[GlowGale]

Okay, Saturday is schematic day.
 
I'm still figuring out some tweaks here and there but this one should already be functional. I tried to keep it as close to the original SSMH as possible. So let's dig in!
 

 
My build type number should say everything: 12AU7-1-48DC150. So we have a 12AU7 build using a single tube with an 48V PSU and a DC-coupled output stage biased with 150mA of current per channel. These are the original specs of the SSMH, except for the single tube. For the record, the heater is configured with the center tap grounded and a split heater at 6,3V and 150mA. This should give the same 300mA total quiescent current as in the original build.
 
I'll be working my way from input to output. Since I'm running on a single tube I added a dual gang (linear) balance pot to compensate for any channel imbalance. This then connects to a resistor and the typical dual gang (log) volume pot of your preference. Credit goes to Rod Elliott from ESP for this input network. The whole thing has 3dB attenuation with the balance pot centered and should give quasi constant loudness when turning it. The input impedance is somewhere around 10k instead of the 100k of the standard 12AU7-mod. This should be both less noisy and has enough juice to push the tube grid closer to 0V without distorting. The output of the volume pot then runs into a low pass filter composed of some grid stoppers and an extra capacitor to account for RF interference. The grid stoppers and the cap should be soldered directly onto the tube socket for maximum effectivity.
 
The tube gain stage and mosfet output stage are next. The cathode and plate resistors are approximately doubled in value because I'm only using one tube half per channel. Again, I added some DC bias pots for any channel imbalance due to the single tube. I decided to recycle the original SSMH mosfet source follower output stage and simply stack another voltage follower on top. The lower one is connected to a soft-start DC voltage reference, which sets the heater voltage. The top one follows the tube plate voltage. AC couple the audio out with some output caps and done.
 
The voltage follower mosfets are the logic level versions (IRL510, not IRF510!) and have a gate stopper and a 5V protection zener to prevent any oscillations and ESD casualties, while also preventing a large turn on thump from charging the output caps too fast. Again, solder these directly on the mosfet pins. In fact, this should be done before anything else to prevent dead mosfets. The input impedance of the mosfet is practically one zillion ohms so maybe solder the plate resistor directly to the gate stopper of the mosfet and run a wire to the tube socket.
 
I added some typical expected voltages for debugging. As I’m still guessing for the exact tube operating point, the plate voltage is just a guideline for optimal bias of the voltage follower. This stuff is still experimental so proceed only if you have some crappy test headphones and know how Ohm's law works.
 
Have fun prototyping!

GG
 
Edit list:

1. Anode voltage tweak
2. Balance pot polarity (pot wiper arrows point to the clockwise setting)
3. Soft-start time constant

 

[jshaffe]

I've been looking all over for someone who has tested the total per channel output of the starving student in mW at say 32 and/or 300 ohms.  It seems that a dummy load would be required for such a test, of which I do not own.  Anyone know if this has been done and published somewhere? 

 

[DingoSmuggler]

  I've been looking all over for someone who has tested the total per channel output of the starving student in mW at say 32 and/or 300 ohms.  It seems that a dummy load would be required for such a test, of which I do not own.  Anyone know if this has been done and published somewhere? 

Get a couple of 2W or 5W resistors in each required value and then you will have some dummy loads, cheap as chips.

 

[jshaffe]

Forgive my ignorance please, but if it only requires a single resistor, why do all the diy dummy loads look like this?

 

[ximamp]

That looks like a fairly large dummy load that is probably immersed in oil. each one of those resistors is prob 2W or more. You don't need that much power dissipation for this amp. 
They are made like that because you need to use a purely resistive load (i.e. no induction that wirewound resistors have). So easier to put together many smaller carbon resistors. Immersed in oil so you can dissipate more heat (with large amps). I believe those dummy loads are used also in transmitters (ham radios) but in that field I know half of someone who knows nothing. 
Others should pitch in with a better explanation but I think in a nutshell this is accurate. 

