I hope that CFA connoisseurs will tell something more about that, and I will do the sim latter this evening.
Overshoot
Here is the simulations. First with C29 of 220p and there is an overshoot, but with 470p it is completely disappeared. In my oppinion it is nothing to wary about, but, please, could some from the CFA specialist give more explanation.
Here is the simulations. First with C29 of 220p and there is an overshoot, but with 470p it is completely disappeared. In my oppinion it is nothing to wary about, but, please, could some from the CFA specialist give more explanation.
Attachments
My thick skull is begging to understand and is catching on slowly but surely. It took me a little bit to start thinking in pico seconds that I dealt with on ECL circuits when transitioning from mostly tube based analog computers then on to TTL based in the low nano second switching times.
So I will get some LED's on and call it a day.
Overshoot
Perhaps the compensation circuit has some impact to this behavior. I removed the 2.5nF-1k R-C branch and the overshoot was completely gone. Then I disabled the input LPF and let the square wave slam its 400V/us transition into the amp and the overshoot at the output was negligibly barely visible. This is not to say removing that part of circuit was the solution by any means though, but perhaps the compensation could probably get some fine tune for a bit more balanced performance?
Perhaps the compensation circuit has some impact to this behavior. I removed the 2.5nF-1k R-C branch and the overshoot was completely gone. Then I disabled the input LPF and let the square wave slam its 400V/us transition into the amp and the overshoot at the output was negligibly barely visible. This is not to say removing that part of circuit was the solution by any means though, but perhaps the compensation could probably get some fine tune for a bit more balanced performance?
Attachments
Yes, you are correct, removing 2.5nF - 1K R-C branch did remove the overshoot but we loose something with this. I tried to optimize the compensation circuit as it became now pure Chery(as krglee call it).
I changed next:
Removed R1, C27
Decreased C6, C9 to 47Pf
Increased C13 to 22pF
Decreased R24 to 10 ohm
Result:
THD20k increased from 22ppm to 26ppm at 50W/8R
THD20k incresead from 50ppm to 66 ppm at 200W/8R
PM 76 degree
GM droped from 22 dB to 15 dB
Stable up to 50 nF without output inductor, before it was 200nF
Does that overshoot is realy a problem??
If we want to change the schematic then that does not provoke any change in the layout, just some components are not inserted some are with diferent values.
Anyhaw during real life test some fine tunig could be made.
BR Damir
Overshoot - resolved
Hi Damir,
I think your idea is good by leaving the compensation topology unchanged and options open, and letting the real life test decide the fine tune result.
"Does that overshoot is realy a problem??" --- Perhaps different people have different views. Honestly, I do not know, it may or may not be a problem.
During my simulation earlier today (circuit unchanged with LPF in) about this overshoot, I tried with much slower edge square waves, at 0.5us rise time or 0.8V/us rate of transition, and the overshoot was still there. It indicated that the transition speed itself on the edges hardly matters, it was at the moment the transition stops where the amp loses grip and keeps going. This behavior, in my view, is similar to a slow amp losing control when it is out-slewed by a fast rising input at a transition edge. The difference is one is too slow to catch up speed, the other is fast enough but has hard time to stop and make a turn. (Perhaps it was exactly because your amp is fast enough that the hard-to-turn problem shows up.)
To me, time domain performance is more critical than frequency domain figures, as an amp should be designed and built to follow the input precisely at any point in time. And for this reason, I think it is worthy to budget some THD and stability margin surplus(?
) for an overshoot free transient response.
Hi Damir,
I think your idea is good by leaving the compensation topology unchanged and options open, and letting the real life test decide the fine tune result.
"Does that overshoot is realy a problem??" --- Perhaps different people have different views. Honestly, I do not know, it may or may not be a problem.
During my simulation earlier today (circuit unchanged with LPF in) about this overshoot, I tried with much slower edge square waves, at 0.5us rise time or 0.8V/us rate of transition, and the overshoot was still there. It indicated that the transition speed itself on the edges hardly matters, it was at the moment the transition stops where the amp loses grip and keeps going. This behavior, in my view, is similar to a slow amp losing control when it is out-slewed by a fast rising input at a transition edge. The difference is one is too slow to catch up speed, the other is fast enough but has hard time to stop and make a turn. (Perhaps it was exactly because your amp is fast enough that the hard-to-turn problem shows up.)
To me, time domain performance is more critical than frequency domain figures, as an amp should be designed and built to follow the input precisely at any point in time. And for this reason, I think it is worthy to budget some THD and stability margin surplus(?
) for an overshoot free transient response.
Last edited:
Which schematic? If it is from #258, the original is sorta TPC with an added Cherry cap which may not do much at all.
Your changes turn it into sorta Miller with another added Cherry which again may not do much at all.
Both put load the Holy HiZ VAS collector so are sorta evil.
