Here is my latest Tokin SIT amplifier. It is a DEFISIT amp (Depletion Enhancement Follower SIT) output stage based on Nelson's First Watt SIT-5 output stage. Diyers more knowledgeable than me picked apart what information Nelson provided in his SIT-5 owners' manual (https://www.diyaudio.com/community/threads/first-watt-sit5.418023/), and I rode on their coattails and came up with this version.
My build is only the output stage, and it incorporates the duo output capacitor arrangement which can be tweaked to adjust the amount of current amplification that the mosfet part of the DEFISIT push-pull output contributes to the total current amplification. This is a departure from the regular DEFISIT amplifier, where the SIT and mosfet both fully contribute to the total current amplification.
Here are a couple of LTSpice simulations. I did many simulations, but these two are what I based my design on. I chose -47VDC for one sim since I intended to build it using one of my existing amplifier chassis and power supply. I chose the -63VDC for the other since that is a reasonable voltage that will take advantage of the power capability of the THF-51S and still keep the power supply voltage reasonable for capacitor voltage and heat sink heat dissipation.
As shown in the simulation results (47VDC power supply and 2.0A Iq), the measured current through C4 connected to the SIT drain was 361mA peak, the measured current through C7 connected to the mosfet drain was 150mA, for a total of 511mA. As a check the current measured through the speaker, R8, was 509mA - close enough. So the contribution of the mosfet was 150mA/511mA = 29%.
I have not included the LTSpice screen shots of power output but the 8 Ohm power output was nearly 23W at 1% THD and 4 Ohm power output was nearly 40W at 1% THD.
63VDC PS 2A 8R 1Watt:
63VDC PS 2A AC signal current through C4, C7, and 8R Speaker:
47VDC PS 2A 8R 1Watt:
47VDC PS 2A AC signal current through C4, C7, and 8R Speaker:
As mentioned I decided to build the 47VDC version. The chassis and power supply is from my BAF2015 Amplifier - A SIT Mu Follower Revisited With Feedback (https://www.diyaudio.com/community/...t-mu-follower-revisited-with-feedback.408193/). The power supply is CLC, comprising 2x22mF - Hammond 156B - 2x22mF, and an Antek AS-3218 transformer per channel for full dual mono construction.
I put together a PCB design and sent it off to JLCPCB for fabrication. In addition to stuffing the PCB once I received it, I also changed the CLC filter from V+ supply to V- supply.
Initial power-up and subsequent Iq and SIT Vds adjustments went smoothly, and Iq and SIT Vds were stable.
Next up was testing with an 1kHz signal and distortion measurements. Unfortunately when powered up with an AC line powered 1kHz oscillator connected, the amp immediately blew the powerline fuse. So it was trouble shooting time. After checking the schematic and pcb, checking that the input capacitor was not faulty, and not finding any visible ground shorts with the meter, I was baffled.
I knew that the issue was mostly likely a grounding problem and that it only manifested itelf when an input device was connected to the amplifier. Luckily a thought came to me fairly quickly - when I changed the CLC filter from V+ to V- supply, I forgot to move the power supply connection to safety ground on the CLC filter board. The correction was made and success. Only one fuse was permanently damaged.
The dim bulb tester came in handy here as once the fuse blew, the dim bulb tester went in and I was able to probe around with power and not blow anything. In addition to using it during trouble shooting, I always use it for first power-up testing of power supplies during various stages of construction and first power-up testing of audio circuits.
As for the amplifier supply not shorting with no input device connected but shorting with an input device connected, I gave it some thought. The power supply was for V-, with V- from the supply connected as power to the amplifier circuit. V+ from the supply was connected to the amplifier circuit board as ground. The V- from the supply was ground in the previous amplifier but I forgot to change it so it was now incorrectly connected to chassis safety ground. With no input device connected to the amplifier circuit, the power to the amplifier board was floating. Although the V- was also connected to the chassis safety ground and to the powerline/IEC ground, there was no other direct connection to V+, so no short. The live from the powerline/IEC was connected to the transformer primary, which was isolated from the transformer secondary. Also the speaker and input jacks were isolated from the chassis. So with no complete direct connection of V- to V+, the current can only flow through the audio circuit. The chassis was connected to V- but there was no path for the current to flow from the chassis to power supply V+.
