Once finished my LME49810 amp i found that it was rather dry, so i decided to add a tube buffer to it. The thing was a very simple PCC88 with a 6K resistor on its cathode.
As expected, some rock and jazz songs did sound much better, but i noticed that the amp had losen some of its original clarity, which was a fault when listening to some good records.
Then i did replace the resistor with a constant current sink, which gave almost the original sound, which made me hate songs i did like with the resistor-loaded cathode follower and i was close to remove the tube buffer when i had that idea: Create a variable cathode load with fixed bias point.
Some (me included) find that deliberately adding THD is something bad to do in an amp, since its a sort of effect rather than amplification, but what about adding distortion when you want it? Douglas Shelf mentioned that tube amplifiers should have a "niceness knob" to choose between an accurate reproduction or a warm one, and that's what my buffer does.
I attach the schematic of the bias circuit and some graphs of simulated distortion at different settings. NICENESS_KNOB should be a reversed log potentiometer of about 100K. The TRIM potentiometer is intended to adjust the voltage of the cathode with the NICENESS_KNOB open-circuit to match the one across R2 with the NICENESS_KNOB at zero resistance.
I love it!
As expected, some rock and jazz songs did sound much better, but i noticed that the amp had losen some of its original clarity, which was a fault when listening to some good records.
Then i did replace the resistor with a constant current sink, which gave almost the original sound, which made me hate songs i did like with the resistor-loaded cathode follower and i was close to remove the tube buffer when i had that idea: Create a variable cathode load with fixed bias point.
Some (me included) find that deliberately adding THD is something bad to do in an amp, since its a sort of effect rather than amplification, but what about adding distortion when you want it? Douglas Shelf mentioned that tube amplifiers should have a "niceness knob" to choose between an accurate reproduction or a warm one, and that's what my buffer does.
I attach the schematic of the bias circuit and some graphs of simulated distortion at different settings. NICENESS_KNOB should be a reversed log potentiometer of about 100K. The TRIM potentiometer is intended to adjust the voltage of the cathode with the NICENESS_KNOB open-circuit to match the one across R2 with the NICENESS_KNOB at zero resistance.
I love it!
I post the pdf version of a schematic with the tube. The trim resistor needs to be tweaked since the value of 700 was to make the bias point stable in simulation. In the actual circuit i need to use a higher value.
Low currents are handy because they can be powered from smps inverters (mine is working from a digital camera flash). And they add quite a bit of tubey signature when resistor-loaded.
With the CCS load distortion is inaudible ( < 0.05% second harmonic). The square wave respone is plain perfect. It seems that CF circuits don't know what ringing or overshot means, and that's good because it keeps the dynamics from high slew rate chipamps.
*The thing misses the grid stopper resistor. It should be something between 100R-470R. The PCC88 loves to oscilate.
I havn't placed any extra decoupling transistors (assumed to have one per rail on the board) because these like to load with spiky currents with the pulses from the rectifier diodes.
The current buffer "cascoding" transistor is important because it isolates the high-transistor-density region from the tube region. 2 transistor ccs are a must because they provide very high power supply rejection and dc operation will give even better performance.
Low currents are handy because they can be powered from smps inverters (mine is working from a digital camera flash). And they add quite a bit of tubey signature when resistor-loaded.
With the CCS load distortion is inaudible ( < 0.05% second harmonic). The square wave respone is plain perfect. It seems that CF circuits don't know what ringing or overshot means, and that's good because it keeps the dynamics from high slew rate chipamps.
*The thing misses the grid stopper resistor. It should be something between 100R-470R. The PCC88 loves to oscilate.
I havn't placed any extra decoupling transistors (assumed to have one per rail on the board) because these like to load with spiky currents with the pulses from the rectifier diodes.
The current buffer "cascoding" transistor is important because it isolates the high-transistor-density region from the tube region. 2 transistor ccs are a must because they provide very high power supply rejection and dc operation will give even better performance.
Very low distortion would make the tube totally transparent?
I've been simulating a much improved version of the buffer, which i hope to build soon. The thing dips into the 3 ppm THD LM4562 realm on spice. Despite i don't expect to get that performance on the actual circuit, it will not distort much for sure. This has been done by ensuring that there is no signal across the tube.
Please note that there is a transistor named "THIS_TRANSISTOR_WILL_EXPLODE". This transistor should be changed by a high voltage, low current one. Vas transistors will do just fine. Then R12 should be choosen to match Rpi||(1/gm) of the vas transistor and a bit more to compensate for the base current of the chosen transistor.
The main idea of the circuit is to provide a "dancing floor" for the tube to do it's best, then decide with how much signal we are going to feed it. The "dancing floor" is pure solid state since its use is to bias the tube as good as possible and all we know that transistors have better electrical characteristics, This should let the tube shine when we need it.
I would love to read your feedback about the circuit.
I've been simulating a much improved version of the buffer, which i hope to build soon. The thing dips into the 3 ppm THD LM4562 realm on spice. Despite i don't expect to get that performance on the actual circuit, it will not distort much for sure. This has been done by ensuring that there is no signal across the tube.
Please note that there is a transistor named "THIS_TRANSISTOR_WILL_EXPLODE". This transistor should be changed by a high voltage, low current one. Vas transistors will do just fine. Then R12 should be choosen to match Rpi||(1/gm) of the vas transistor and a bit more to compensate for the base current of the chosen transistor.
The main idea of the circuit is to provide a "dancing floor" for the tube to do it's best, then decide with how much signal we are going to feed it. The "dancing floor" is pure solid state since its use is to bias the tube as good as possible and all we know that transistors have better electrical characteristics, This should let the tube shine when we need it.
I would love to read your feedback about the circuit.
Today i've bought the transistors, but it will take at least one month to build because i'm on exams now. I posted here for public review to be alerted if someone founds some mistake and to give ideas to people doing theyr own designs.
I belive that the low distortion of the design when ccs-loaded is in part due to the PCC88 being linear to boredom when ccs loaded even at low currents. The two only requisites i see for a tube to work well here are:
· High grid voltages at low currents (This may be overcomed by changing the 6K resistor values and the ccs current).
· Cathode voltage within the SOA of the "hot" transistor, which excludes tubes that go 600V or so.
Maybe this is a situation where the least linear tube will do the best since it's possible to trim how much signal does the tube actually see and the linearity of the tube sets the maximum tubeness.
Solid state section has a lot of parts cost but it accounts for a grand total of less than $5 excluding the quad potentiometer.
It still requires some way of reducing psu noise, in my prototypes (no niceness-know, either maximum or minimum settings) i use a lowpass filter done with two 1K resistors and two 470uF 200V* || 1uF poly which is not the best way to get it for sure. *Taken from a computer smps.
If i have time i will do a conceptual schematic replacing discrete ccs and voltage sources with theyr symbol.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.