Hi Hugh,
I should have said "the only powerful P-Channel fets available for reasonable price/parameters" 🙂
Have you tried these IXSYS fets? I'm not seeing too many amps built with them..
I remember XRK was using them at some point, but then he went back to IRFPs/FDAs..
FQAs can be bought for $1.60, IXSYS fets better be really good at these prices..
I used FQS, FDA, IRFPs in quasi amps, but in complementary ones - only IRFPs.
I should have said "the only powerful P-Channel fets available for reasonable price/parameters" 🙂
Have you tried these IXSYS fets? I'm not seeing too many amps built with them..
I remember XRK was using them at some point, but then he went back to IRFPs/FDAs..
FQAs can be bought for $1.60, IXSYS fets better be really good at these prices..
I used FQS, FDA, IRFPs in quasi amps, but in complementary ones - only IRFPs.
>Chassis will be (most likely) finished tomorrow.
>Time for music test coming..
Unfortunately - 'LMK Max' (post amp #1320) is terribly oscillating when mounted on the chassis. Zobel resistor gets red hot immediately.
With or without load, with no input signal, or with input shorted to ground.
Both channels behave the same.
It was working fine on the test bench, but it was 'fragile'..
This sucks..
>Time for music test coming..
Unfortunately - 'LMK Max' (post amp #1320) is terribly oscillating when mounted on the chassis. Zobel resistor gets red hot immediately.
With or without load, with no input signal, or with input shorted to ground.
Both channels behave the same.
It was working fine on the test bench, but it was 'fragile'..
This sucks..
Last edited:
160 dB gain is hard to tame. No wonder I used inductance. It's a pity that failure.
Maybe it can be fixed, but it'll not be easy. It works without the chassis,
but it's extremely difficult to play with it and try different things while mounted on the chassis.
I guess it's easier to build a new amp 🙂
but it's extremely difficult to play with it and try different things while mounted on the chassis.
I guess it's easier to build a new amp 🙂
It is possible to fix the work .Throw away the extra parts, and that's it .
The emitter repeater is discarded and the propensity to excite is reduced
Yeah, but I don't need another amp like this.
Already have enough 'normal' LMKs and Wiederholds working perfectly.
I prefer to re-use the chassis for something new..
Chassis is much more work/expenses than amp/boards 🙂
Already have enough 'normal' LMKs and Wiederholds working perfectly.
I prefer to re-use the chassis for something new..
Chassis is much more work/expenses than amp/boards 🙂
Yes, Minek, the IXYS 48A version, which is rated to 462W.
IXYS is more expensive than IRF, as you'd expect, and if you are going to all pmos the best buy is the FQA36P15, rated to 280W. These should be Polar fets, NOT the TRENCH fets, which are not happy with linear use at all.
HD
IXYS is more expensive than IRF, as you'd expect, and if you are going to all pmos the best buy is the FQA36P15, rated to 280W. These should be Polar fets, NOT the TRENCH fets, which are not happy with linear use at all.
HD
I bought FQAs for $1.60 from
tevatronix on eBay
Usually I don't buy semiconductors on Ebay, but this seller is OK.
There is also other, mostly Asian sources:
FQA36P15 Price & Stock | DigiPart
tevatronix on eBay
Usually I don't buy semiconductors on Ebay, but this seller is OK.
There is also other, mostly Asian sources:
FQA36P15 Price & Stock | DigiPart
Well behaved Quasi MOS
I promised to post my folded cascode quasi MOS circuit. Note that this is well behaved:
1. Good rail sticking, little to none.
2. Good THD (0.001 @ 1KHz) at low bias (20mA, 15mA)
3. Clips < 2V from both rails.
4. Includes current limit protection and gate clamps.
5. Easy to stabilize.
This example was done with cheap bjts so the supply is only +/-24 but you could easily go higher with better BJT transistors. At +/-24V you could probably use smaller TO220 FETs, ie real cheap.
I promised to post my folded cascode quasi MOS circuit. Note that this is well behaved:
1. Good rail sticking, little to none.
2. Good THD (0.001 @ 1KHz) at low bias (20mA, 15mA)
3. Clips < 2V from both rails.
4. Includes current limit protection and gate clamps.
5. Easy to stabilize.
This example was done with cheap bjts so the supply is only +/-24 but you could easily go higher with better BJT transistors. At +/-24V you could probably use smaller TO220 FETs, ie real cheap.
Attachments
The idea for R30 is to create a cascode function in Q10 and local feedback around M1 because M1+Q8 have more loss vs the quasi side. It is a partial feedback vs total feedback used in some amps to control idle current like the old "Specrta Sonic 701"
http://www.waltzingbear.com/Schematics/SpectraSonics/SpectraSonics701amp.pdf
I didn't test it much so it may not help. I actually did this a while ago and I forget where I left it. There is a little glitch on the falling crossover that I'm still working on.
http://www.waltzingbear.com/Schematics/SpectraSonics/SpectraSonics701amp.pdf
I didn't test it much so it may not help. I actually did this a while ago and I forget where I left it. There is a little glitch on the falling crossover that I'm still working on.
Improvement
This is better. R3 was a mistake because it causes phase reversal/clipping on positive peaks. And R13/R9 is supply voltage sensitive so I have replaced it with a CCS. The same is true of R8 because the VBE is not rock solid, but this post is about the quasi driver so I just changed R8 for 36V. I changed a couple BJT for higher voltage but note that Q8 is still a cheapo because there is only 8V on it. Note that the 1KHz THD is still ~0.001% and the bias is even lower, almost class-B.
