Indeed? Those 1Ks in each cathode are higher than the cathode impedances at those points, if I understand correctly. They render the action of the CCS as 'infinitely' high common impedance and balance creator useless - you are closer to simply a common cathode resistor than any compensating action a CCS might have. I would respectfully suggest that you refresh your memory with a look at the associated formulae, in which a common cathode resistor (Rc) features. It will show the well-known difference necessary in the anode loads which you mentioned earlier for finite Rc, and also that should Rc approach infinity, balance with equal Ra's is perfect (well, let us not nit-pick regarding stray Cs etc!).
I will leave it to the industrious SY to 're-invent' the wheel here!
Then:
It is more difficult to disagree with Eli (no disrespect to Geek1) on the presence of grid stoppers here. I respect his experience and he has helped me in PMs on several occasions.
But having said that, Eli (and meant it!) - the fact that a tube is made for use at VHF under specific conditions, does not mean that it will behave "VHF-fy" under totally different conditions. I get the feeling that general comments regarding grid stoppers are all too often 'fuelled' by r.f. experiences and not audio practices. With Eli I am more careful, but I can honestly state that after much tube experience (I come from the era - ouch
) I have never found any necessity for stoppers in these applications. Appliccable inter-electrode capacitances and using R's for loads contra-indicate this. I do not know if this could be different on locations in the vicinity of radio stations; I am talking mainly of internal tendencies to go unstable. I am also used to r.f. work, but my audio investigations up to 1MHz for certain research has never revealed any tendency to deviate from a smooth roll-off - the first indication of possible instability - in voltage amplifiers. (In some cases with transistors - but that is irrelevant here.)Not to make too much of this; I would hate to cause offence over two measly 10c resistors
, and perhaps Eli can broaden my horizons with his experience of which he has quite some.THEN!
Apology if this is getting long ... but what am I GROSSLY missing?? To trim a 1K resistor for 'fine balance of output' in series with a CCS of some 100K+ in an LTP in audio?? (See above-mentioned formulae.)
Now I am worried. Where in &%^$#@* does this come from, and what learned Einstein-baffling dissertation is provided to support the same??
Uh-huh
The 470K is bypassed at AC, so you're looking at it being signal --> 47K --> gate. Now see why this only works with MOSFETS with small gates...
I like using this with the CCS on the Pinkmouse board, but the higher Z of the gate gave this one an edge...
The 1K resistors in those cathodes are designed to be modded to adjust the overall sensitivity of the circuit.
The 1K value was actually granfathered from the Kevin Kennedy design, which performs quite well, but uses differing anode load resistors.
The CMFB+CCS is maintaining AC balance and performance throughout the spectrum of reasonable cathode resistor values from 0 to ~1.5K.
And FYI - formula gives me the ballpark, the rest I design on pure sound.
Cheers!
Last edited:
Respecting your right to do so (and not pursuing this much deeper as it is off-topic!) - but one has a problem when someone's ears tell him that Ohm's Law no longer works ....?
The equation for the ratio of the respective outputs is:
A1/A2 = 1 + [(Rl + ra)/{Rk(µ + 1)}] with Rl, Rk and ra being the usual parameters.
From which it is evident that, when Rk is sufficiently high, the term in [ ] can be ignored and the balance is perfect. Such a condition pre-supposes that the CCS is a C-CS, thus really a very high value, not just a simple BJT device. It would also pre-suppose that the two µ-values are not that far apart (say 30%; if it is more with any twin triode, time to throw the tube away!). If for some heroic reason one wants to pair a high gm pentode with a 12AU7, this equation will become more complex. That is not usual practice. (The Radford phase inverter doing that is not a LTP; the analysis is different!)
But enough of that here - back to topic!
You're right. It's gone. Must have been a bad solder joint... I'll fix on the next rev.
That will be the same revision that includes all the remaining things that need to be updated; all the missing resistor values, the topology decisions, etc. assuming of course that someone bothers to tell me what those items would be.
It's getting frustrating to see this design stalled at 99% done. TubeMack is itching to buy parts and start generating some smoke.
..Todd
Geek,
may I ask where is the scope picture with only CCS and no CMFB, did you ever check that?
Driver Board for the ST-70 Using Common Mode Feedback PI
Cheers Michael
may I ask where is the scope picture with only CCS and no CMFB, did you ever check that?
Driver Board for the ST-70 Using Common Mode Feedback PI
Cheers Michael
FWIW, I subscribe to the notion that parasitic oscillation is presumed to be present in DIY projects and that appropriate suppression techniques are in order. Yes, I call for 100 Ω stoppers on 'X7 grids.
The late John (Buddha) Camille had access to instruments we can only fantasize about. He noted that high gm VHF types could exhibit parasitic oscillation in the hundreds of MHz. I take that caveat quite seriously.
Perhaps the very worst culprits are the Russian 6С45П and the 5842. Those types were designed for grounded grid RF service and extraordinary precautions are absolutely necessary, when they are set up as common cathode gain blocks. I say only 1/2 in jest that the 6С45П and 5842 come annoyingly close to oscillating in the cardboard box.
The late John (Buddha) Camille had access to instruments we can only fantasize about. He noted that high gm VHF types could exhibit parasitic oscillation in the hundreds of MHz. I take that caveat quite seriously.
Perhaps the very worst culprits are the Russian 6С45П and the 5842. Those types were designed for grounded grid RF service and extraordinary precautions are absolutely necessary, when they are set up as common cathode gain blocks. I say only 1/2 in jest that the 6С45П and 5842 come annoyingly close to oscillating in the cardboard box.
Not eally, U.T. - as I said somewhat earlier, Miller effect in the first triode quashes that; in the second triode .... nah, not there either. (See earlier, as you probably have.)
