework these numbers. I just adjusted this divider to 200k + 8k, but anything in the ballpark should work fairly well I think.
This divider sets the grid bias voltage; in conjuction with the current drawn by the JFET, it sets the cathode voltage (which is Vds for the JFET).
Only inject a small portion of the ripple so that when it's amplified, it matches the amplitude of the anode ripple. If your gain is 175x as you calculated, then this is approx how much you should attenuate the ripple before injecting it to the gate. A trimpot to adjust for the best PSRR is easy to implement. Check out the Aikido Gnd-K MC Prepreamp schematic for an idea of how to.implement it: https://www.tubecad.com/2015/08/blog0328.htm
When you inject at the triode grid, you are relying on the gain to be as close as possible to -1 to counter the anode ripple. This can be rather unreliable as it depends on the jFET Rd as you calculated. Should Rd becomes even slightly more than 12k, you are getting further away from deep ripple null which is impossible to correct as you need to inject even larger ripple. More explanation here
Improving the Cascode's PSRR (page 2)
5AM in the morning here, i hope i didn't miss anything.
No, I think you're right on. Looking at my spreadsheet output, the gain is pretty sensitive to the Rd of the JFET. Much higher than 12k Rd, and the divider would have to inject more than the ripple value to be amplified and cancelled completely (though it would still be partially cancelled). Lower than 12k Rd and a cap across the lower resistor in the divider would help tune it.
I'll read that TubeCAD article. I think I have been through it once, but JB always has some good noise rejection tricks (which is where I originally got the idea to inject ripple to the upper triode).
I'll read that TubeCAD article. I think I have been through it once, but JB always has some good noise rejection tricks (which is where I originally got the idea to inject ripple to the upper triode).
Not desiring to be pedantic, but where do you get that, Kodabmx? What 20 Hz pole?
Osvaldo was talking about the resistor in parallel with the 4,7µF cap to the grid of V1. For signal the 241,5K resistor is effectively in parallel with the 8,5K resistor, presuming the upside-down electrolytic bypass cap is large. Whatever capacitor gives a pole in conjunction with these resistors, the difference in value between 220K and 241K in parallel with 8,5K will amount to 0,028 dB, not 0,75 dB.
But perhaps OT.
.5% resistors will reduce RIAA gain error by 10x. Using multiple Capacitors to get the proper cap values will work to produce the most accurate RIAA curve response.
The problem is that the system is not considered, this consists of the Phono Cartridge as well. When this is taken into account, resistor and capacitor values can vary more.
I "Tuned" my preamp to work with the Cartridge on my LAB 500 turntable to sound the same perceptively as the CD version of the LP. Mine varies from RIAA by more then 5 DB. RIAA sounds dead and lacks Bass punch.
I will never go back to RIAA. the only RIAA spec that holds up is gain at 1khz is 36 Db.
The problem is that the system is not considered, this consists of the Phono Cartridge as well. When this is taken into account, resistor and capacitor values can vary more.
I "Tuned" my preamp to work with the Cartridge on my LAB 500 turntable to sound the same perceptively as the CD version of the LP. Mine varies from RIAA by more then 5 DB. RIAA sounds dead and lacks Bass punch.
I will never go back to RIAA. the only RIAA spec that holds up is gain at 1khz is 36 Db.
You probably mean 10%? Even then .... but let us leave it at that.
I think the general aspiration to RIAA accuracy to far better than human hearing can detect is overrated. All capacitors have tolerances, and simply 'piling it up' as per marked values does not necessarily eventually give an exact value. For that one rather needs a capacitance meter and select.
Ah, kind Sir! There you have it! Indeed.
I sense your frustration, and it does not really matter how you get the desired response in the end. But it might serve the purpose better to cure the trouble at its source. The RIAA characterisitc has for decades been used so often that it has to be accepted almost as much as Ohm's Law. That is simply the way records are cut.
To go where you are going measuring instruments are essential; day-to-day hearing is not an accurate measuring tool. And part of this chain will have to be a frequency record (if still obtainable).
Please note that I am not saying hearing is not the final arbiter (your hearing, that is). But simply 'getting there' is easier with measuring tools/techniques than listening.
I would like to know where exactly in your system the deviation(s) occur. Happy hunting!
Johan, Thanks for the kind words, You are absolutely correct that tuning a system by ear is completely subjective and prone to much higher mathematical error then frequency analysis via controlled systems. That being said, I had no other way to test then by using the LP version of Dire Straits Brothers in Arms vs it's CD version. I dropped the needle o the record and released the pause on CD at first sound. I then used a tape monitor input for the CD to flip back and forth. The result is that the first point at 500Hz is now 435Hz and the second RIAA break point of 2122 Hz is now at 2350. the gain resistor is 50% larger giving more gain below 80hz. the cut from 2350Hz is leveled off at 15db giving a smooth rising of the top end.
The Cartridge is loaded 75K and 47pf at the input.
The Cartridge is loaded 75K and 47pf at the input.
Last edited:
Hi Soulmerchant,
I have a couple of matched 2SK170 pairs at around 8.5ma Idss and a pair of LSK170 (Linear Systems) at 10ma Idss. I think I'll build this with the Linear Systems part because it is available as new production. They sent me a sample of six Jfets and two were within 1% at 10ma (which was what I was designing around). Lucky!
I have a couple of matched 2SK170 pairs at around 8.5ma Idss and a pair of LSK170 (Linear Systems) at 10ma Idss. I think I'll build this with the Linear Systems part because it is available as new production. They sent me a sample of six Jfets and two were within 1% at 10ma (which was what I was designing around). Lucky!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.