This is better. R3 was a mistake because it causes phase reversal/clipping on positive peaks. And R13/R9 is supply voltage sensitive so I have replaced it with a CCS. The same is true of R8 because the VBE is not rock solid, but this post is about the quasi driver so I just changed R8 for 36V. I changed a couple BJT for higher voltage but note that Q8 is still a cheapo because there is only 8V on it. Note that the 1KHz THD is still ~0.001% and the bias is even lower, almost class-B.
Attachments
How about replacing the upper bootstrap with CCS as well?
And then, using an op-amp instead of LTP 🙂
Actually TL072 might be cheaper than LTP.. and you are getting JFET input, perfectly matched.
I wonder if I could use biasing, lower driver+ccs, and upper driver from your schematic, and plug it into one the op-amp amps from this thread..
And then, using an op-amp instead of LTP 🙂
Actually TL072 might be cheaper than LTP.. and you are getting JFET input, perfectly matched.
I wonder if I could use biasing, lower driver+ccs, and upper driver from your schematic, and plug it into one the op-amp amps from this thread..
Last edited:
Or even using a DN2535 as a CCS.........
That drive on the quasi inverter is very clever, I really like the improvement of loop gain it gives the half cycle. Congratulations, SteveEU.
Hugh
That drive on the quasi inverter is very clever, I really like the improvement of loop gain it gives the half cycle. Congratulations, SteveEU.
Hugh
I was checking current in source resistors, there is little kink,
not sure if this is big deal..
But it gets worse at 10kHz.. Cross-conduction reaches 300mA at some point...
not sure if this is big deal..
But it gets worse at 10kHz.. Cross-conduction reaches 300mA at some point...
Attachments
Last edited:
improvements
Ya, a VAS CCS is better in a couple ways. Mostly I was concerned about supply voltage sensitivity. But I was trying to keep it simple. I would probably still bootstrap the CCS in order to drive M1 gate above the +rail. The LTP could use a CCS too.
You could replace the whole IPS with a folder cascode VAS and op-amp but you would probably have stability problems. You are better at that than I am. I usually give up and stick with a LTP. CFA is also an option.
You can tweak the idle current and R9 etc to optimize the XO. It's not perfect but better than many amps. I will continue to fiddle with it but I thought the CCS vs R3 was a significant improvement that I should share.
This driver circuit is "fast enough" but not a slew champion. I think we know how to improve that but do I want/need to?
Ya, a VAS CCS is better in a couple ways. Mostly I was concerned about supply voltage sensitivity. But I was trying to keep it simple. I would probably still bootstrap the CCS in order to drive M1 gate above the +rail. The LTP could use a CCS too.
You could replace the whole IPS with a folder cascode VAS and op-amp but you would probably have stability problems. You are better at that than I am. I usually give up and stick with a LTP. CFA is also an option.
You can tweak the idle current and R9 etc to optimize the XO. It's not perfect but better than many amps. I will continue to fiddle with it but I thought the CCS vs R3 was a significant improvement that I should share.
This driver circuit is "fast enough" but not a slew champion. I think we know how to improve that but do I want/need to?
Last edited:
>I wonder if I could use biasing, lower driver+ccs, and upper driver from your schematic, and plug it into one the op-amp amps from this thread..
I tried. Mixed results so far; it works but has it's own issues. Will play with it more..
I tried. Mixed results so far; it works but has it's own issues. Will play with it more..
Minek,
The LT1056A is now obsolete. Do you have any ideas of replacements?
I'm thinking the OPA1656 - another fet input with high loop gain and very low noise.
The LT1056A is now obsolete. Do you have any ideas of replacements?
I'm thinking the OPA1656 - another fet input with high loop gain and very low noise.
Are you sure it's obsoleted?
Mouser.com still has it, and there is no mention of it being obsolete.
https://www.mouser.com/ProductDetail/Analog-Devices/LT1056CN8PBF/?qs=ytflclh7QUWa9djvSkefEg%3D%3D
Here are similar op-amps I've been trying in the past (all JFET):
TLE2071
TLE2081
LT1357
LF356
OPA134
To be honest, I'm usingLT1056 in all sims, because its model is included in LTSpice, so everyone can run sims, but in actual builds I usually use TL071.
I tried all these op-amps, and I don't see/hear/measure ANY difference in the final amp..
So far I built 10 different amps with op-amp input, and LT1056 was actually used in only 1 amp..
Update: Digikey also shows it as 'active'.
Mouser.com still has it, and there is no mention of it being obsolete.
https://www.mouser.com/ProductDetail/Analog-Devices/LT1056CN8PBF/?qs=ytflclh7QUWa9djvSkefEg%3D%3D
Here are similar op-amps I've been trying in the past (all JFET):
TLE2071
TLE2081
LT1357
LF356
OPA134
To be honest, I'm usingLT1056 in all sims, because its model is included in LTSpice, so everyone can run sims, but in actual builds I usually use TL071.
I tried all these op-amps, and I don't see/hear/measure ANY difference in the final amp..
So far I built 10 different amps with op-amp input, and LT1056 was actually used in only 1 amp..
Update: Digikey also shows it as 'active'.
Last edited: