I have to notice that I'm not questioning the particular choice of the opamp (which is an excellent opamp, I use its sibling LT1055 as DC servo because of its excellent offset specs), I'm trying to figure out the slew rate difference and the other aspecs of the opamp approach that helped the final result which seems the best I've seen on any solid state Amp.Voltage swing on the opamp is small, so 16V SR is more than enough. VAS and output needs to provide much higher SR to accomodate rail to rail voltage swing.
Opamp doesn't have to, we are not using it as vas.
The higher voltage swing, the higher slew rate is needed.
1W amp doesn't need the same slew rate as 100W amp.
Reasons to use JFET op-amp:
a) high input impedance
b) low input bias current
c) low noise characteristics
d) good thermal stability
Good comparison on BJT and JFET op-amps:
https://www.ti.com/lit/ab/sboa355/sboa355.pdf
https://eelectronics.medium.com/exploring-the-features-of-jfet-input-op-amps-84773f3e4ebb
Basically, the time domain doesn't change but the amplitude changes so slew rate increases by the amount of gain, right?
Thank you for bringing up this new stuff.
Basically, the time domain doesn't change but the amplitude changes so slew rate increases by the amount of gain, right?
Exactly! That's the way I understand it.
So far I had no luck with simulating hexfet version, it kind of works, but lots of problems.
When I'm back home, will post my sim, perhaps you can help to make it work? Any help will be appreciated..
Last edited:
It's been for a while since doing things with LTSpice. I have to revise my memory first. Such things happen to me so often because I have many hobbies besides electronics. I will try to help indeed.
The first thing I would do is omitting C5, R22 (in your schematic) compensation network because they cause overshoot in Hexfet version. Hexfet version is best without those. They also may do other nasty things on your approach.
The first thing I would do is omitting C5, R22 (in your schematic) compensation network because they cause overshoot in Hexfet version. Hexfet version is best without those. They also may do other nasty things on your approach.
Here is the half-baked HexFet version of op-amp based circlophone.
The whole schematic had to be re-drawn upside down with opposite polarity devices.
Irfp260 were chosen for the output devices, because I have lots of them, and they need to go....
Feel free to replace them with something else, models are available in my spice lib included in the sim.
The whole schematic had to be re-drawn upside down with opposite polarity devices.
Irfp260 were chosen for the output devices, because I have lots of them, and they need to go....
Feel free to replace them with something else, models are available in my spice lib included in the sim.
Attachments
Last edited:
Here is seemingly final BJT simulation that looks promising to test it in real build.
I redrawn some portions of the schematic to make more readable, and renumbered all transistors for backward compatibility with
original Circlophone numbering scheme. One cap was also tweaked compared to the previous version.
SlewRate is estimated at 90 V/us, Phase Margin: 80, Gain Margin: 12, Thd: 0.000018% at 1kHz, 0.00084% at 20kHz.
Square waves looking good, and clipping also is not bad - there is a little bit of latching, but it looks clean, and as far as I can tell, it's
unavoidable with op-amp driven amp.
Any suggestions for further changes?
I redrawn some portions of the schematic to make more readable, and renumbered all transistors for backward compatibility with
original Circlophone numbering scheme. One cap was also tweaked compared to the previous version.
SlewRate is estimated at 90 V/us, Phase Margin: 80, Gain Margin: 12, Thd: 0.000018% at 1kHz, 0.00084% at 20kHz.
Square waves looking good, and clipping also is not bad - there is a little bit of latching, but it looks clean, and as far as I can tell, it's
unavoidable with op-amp driven amp.
Any suggestions for further changes?
Attachments
Re: HexFet version from post 2485: if I replace IRFP260 with IRFP240_ model, suddenly sim looks much better....
Of course the compensation caps: C10 and C4 have weird values for now, the best would be to get rid of C4 completely, and keep/adjust
C7 and C10 only. C10 is way too high judging from my previous attempts.
Of course the compensation caps: C10 and C4 have weird values for now, the best would be to get rid of C4 completely, and keep/adjust
C7 and C10 only. C10 is way too high judging from my previous attempts.
Attachments
Regarding the Hexfet version, I think I have proposed a bootstrapped version, to allow the erasure of the Vgs and arrive at a swing amplitude comparable to the Bjt version.
It must be somewhere up there ^ . Maybe someone remembers where to find it.
It would be an useful inclusion for a FET super-circlo
It must be somewhere up there ^ . Maybe someone remembers where to find it.
It would be an useful inclusion for a FET super-circlo
Elvee,
Question (referring schematic from post 2486):
C2 is connected between B of Q3, and output.
