Hey everyone 🙂
A few days ago I finally finished up my first chip amp - A tube-buffered inverted LM3875 Gainclone based off of the work of Joe Rasmussen. My question doesn't deal with the chip amp section so much as the cathode follower that is buffering it.
Here are some details regarding the circuit in its current state. The tube is a
cryo-treated Mullard from Pearl Cryo and is fed from a bipolar supply with CRCRC filtering. Voltages are +68/-68V. The CF is configured with a cascoded FET current sink at 7mA using a IXCP10M45s as the upper device and a DN2540 as the lower device. Cathode is at 0.6V in reference to ground. After running the amp like this for a few days, I implemented a MOSFET voltage bootstrap based on Shoog's iteration of the Super Linear Cathode Follower. IRF510 is used and results in an anode voltage of 56V.
The sound with the FET bootstrap fitted is much more detailed and transparent. The boost in performance is very audible. The problem is the same as that experienced by Allen Wright, Joe Rasmussen and many others. The sound is more detailed, yes, but it added a harsh, fatiguing edge to the sound. This is why I want to replace the FET with a BJT. Joe Rasmussen has said that he's used both single BJTs and Darlington Pairs with great results.
I have a few different BJTs at my disposal. Both NPN and their PNP compliments - 2SC4793, 2SA1837, BD139, BD140, MJE340, and MJE350. The 2SC/2SA and the BD139/140 have high hfe values whereas the MJE340/350 have comparably low hfe. Right now I I've been searching for days to find an example of a BJT implemented as a voltage strap but haven't seen any, with the exception of Joe's JLTi. I've never designed a circuit using a BJT before and I'm struggling with calculating the values for the bias resistors. I see that Joe's schematic uses a fixed bias whereas a lot of the tutorials for BJT biasing show resistive voltage dividers to set the base bias. My problem is that I don't know where to set the base voltage for proper functioning without causing a huge voltage drop from B+ to the emitter. I know it's gotta be simple but I just can't seem to wrap my head around setting transistors operating points.
It would be awesome if you guys can give me some guidelines for setting the bias voltage and the value for the base resistor along with choosing the most appropriate transistor for the job.
Thanks guys!
PS, I'd also like to thank both Shoog and Joe Rasmussen for their very informative posts and clever circuit designs. I've gained tons of invaluable information from you guys! Thank you both 🙂
A few days ago I finally finished up my first chip amp - A tube-buffered inverted LM3875 Gainclone based off of the work of Joe Rasmussen. My question doesn't deal with the chip amp section so much as the cathode follower that is buffering it.
Here are some details regarding the circuit in its current state. The tube is a
cryo-treated Mullard from Pearl Cryo and is fed from a bipolar supply with CRCRC filtering. Voltages are +68/-68V. The CF is configured with a cascoded FET current sink at 7mA using a IXCP10M45s as the upper device and a DN2540 as the lower device. Cathode is at 0.6V in reference to ground. After running the amp like this for a few days, I implemented a MOSFET voltage bootstrap based on Shoog's iteration of the Super Linear Cathode Follower. IRF510 is used and results in an anode voltage of 56V.
The sound with the FET bootstrap fitted is much more detailed and transparent. The boost in performance is very audible. The problem is the same as that experienced by Allen Wright, Joe Rasmussen and many others. The sound is more detailed, yes, but it added a harsh, fatiguing edge to the sound. This is why I want to replace the FET with a BJT. Joe Rasmussen has said that he's used both single BJTs and Darlington Pairs with great results.
I have a few different BJTs at my disposal. Both NPN and their PNP compliments - 2SC4793, 2SA1837, BD139, BD140, MJE340, and MJE350. The 2SC/2SA and the BD139/140 have high hfe values whereas the MJE340/350 have comparably low hfe. Right now I I've been searching for days to find an example of a BJT implemented as a voltage strap but haven't seen any, with the exception of Joe's JLTi. I've never designed a circuit using a BJT before and I'm struggling with calculating the values for the bias resistors. I see that Joe's schematic uses a fixed bias whereas a lot of the tutorials for BJT biasing show resistive voltage dividers to set the base bias. My problem is that I don't know where to set the base voltage for proper functioning without causing a huge voltage drop from B+ to the emitter. I know it's gotta be simple but I just can't seem to wrap my head around setting transistors operating points.
