The SRPP as a LTP to balance any variation in the input signal was an interesting idea...
According to LTSpice, a variation in the input voltage to the inverting input balances out in the end. Same concept that is used to balance the bias LED's should work here and it appears it does.
According to LTSpice, a variation in the input voltage to the inverting input balances out in the end. Same concept that is used to balance the bias LED's should work here and it appears it does.
I designed this to use my equipment only... I have an Yggdrasil DAC which is fully balanced input to output. In keeping with minimizing gain stages, I built a JFET passive preamp using Nelson Pass's design, modified for fully balanced. I used a pair of 23-position Electroswitches for the channels with 0.1% resistors ordered in batches of 50 that I picked out of exact matches. Even though nothing is certain, I am comfortable that I will not see mismatched +/- voltages on the input.
I posted below the SRPP's tied in LTP with a ccs... and it appears to correct for imbalance. If this tests out on the prototype...
Umm the top bit is CCS and the bottom too could be tricky to get the mid voltage correct. You could use a current mirror referenced to the mid point. Think I would go CCS LPT with passive load and a cathode follower - still the same number of tubes. However that's quite conventional.
You now have Approximately Infinite gain! You will need to at least remove C3 and C4 -- the more perfect a CCS is (the Plate loads), the closer the stage gain approaches infinity.
Unless you can apply a Monte Carlo analysis in LTSpice (don't ask me -- I don't know how . . 😉 ), the design will balance perfectly, because the components are exact matches! I might be completely wrong, but I'm pretty sure you don't need that much gain.
I'm also not sure you're getting the full 'figure-of-merit' out of I1. The 220R common-Cathode resistor is surely not showing its true colors, fed from a current source. At the very least I'd recommend changing the ratio between it (R8) and R7 and R13.
It would help some if you included the files LTSpice emits -- squinting at a screen cap trying to sort out what you're doing takes a little extra patience; but it's OK. 🙂
Best Regards
edit: Almost completely forgot: The 2N5551's comparing output Cathode currents may be dissipating too much power. Unless I bodged the arithmetic, if one of the tubes takes off, its transistor may need to dissipate 1,25W ! (And that's if the bias supply holds up -- unlikely . .) Also, according to "KT120_FIT OUT_v006", sucked down 17:00-something, 28May, you have no current-setting resistors for Q202, and Q203 and the LED's they drive -- either transistor conducts heavily AND THAT TOO will suck down the bias supply.
Unless you can apply a Monte Carlo analysis in LTSpice (don't ask me -- I don't know how . . 😉 ), the design will balance perfectly, because the components are exact matches! I might be completely wrong, but I'm pretty sure you don't need that much gain.
I'm also not sure you're getting the full 'figure-of-merit' out of I1. The 220R common-Cathode resistor is surely not showing its true colors, fed from a current source. At the very least I'd recommend changing the ratio between it (R8) and R7 and R13.
It would help some if you included the files LTSpice emits -- squinting at a screen cap trying to sort out what you're doing takes a little extra patience; but it's OK. 🙂
Best Regards
edit: Almost completely forgot: The 2N5551's comparing output Cathode currents may be dissipating too much power. Unless I bodged the arithmetic, if one of the tubes takes off, its transistor may need to dissipate 1,25W ! (And that's if the bias supply holds up -- unlikely . .) Also, according to "KT120_FIT OUT_v006", sucked down 17:00-something, 28May, you have no current-setting resistors for Q202, and Q203 and the LED's they drive -- either transistor conducts heavily AND THAT TOO will suck down the bias supply.
Last edited:
Oh, oh, and the electrolytics filtering the Base signals to Q202, an Q203, will be subject to reversed polarity . . they're 6,3V parts with the '+' to the Base of two transistors that have their Emitters biased to -18V ! If you keep them (and I'm not sure that is a good idea), the '-' terminal should be connected to either the Emitters, or the -Vee supply that feeds them.
I think baudouin0 is spot on -- it'll be hard to balance!
In sum, there are a whole host of oopsi-doodles to explain the difference between the 1st amp and this one. Keep posting, we'll sort this out!
Cheers
I think baudouin0 is spot on -- it'll be hard to balance!
In sum, there are a whole host of oopsi-doodles to explain the difference between the 1st amp and this one. Keep posting, we'll sort this out!
Cheers
Last edited:
Oh, and another thing (if I haven't already gotten on your last nerve 😉 ), you can dead-spot-on match R220 and R221 -- to even greater precision than the selected-from-50-count 0,1% hand-matched -- and the errors will still be swamped by the Vbe and hFE mismatch of QQ202 and Q203! Sorry, but passives are still much cheaper to produce to 0,1% tolerances than semiconductors.
Last edited:
I've seem an amp where a LTP input stage drives two SRPP stages. There's then enough gain to apply global NFB to the other input of the LTP.
O.K., sorry -- I just found R222, the 680R that controls current for the Cathode balance LED's.😳
Ya' know, there are much simpler, easier, cheaper ways to do the switching you're doing without the $6, $8, and $15 opto-isolated MOS 'relays'.
Cheers
Ya' know, there are much simpler, easier, cheaper ways to do the switching you're doing without the $6, $8, and $15 opto-isolated MOS 'relays'.
Cheers
Last edited:
Couple o' other bits, if you aren't already sick of me (guess I have gotten a little too caught up in your project!).
The extra precision of the selected R220 / R221 still can't overcome a difference between R114 / R115 -- unless you're planning on also selecting those from a 50-device pool of 0,1% parts. Then the only error term remaining will be the Vbe and hFE of Q202 / Q203 -- which could be considerable.
Almost any op-amp whose common-mode input range includes its negative supply would be considerably more accurate. And it would avoid involving the -80V bias supply. A separate 'bias loss detector' circuit might be a good thing -- since you already have HT switching.
By the way, R214 will smoke if anything like the -18V shown on the schematic is really on the Q202 / Q203 Emitters. 60V across 1k8 is 2 watts!
Best Regards
The extra precision of the selected R220 / R221 still can't overcome a difference between R114 / R115 -- unless you're planning on also selecting those from a 50-device pool of 0,1% parts. Then the only error term remaining will be the Vbe and hFE of Q202 / Q203 -- which could be considerable.
Almost any op-amp whose common-mode input range includes its negative supply would be considerably more accurate. And it would avoid involving the -80V bias supply. A separate 'bias loss detector' circuit might be a good thing -- since you already have HT switching.
By the way, R214 will smoke if anything like the -18V shown on the schematic is really on the Q202 / Q203 Emitters. 60V across 1k8 is 2 watts!
Best Regards