When the oscillator was connected to the amplifier input, the circuit was completed, current flowed, and the fuse blew. That was because the oscillator was AC powered and had a safety ground connection. So the oscillator line safety ground connected to the oscillator power supply ground, then connected to the oscillator signal ground, then connected to the amplifier signal ground, then connected to the amplifier V+. The oscillator safety ground is also connected to the amplifier safety ground through the line ground, and the amplifier safety ground is connected to amplifier V-. The net result was amplifier V+ and amplifier V- were connected together - a short circuit. There was a CL60 thermistor at the amplifier power supply ground connection. That would have limited the short circuit current: 47V/10R = 4.7A. Fuse was 2.5A slow blow.
I have swapped power supply polarity before but had always remembered to switch the safety ground connected, until this time.
So always be careful. Electricity can kill.
My build is only the output stage, and it incorporates the duo output capacitor arrangement which can be tweaked to adjust the amount of current amplification that the mosfet part of the DEFISIT push-pull output contributes to the total current amplification. This is a departure from the regular DEFISIT amplifier, where the SIT and mosfet both fully contribute to the total current amplification.
Here are a couple of LTSpice simulations. I did many simulations, but these two are what I based my design on. I chose -47VDC for one sim since I intended to build it using one of my existing amplifier chassis and power supply. I chose the -63VDC for the other since that is a reasonable voltage that will take advantage of the power capability of the THF-51S and still keep the power supply voltage reasonable for capacitor voltage and heat sink heat dissipation.
As shown in the simulation results (47VDC power supply and 2.0A Iq), the measured current through C4 connected to the SIT drain was 361mA peak, the measured current through C7 connected to the mosfet drain was 150mA, for a total of 511mA. As a check the current measured through the speaker, R8, was 509mA - close enough. So the contribution of the mosfet was 150mA/511mA = 29%.
I have not included the LTSpice screen shots of power output but the 8 Ohm power output was nearly 23W at 1% THD and 4 Ohm power output was nearly 40W at 1% THD.
63VDC PS 2A 8R 1Watt:
63VDC PS 2A AC signal current through C4, C7, and 8R Speaker:
47VDC PS 2A 8R 1Watt:
47VDC PS 2A AC signal current through C4, C7, and 8R Speaker:
As mentioned I decided to build the 47VDC version. The chassis and power supply is from my BAF2015 Amplifier - A SIT Mu Follower Revisited With Feedback (https://www.diyaudio.com/community/...t-mu-follower-revisited-with-feedback.408193/). The power supply is CLC, comprising 2x22mF - Hammond 156B - 2x22mF, and an Antek AS-3218 transformer per channel for full dual mono construction.
I put together a PCB design and sent it off to JLCPCB for fabrication. In addition to stuffing the PCB once I received it, I also changed the CLC filter from V+ supply to V- supply.
Initial power-up and subsequent Iq and SIT Vds adjustments went smoothly, and Iq and SIT Vds were stable.
Next up was testing with an 1kHz signal and distortion measurements. Unfortunately when powered up with an AC line powered 1kHz oscillator connected, the amp immediately blew the powerline fuse. So it was trouble shooting time. After checking the schematic and pcb, checking that the input capacitor was not faulty, and not finding any visible ground shorts with the meter, I was baffled.
I knew that the issue was mostly likely a grounding problem and that it only manifested itelf when an input device was connected to the amplifier. Luckily a thought came to me fairly quickly - when I changed the CLC filter from V+ to V- supply, I forgot to move the power supply connection to safety ground on the CLC filter board. The correction was made and success. Only one fuse was permanently damaged.
The dim bulb tester came in handy here as once the fuse blew, the dim bulb tester went in and I was able to probe around with power and not blow anything. In addition to using it during trouble shooting, I always use it for first power-up testing of power supplies during various stages of construction and first power-up testing of audio circuits.
As for the amplifier supply not shorting with no input device connected but shorting with an input device connected, I gave it some thought. The power supply was for V-, with V- from the supply connected as power to the amplifier circuit. V+ from the supply was connected to the amplifier circuit board as ground. The V- from the supply was ground in the previous amplifier but I forgot to change it so it was now incorrectly connected to chassis safety ground. With no input device connected to the amplifier circuit, the power to the amplifier board was floating. Although the V- was also connected to the chassis safety ground and to the powerline/IEC ground, there was no other direct connection to V+, so no short. The live from the powerline/IEC was connected to the transformer primary, which was isolated from the transformer secondary. Also the speaker and input jacks were isolated from the chassis. So with no complete direct connection of V- to V+, the current can only flow through the audio circuit. The chassis was connected to V- but there was no path for the current to flow from the chassis to power supply V+.
