Mo' Power - 100W / 8 ohms From Tokin 2SK182ES SIT? Can't Resist the Siren's Song

Member
Joined 2010
Paid Member
I don't need a hundred watt per channel amplifier but I want one. The siren's song is very alluring and I can't resist, so my next project is a Tokin 2SK182ES SIT amp which I hope wiil be able to output 100 watts into 8 ohms.

Not only do I not need a hundred watt amplifier, I really don't need another amplifer, so I am planning to modify my BAF 2015 THF-51S monoblocks. I like these amps and I've got them currently in my bi-amped system. But I overbuilt the cooling system in them and they barely get warm after extended operation, which makes them the perfect platform for the high power dissipation requirements of the new design.

The other plus is that with some juggling around, I can use a lot of the existing pricey parts. I really don't want to spend a lot of money up front on a design that is unproven.

I knew I needed more voltage than the 62V power supply of my BAF2015 amps. Each monoblock is powered an Antek AN-6450 toroidal transformer. That is a 600VA transformer with 2 x 50V secondaries. Currently, they are wired in parallel. If I wired them in series, PSUD Designer predicts about 135 volts. That's the value that I used for the basis of my new design.

I spent many hours with LTspice trying various circuit topologies. I tried mu follower, choke loading, ccs loading, common drain, and common source. The most promising was the mu follower. I really wanted the common source mu follower to work, but I could not get as much power out as with common drain (follower). The common source version would be easier to implement as it has voltage gain.

So although LTspice said that a mu follower follower could produce 100 watts into 8 ohms, I wasn't sure if that was worth anything if I couldn't come up with a voltage amplifier to drive the output stage to full power. A hundred watts at 8 ohms is about 40 volts peak or 80 volts peak-to-peak. And, the 80 volts peak-to-peak should be fairly low distortion.

Luckily, I remembered that I had been exploring the use of the United SiC power JFETs in preamplifier and amplifier duties and that they performed quite well in those tasks (https://www.diyaudio.com/community/...loring-the-uj3n065080k3s.409636/#post-7624728). In fact, I still had a working prototype of a United SiC JFET preamp sitting on my workbench. So naturally, I decided to explore the upper limits of this preamp.

My previous experiments showed that I could get a clean 50 volt peak-to-peak signal out of the preamp when driven by my test oscillator at maximum output. I have an old Acurus preamp that I know to have low distortion so I set up a test with my test oscillator driving the Acurus preamp which fed into the United SiC preamp.

On checking the output with my oscilloscope, I was very pleased to see that the 1 kHz sine was amplified to about 190 volts peak-to-peak before noticeable distortion was visible. The Acurus preamp supplied about 26 volts peak-to-peak to the United SiC preamp. That encouraged me enough to get out my distortion test rig and take some measurements.

To make things simple, I set the preamp chain to output about 100 volts peak-to-peak for the FFT test. I figured that would be enough voltage swing to drive the follower output stage to 100 watts output with some safety margin. The results show a nice distortion profile with 0.029% Total Harmonic Distortion. This is significantly lower than the expected greater than 1% THD from the amp output stage at full power. So, this was good news.

With all this preliminary work done, I think that it will be possible to reach the goal of 100 watt into 8 ohms with this Tokin SIT. However, the SIT and CCS MOSFET will be operating near their limits, so there are unknowns and rough waters ahead.

But the sirens sing so sweetly.
 

Attachments

  • 100W_2SK182ES_Amp.jpg
    100W_2SK182ES_Amp.jpg
    239.2 KB · Views: 391
  • VA_188VppOut.png
    VA_188VppOut.png
    47.4 KB · Views: 369
  • VA_103VppOut.png
    VA_103VppOut.png
    40.1 KB · Views: 165
  • Voltage Amp Out.jpg
    Voltage Amp Out.jpg
    85.2 KB · Views: 372
On checking the output with my oscilloscope, I was very pleased to see that the 1 kHz sine was amplified to about 190 volts peak-to-peak before noticeable distortion was visible.
DANG! I mean c'mon why not shoot for the moon and hit 400W into 8R?

Seriously though - I enjoy reading the thought process behind projects like this. You have no idea how it spurs creativity and curiosity seeing how others approach choosing appropriate 'building blocks' and the design process.

Can't wait to see the results and progress along the way.

:cheers: :worship:
 
  • Like
Reactions: birdbox
YES, you're talking my language now! The 'easiest' (not cheapest) way to do it, would be common source choke loaded. I have similar intentions of sending it to the moon. I used the BAF2015 circuit and just replaced the current source with the choke, simple enough. Keep it alive with water cooling maybe...?
 

Attachments

  • 222.png
    222.png
    4.8 MB · Views: 2,198
  • 111.png
    111.png
    4.9 MB · Views: 823
That's going to be an expensive amplifier power supply!

The arms race continues. 🤓
Well, Ben, I'm going to re-use my power supply filter capacitors. I have 4 x 22mF 80V capacitors per channel. My plan is to use pairs in series to double the voltage rating to 160V. This will yield each unit 11mF of capacitance, which in a CLC filter with a Hammond 159ZJ (10mH) (which I have in the BAF 2015) willl result in about 12mV of ripple according to PSUD. Adding an additional RC filter stage would reduce this without too much difficulty.