 

[jshaffe]

Awesome. Thanks for the reply. I'm picking up 33, 120, & 300 ohm 5w resistors from digikey.  
 
Process;
-----Amp Output----Multimeter-----Resistor-----Ground-----
Measure peak voltage with DMM, divide by the square root of 2 to get RMS voltage. 
(RMS^2)/load impedance = Wattage
 
Correct?

 

[the_equalizer]

  Okay, Saturday is schematic day.
 
I'm still figuring out some tweaks here and there but this one should already be functional. I tried to keep it as close to the original SSMH as possible. So let's dig in!
 
<snip>
 
My build type number should say everything: 12AU7-1-48DC150. So we have a 12AU7 build using a single tube with an 48V PSU and a DC-coupled output stage biased with 150mA of current per channel. These are the original specs of the SSMH, except for the single tube. For the record, the heater is configured with the center tap grounded and a split heater at 6,3V and 150mA. This should give the same 300mA total quiescent current as in the original build.
 
I'll be working my way from input to output. Since I'm running on a single tube I added a dual gang (linear) balance pot to compensate for any channel imbalance. This then connects to a resistor and the typical dual gang (log) volume pot of your preference. Credit goes to Rod Elliott from ESP for this input network. The whole thing has 3dB attenuation with the balance pot centered and should give quasi constant loudness when turning it. The input impedance is somewhere around 10k instead of the 100k of the standard 12AU7-mod. This should be both less noisy and has enough juice to push the tube grid closer to 0V without distorting. The output of the volume pot then runs into a low pass filter composed of some grid stoppers and an extra capacitor to account for RF interference. The grid stoppers and the cap should be soldered directly onto the tube socket for maximum effectivity.
 
The tube gain stage and mosfet output stage are next. The cathode and plate resistors are approximately doubled in value because I'm only using one tube half per channel. Again, I added some DC bias pots for any channel imbalance due to the single tube. I decided to recycle the original SSMH mosfet source follower output stage and simply stack another voltage follower on top. The lower one is connected to a soft-start DC voltage reference, which sets the heater voltage. The top one follows the tube plate voltage. AC couple the audio out with some output caps and done.
 
The voltage follower mosfets are the logic level versions (IRL510, not IRF510!) and have a gate stopper and a 5V protection zener to prevent any oscillations and ESD casualties, while also preventing a large turn on thump from charging the output caps too fast. Again, solder these directly on the mosfet pins. In fact, this should be done before anything else to prevent dead mosfets. The input impedance of the mosfet is practically one zillion ohms so maybe solder the plate resistor directly to the gate stopper of the mosfet and run a wire to the tube socket.
 
I added some typical expected voltages for debugging. As I’m still guessing for the exact tube operating point, the plate voltage is just a guideline for optimal bias of the voltage follower. This stuff is still experimental so proceed only if you have some crappy test headphones and know how Ohm's law works.
 
Have fun prototyping!

GG
 
Edit: Did a re-upload due to incorrect plate voltage. Should be right this time.

Great post, thanks for that!  I'm looking at prototyping this. I have just one question, about the output section, why are the logic version of the MOSFETs required?
 
cheers!

 

[GlowGale]

 
Great post, thanks for that!  I'm looking at prototyping this. I have just one question, about the output section, why are the logic version of the MOSFETs required?
 
cheers!

 
No prob, you're welcome. The lower doesn't actually need to be the logic version as it's Vg (and thus Vs) is static and defined by the voltage divider in the PSU. Buying in bulk is cheaper than having 2 types of mosfets in a design and the logic version costs only a few cents more. The Vgs can be compared to the "width" of the mosfet. The wider the mosfet is, the less room it has to wiggle around between ground and 48V, so it will clip at lower AC voltage levels than with a logic type. Standard types have a Vgs somewhere around 5V, while logic types have a Vgs around 2,5V. So you gain 2,5V of clean headroom for any high impedance headphones. If you want to use the standard IRF510's, recalculate R4 and R5 to get Vg 2,5V higher and rebias the plate voltage 1,25V higher than on the schematic. You'll know the mosfet Vgs better than me as I don't have anything laying around for the moment.
 
