taylorCS follower
Maybe this is a good spot to chip in my current project:
A source follower buffer with a Taylor current source. EUVL brought it up in his thread on source followers. It is also explained in John Broskie's tubecad.
The circuit is push-pull, but instead of transmitting the voltage from the sense resistor R1 to the current source J2 by a capacitor, it uses a pnp-transistor. That way you can use a capacitor C5 to keep the power supply ripple away from the gate of the current source. A second extra is the opportunity to add a servo to set the base voltage of the transistor. With two low pass filters R4/C1 and R11/C5 there is virtually no signal left at Q1.
Actually it doesn't need a servo because you can also set the Q5 base voltage, and thus output DC, with a trimpot and it will pretty much stay there. If J1-J4 are pairwise matched and thermally coupled they only thing that makes DC drift is the temperature of Q1. And you can keep the heat on Q1 low with choosing the right R6.
The circuit is not CFP but it still sports a Zout of only 12 Ohm.
Maybe this is a good spot to chip in my current project:
A source follower buffer with a Taylor current source. EUVL brought it up in his thread on source followers. It is also explained in John Broskie's tubecad.
The circuit is push-pull, but instead of transmitting the voltage from the sense resistor R1 to the current source J2 by a capacitor, it uses a pnp-transistor. That way you can use a capacitor C5 to keep the power supply ripple away from the gate of the current source. A second extra is the opportunity to add a servo to set the base voltage of the transistor. With two low pass filters R4/C1 and R11/C5 there is virtually no signal left at Q1.
Actually it doesn't need a servo because you can also set the Q5 base voltage, and thus output DC, with a trimpot and it will pretty much stay there. If J1-J4 are pairwise matched and thermally coupled they only thing that makes DC drift is the temperature of Q1. And you can keep the heat on Q1 low with choosing the right R6.
The circuit is not CFP but it still sports a Zout of only 12 Ohm.
Attachments
Sorry for creating errors guys,I guess sleep deprivation.
So here is my understanding about Calvin's buffer with servo. All part numbers I describe are based on Calvins schematic in post 15.
Servo being used because there is voltage offset between input and output.
In original design without servo a Trim pot is used across R8 to correct it. But my understanding is ,over the time and due to temperature offset can change. I guess that means if I set trimpot , after circuit has been running for sometime, there still will be an offset when I turn it on next time until fets are warmed up. Also over the time I need to make sure there is no drift due to aging.
The current through buffer is determined by two resistors R7 and R12. In servoed version the R12 is about 10% lower made by R12b to create an offset to trigger the servo. Calvin described in post #15 how to calculate a suitable resistor to control current through R20. To simplify it is the positive supply voltage/ (1/10th of buffer idle current). My schematic idle current is near 33ma. so 10/3.3=3K.
Idle current is measured across R7.
My last schematic has mistakes.Attached is the corrected one. What do y'all think about this one?
To ElFishi,
I have simulated Taylor follower and found to have higher THD. With Calvins design THD is very low with Z out close to 9Ohms. so why go after TCS follower?
So here is my understanding about Calvin's buffer with servo. All part numbers I describe are based on Calvins schematic in post 15.
Servo being used because there is voltage offset between input and output.
In original design without servo a Trim pot is used across R8 to correct it. But my understanding is ,over the time and due to temperature offset can change. I guess that means if I set trimpot , after circuit has been running for sometime, there still will be an offset when I turn it on next time until fets are warmed up. Also over the time I need to make sure there is no drift due to aging.
The current through buffer is determined by two resistors R7 and R12. In servoed version the R12 is about 10% lower made by R12b to create an offset to trigger the servo. Calvin described in post #15 how to calculate a suitable resistor to control current through R20. To simplify it is the positive supply voltage/ (1/10th of buffer idle current). My schematic idle current is near 33ma. so 10/3.3=3K.
Idle current is measured across R7.
My last schematic has mistakes.Attached is the corrected one. What do y'all think about this one?
To ElFishi,
I have simulated Taylor follower and found to have higher THD. With Calvins design THD is very low with Z out close to 9Ohms. so why go after TCS follower?
Attachments
Last edited:
Go with what suits you. The TCS follower has far less parts 🙂
The differences in simulated THD I find negligible, we are already far out.
Real life THD will be different anyway since the JFET model in LTSPICE is linear while real JFETs are not.
When I measure my test circuit with a behringer UCA202 (admittedly not a high end device) and RightMark Audio Analyzer I find it indistinguishable from the plain loopback.
Your input has a .22uF cap and a 10k pot: makes for a highpass filter with a cut-off frequency of 72Hz. May sound a bit dull in the low end...
Thanks Elfishi. Well don't mind having few more parts. What would be a Cap value that won't cut above 15-20 Hz?does 1 uF work?
