For those that are not afraid of a little (OK, more than a little) sand in their diyAudio diet, my ADHD powered brain wandered far off its normal path.
Some readers of this thread have already noticed the diversion. Another diyAudio forum member made a mosfet based "preamp" using UNSET technology. He even laid out a board for the design and did a group buy. I answered several questions about the circuit during the design cycle and about shipping boards all over the world for the GB. For that I got a couple boards. One of which I built and applied the Tubelab "turn it up till it blows and back up a little" treatment to. I also cranked up the gain and used it to drive some vacuum tubes in an UNSET output board.
My built board is seen in post #761 and it went live in post # 765 here:
https://www.diyaudio.com/community/threads/schade-common-gate-scg-preamp.380487/
The board was used to drive two vacuum tubes in UNSET configuration in post #767 and to drive two parallel pairs to 35 WPC in post #779.
FETSET was born in post #787 FETSET V1.0 came to life in post #872 where it made a clean, nice sounding 30 WPC at the edge of thermonuclear meltdown.
If one output fet is good, two must be better, post #903
In post #950 I had built a two stage FETSET on one small perf board. It was intended to be the ultimate driver board and did make over 400 volts peak to peak on 450 volts of B+.
It should be capable of operation on 600 volts or more for a big cathode or source follower.
All of this got put on hold a few weeks back due to other priorities. That thread is centered around solid state devices, but the underlying technology is very similar. I learned a lot during this diversion, and the thread may be worth reading to anyone interested in the UNSET technology.
Some readers of this thread have already noticed the diversion. Another diyAudio forum member made a mosfet based "preamp" using UNSET technology. He even laid out a board for the design and did a group buy. I answered several questions about the circuit during the design cycle and about shipping boards all over the world for the GB. For that I got a couple boards. One of which I built and applied the Tubelab "turn it up till it blows and back up a little" treatment to. I also cranked up the gain and used it to drive some vacuum tubes in an UNSET output board.
My built board is seen in post #761 and it went live in post # 765 here:
https://www.diyaudio.com/community/threads/schade-common-gate-scg-preamp.380487/
The board was used to drive two vacuum tubes in UNSET configuration in post #767 and to drive two parallel pairs to 35 WPC in post #779.
FETSET was born in post #787 FETSET V1.0 came to life in post #872 where it made a clean, nice sounding 30 WPC at the edge of thermonuclear meltdown.
If one output fet is good, two must be better, post #903
In post #950 I had built a two stage FETSET on one small perf board. It was intended to be the ultimate driver board and did make over 400 volts peak to peak on 450 volts of B+.
It should be capable of operation on 600 volts or more for a big cathode or source follower.
All of this got put on hold a few weeks back due to other priorities. That thread is centered around solid state devices, but the underlying technology is very similar. I learned a lot during this diversion, and the thread may be worth reading to anyone interested in the UNSET technology.
Thanks @Francois G ,
you’ll be ready earlier than me. I still need to give my version of the BHEL34 a case with the new feedback.
Thanks to @Tubelab_com @spiggs and everyone has tried: in sims the driver can have very low local feedbacks, and this works with VP0106 grounded on both drain and gate, with the source raised a few volts above ground thanks to the driver’s cathode current and Rdson.
So the input is dc coupled to VP0106 gate and there’s only one coupling cap between driver and output stage.
Is it something feasable in reality as well, or just in LTSpice wonderland?
Is there a pmosfet for the output tube capable of 400 V? GU-50 works better with around 20% local feedback, and it needs around 200 Vpp on its cathode.
Thanks,
Roberto
you’ll be ready earlier than me. I still need to give my version of the BHEL34 a case with the new feedback.
Thanks to @Tubelab_com @spiggs and everyone has tried: in sims the driver can have very low local feedbacks, and this works with VP0106 grounded on both drain and gate, with the source raised a few volts above ground thanks to the driver’s cathode current and Rdson.
So the input is dc coupled to VP0106 gate and there’s only one coupling cap between driver and output stage.
Is it something feasable in reality as well, or just in LTSpice wonderland?