By the way. I'm in need of some measurements from the SSMH as I'm basically flying blind on where the 12AU7 is operating (currently no prototyping rig here). I doubled the plate and cathode resistor values from the original design but I'd like to know the plate voltage, plate current and cathode voltage of the 12AU7-modded SSMH (and a resistance measurement of the cathode and anode resistor values). This post is my best guess for the moment. The lower one gives me something around 642µA per tube.
 
Let me know if it works,
GG
 
PS: Mind the turn-off thump, it will be larger as the output capacitors will charge to 26V instead of 12,6V.

 

[the_equalizer]

   
No prob, you're welcome. The lower doesn't actually need to be the logic version as it's Vg (and thus Vs) is static and defined by the voltage divider in the PSU. Buying in bulk is cheaper than having 2 types of mosfets in a design and the logic version costs only a few cents more. The Vgs can be compared to the "width" of the mosfet. The wider the mosfet is, the less room it has to wiggle around between ground and 48V, so it will clip at lower AC voltage levels than with a logic type. Standard types have a Vgs somewhere around 5V, while logic types have a Vgs around 2,5V. So you gain 2,5V of clean headroom for any high impedance headphones. If you want to use the standard IRF510's, recalculate R4 and R5 to get Vg 2,5V higher and rebias the plate voltage 1,25V higher than on the schematic. You'll know the mosfet Vgs better than me as I don't have anything laying around for the moment.
 
By the way. I'm in need of some measurements from the SSMH as I'm basically flying blind on where the 12AU7 is operating (currently no prototyping rig here). I doubled the plate and cathode resistor values from the original design but I'd like to know the plate voltage, plate current and cathode voltage of the 12AU7-modded SSMH (and a resistance measurement of the cathode and anode resistor values). This post is my best guess for the moment. The lower one gives me something around 642µA per tube.
 
Let me know if it works,
GG
 
PS: Mind the turn-off thump, it will be larger as the output capacitors will charge to 26V instead of 12,6V.

Thanks, got it. I'll probably prototype with the IRF; getting electronic components where I live is not at all easy and they barely know what I'm talking about when I ask for an IRF MOSFET.
 
The current you're calculating per tube for the 12AU7 SSMH is in the ballpark, yes. I seem to remember measuring mine and getting ~700µA per tube, cathode voltage ~0.9 Volts.  Unfortunately I don't have a "standard" 12AU7 build anymore as mine is running with CCS plate loads hooked directly to the 48 volt line. If I could find the original plate load resistors, I might replace the CCSs and measure but I'm away from home now and will be unable to get to my build for another 2 weeks.
 
About the operating point, I'm not sure how feasible it is to choose one for the 12AU7 in this build; given that the voltage and current are sooo low compared to what the optimum environment would be for the tube, I'm sure it's operating in a very unlinear area of it's characteristics anyway (a compelling reason to use CCS plate loads and get improved linearity from the tube). I think it was in Morgan Jones' "Valve Amplifiers", 3rd. ed. where I read that around 200 volts at the plate are required to bias a 12AU7 out of grid current...
 
Back when I did the mod, I asked Dsavitsk about drawing a load line for the 12AU7 at such low voltage and current; you can see his answer below that post of mine.
 
That turn-off thump does not look good, power failures are not infrequent where I live (one of the main reasons why I moved to a Millet Mini-Max).
 
cheers!

 

[GlowGale]

 
Thanks, got it. I'll probably prototype with the IRF; getting electronic components where I live is not at all easy and they barely know what I'm talking about when I ask for an IRF MOSFET.
 
The current you're calculating per tube for the 12AU7 SSMH is in the ballpark, yes. I seem to remember measuring mine and getting ~700µA per tube, cathode voltage ~0.9 Volts.  Unfortunately I don't have a "standard" 12AU7 build anymore as mine is running with CCS plate loads hooked directly to the 48 volt line. If I could find the original plate load resistors, I might replace the CCSs and measure but I'm away from home now and will be unable to get to my build for another 2 weeks.
 
About the operating point, I'm not sure how feasible it is to choose one for the 12AU7 in this build; given that the voltage and current are sooo low compared to what the optimum environment would be for the tube, I'm sure it's operating in a very unlinear area of it's characteristics anyway (a compelling reason to use CCS plate loads and get improved linearity from the tube). I think it was in Morgan Jones' "Valve Amplifiers", 3rd. ed. where I read that around 200 volts at the plate are required to bias a 12AU7 out of grid current...
 
Back when I did the mod, I asked Dsavitsk about drawing a load line for the 12AU7 at such low voltage and current; you can see his answer below that post of mine.
 
That turn-off thump does not look good, power failures are not infrequent where I live (one of the main reasons why I moved to a Millet Mini-Max).
 
cheers!

Too bad they don't make any current production car audio tubes (at least that I know of). Don't dismantle your build just for some resistors! I'll just leave those 66-68k resistors like in the schematic and will fiddle around myself for the optimal value. Thanks for the reference, should come in handy.
 
The only thing I'm a bit afraid of is running hot audio sources with a cathode voltage of only 0,9V. As far as I read, you should have a nice low input impedance to drive the grid close to 0V without grid current distortion, so that's why I used such a wacky input network.
 
The best way to avoid the turn-off thump alltogether (without any propagation delay) is by wiring your audio output through a 3PDT or 4PDT power switch. Problem solved. If you're using the Cisco PSU (and the PSU is always plugged in), use a 48V or 2x24V relay in series (with some freewheeling diodes) to remotely switch on the power to the system without having PSU and audio wiring in the same switch. So the power switch only has a connection to the relay and across the audio output. The relay coil should actually work a bit noise reducing so you have no PSU hash on the switch.
 
The turn-on thump shouldn't be too high due to the RC low pass filter on the mosfet gate controlling the heater voltage.
 
Greetings,
GG

 

[Jo-Sur]

hi im new here.i want to ask if i can use irf520 without modifying the schematic parts.because is difficult to find in Greece irf510.thnks to all

 

[vixr]

  Too bad they don't make any current production car audio tubes (at least that I know of). Don't dismantle your build just for some resistors! I'll just leave those 66-68k resistors like in the schematic and will fiddle around myself for the optimal value. Thanks for the reference, should come in handy.
 
The only thing I'm a bit afraid of is running hot audio sources with a cathode voltage of only 0,9V. As far as I read, you should have a nice low input impedance to drive the grid close to 0V without grid current distortion, so that's why I used such a wacky input network.
 
The best way to avoid the turn-off thump alltogether (without any propagation delay) is by wiring your audio output through a 3PDT or 4PDT power switch. Problem solved. If you're using the Cisco PSU (and the PSU is always plugged in), use a 48V or 2x24V relay in series (with some freewheeling diodes) to remotely switch on the power to the system without having PSU and audio wiring in the same switch. So the power switch only has a connection to the relay and across the audio output. The relay coil should actually work a bit noise reducing so you have no PSU hash on the switch.
 
The turn-on thump shouldn't be too high due to the RC low pass filter on the mosfet gate controlling the heater voltage.
 
Greetings,
GG

Here is something I've been fooling around with... For a single side etch.
 

 

[GlowGale]

  Here is something I've been fooling around with... For a single side etch.
 

I almost spilled my coffee when I saw this! Cool design, I can tell it's not your first pcb. You get better heat dissipation with 4 heatsinks so lower temperatures. You can also put 2 mosfets on one of those heatsinks (front & back) and isolate them completely from each other and from the heatsink. This should give a more compact design, albeit with higher temperatures.
 