It would be awesome if you guys can give me some guidelines for setting the bias voltage and the value for the base resistor along with choosing the most appropriate transistor for the job.
Thanks guys!
PS, I'd also like to thank both Shoog and Joe Rasmussen for their very informative posts and clever circuit designs. I've gained tons of invaluable information from you guys! Thank you both 🙂
Sounds like odd order harmonics to me. A BJT might not improve things over a FET.
When I first listened to the amp after adding the MOSFET bootstrap I had the same thought regarding odd-order harmonics. It sounds clean and detailed but is lacking the calmer tone of the basic CCS-loaded CF. Still, I won't go back to the bare CF after hearing it with the bootstrap circuit. I love the detail, just not it's overall tonal quality.
From the testimonies of Allen Wright, he stood in total opposition to FETs anywhere near the signal path and greatly preferred the sound produced by using another triode as the voltage bootstrap. Allen theorized that the non-linear gate capacitance was the culprit. Unfortunatly I simply don't have the voltage overhead to add another valve 🙁
Joe Rasmussen developed the SLCF with the BJT bootstrap and used it in his JLTi Hybrid to great effect. I want to give it a try with a BJT to see how the sound changes. If nothing else, it'll be a great learning experience to see how a BJT is implemented into a circuit in place of a FET or a valve.
From the testimonies of Allen Wright, he stood in total opposition to FETs anywhere near the signal path and greatly preferred the sound produced by using another triode as the voltage bootstrap. Allen theorized that the non-linear gate capacitance was the culprit. Unfortunatly I simply don't have the voltage overhead to add another valve 🙁
Joe Rasmussen developed the SLCF with the BJT bootstrap and used it in his JLTi Hybrid to great effect. I want to give it a try with a BJT to see how the sound changes. If nothing else, it'll be a great learning experience to see how a BJT is implemented into a circuit in place of a FET or a valve.
I ran through some formulas and got some values for a bias network for the MJE340. I'm going to try to grab some resistors from work tomorrow and give it a try. I'm going to have the transistor biased at 10 volts under the supply voltage going to the collector. So collector sees 68VDC and the base 58VDC. I have a few poly film cap values to play with too.
Is the gate of the IRF510 adequately damped. A 1k resistor tied straight to the gate is adequate.
The reason I say this is that I still occasionally use my original SLCF and never found it fatiguing. I can only assume that yours is oscillating. Compared to my current version, with a triode on top, the sound is very similar and the difference in tone I attribute to the different triodes used more than the FET itself.
The purpose of the FET is to push back against the plate voltage swing of the CF in order to maintain a constant voltage across the triode. Since the primary voltage swing is in the bass region rather than the treble the none linear capacitance should have almost no impact on the linearity of the CF since the FET is linear over the bass region. I think you are drawing the wrong conclusions and advise you to spend a bit more time discovering what is really going on.
On this I feel that Allen was been a bit pedantic.
Shoog
The reason I say this is that I still occasionally use my original SLCF and never found it fatiguing. I can only assume that yours is oscillating. Compared to my current version, with a triode on top, the sound is very similar and the difference in tone I attribute to the different triodes used more than the FET itself.
The purpose of the FET is to push back against the plate voltage swing of the CF in order to maintain a constant voltage across the triode. Since the primary voltage swing is in the bass region rather than the treble the none linear capacitance should have almost no impact on the linearity of the CF since the FET is linear over the bass region. I think you are drawing the wrong conclusions and advise you to spend a bit more time discovering what is really going on.
On this I feel that Allen was been a bit pedantic.
Shoog
Last edited:
Hey Shoog, thanks for the reply 🙂
Yep, the gate is damped with a 1k gate-stopper just like in your schematic. Still, the bootstrap circuit was thrown together on a cheap breadboard so I can't be 100% sure everything is stable. The whole amp is temporarily thrown together on a piece of MDF so I'm gonna bring it to work sometime this week and do some probing with the oscilloscope.