I can't find R24
______________
The sensible cure is to increase the input HF roll-off cap C29 to 470p as Damir shows.
Last edited:
Thanks for your comments!
You're right. I didn't realize the 1k-ohm and 220pF makes a high-ish 723KHz cutoff.
#258 was the vertical MOSFET version. I don't recall the number of the post but the schematic in question is attached. R24 (22R) is next to the 480R feedback resistor.
Which schematic? If it is from #258, the original is sorta TPC with an added Cherry cap which may not do much at all.
Your changes turn it into sorta Miller with another added Cherry which again may not do much at all.
Both put load the Holy HiZ VAS collector so are sorta evil.
I can't find R24
______________
The sensible cure is to increase the input HF roll-off cap C29 to 470p as Damir shows.
Thank you kgrlee for your comment, and try to simulate as described below, please.
First there is small part of Miller, C7 R22, C14 R30, but those caps are 10 pF only.
If I use pure Miller and connect C6 and C9 to the corresponding VAS collectors(decreased to 68 pF to keep similar PM, and C7 R22, C14 R30 connected as shunt compensation) THD20k increases from 22 ppm to 128 ppm. If I use pure TPC(C8 connected to the one of the VASes collector) THD20k increases to 33 ppm, so it looks that added Charry is doing something useful.
With my combination of Miller and Cherry I've got lowest distortion, very good PM and excellent GM, insensitive to the capacitive loading even without output inductor, so I put 0.5 uH inductor as J. Curl said to be maximum acceptable.
Damir
p.s. Increasing the input cap(C29) from 220 pF to 470 pf decrease -1 dB HF from 500 kHz to 200 kHz.
Last edited:
Looking forward to testing a built circuit.
-RNM
Look for a board to do just that later today, I am polishing up the board as we speak.
I hope it will do for it intended purpose, "development and testing" not for production and final. Maybe a small step in the right direction.
The OPS is done, working on the IPS.
Attachments
Last edited:
Thanks Keith, I got those done, you are correct. Sharp eyed you are, so says Yoda.
Ready for Review
A few very minor ones.....
R10 dissipates about a constant 0.48W so its layout footprint could be made for a larger power rating one, at least 1W. Of course you can always make fit a big one in there for a non-production, development & testing build. No biggie.
You seem to be missing a 2SC2240, perhaps you overlooked and substituted it with something else like BC550, which may not be able to withstand the operating voltage, as Dadod pointed out at one time. Again this is not anything serious since it is a non production design.
Q8 has wrong pin-out but I guess you can mount the transistor front-side-back to the heat sink. No problems there.
The layout does not seem to match the schematic in that the two A1381 driver transistors have their base hooked up to the wrong pin of the VAS transistor. It probably doesn't worth a mention as it can always be easily fixed with a razor blade, a tiny piece of wire, and some soldering.
Perhaps I do not quite understand what it means by "development and testing", and how it is different from a "production" design.
Oh, I wouldn't put the "CE" mark on a PCB. PCB shop will put safety certificate markings in the silk screen labeling according to the materials being ordered and used and their specific safety certification, which may not be of CE. For example, a PCB with tin-lead HAL finish most popular to DIY community is almost certainly non CE.
Hi Dadod. I have not had big chunk of time for the layouts recently. I'm also attempting to put in some protection on the PCB so that one does not have to rely on an off-board protection. I for one don't have room in the chassis (already made) for a regulated OPS power supply/protection board.
Nattawa
You being the true professional that you are understand the difference between the two, a little humor there, very little. Forgive my totally amateurish rendition of Dadod design, but trust me ANY and all mistakes, oversights and goof ups I might make will be corrected once everyone has had a chance to observe and report them. Then and only then will this board be fabricated for the sole purpose of Development, testing, enhancing, correcting and making the circuit ready for a production of a quantity for many DIY's to build. Take minute and look at some other threads and you will see a strong similarity in the grooming of the circuit and its initial board(s). Your input is needed and wanted, along with others to make this project advance. Maybe you can bring forth your board and SMD's so we can smile too. As you can see I have incorporated several of your previous observations into this board. Again Thanks.
You being the true professional that you are understand the difference between the two, a little humor there, very little. Forgive my totally amateurish rendition of Dadod design, but trust me ANY and all mistakes, oversights and goof ups I might make will be corrected once everyone has had a chance to observe and report them. Then and only then will this board be fabricated for the sole purpose of Development, testing, enhancing, correcting and making the circuit ready for a production of a quantity for many DIY's to build. Take minute and look at some other threads and you will see a strong similarity in the grooming of the circuit and its initial board(s). Your input is needed and wanted, along with others to make this project advance. Maybe you can bring forth your board and SMD's so we can smile too. As you can see I have incorporated several of your previous observations into this board. Again Thanks.