The balance pot just looks like a copy paste from the volume control, which is my fault. I copy pasted the input control for the other channel and forgot to adjust the balance pot polarity. I will do a reupload in a few minutes. If the balance control is turned clockwise, the volume on the right channel should go up, while the volume on the right channel should go down. If I'm right to understanding the balance pot, swap the left-in and the ground on your balance pot. If looks matter, I designed the balance pot to actually be a trimmer, so it will not be centered perfectly if your tube is unbalanced. Quick 'n dirty solutions: use a potentiometer knob without center marking or first adjust for balance and then mount the knob so the marking is centered. Btw, does the Alps pot also have a linear 25k version?
 
For preliminary testing without a tube: Use two 2W power resistors instead of the filaments. R12 and R14 will act as pull-up resistors for Q1 and Q3. This dumps most of the voltage on Q2 and Q4 which should get warm. Check that the voltage across the power resistors is approximately 6,3V. Use 47 ohm resistors to get the same current, or 100 ohm to get the same power dissipation as with the tube inserted.
 
Otherwise this looks quite neat. Go for it!
Greetings,
GG
 
PS: You should be able to put some Wima FKP's in for C4 and C5.

 

[vixr]

  I almost spilled my coffee when I saw this! Cool design, I can tell it's not your first pcb. You get better heat dissipation with 4 heatsinks so lower temperatures. You can also put 2 mosfets on one of those heatsinks (front & back) and isolate them completely from each other and from the heatsink. This should give a more compact design, albeit with higher temperatures.
 
The balance pot just looks like a copy paste from the volume control, which is my fault. I copy pasted the input control for the other channel and forgot to adjust the balance pot polarity. I will do a reupload in a few minutes. If the balance control is turned clockwise, the volume on the right channel should go up, while the volume on the right channel should go down. If I'm right to understanding the balance pot, swap the left-in and the ground on your balance pot. If looks matter, I designed the balance pot to actually be a trimmer, so it will not be centered perfectly if your tube is unbalanced. Quick 'n dirty solutions: use a potentiometer knob without center marking or first adjust for balance and then mount the knob so the marking is centered. Btw, does the Alps pot also have a linear 25k version?
 
For preliminary testing without a tube: Use two 2W power resistors instead of the filaments. R12 and R14 will act as pull-up resistors for Q1 and Q3. This dumps most of the voltage on Q2 and Q4 which should get warm. Check that the voltage across the power resistors is approximately 6,3V. Use 47 ohm resistors to get the same current, or 100 ohm to get the same power dissipation as with the tube inserted.
 
Otherwise this looks quite neat. Go for it!
Greetings,
GG
 
PS: You should be able to put some Wima FKP's in for C4 and C5.

Sir,
    thanks for the tips...I figured I would have to wire the balance control off board anyway, so I did copy and paste the library volume pot. It didn't occur to me to stack the mosfets! Great idea! I have found a few 25k balance controls but some dont list if they are linear or not. I will tweak the PCB a bit and see if I can get it smaller. As it is, the board would be 3.5" x 5.5". Thanks for the kind comments.

 

[GlowGale]

  Sir,
    thanks for the tips...I figured I would have to wire the balance control off board anyway, so I did copy and paste the library volume pot. It didn't occur to me to stack the mosfets! Great idea! I have found a few 25k balance controls but some dont list if they are linear or not. I will tweak the PCB a bit and see if I can get it smaller. As it is, the board would be 3.5" x 5.5". Thanks for the kind comments.

Sir, I salute you! Doing single sided pcb design without any jumper wires is a black art in itself.
 
If you can find a dual gang trimmer or 2 simple 25k trimmers, go ahead and use them on the board. The main idea of the original circuit was to get constant loudness while tweaking the balance of your typical hifi speaker amp. Just set both to 50% with a DMM and turn the more silent side up untill both channels match in loudness. Just make sure you use the right tongue angle when tweaking those trimmers.
 
Greetings,
GG