So, I will need to buy some higher voltage rated capacitors but I don't think it will be too bad. At least, for development and testing, I can live with a bit of hum. There will be some capacitors that will need to be upgraded, though.
 
YES, you're talking my language now! The 'easiest' (not cheapest) way to do it, would be common source choke loaded. I have similar intentions of sending it to the moon. I used the BAF2015 circuit and just replaced the current source with the choke, simple enough. Keep it alive with water cooling maybe...?

This was one of the topologies that I modeled with LTspice. Unfortunately, I could only get power output in the area of 90 - 100 watts with much higher current which would have exceeded the capabilities of the Tokin SIT. The 2SK182ES is rated at a maximum power dissipation of 500 watts. I think this device was designed for switching applications and I don't know what the DC Safe Operating Area is. As it is, I hope it can handle 200W with fan cooling.
 
Wait, what? Did you write 135 volts?

Grrr! Well this means Tokin SIT amplifier war! Now i am going to run my two 1000VA 85V DC power supplies in series!

That now means a 2000VA 170V PSU to my 2SK182ES and THF-51S SITs - in my SB an LSB amplifier builds… 🙂👍




…. 170V!



…… No, i am just joking 🙂🙂🙂

I like your amplifier plans! That project looks exiting - please continue sharing your build progress 🙂👍
 
Well, Ben, I'm going to re-use my power supply filter capacitors. I have 4 x 22mF 80V capacitors per channel. My plan is to use pairs in series to double the voltage rating to 160V. This will yield each unit 11mF of capacitance, which in a CLC filter with a Hammond 159ZJ (10mH) (which I have in the BAF 2015) willl result in about 12mV of ripple according to PSUD. Adding an additional RC filter stage would reduce this without too much difficulty.

So, I will need to buy some higher voltage rated capacitors but I don't think it will be too bad. At least, for development and testing, I can live with a bit of hum. There will be some capacitors that will need to be upgraded, though.
I knew I needed more voltage than the 62V power supply of my BAF2015 amps. Each monoblock is powered an Antek AN-6450 toroidal transformer. That is a 600VA transformer with 2 x 50V secondaries. Currently, they are wired in parallel. If I wired them in series, PSUD Designer predicts about 135 volts. That's the value that I used for the basis of my new design.

With the 50VAC secondaries in series, perhaps you will only get 125VDC, but that will still be more than enough voltage to get 100W output. There will be enough voltage for you to spare some for a capacitance multiplier to save on large and expensive 150VDC electrolytic capacitors.
 
Good idea about using a capacitance multiplier. I hadn't thought of that.

My current favourite solution, though, is to add another LC filter stage. I'm thinking Hammond 157E (6mH, 133 mohms) and 3300uF of capacitance. According to PSUDII, adding this will reduce the ripple to less than 1.5 mV.

Digikey has 3300 uF / 160V capacitors for about $8 Canadian and the Hammond inductor is around $35. I'm not sure about the price for the Hammond, though, since it looks like the American price, but the digikey.ca website shows that price. May actually be around $55 Canadian. Either way, it appeals to me due to its simplicity.

The power supply will feed the voltage amplifier and bias for the SIT as well, so it's like three power supplies in one.
 
I've decided to use a positive ground for the output stage so the drain of the SIT is grounded. This requires a negative voltage power supply. I intend to use the same power supply for the voltage amplifier stage, since the voltage for it and the output stage are pretty well the same. So, I redesigned the United SiC cascode voltage amp for positive ground.

I fired up LTspice and after a bit of playing around, I got it figured out. Of course, I had to verify the design in real-life, so I disassembled my working negative ground prototype and reconfigured it according to my LTspice design. I slowly fired it up using my lab power supply with current limiting and all went well. Connecting my 1 kHz oscillator to the input confirmed that it was functioning. Voltage measurements more or less agreed with LTspice.

For the subsequent tests, I set the power supply voltage to 120 volts. This will be the realistic achievable voltage that my existing power transformer will supply. I conducted tests to check the maximum output voltage swing and measured distortion at 10 volts peak-to-peak and at 100 volts peak-to-peak with a 1 kHz signal.

Observing the output waveform on my oscilloscope, I saw beginnings of second order harmonic distortion at about 180 volts peak-to-peak. This is a bit lower than my initial test using the negative ground prototype, but I think the power supply voltage was higher in that case, somewhere around 130 volts.

The FFT result for 100 Vpp is better than the first go-around, 0.0058% versus 0.029%, but there are enough variables in the test setup using REW software, soundcard, voltage attenuator, and various cables that all I can say is that the new design appears to have lower distortion.

This time around, I also measured the THD for an output of 10 Vpp, and that came out to 0.0027%.

Overall, I'm happy with the results. Especially since my test prototype is a pretty dodgy affair with poor layout which would contribute to noise.

Another thing that I'm very happy about is the availability of tools such as LTspice, KiCad, and Room Equalization Wizard which make this hobby accessible to the ordinary person. I know without LTspice, my effort to convert the voltage amp from negative ground to positive ground would have been a huge exercise for me. I likely would have let out some magic smoke along the way, too. Anyways, that's my contemplative thought for New Year's Day.