Thanks for the information regarding the frequencies that the FET is having the most effect on.
Also I have some values for the voltage divider and base resistor and tomorrow I'm gonna snag some resistors from work and give it a try. Even if there isn't any significant change in the sound I won't be disappointed. This is largely just a learning experience for me and I'm curious to see how a BJT compares to the FET in this circuit. Aside from Joe, I haven't seen anyone else implement a BJT for the bootstrap device.
I'll report back later this week when I get the chance to examine things under the scope. Thanks again!
-Mike
Yep, the gate is damped with a 1k gate-stopper just like in your schematic. Still, the bootstrap circuit was thrown together on a cheap breadboard so I can't be 100% sure everything is stable. The whole amp is temporarily thrown together on a piece of MDF so I'm gonna bring it to work sometime this week and do some probing with the oscilloscope.
Thanks for the information regarding the frequencies that the FET is having the most effect on.
Also I have some values for the voltage divider and base resistor and tomorrow I'm gonna snag some resistors from work and give it a try. Even if there isn't any significant change in the sound I won't be disappointed. This is largely just a learning experience for me and I'm curious to see how a BJT compares to the FET in this circuit. Aside from Joe, I haven't seen anyone else implement a BJT for the bootstrap device.
I'll report back later this week when I get the chance to examine things under the scope. Thanks again!
-Mike
you are struggling with voltage here a little.
For a MOSFET in a source follower role (as in the bootstrap device at the top of your cathode follower triode) you want to minimise the modulation of the internal capacitance by the signal. That signal will appear at gate and source.
Look at Fig 5 of the MOSFET datasheet
http://www.irf.com/product-info/datasheets/data/irf510.pdf
Note how the capacitances change with the Drain to Source Voltage Vds. The signal is on the source. To keep those capacitance low and with lowish variation with signal you want to arrange the voltages such that at max positive signal swing on the source you still have at least 10 and preferrably more like 20 volts from drain to source. In normal amps where there is plenty of voltage avaliable then I always make sure I have at least 25 Volts left across the MOSFET at max positive signal swing at the source. I also would not use an IRF510. Smaller devices have smaller capacitances. I use Zetex ZVN0545A whenever teh 600mW power rating is enough. Else look fo a 1.5w to 2W rated device.
EDIT: When looking at those capacitances don't get hung up on gate capacitance since there is no real signal voltage difference from gate to source (that why it called aa source follower), it is mostly Crss that you are interested in.
Cheers,
Ian
For a MOSFET in a source follower role (as in the bootstrap device at the top of your cathode follower triode) you want to minimise the modulation of the internal capacitance by the signal. That signal will appear at gate and source.
Look at Fig 5 of the MOSFET datasheet
http://www.irf.com/product-info/datasheets/data/irf510.pdf
Note how the capacitances change with the Drain to Source Voltage Vds. The signal is on the source. To keep those capacitance low and with lowish variation with signal you want to arrange the voltages such that at max positive signal swing on the source you still have at least 10 and preferrably more like 20 volts from drain to source. In normal amps where there is plenty of voltage avaliable then I always make sure I have at least 25 Volts left across the MOSFET at max positive signal swing at the source. I also would not use an IRF510. Smaller devices have smaller capacitances. I use Zetex ZVN0545A whenever teh 600mW power rating is enough. Else look fo a 1.5w to 2W rated device.
EDIT: When looking at those capacitances don't get hung up on gate capacitance since there is no real signal voltage difference from gate to source (that why it called aa source follower), it is mostly Crss that you are interested in.
Cheers,
Ian
Last edited:
Thanks for the info gingertube! When I go to work tomorrow I'm gonna grab some more resistors and play with the voltage divider a bit and increase Vds. I still have some voltage headroom to play with but looking back, I wish I had gone with a transformer with slightly higher secondaries. I also am going to try and grab some trimmers so I can easily try small steps without having to switch out resistors.