Alternatively, it can be connected between B of Q3 and B of Q11. Sim results are the same as far as I can tell.
Which option is better/correct?
Question (referring schematic from post 2486):
C2 is connected between B of Q3, and output.
Alternatively, it can be connected between B of Q3 and B of Q11. Sim results are the same as far as I can tell.
Which option is better/correct?
Yes, it is. However, the values should probably be adapted to the gate resistors.Was it this one?
If C2 is connected to the output, the compensation is output-inclusive, which should benefit the linearity, and to B/Q11, the loop is shorter which should benefit the HF stability.
If the stability is OK, the output-inclusive option is better (in principle), but in practice I doubt it will make any significant difference
I have used the Bootstrap circuit as suggested by you and have found that the output swing is slightly lower to Rail supply voltage almost compensating the voltage swing drop due to Vgs.Regarding the Hexfet version, I think I have proposed a bootstrapped version, to allow the erasure of the Vgs and arrive at a swing amplitude comparable to the Bjt version.
For me stability of the amplifier is more important than other parameters which I cannot distinguish with my ears. The present MOSFET version is fully stable even with mains voltage fluctuations. Quiescent current and Heat sink temperature are stable.
Last edited:
Thats right Minek. Initially I used 220 Ohm gate resisters but on Mr Elvee's advice I removed them.
Mr Minek,
Thats what Me Elvee suggested at post 1903 regarding Gate resistors
Quote
"Gate stoppers are an easy solution, but they are best avoided whenever possible: they introduce parasitic poles in the global response, eroding the phase margin.
If the MOSFets are wired to a remote sink rather than directly on the PCB, they might be unavoidable, but in this case, ferrite beads are preferable, as their influence on the GNFB is smaller at frequencies of interest"
Unquote.
Thats what Me Elvee suggested at post 1903 regarding Gate resistors
Quote
"Gate stoppers are an easy solution, but they are best avoided whenever possible: they introduce parasitic poles in the global response, eroding the phase margin.
If the MOSFets are wired to a remote sink rather than directly on the PCB, they might be unavoidable, but in this case, ferrite beads are preferable, as their influence on the GNFB is smaller at frequencies of interest"
Unquote.
I checked HexFet sim and it looks the same with or without gate stoppers.....
Anyway, HexFet sim still needs more work, it is not satisfactory yet, and overall results so far are not as good as BJT version.
I'm little bit disappointed, normally I would expect better results with Hexfets.....
Anyway, HexFet sim still needs more work, it is not satisfactory yet, and overall results so far are not as good as BJT version.
I'm little bit disappointed, normally I would expect better results with Hexfets.....
Last edited:
In fact I was not talking about gate stoppers (which should be minimal or completely eliminated if at all possible) but the G-S resistors.
Now that I have located the bootstrapped schematic (thanks to the help of the community), I see that they are 390 ohm, but the bootstrap components are 1K-100µF, which is more reasonable (but there is certainly room for improvement)
Now that I have located the bootstrapped schematic (thanks to the help of the community), I see that they are 390 ohm, but the bootstrap components are 1K-100µF, which is more reasonable (but there is certainly room for improvement)
I tried that bootstrap, and I don't see any difference in output swing....
Perhaps because of the op-amp? Or I did something wrong....
Anyway, here is the latest sim of HexFet version, and everything looks good.
PM: 77, GM: 8, SR: 55
Good square waves and stable. No gate stoppers.
Overall results slightly worse than BJT version.
Perhaps you guys can make it better?
I'll focus on the PCB for the BJT version.
PS. Ignore these 1u Ohm resistors in the output. They should be removed.
PS2. I'll still try to re-do this sim with IRFP260 outputs; have to get rid of them somehow....
Perhaps because of the op-amp? Or I did something wrong....
Anyway, here is the latest sim of HexFet version, and everything looks good.
PM: 77, GM: 8, SR: 55
Good square waves and stable. No gate stoppers.
Overall results slightly worse than BJT version.
Perhaps you guys can make it better?
I'll focus on the PCB for the BJT version.
PS. Ignore these 1u Ohm resistors in the output. They should be removed.
PS2. I'll still try to re-do this sim with IRFP260 outputs; have to get rid of them somehow....
Attachments
Last edited:
I suspect that it is caused by Q3, limiting the output to Vto at most. Needs confirmation, I'll make some sims to be sureI tried that bootstrap, and I don't see any difference in output swing....
- Home
- Amplifiers
- Solid State
- ♫♪ My little cheap Circlophone© ♫♪