Happy New Year to all.
 

Attachments

  • 10Vpp-PosGroundPreamp.jpg
    10Vpp-PosGroundPreamp.jpg
    84.5 KB · Views: 167
  • 104Vpp-PosGroundPreamp.jpg
    104Vpp-PosGroundPreamp.jpg
    83.1 KB · Views: 130
  • UnitedSiC_Preamp_PosGnd.JPG
    UnitedSiC_Preamp_PosGnd.JPG
    255.4 KB · Views: 181
  • UnitedSiC_PreampNegPS.jpg
    UnitedSiC_PreampNegPS.jpg
    127.6 KB · Views: 185
I changed the LTspice output stage to 120 volt positive ground power supply and reran the simulation. After adjusting some values a bit, I managed to get about 90 volts peak-to-peak out into 8 ohms at 1.38% THD. This works out to about 128 watts. Other simulated power outputs: 100W at 0.32% THD and 1W at 0.0174% THD.

Given these results, I think that the goal of 100 watts out into 8 ohms is a real possibility. Time to buy some parts and test the output stage design.
 

Attachments

  • 100W_2SK182ES_AmpNegPS.jpg
    100W_2SK182ES_AmpNegPS.jpg
    140.4 KB · Views: 81
  • LTspice_90VppSimulatedOut.jpg
    LTspice_90VppSimulatedOut.jpg
    190.4 KB · Views: 83
I disassembled one of my BAF 2015 THF-51S monoblocks and started the conversion to the new 100 watt design.

First, the two 50 VAC power transformer secondaries were rewired in series instead of parallel connections to yield 100 VAC out.

Next up, the CLC filter, which uses 4 x 22mF 80V capacitors, was reconfigured with pairs of capacitors wired in series to get 160V DC rating. 16k ohm balancing resistors were added across each capacitor. The end result is a new CLC filter of 11 mF - 10 mH - 11 mF.

After cautiously powering up the new power supply and verifying that it was working correctly, I connected a 40 ohm load to it for a real test. Under this load, the voltage settled in at 120.5V. The current through the load works out to 120.5V / 40 ohms = 3.0 amps. The new amp is designed for about 2.5 amps, so with another filter section to come, the end result should be close to 120 volts.

I was tending towards using a capacitance multiplier for the final filter section, but I managed to find a good deal on some Hammond inductors so I will use another LC filter. This should bring the final ripple to below 1 mV.

I'm trying to reuse as much as possible from the BAF 2015 amp, and it turns out that I won't have to make a new board for the CCS. Because I'm going to a negative power supply, I will be changing from an N channel MOSFET to a P channel MOSFET. However, after studying the two schematics, the only difference is in the wiring of the opto-isolator. I had been expecting to have to totally redo this circuit, so I'm very happy about this.

I've got most, if not all of the critical parts now, so I can proceed to getting the rest of the amp built. To be on the safe side, I've ordered two more 2SK182ES SITs from pras1170. I've still got two pieces but you never know. This is uncharted territory for me, so it's better to be prepared. I believe in the 6 Ps: Proper Preparation Prevents Pss Poor Performance.

Attached is a full schematic including the UnitedSiC voltage amplifier.
 

Attachments

  • 120V_PS_PosGnd.JPG
    120V_PS_PosGnd.JPG
    390.5 KB · Views: 168
  • 100W_2SK182ES_Amp_120V_NegPS.jpg
    100W_2SK182ES_Amp_120V_NegPS.jpg
    145.9 KB · Views: 170
With the ratio you have there for the source resistors you will have more of mosfet sound signature. In the experiments I done recently I noticed I like it more with a bigger source resistor for the mos. Iirc it gave also better measured performance (but here my short memory doesn’t help much) though it doesn’t take much to investigate in the sim.
 
  • Like
Reactions: Ben Mah
Wow! This is cool. Is the fan running full blast?
Yes, the 4 inch diameter 120 VAC fan was running full blast, and the resistors were getting very hot. They were almost too hot to touch. The resistors are rated at 100 watts each heatsinked, but my heatsinking was not up to par. That's around 360 watts dissipated so it's a lot of heat. I only ran the power supply full bore for about 5 minutes. Just long enough for it to stabilize and take measurements.

The amp will have better heatsinking. My BAF 2015 amp barely warmed up the fan-cooled heatsink when dissipating about 150 watts, so I don't think 300 watts will be a problem.
 

Attachments

  • Heatsink_fans.JPG
    Heatsink_fans.JPG
    250.7 KB · Views: 70
With the ratio you have there for the source resistors you will have more of mosfet sound signature. In the experiments I done recently I noticed I like it more with a bigger source resistor for the mos. Iirc it gave also better measured performance (but here my short memory doesn’t help much) though it doesn’t take much to investigate in the sim.
It will be easy to change resistor configuration, so I'll try different combinations when the amp is built. I have investigated in the sim, but the true test will be listening. Sim is one thing, but the proof of the pudding is in the tasting.
 
  • Like
Reactions: schultzsch