Thanks mcd.
I've tried the OPS but I'm not very successful. In order to get it stable initially I had to up all the Cdoms significantly. When I had it stable enough I went to set the OPS bias. THis is where I run into weird stuff. It remains 0A, no matter how I change the bias resistors. It may be that the slight difference in supply voltages has upset the OPS, although there's only two CCS really that would need readjustment.
I've tried the OPS but I'm not very successful. In order to get it stable initially I had to up all the Cdoms significantly. When I had it stable enough I went to set the OPS bias. THis is where I run into weird stuff. It remains 0A, no matter how I change the bias resistors. It may be that the slight difference in supply voltages has upset the OPS, although there's only two CCS really that would need readjustment.
Shame, the difference in the supply voltages may have upset the driver bias current and the bootstrapping. The voltage at the driver collectors is normally about 12v. What was the driver bias current? In my design it is approx 12mA. Please could you post the failed asc file if you still have it? I'll try and find some time to have a play over the next week. The two CCS are set at 10mA.
Thank you for trying it.
Thank you for trying it.
Hi MagicBox,
Are you sure your "cherry only" and "TPC only" plots from post #174 are labelled the correct way around?
"TPC" should strictly speaking be called "TPOZC" (two-pole, one zero compensation) because it entails having two poles followed by one zero, to bring the loop-gain slope back to -20 dB/decade in the vicinity of the ULGF. Your "cherry only" plot has a zero, but the "TPC only" plot does not.
Are you sure your "cherry only" and "TPC only" plots from post #174 are labelled the correct way around?
"TPC" should strictly speaking be called "TPOZC" (two-pole, one zero compensation) because it entails having two poles followed by one zero, to bring the loop-gain slope back to -20 dB/decade in the vicinity of the ULGF. Your "cherry only" plot has a zero, but the "TPC only" plot does not.
Using 2EF output stage ...
Attached your VAS with TMC compensation and my 2EF outputstage seems to be a good starting point for further optimizations.
THD20k@100W@8R:
0.0002% / PM 63 degree / GM 17 dB
BR, Toni
P.S.: SOA protection has to be recalculated!
Attached your VAS with TMC compensation and my 2EF outputstage seems to be a good starting point for further optimizations.
THD20k@100W@8R:
0.0002% / PM 63 degree / GM 17 dB
BR, Toni
P.S.: SOA protection has to be recalculated!
Attachments
Last edited:
Hi Harry,
Yeah I labeled them correctly. A side note though is that the TMC network isn't as 'strong' as you'd dimension it when used as the only form of compensation (caps and resistors are smaller in my case). Maybe that causes the slope to be more unfamiliar.
@mcd99uk
Here's the asc file with your OPS. Forgive my attempts to making it work. As it is now, you don't need to modify the compensation networks.
Yeah I labeled them correctly. A side note though is that the TMC network isn't as 'strong' as you'd dimension it when used as the only form of compensation (caps and resistors are smaller in my case). Maybe that causes the slope to be more unfamiliar.
@mcd99uk
Here's the asc file with your OPS. Forgive my attempts to making it work. As it is now, you don't need to modify the compensation networks.
Thank you!
Nothing to forgive. I'll try and get it working. If I succeed I'll post the asc file. I think the zero bias current would be messing up the PM/GM margins. Have noticed that within reason upping the bias current in my amplifier design improves the PM.
With all these different OP stage experiments (successful or not) I think we will all learn something.
Something must be wrong somewhere, a outputstage with a diamond topology will always outperform a EF2. Try using same outputs, BJTs instead of Fets.
Sorry guys, I haven't been able to experiment more the past day. Tomorrow I can work some more on a different OPS. I think it's too early to discount the 'simple' buffered mosfet stage. One issue with a more advanced OPS is that there are even more devices in the Cherry loop and that may get problematic. More experimenting to come.
And yes it's all a learning experience
And yes it's all a learning experience
I have a few ALFET ALF16N16W/ALF16P16W parts that I could use for this application, that and a spare/used Pioneer SX-1250 toroid (38Vrms/8ohm), working on the other necessary bits as well.