Is there a pmosfet for the output tube capable of 400 V? GU-50 works better with around 20% local feedback, and it needs around 200 Vpp on its cathode.
Thanks,
Roberto
I was looking to try one of Rod Coleman's Shunt Cascode CCS drivers but he is out of stock of the current version and is planning to offer a new version since some parts became obsolete. I have not got the courage up yet to try and recreate one of the various schematics floating around on a breadboard.
I have been following your posts on the SCG preamp thread however seems the first run of PCBs sold out quickly. Inspired though I did bypass the driver stage on my 6DQ5 UNSET pcb and hooked up the pre with the most swing in my collection, a AKSA Lender fed by a 50v power supply, to see if it would do anything. I got about 1W with poor THD but it did play music and at low volume did not sound bad. I plan to try some other driver stages as I can and have been on the lookout for designs within my ability to implement. Currently have a PCB from user kodabmx working it's way through the postal system.
However besides normal life I am a bit distracted by speakers. I had recently completed a new set but have been disappointed with their performance in my living room so am on the hunt for what to build to fill the gap.
I asked Rod about the new version a few weeks ago as well, but apparently they are still not available.
But what I meant is if you have tried the driver in UNSET mode with a simple or cascoded CCS as load (one single mosfet or two).
Thanks
Roberto
The first run was a group buy that took reservations in advance, so in theory all were spoken for before they were ordered. After people started building them, there were requests for more so he got another batch made. Posts number 978 and 979 seem to indicate that there might be a few more available from the second batch.
If you wind up with a board post here or PM me and I see what mods I did on my board to make it capable of driving the UNSET output stage into the high power regions. All of the audio stuff is in storage for the moment.
I got this board out of its envelope one day and had music playing later the same day. My mission was to spend no money therefore buy no parts. Most of my "mods" were based on using what I had and mad no real difference in performance. There are two details that need to be addressed if you want to drive some decent sized tubes to full power output. You need to get more gain and output voltage capability.
I now know that the circuit is capable of operating at 450 volts without failure, but I didn't know that when I built this board, so I boosted the B+ voltage a bit by raising the zener voltages to get my B+ to about 120 volts. I run the board on the same power supply that I use for the vacuum tube's screen voltage which is usually 150 volts, but I have been from 130 volts to 200 volts in search of the optimum screen voltage for the tube in the output stage. The board will regulate it down to about 120 volts. All of my zeners have house numbers on them, but I think there are two 51 volt diodes and one 25 volt diode in the series string. The regulator mosfet in the power supply (Q11 and Q21) is an unknown brand marked 11N80C3. Just about any N channel TO-220 mosfet capable of eating 200 or more volts will work. An enhancement fet is preferred but a depletion fet will work with a slightly different output voltage. As long as your B+ is between 120 and 150 volts you will be fine. Build and test the power supplies first before populating the rest of the board. That will avoid blowing lots of parts if the supplies are not right. I used what I had for the big caps, so I have 100 uF 250 volt parts for C11, C21, C14, and C24. I have 1000 uF 200 volt parts for C12 and C22. C13 and C23 are 33 uF 200 volts. Use a 47 uF if you are buying parts. None of these parts are really critical, use what you have if you have something close, but make sure the voltage rating is well above what you are feeding into the board. The board will run fine from a Triad N-68X isolation transformer which will produce a raw B+ (before the regulator) of about 160 to 170 volts. Use 200 volt caps or higher.
Build the CCS next and set the current with the clip lead to ground as outlined in the build docs. The CCS fet must be a depletion mode mosfet. The usual IXYS IXCP10M45S will work as will a DN2540, DN2535, IXCP10M90, IXTP08N100D2, and a few others.
Just about every N channel part I tried for the gain fet works fine, but the designer rolled through lots of different fets and stated that the ST30N10F7 sounded the best. That's what's in my board.
The input must be a P channel mosfet. Most everything should work. I had a bunch of Fairchild FQP3P20's so I used them. The FQP9P25 that's used in the UNSET board works too.
Once the board is built, tested and working you need to boost the gain a bunch to drive tubes. I simply tacked a 750 ohm resistor in parallel with the 1K ohm gain resistor, Rg. This boosts the gain high enough so that I could drive a 26HU5 to clipping (25 watts) with a CD player. Board gain is just over 20.
I'm sure that there is some details that I'm missing, so ask and I'll look at the board to see.
You may have noticed the breadboard creation I called the FETSET in post #787 of the SCG thread. It is an UNSET output stage made with mosfets and a 600 ohm Toroidy SE OPT designed for a 6C33 tube. I think it may be the best sounding thing that I have built in some time. Same circuit, bigger parts running from 300 volts and over 300 mA.
I now know that the circuit is capable of operating at 450 volts without failure, but I didn't know that when I built this board, so I boosted the B+ voltage a bit by raising the zener voltages to get my B+ to about 120 volts. I run the board on the same power supply that I use for the vacuum tube's screen voltage which is usually 150 volts, but I have been from 130 volts to 200 volts in search of the optimum screen voltage for the tube in the output stage. The board will regulate it down to about 120 volts. All of my zeners have house numbers on them, but I think there are two 51 volt diodes and one 25 volt diode in the series string. The regulator mosfet in the power supply (Q11 and Q21) is an unknown brand marked 11N80C3. Just about any N channel TO-220 mosfet capable of eating 200 or more volts will work. An enhancement fet is preferred but a depletion fet will work with a slightly different output voltage. As long as your B+ is between 120 and 150 volts you will be fine. Build and test the power supplies first before populating the rest of the board. That will avoid blowing lots of parts if the supplies are not right. I used what I had for the big caps, so I have 100 uF 250 volt parts for C11, C21, C14, and C24. I have 1000 uF 200 volt parts for C12 and C22. C13 and C23 are 33 uF 200 volts. Use a 47 uF if you are buying parts. None of these parts are really critical, use what you have if you have something close, but make sure the voltage rating is well above what you are feeding into the board. The board will run fine from a Triad N-68X isolation transformer which will produce a raw B+ (before the regulator) of about 160 to 170 volts. Use 200 volt caps or higher.
Build the CCS next and set the current with the clip lead to ground as outlined in the build docs. The CCS fet must be a depletion mode mosfet. The usual IXYS IXCP10M45S will work as will a DN2540, DN2535, IXCP10M90, IXTP08N100D2, and a few others.
Just about every N channel part I tried for the gain fet works fine, but the designer rolled through lots of different fets and stated that the ST30N10F7 sounded the best. That's what's in my board.
The input must be a P channel mosfet. Most everything should work. I had a bunch of Fairchild FQP3P20's so I used them. The FQP9P25 that's used in the UNSET board works too.
Once the board is built, tested and working you need to boost the gain a bunch to drive tubes. I simply tacked a 750 ohm resistor in parallel with the 1K ohm gain resistor, Rg. This boosts the gain high enough so that I could drive a 26HU5 to clipping (25 watts) with a CD player. Board gain is just over 20.
I'm sure that there is some details that I'm missing, so ask and I'll look at the board to see.
You may have noticed the breadboard creation I called the FETSET in post #787 of the SCG thread. It is an UNSET output stage made with mosfets and a 600 ohm Toroidy SE OPT designed for a 6C33 tube. I think it may be the best sounding thing that I have built in some time. Same circuit, bigger parts running from 300 volts and over 300 mA.
George, is there any chance to have specs of that amp? My 3k Toroidy SE transformers are under construction and I could add a pair of those: https://sklep.toroidy.pl/en_US/p/TTG-6C33CSE-Tube-output-transformer-600-Ohm-6C33C/610 with the same shipping costs.
Thanks George, so the drive stage for the 600 Ohm transformer could be:
on top: https://www.mouser.it/datasheet/2/827/DS_UJ3N065080K3S-1530401.pdf
on bottom: https://www.mouser.it/datasheet/2/308/1/FQP9P25_D-1809667.pdf
Could the PSU be this one? https://www.ebay.it/itm/155281091654
One per channel, to have some margin.
on top: https://www.mouser.it/datasheet/2/827/DS_UJ3N065080K3S-1530401.pdf
on bottom: https://www.mouser.it/datasheet/2/308/1/FQP9P25_D-1809667.pdf
Could the PSU be this one? https://www.ebay.it/itm/155281091654
One per channel, to have some margin.
Not sure what specs you need.
Those are the same OPT's that I have, the TTG6C33CSE. The schematic of the amp is in post #789 of the SCG thread. You can squeeze over 35 watts from this amp if you can figure out how to adequately get the heat out of the mosfets. It made 35 watts for several minutes until my big heatsink hit 60C with the mosfets over 90C and I cut the power (post #872). I learned the hard way that you need to add a 15 volt zener from gate to source on both fets to avoid blowing mosfets on transients.
I modified the board to use two parallel output devices, but they don't share current well with each other and this led to a meltdown trying to hit 50 WPC. No further work had been done on this board, and all is currently in a box waiting for some time on my bench. I cleared the bench to fix my dead Fluke 407D power supply and I have a few other tasks to finish before any audio work resumes. I will resume with the big source follower fed by two SCG stages as this is needed to continue the unfinished high efficiency SE work that I started back in 2007. That's the last thing I want to conquer before Tubelab Inc goes away.
The P fet is FQA9P25 (FQP9P25 will work too) and the N-fet is IXTH6N100D2. My brief experiment with the SiC J-fet did not work, but I didn't try to figure out why it didn't.
I have no idea if that power supply will work. Somewhere in diyAudio I remember reading about someone testing something similar and finding that it did not meet specs. My test amp consumed 224 watts to make two 25 watt channels. A higher powered amp will need more input and a bigger heat sink. I have two possible power supply plans for the final amp build. One involves several surplus 48 volt 3 amp SMPS supplies wired in series, and the other involves a huge surplus toroid I just bought.
Those are the same OPT's that I have, the TTG6C33CSE. The schematic of the amp is in post #789 of the SCG thread. You can squeeze over 35 watts from this amp if you can figure out how to adequately get the heat out of the mosfets. It made 35 watts for several minutes until my big heatsink hit 60C with the mosfets over 90C and I cut the power (post #872). I learned the hard way that you need to add a 15 volt zener from gate to source on both fets to avoid blowing mosfets on transients.
I modified the board to use two parallel output devices, but they don't share current well with each other and this led to a meltdown trying to hit 50 WPC. No further work had been done on this board, and all is currently in a box waiting for some time on my bench. I cleared the bench to fix my dead Fluke 407D power supply and I have a few other tasks to finish before any audio work resumes. I will resume with the big source follower fed by two SCG stages as this is needed to continue the unfinished high efficiency SE work that I started back in 2007. That's the last thing I want to conquer before Tubelab Inc goes away.
The P fet is FQA9P25 (FQP9P25 will work too) and the N-fet is IXTH6N100D2. My brief experiment with the SiC J-fet did not work, but I didn't try to figure out why it didn't.
I have no idea if that power supply will work. Somewhere in diyAudio I remember reading about someone testing something similar and finding that it did not meet specs. My test amp consumed 224 watts to make two 25 watt channels. A higher powered amp will need more input and a bigger heat sink. I have two possible power supply plans for the final amp build. One involves several surplus 48 volt 3 amp SMPS supplies wired in series, and the other involves a huge surplus toroid I just bought.
The best you can do with a class A SE amp is 50% according to theory. My sim shows 31.4% at 33 watts out. Drop the power output to 25 watts with the same power input and you get 23.8% so reality is not too far from the sim world. Remember that the Toroidy 600 ohm OPT has a DCR of 50 ohms. That will dissipate over 6 watts of your DC as heat and cost you 50/600 or about 8% of your audio output power.
Thanks George,
considering the current capabilities of that transformer, could be 250V an easier voltage for this circuit?
Idle current drops, wattage to be dissipated drops as well, and we still can get around 35 Wrms.
Can you comment the attached circuit? Due to the fact that the driver has high gain and low distortion, I've pushed the local feedback on the output stage to 20%. This should bring towards 35 Wrms before clipping.
Thanks!
Roberto
considering the current capabilities of that transformer, could be 250V an easier voltage for this circuit?
Idle current drops, wattage to be dissipated drops as well, and we still can get around 35 Wrms.
Can you comment the attached circuit? Due to the fact that the driver has high gain and low distortion, I've pushed the local feedback on the output stage to 20%. This should bring towards 35 Wrms before clipping.
Thanks!
Roberto
Yes, the sim shows 34.something watts at the onset of clipping but it takes 1.05 volts peak (2.1 V peak to peak) to get there. My DVD player would not be able to drive this amp to clipping without another gain stage. There is a segment of users out there that will attempt to drive a nice HiFi amp with an iPhone or other similar device. My testing reveals that I need a LY Spice sim that can hit clipping with 0.6 to 0.7 volts for my iPhone to be able to make it clip on at least half of the songs stored in it in my testing. My cheapie Motorola Android phone seems to have a bit more output, but not much more.
The last amp I had running on the bench used a two stage SCG design to generate lots of voltage gain. It is follower by a large source follower for current gain. It was running on the bench successfully when a slip of the screwdriver turned some of the fets into charcoal. All has been torn down, and today's news may spell the end of my testing for a long time.
The retirement house that we had custom built 9 years ago has extremely high levels of Radon gas especially in the basement where my lab is. Radon gas is a major cause of lung cancer, the number one cause among non smokers. Mitigation will likely require the dismantling of the entire lab to get to the concrete floor which has developed some serious cracks. It took several years to build this lab, and much of the wood based benches are glued and cannot be easily dismantled. The game might be over forever.
The last amp I had running on the bench used a two stage SCG design to generate lots of voltage gain. It is follower by a large source follower for current gain. It was running on the bench successfully when a slip of the screwdriver turned some of the fets into charcoal. All has been torn down, and today's news may spell the end of my testing for a long time.
The retirement house that we had custom built 9 years ago has extremely high levels of Radon gas especially in the basement where my lab is. Radon gas is a major cause of lung cancer, the number one cause among non smokers. Mitigation will likely require the dismantling of the entire lab to get to the concrete floor which has developed some serious cracks. It took several years to build this lab, and much of the wood based benches are glued and cannot be easily dismantled. The game might be over forever.
With a small mod to the driver stage I targeted your suggestion of clipping with 600 mVp at input.
Ciss of the IRF9640 is 1.2 nF ( https://www.vishay.com/docs/91086/irf9640.pdf ), so to swing 93Vpp at 30kHz it is needed to have:
2 x pi x F x Vpp = 17 V/uS
I = Ciss x slewrate = 21 mA
Is it correct? On the driver I have 1.5 mA, do I need more current?
Ciss of the IRF9640 is 1.2 nF ( https://www.vishay.com/docs/91086/irf9640.pdf ), so to swing 93Vpp at 30kHz it is needed to have:
2 x pi x F x Vpp = 17 V/uS
I = Ciss x slewrate = 21 mA
Is it correct? On the driver I have 1.5 mA, do I need more current?
I do not actually use an IRF9640 in my builds. I use an FQA9P25 in the output stage because they fit my quick change sockets. They have the same die as the TO-220 FQP9P25 part. I have not found a model for the 9P25, so the sim uses a 9640.
The p-fet operates as a follower so the gate to source capacitance, Ciss is effectively bootstrapped since they operate at nearly the same AC potential. The important capacitance in a follower circuit is the gate to drain capacitance, Crss since this is from the input (gate) to AC ground (drain). The Crss of the 9P25 is 27 pF. Note that all of the p-fets in this and other similar series have been discontinued and are no longer being made.
My test board has a 330 ohm resistor in the CCS of the first stage with an IXYS IXCP10M45S CCS "chip." That generates about 8 ma of current. It's basically copied directly from my modified UNSET board with an LND150 in place of the tube. Measured frequency response was limited only by the OPT and went out beyond 30 KHz. Didn't record the actual number, just turned the frequency knob on the generator. It was the LF response that surprised me the most, as I could get a clean 20 watts at 20 Hz.