Last edited:
Low parts count gives you a better chance to understand how the circuit works, I'd say.
High-pass filter - Wikipedia
But the fun about this circuit is that it doesn't need an input cap.
If you are happy to use an input cap (and you can have a small one after the volume pot) then you can actually use a more regular (and tested) servo circuit. See e.g. chapter 8 in Cordell, Designing Audio Amplifiers.
Can you post your asc file with the non-standard models you use?
High-pass filter - Wikipedia
But the fun about this circuit is that it doesn't need an input cap.
If you are happy to use an input cap (and you can have a small one after the volume pot) then you can actually use a more regular (and tested) servo circuit. See e.g. chapter 8 in Cordell, Designing Audio Amplifiers.
Can you post your asc file with the non-standard models you use?
I modified one of Calvin's design so if I have the permission from Calvin I can post it here.
" but the fun about this circuit is that it does not need an input cap" which circuit are you referring to?
" but the fun about this circuit is that it does not need an input cap" which circuit are you referring to?
Well the problem is we can't be sure about it always, right? So Won't it be wise to add a blocking cap. Still it is better than having two, may be?
Thanks for the book ref, reading it now.
I don't have the skill set to do it , sadly. All I can do is a little bit more than following instructions🙁. So if you can help me out by any means by improving the servo please do.
Last edited:
Your choice. There are people who want to avoid caps at any rate.
Playing it safe is another valid design objective.
The high input impedance of a JFET input stage makes it particularly undemanding to use a cap, but it has to go after the volume pot. After the cap you need to tie the JFTE gate to ground or it will latch up.
So, source - pot - cap - resistor to ground - JFET. With 0.1 uF and 1MOhm you get a decent cutoff frequency and you can still afford your favourite fancy boutique cap.
Reading the servo chapter from Cordell. Is there any benefit in using double OPAMP to make a servo as such rather than using one cap less?i would assume two will be more noisy?
Attachments
Last edited:
Normally you wouldn't worry b/c all opamps go behind a largish R4 in your snippet. I'm not sure with Calvin's emplyoment of an opamp, though.
Hi,
I´d recommend to use the dc-blocking cap at the input.
You never know what the device ahead is doing at start-up and shut-down, not to mention what it might do in case of a break-down.
A cap functions well under all conditions.
The servo was designed to reduce its effect on the audible bandwidth in that it doesn´t utilizes a voltage fed back into a (possibly very sensitive) input node, but in that it controls the idle current with a corrective current.
Also the post filtering between the OPAmp and the bipolar adds in reducing audible artefacts that may come from the servo as it makes the filter curve second order.
jauu
Calvin
I´d recommend to use the dc-blocking cap at the input.
You never know what the device ahead is doing at start-up and shut-down, not to mention what it might do in case of a break-down.
A cap functions well under all conditions.
The servo was designed to reduce its effect on the audible bandwidth in that it doesn´t utilizes a voltage fed back into a (possibly very sensitive) input node, but in that it controls the idle current with a corrective current.
Also the post filtering between the OPAmp and the bipolar adds in reducing audible artefacts that may come from the servo as it makes the filter curve second order.
jauu
Calvin
Hi,
Beeing 1 with a noninverting integrator, while the inverting integrator drops down to 0.
Doesn´t count much though, as the noninverting integrators input is pre-filtered by R17/C5 (in #15), thereby generating the 0-output not by feedback action but ´nulling´ the input signal.
jauu
Calvin
The minimum gain makes a difference.
Beeing 1 with a noninverting integrator, while the inverting integrator drops down to 0.
Doesn´t count much though, as the noninverting integrators input is pre-filtered by R17/C5 (in #15), thereby generating the 0-output not by feedback action but ´nulling´ the input signal.
jauu
Calvin
Calvin, just out of curiosity , how would I implement that dual opamp design in your circuit. What changes have to be made? The second opamp can be an OPA 140?
I don't recommend the dual opamp version.
Use a +ve servo where it works alone.
Use a -ve version where it works alone.
Don't use a -ve plus inverter, where a +ve does the job.
A pair of 1uF 63V MKT are <40pence (50cents) and not worth saving 20p.
There is a very good DC servo Thread from about 6 to 10years ago where various topologies were examined (LTspice analysis) and discussed at length.
Use a +ve servo where it works alone.
Use a -ve version where it works alone.
Don't use a -ve plus inverter, where a +ve does the job.
A pair of 1uF 63V MKT are <40pence (50cents) and not worth saving 20p.
There is a very good DC servo Thread from about 6 to 10years ago where various topologies were examined (LTspice analysis) and discussed at length.
- Status
- Not open for further replies.