I do have one more question but this time its in regards to the CCS and the negative rail voltage. Since the cathode self-biases itself just a little above ground potential, does the negative rail voltage make a difference in the Vak of the tube? What I mean is does the operating point of the tube change if the negative rail was reduced to, say, -15V? Will the circuit not self-bias itself with asymmetrical B+/B-? I ask this because the negative rail sits at -68V, the CCS-loaded cathode sits around .6V above ground and the voltage applied to the anode (from the source of the FET) is about 56V. Even with 136V between rails, there is only 67.4V between anode and cathode. Does the -68V contribute to Vak? I know this sounds like a really dumb question but I'm having a little trouble grasping this.😱
I do have one more question but this time its in regards to the CCS and the negative rail voltage. Since the cathode self-biases itself just a little above ground potential, does the negative rail voltage make a difference in the Vak of the tube? What I mean is does the operating point of the tube change if the negative rail was reduced to, say, -15V? Will the circuit not self-bias itself with asymmetrical B+/B-? I ask this because the negative rail sits at -68V, the CCS-loaded cathode sits around .6V above ground and the voltage applied to the anode (from the source of the FET) is about 56V. Even with 136V between rails, there is only 67.4V between anode and cathode. Does the -68V contribute to Vak? I know this sounds like a really dumb question but I'm having a little trouble grasping this.😱
Where the cathode sits will depend upon where you tied the grid. If (as is normal) you tied the grid to 0V with a grid leak resistor then the cathode will sit just above 0 to establish the Vg1k (grid 1 to cathode voltage) required for tube to pass the current set by the CCS.
The negative rail then needs to be just enough for the CCS to operate correctly. This is often called the "Compliance" Voltage of the CCS.
Depending upon which CCS you have used this Compliance voltage could be as low as 2V which would mean a -2V rail would work. Most CCS will need more than that but a -12V rail would be ample.
The Vak is just the Voltage from Anode to Kathode (Deutch Spellling) so the voltage dropped across the CCS does not contribute to it.
If you wanted to get higher Vak you could tie the grid to some negative voltage, say -40V by generating a well filtered -40V supply node and connecting the bottom of the grid leak resistor there. The Cathode would then sit at -40 + 0.6 = -39.4V and Vak would be increased by that 40V. You would have to use an input coupling (DC Block) capacitor but it maybe an acceptable trade off.
Cheers,
Ian
The negative rail then needs to be just enough for the CCS to operate correctly. This is often called the "Compliance" Voltage of the CCS.
Depending upon which CCS you have used this Compliance voltage could be as low as 2V which would mean a -2V rail would work. Most CCS will need more than that but a -12V rail would be ample.
The Vak is just the Voltage from Anode to Kathode (Deutch Spellling) so the voltage dropped across the CCS does not contribute to it.
If you wanted to get higher Vak you could tie the grid to some negative voltage, say -40V by generating a well filtered -40V supply node and connecting the bottom of the grid leak resistor there. The Cathode would then sit at -40 + 0.6 = -39.4V and Vak would be increased by that 40V. You would have to use an input coupling (DC Block) capacitor but it maybe an acceptable trade off.
Cheers,
Ian
Bare in mind that the ECC88 is not an audio valve despite it been the most popular valve for audio. It was designed for VHF frequencies. People build with it and often complain about its hard edgy sound and this is because they have not paid enough attention to radio freq interference and oscillation. Used correctly it is a very detailed and neutral valve.
I would suggest adding a 10R resistor to the plate of the ECC88 as an anode stopper and making sure your grid stopper is at least 100R and right on the valve pin. These tips may just help with the edgy sound.
Have you implemented Joe's low pass filter between the buffer and the gainclone - it helps.
Shoog
I would suggest adding a 10R resistor to the plate of the ECC88 as an anode stopper and making sure your grid stopper is at least 100R and right on the valve pin. These tips may just help with the edgy sound.
Have you implemented Joe's low pass filter between the buffer and the gainclone - it helps.
Shoog
Thanks gingertube; That clears up a lot of the uncertainties I had regarding the negative rail and CCS. It would be a shame to lose the DC coupling to the grid but when I get a chance I may try AC coupling and pulling the grid down some volts.
And yup, I have Joe's low-pass filter implemented after the coupling cap. I swapped the cap values out several times and decided that 1.4nF sounded the best after I got the amp running. I may try 1.5nF again and see if that softens things up a bit. I have a 3.8K grid-stopper resistor like Joe shows in his schematic. The resistor body is right up against the socket pin. I'll try the 10 ohm anode stopper and see where that leads me. Good tips!
I bought a carbon film resistor assortment pack from Radio Shack 🙂shutup🙂 so I have a bunch of cheap resistors to have on hand. We just got some snow here today so I haven't had a chance to tinker but my next move is to increase the voltage across the FET and see how that sounds.
Thanks again for all of the advise!🙂
And yup, I have Joe's low-pass filter implemented after the coupling cap. I swapped the cap values out several times and decided that 1.4nF sounded the best after I got the amp running. I may try 1.5nF again and see if that softens things up a bit. I have a 3.8K grid-stopper resistor like Joe shows in his schematic. The resistor body is right up against the socket pin. I'll try the 10 ohm anode stopper and see where that leads me. Good tips!
I bought a carbon film resistor assortment pack from Radio Shack 🙂shutup🙂 so I have a bunch of cheap resistors to have on hand. We just got some snow here today so I haven't had a chance to tinker but my next move is to increase the voltage across the FET and see how that sounds.
Thanks again for all of the advise!🙂
Hey gentlemen, sorry for reviving this old thread but I wanted to share my experience with you guys. I finally got around to placing another order with Mouser so I can try out the ZTX0545A FET as the bootstrap device. I cobbled together a bootstrap module on a small piece of perfboard with the ZTX0545A FET and a resistive voltage divider with revised values. My B+ is 62V and I'm dropping about 18 volts through the FET with 180K/470K ohm voltage divider. My anode voltage is now around 44V and the CCS is currently set to 7mA. This thing sings! I honestly didn't expect a really significant difference between this setup and the old prototype jigged up on a breadboard with Radio Shack resistors (with ferrous leads) and IRF610 FETs but wow, it made a great improvement! I still want to experiment with the operating point by varying the current through the tube but for now I'm very content.
I was bored last night so I was playing with Eagle and whipped up a little board that has the bootstrap circuit and cascoded CCS onboard, complete with a ground plane, terminal blocks for the heater supply, B+, B- and signal ground. Also included is appropriations for proper decoupling of the B+, B- and heater supplies. I found that decoupling the B+ and B- supplies to ground with good quality polypropylene caps made the most audible difference in my current setup. I'm on my phone right now but I'm going to post up some pics later.
I'm very new to Eagle but I've been experimenting and learning a lot lately. For fun I'm designing several "modules" based on my personal implementation of Joe's JLTi so that I can cram all of the circuitry into small boards for easy installation into a much smaller chassis than what I have currently. Next up is a board for the bipolar power supply with the option of on-board resistors or off-board chokes for filtering. After that is a board for the LM3875, low-pass filter, and optional Zobel network and output inductor. I'm debating on including a DC servo circuit using SMT or just keeping it simple.
Sorry for the long-winded post. I'm just super excited that it's running and sounding great! I thank all of you that participated in this thread and helped steer me in the right direction.
I was bored last night so I was playing with Eagle and whipped up a little board that has the bootstrap circuit and cascoded CCS onboard, complete with a ground plane, terminal blocks for the heater supply, B+, B- and signal ground. Also included is appropriations for proper decoupling of the B+, B- and heater supplies. I found that decoupling the B+ and B- supplies to ground with good quality polypropylene caps made the most audible difference in my current setup. I'm on my phone right now but I'm going to post up some pics later.
I'm very new to Eagle but I've been experimenting and learning a lot lately. For fun I'm designing several "modules" based on my personal implementation of Joe's JLTi so that I can cram all of the circuitry into small boards for easy installation into a much smaller chassis than what I have currently. Next up is a board for the bipolar power supply with the option of on-board resistors or off-board chokes for filtering. After that is a board for the LM3875, low-pass filter, and optional Zobel network and output inductor. I'm debating on including a DC servo circuit using SMT or just keeping it simple.
Sorry for the long-winded post. I'm just super excited that it's running and sounding great! I thank all of you that participated in this thread and helped steer me in the right direction.
- Status
- Not open for further replies.