When the oscillator was connected to the amplifier input, the circuit was completed, current flowed, and the fuse blew. That was because the oscillator was AC powered and had a safety ground connection. So the oscillator line safety ground connected to the oscillator power supply ground, then connected to the oscillator signal ground, then connected to the amplifier signal ground, then connected to the amplifier V+. The oscillator safety ground is also connected to the amplifier safety ground through the line ground, and the amplifier safety ground is connected to amplifier V-. The net result was amplifier V+ and amplifier V- were connected together - a short circuit. There was a CL60 thermistor at the amplifier power supply ground connection. That would have limited the short circuit current: 47V/10R = 4.7A. Fuse was 2.5A slow blow.
I have swapped power supply polarity before but had always remembered to switch the safety ground connected, until this time.
So always be careful. Electricity can kill.
Wonderful Ben. That heatsink is not huge; what are its dimensions? Any listening tests so far? Enjoy!!!!
The board is 100mm x 130mm and the heat sink is 150mm x 400mm, so it is not a small heat sink. In the past I have had 90W on it with no overheating issues.
Yes, I have been listening to this completed channel, with the left channel amplification provided by my choked loaded 2SK180 follower. I am liking the sound so far. What's not to like with SIT amplification. I will be able to have a better idea once I get the other channel built. The PCB is at the airport waiting to clear customs so I should have it later this week.'
Yes, I have been listening to this completed channel, with the left channel amplification provided by my choked loaded 2SK180 follower. I am liking the sound so far. What's not to like with SIT amplification. I will be able to have a better idea once I get the other channel built. The PCB is at the airport waiting to clear customs so I should have it later this week.'
Here is my KiCad schematic and right channel PCB layout.
The PCB has a ground plane on the back, and I was able to place all of the circuit traces on the front side except for one short trace.
As mentioned previously, the power supply outputs 47VDC under load.
The only difference between the schematic is that for R12 to R16, I replaced the five 0.3R 3W resistors with three 0.5R 3W resistors and two jumper wires. There is no DC through those resistors and only signal AC, so no constant heat dissipation. In normal operation they feel only slightly warm.
The PCB has a ground plane on the back, and I was able to place all of the circuit traces on the front side except for one short trace.
As mentioned previously, the power supply outputs 47VDC under load.
The only difference between the schematic is that for R12 to R16, I replaced the five 0.3R 3W resistors with three 0.5R 3W resistors and two jumper wires. There is no DC through those resistors and only signal AC, so no constant heat dissipation. In normal operation they feel only slightly warm.
KiCad is fun, LTSpice is fun. I look at them as nerds' computer games.
I have never used auto routing. I prefer to place and route on my own based on my understanding of current flows and loops. So far the boards that I have designed have performed to my expectations. 🤓
I have never used auto routing. I prefer to place and route on my own based on my understanding of current flows and loops. So far the boards that I have designed have performed to my expectations. 🤓
I guess you wanted to say at source pins..
If you plot the source current of sit and source current of mos, how do the ratio of each looks when compared to current through load?
Whoops, my mistake. The capacitors are obviously connected to the source pins.
I have also looked at the signal current at the source pin of the mosfet. The current is much higher than the signal current at the SIT pin.
C4: Ip-p=350mA+345mA=695mA
C7: Ip-p=162mA+165mA=327mA
R8 (speaker) Ip-p=510ma+512mA=1022mA
U1 (SIT): Ip-p=2081mA-1869mA=212mA
M1 (Mosfet): Ip-p=2386mA-1571mA=815mA
The numbers show that there is no signal loss. The resistor in series with the capacitor controls the current through C7, and the remainder of the current goes through C4.
The numbers show the current at the mosfet source is much higher than the current at the SIT source - 21% at the SIT and 79% at the mosfet.
So according to the simulation numbers and my interpretation of the numbers, the mosfet signal current at its source partially flows through C7 and the remainder flows through C4. C4 has current from the SIT and mosfet.
I have also measured the AC current across an 8Ohm load connected to the amplifier speaker and also across the string of resistors in series with C7. Real life measurements confirm the distribution of AC current through the two capacitors.
I am not an EE, just a hobbyist who is having fun trying to understand the the operation of the circuit. This is a science experiment for me, and any insight would be welcome. 🤓
I have also looked at the signal current at the source pin of the mosfet. The current is much higher than the signal current at the SIT pin.
