I assumed 1oz is standard and 2oz wasn’t specifically called out, so I took a leap of faith and went with 1oz pours on the Output and PSFILT boards along with the input boards I ordered.
I haven’t analyzed currents in the circuits as related to trace widths on the boards, but your question now makes me wonder how the 1oz boards will perform. Enough so to attempt this analysis on my own, hoping my initial assumption mitigates my lazy approach 😊.
Late to the party but thanks @Mark Johnson !! I thought there were a myriad possibilities with the M2X thread…and now there is more! Infinity Myriad! Haha!
I always wondered how you came up with the names, so thanks for explaining! Ship of Theseus! It’s a journey!
I’ll look at the Nimitz first since I have the M2X…
Best,
Anand.
I always wondered how you came up with the names, so thanks for explaining! Ship of Theseus! It’s a journey!
I’ll look at the Nimitz first since I have the M2X…
Best,
Anand.
@Mark Johnson
Thanks! I’ll definitely look at that (Bon Homme Richard) as well. I am spinning with how many possibilities you have offered for free! Very altruistic of you! FWIW, I am currently greatly enjoying Hugh Dean’s Alpha Nirvana which is a SEPP design.
I’ll ask the question that some folks may be wondering or others already know. The elephant in the room with this design is the output stage which appears to be SE with a CCS running at 1.6A. Instead of a VFET, we are using a Pchannel MOSFET - great for the proletariat! As such, is it right to assume that the max output power/8 ohms is roughly 10 Watts Class A? And I am sure the gains are somewhat variable depending on what input card you use, but probably in the 10-15dB range as well?
I figure I can build them all and measure on my Keysight DSOX which I love using
Best,
Anand.
Thanks! I’ll definitely look at that (Bon Homme Richard) as well. I am spinning with how many possibilities you have offered for free! Very altruistic of you! FWIW, I am currently greatly enjoying Hugh Dean’s Alpha Nirvana which is a SEPP design.
I’ll ask the question that some folks may be wondering or others already know. The elephant in the room with this design is the output stage which appears to be SE with a CCS running at 1.6A. Instead of a VFET, we are using a Pchannel MOSFET - great for the proletariat! As such, is it right to assume that the max output power/8 ohms is roughly 10 Watts Class A? And I am sure the gains are somewhat variable depending on what input card you use, but probably in the 10-15dB range as well?
I figure I can build them all and measure on my Keysight DSOX which I love using
Best,
Anand.
Max output power is the same as the diyAudio Store VFET kits designed by Nelson Pass -- see page 12 of his .pdf (linked in post #1 of this thread)
Voltage gain is about 3-4 dB higher, because the PMOSFET follower's gain is unity ( 0 dB ), but the VFET followers have a voltage "gain" of -3 to -4 dB. A downside of triode-like output devices.
Voltage gain is about 3-4 dB higher, because the PMOSFET follower's gain is unity ( 0 dB ), but the VFET followers have a voltage "gain" of -3 to -4 dB. A downside of triode-like output devices.
The Bon Homme Richard is very close to the front end I made, where I basically used the datasheet app for the JFE2140 dual JFET I chose (and not paralleled), and then modified to suit. My op amp is an OPA604, but second choice was LME49860. Either should make the required output swing (these being for my N-VFET).
I just need to build the other channel and wire it all together with my DC-DC converter boards which gets almost -24V. Then getting everything mounted, which is one of my toughest challenges.
My schematic (and PCB) tools are limited for now, but I plan to share it in some form.
I just need to build the other channel and wire it all together with my DC-DC converter boards which gets almost -24V. Then getting everything mounted, which is one of my toughest challenges.
My schematic (and PCB) tools are limited for now, but I plan to share it in some form.
Keep in mind that your N-VFET amp has an end-to-end voltage gain of +10 dB (i.e. 3.2 volts per volt). When operating it at full output power (8.9V RMS at the speaker terminals), the input to your front end will be 2.8V RMS. That's 8 volts peak-to-trough. Some of the application circuit examples in the JFE2140 datasheet, will misbehave pretty badly with input signals this large. The one in Figure 9-6 is a prime example.
I currently have the FE gain at 6.9, so ~2.2V on input. Are you talking Vds headroom or Vgs limit?
As long as your design comfortably handles 2.2V RMS (6.2V peak-to-trough) on the input, without degraded output waveform shape, you're probably good to go. I'd suggest testing the front end board in isolation, with a 20 kHz triangle wave input signal whose amplitude is 7V peak-to-trough, before hooking it up to the amp channel boards. If all performs well in isolation, that means the risk is lower when you connect it up to the amp channel boards with their irreplaceable, unobtanium VFETs.
Actually, 2.2V may be too close for comfort. I had a decent looking 20Vpk sine output (with some other supply issues I was still working on) but now I see clipping at the gate of the feedback JFET with about 19V on output which limits to 8W on output. (I haven't done distortion measurements yet.) More tweaking!
Definitely won't hook up to output stage until it passes bench tests not yet finished. Thanks for your caveats.
Definitely won't hook up to output stage until it passes bench tests not yet finished. Thanks for your caveats.
EDIT: The clipping seen prior to earlier post was from the power supply sag I was already working on (an oversized RC filter on my DC-DC converter output I've been in the process of fixing). It was just divided down at the diff pair and spooked me.
The FETs look fine to me. All is right in the world. (At least here.) So far.
The FETs look fine to me. All is right in the world. (At least here.) So far.
We always want to keep in mind that the low Drain impedance of VFETs operated as followers gives a couple dB loss so that the front end has to swing more voltage than you are looking for at the output. Fortunately the output starts to to crap out around the same point as the front end, so it's not such a big deal.
BTW, am puzzled by the triangle wave recommendation. Is because it's less voltage stressful downstream than a sine or square?
Just a personal preference. Sometimes it's easy to visually recognize distortions on a bi-ramp, such as compression or rail sticking. But recall that a sinewave has a factor of pi (3.14) greater slew rate (dV/dt) than a triangle wave at the same frequency. So perhaps a 65 kHz triangle wave input is desirable.
Hi Mark, getting parts together to see what I need to order. Can I ask a quick question - for the Theseus-MOS output stage pcb, the schematic shows R5 to be 2.26K, however on your parts list for this board you show R5 to be 2.08K and for the Mouser part number for R5 it says 2K05 (2.05K). Can you confirm the correct value for R5 thanks.
I have just read post #35 after writing the above - I assume R5 can be 2.05K?
I have just read post #35 after writing the above - I assume R5 can be 2.05K?
Thanks for the heads-up; not one but two irregularities slipped through the final checks and into the .zip archive. I apologize for the errors. First, Channel-A and Channel-B schematics got out of sync; to see this clearly, print both, overlay them, and hold up to a bright light source. Whoops!
Second, although "2.08K" is a legitimate member of the E192 standard series of electronic component values (see attachment 1), there are none in stock anywhere. So I used 2.05K in my build, and I copied the Mouser part number of my 2.05K into the Detailed Parts List, leading to confusion: 2.05 is not 2.08! Horrors!. To repair the damage, I have modified the schematics (attachments 2 and 3) to indicate that either 2.08K or 2.05K is perfectly fine. I have also modified the Detailed Parts Lists (attachment 4) to indicate that either 2.08K or 2.05K is perfectly fine, and I removed all Mouser part numbers for R5. Builders can purchase whatever is in stock on the day of their order, without the distraction of someone else's ancient choice of part number, way back in February of 2022.
I'll edit post #1 of this thread so it points here (at least until the NEXT edit...)
_
Second, although "2.08K" is a legitimate member of the E192 standard series of electronic component values (see attachment 1), there are none in stock anywhere. So I used 2.05K in my build, and I copied the Mouser part number of my 2.05K into the Detailed Parts List, leading to confusion: 2.05 is not 2.08! Horrors!. To repair the damage, I have modified the schematics (attachments 2 and 3) to indicate that either 2.08K or 2.05K is perfectly fine. I have also modified the Detailed Parts Lists (attachment 4) to indicate that either 2.08K or 2.05K is perfectly fine, and I removed all Mouser part numbers for R5. Builders can purchase whatever is in stock on the day of their order, without the distraction of someone else's ancient choice of part number, way back in February of 2022.
I'll edit post #1 of this thread so it points here (at least until the NEXT edit...)
_