Any spice models avail for these fets? ALF16N16W/ALF16P16W, I have yet to figure out how to make my own models.
Either way of OPS, the basic pcb layout would be similar. I hope that this is still the end result, a pcb that we can obtain somehow or do ourselves.
Rick
Last edited:
Hi Harrison,Nice Job. Magic box. As the focus is the vas, have you tried to substitute that front end for a diamond
kind regards,
Harrison.
No I have not tried a diamond front-end yet. I've designed this VAS to be married exclusively to a single LTP. A diamond (or double LTP) lets me end up with two forward paths while the phylosophy behind this VAS is to eliminate having two forward paths. This makes for easy compensation schemes of all sorts, amongst making Cherry feedback work well, because there's just one forward path. But, that's the strength of this VAS: While it has a single ended input, it has a full symmetrical push-pull output which is what I wanted for the VAS output.
So.. I don't think I'm going to have a diamond input. In a sense it would then become two VAS' in series if the diamond input would provide a single-ended output.
Meanwhile I got the diamond output to work as you gave it to me. I made some stupid mistake at first, which is why I couldn't set bias. Had to do with supply net labels.
Anyways, I have not been able to get the OPS stable in conjunction with my VAS. The OPS causes a particular gain spike at a little over ULGF, with a severe phase twist which I just couldn't remove at all, no matter the compensation efforts on both the VAS and the OPS itself. I won't say it's impossible, but with all methods I know at my disposal wasn't enough.
The main reason I think is that there are too many devices in the Cherry loop this way. From that perspective, it seems acceptable that a 2EF would perform better; it would be more direct path.
I can give the 2EF stage a try still, but I'm leaning more and more towards the Exicon MOSFETs in a simple buffered OPS. Using very low gate stoppers together with a buffer that can supply the associated peak-current without making the VAS limp is what I would say a very fast OPS.
And that's what this VAS needs. A fast output stage. But also one that doesn't perform any EC itself, as the Cherry loop already takes care of that and active EC in the OPS may turn out to be counter productive.
What do you guys think?
Those trannies would do well yes. I will design a PCB once I finished the amp design as a whole. Right now, I'm still open to an OPS implementation to see what gives best performance for the circuit. Once I made a personal choice, I'll make a PCB and BOM for it
The project will start in a new thread though.Magic,
Yes, it is difficult to get your VAS stable with a diamond. Had noticed the phase shift and peaking you mentioned. I was thinking to myself that this VAS is hard to tame never considered it was an op stage characteristic. Not seen it my own design. So from my amateur perspective it could be the cherry loop. Not giving up yet though. I can deal with the spike but the phase twisting is proving more interesting. But I have managed some stability so far.
If THD is the priority BJTs seem have the edge over MOSFETs at least in simulation. Why do you say that op stage speed is important?
Yes, it is difficult to get your VAS stable with a diamond. Had noticed the phase shift and peaking you mentioned. I was thinking to myself that this VAS is hard to tame never considered it was an op stage characteristic. Not seen it my own design. So from my amateur perspective it could be the cherry loop. Not giving up yet though. I can deal with the spike but the phase twisting is proving more interesting. But I have managed some stability so far.
If THD is the priority BJTs seem have the edge over MOSFETs at least in simulation. Why do you say that op stage speed is important?
It's because at HF, the outputstage is included in the VAS local feedback. That's what the Cherry cap does, it opens a path at HF that 'dampens' changes from the IPS as it's an inverse version of the output signal. Feedback taken from the VAS is the output signal + OPS distortion. That's why Cherry does so well on the THD side of things, because the VAS is compensated with the clean output signal rather than the output signal with OPS distortion.
And that's how I can see that OPS's with own EC would improve THD over regular OPS's when Cherry feedback is not used.
At least, that's my take on looking at all these experiments
And that's how I can see that OPS's with own EC would improve THD over regular OPS's when Cherry feedback is not used.
At least, that's my take on looking at all these experiments
Actually one can use just one of the forward paths of a diamond buffer although I dont see how it could benefit this vas but you never know. Using a diamond buffer in this way has been seen and Sony used it 10 years back in their top of the range amps together with a cascoded vas.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.