C4: Ip-p=350mA+345mA=695mA
C7: Ip-p=162mA+165mA=327mA
R8 (speaker) Ip-p=510ma+512mA=1022mA
U1 (SIT): Ip-p=2081mA-1869mA=212mA
M1 (Mosfet): Ip-p=2386mA-1571mA=815mA
The numbers show that there is no signal loss. The resistor in series with the capacitor controls the current through C7, and the remainder of the current goes through C4.
The numbers show the current at the mosfet source is much higher than the current at the SIT source - 21% at the SIT and 79% at the mosfet.
So according to the simulation numbers and my interpretation of the numbers, the mosfet signal current at its source partially flows through C7 and the remainder flows through C4. C4 has current from the SIT and mosfet.
I have also measured the AC current across an 8Ohm load connected to the amplifier speaker and also across the string of resistors in series with C7. Real life measurements confirm the distribution of AC current through the two capacitors.
I am not an EE, just a hobbyist who is having fun trying to understand the the operation of the circuit. This is a science experiment for me, and any insight would be welcome. 🤓
Quite a bit of difference..
Same thing here, started to get things better when started to play with real os and results were far from sim.
Don’t get me wrong. Wanted to point the right direction..
Another "experiment" - no resistor in series with C7:
C4 Ip-p=276mA+267mA=543mA
C7 Ip-p=746mA+764mA=1510mA
R8(Speaker) Ip-p=505mA+508mA=1013mA
U1 (SIT) Ip-p=2262mA-1730mA=532mA
M1(Mosfet) Ip-p=2752mA-1235mA=1517mA
Looking at the numbers, the currents through C4 and C7 are now out of phase, as are the currents at the SIT and mosfet sources.
The numbers do not add up exactly as the peaks do not line up exactly.
The current at the speaker is now the difference of the two currents.
C4 Ip-p=276mA+267mA=543mA
C7 Ip-p=746mA+764mA=1510mA
R8(Speaker) Ip-p=505mA+508mA=1013mA
U1 (SIT) Ip-p=2262mA-1730mA=532mA
M1(Mosfet) Ip-p=2752mA-1235mA=1517mA
Looking at the numbers, the currents through C4 and C7 are now out of phase, as are the currents at the SIT and mosfet sources.
The numbers do not add up exactly as the peaks do not line up exactly.
The current at the speaker is now the difference of the two currents.
I have the impression that we get at least in Spice with the models flowing around and maybe in real, I do not yet know, the 20% for the MosFet only with a 3R resistor between the sources. At least with the Nelson way to determine the contribution.
use LTSpice for general idea, then fiddle in vivo to get proper values
as is, LTSpice is better dealing with mosfets than Sits, but even as that, conductance of pucks is so wild that reality is much weightier than simulation
pretty much, all you need is properly floating scope, to compare current through load with current through RC on mosfet side
if not scope handy, feed it with 400Hz to 1KHz sine, use DVM
as is, LTSpice is better dealing with mosfets than Sits, but even as that, conductance of pucks is so wild that reality is much weightier than simulation
pretty much, all you need is properly floating scope, to compare current through load with current through RC on mosfet side
if not scope handy, feed it with 400Hz to 1KHz sine, use DVM
As I had mentioned in post #11, had measured the AC current through the resistor at the mosfet side capacitor and through the load resistor in real life. I had fed the amp with a 1kHz signal and measured the voltage drop across the resistors to determine the currents. The calculated currents were in agreement with the LTSpice results.
What is questioned and discussed is what are the AC currents coming out of the SIT and mosfet source. Is the mosfet outputting higher AC current than the SIT? And is part of the mosfet AC current passing through the RC path and the rest of the mosfet AC current passing through the resistor between the sources and then joining with the SIT AC current and pass through the SIT capacitor?
The simulation seems to show that, but is the simulation results to be believed? Is the simulation valid and representative of the actual circuit? Am I misunderstanding the simulation currents at the various nodes? It is way beyond my knowledge, understanding, and ability to answer that. 🤓
What is questioned and discussed is what are the AC currents coming out of the SIT and mosfet source. Is the mosfet outputting higher AC current than the SIT? And is part of the mosfet AC current passing through the RC path and the rest of the mosfet AC current passing through the resistor between the sources and then joining with the SIT AC current and pass through the SIT capacitor?
The simulation seems to show that, but is the simulation results to be believed? Is the simulation valid and representative of the actual circuit? Am I misunderstanding the simulation currents at the various nodes? It is way beyond my knowledge, understanding, and ability to answer that. 🤓
I’m going with matching the curves at 4 and 8ohms with the SIT -5.
The math hurts my head 😀
The math hurts my head 😀