The 5th RIAA time constant by JP?
(I gather that you are probably aware of the 4th "unknown" RIAA time constant at 50 kHz.)
Heavens no! But I notice no sooner had the world become aware of the '4th time constant', or it is splashed as being included in playback designs! (What next to sell...) That time constant at 50kHz actually has little to do with amplifiers. It was devised by Neumann as I understand, to simplify recording matters up there. It has no effect (unless one split hairs - see later) on what is audible. In NFB equalisation there is usually a time constant at about 19 kHz otherwise rising NFB can cause the network to oscillate with NFB going ever higher.
Thanks for the link on RIAA - but brother; does that fellow love maths! Any uninformed reader will really think that this RIAA business requires MIT honours qualifications or Mensa membership. Jeepers! I do two simple calculations, one at 500Hz and the other at some 2,2kHz, and then trim in on breadboard. Folks are obsessed with the nearest 0,1dB, when their loudspeakers hobble up and down over the audio realm by several dB, apart from a cartridge's unsmooth response. As for audiophiles 'preferring not NFB' - once again a rather vocal minority waving their little flag.
Regarding rumble filtering (a term all with expensive turntables hate because it might show that they spent their thousands in vain), a lame single RC down below will have little effect. As said before with little bother and choosing say 24 Hz as the peak frequency (there's not really much there on vinyl), one can be down to 2dB at 10Hz in a straight line. Components expenditure say $5; turntable extra expenditure for the same effect, say >$500. I mercifully never had any desire 'to keep up with the whoevers'.
Last edited:
No worries, i'm just spending my time reading my Morgan Amp book, and buying more test equipment.
For what I have spent on this test gear, I think I could have bought new CJ gear, but where's the fun in that? Just unboxing a pristine HP 339a right now.
What a deal I got on it!
Last edited:
Hi Michael,
Yeah, I'm going to have to redo them because other than the last one in the thread, those are the ones I'm missing. I was taking A/B pics... one to CMFB point and the other to a straight supply divider to reflact the same voltage level, but no correction signal.
@ Johan Potgieter - Yeah, we're drifting off topic a wee When I have time, I was going to make a blog post solely on the LTP PI w/CMFB and we can look at this in real depth... the mathematician's can show us the *why* as to what we are hearing
@ SY - The 47K are really considered parallel, so that's 23.5K. The signal to the gate is really high impedance and the 10K is practically to ground because of the ring-of-two being 100% NFB.
Why do I have a feeling we are in the same book, but aren't on the same page?
Cheers!
Yeah, I'm going to have to redo them because other than the last one in the thread, those are the ones I'm missing. I was taking A/B pics... one to CMFB point and the other to a straight supply divider to reflact the same voltage level, but no correction signal.
@ Johan Potgieter - Yeah, we're drifting off topic a wee
When I have time, I was going to make a blog post solely on the LTP PI w/CMFB and we can look at this in real depth... the mathematician's can show us the *why* as to what we are hearing @ SY - The 47K are really considered parallel, so that's 23.5K. The signal to the gate is really high impedance and the 10K is practically to ground because of the ring-of-two being 100% NFB.
Why do I have a feeling we are in the same book, but aren't on the same page?
Cheers!
I routinely scan my circuits with a 0 to 1250 MHz spectrum analyzer. I can't speak for the Russian tube, but I can tell you that the 5842 will oscillate if you give it even half a chance. Many other high Gm tubes can be guilty also, and often the oscillation is not obvious. The amp just sounds grainy.
What makes you think that they are not oscillating when you aren't watching?
There has been some debate over the balance of the LTP and the need for a means to force balance. I have been using LTP's for years. They are my favorite splitter when tube count is not the primary design factor. I have found that a simple resistor in the tail is sufficient IF there is a large voltage drop across it compared to the signal voltage. A resistor will however not give perfect balance, only a "perfect CCS" can do that. In the real world a 10M45 will give balance that is better than my measurement capability out to past 250 KHz. It is better and simpler than the ring of two BJT's. More comples CCS's may measure better and may be useful for plate load circuits, but aren't needed here.
There are however a few "imbalancers" that I had to find the hard way. The wiring (or PCB layout) should be as symmetrical as possible. The two grid stoppers should be identical and as small (in value) as possible to prevent oscillation. The two plate load resistors should also be identical and matched as close as possible. I pick 2 out of a package of 10. The biggie, the cap from the second grid to ground (C14 in your schematic) should be a high quality cap (same type as the coupling cap). A cheap cap here will lead to all sorts of high frequency anomalies.
I use two LTP's connected together without coupling caps. I do need a means to adjust the DC balance, but AC balance is perfect. 50 KHz square waves lie on top of each other on the scope screen (one channel inverted in the scope).
@ SY - The 47K are really considered parallel, so that's 23.5K. The signal to the gate is really high impedance and the 10K is practically to ground because of the ring-of-two being 100% NFB.
Why do I have a feeling we are in the same book, but aren't on the same page?
Because you're claiming a measurement of an imbalance in a CCSed long tail pair. That means that either your circuit variation is causing this (probably from mismatch of the FB resistors; I don't see the same imbalance in the Curcio ST-70 board which has a very similar scheme) or your measurement is flawed. In any case, it's worthwhile for me to demonstrate the at-first-glance nonintuitive idea that tubes in a conventional LTP with a CCS tail can be grossly mismatched without affecting the balance a bit. It's probably worth starting another thread on this, and after I take the scope photos, I'll do that.
R20 is a gate stopper. It does the same thing that a grid stopper does for a tube. Use a 1K resistor.
There is no mention of this on the IXYS data sheet but trust me it is needed. I started experimenting with the 10M45 when it first came to the market about 5 years ago. When a tube oscillates it can appear normal to audio, distort, misbehave in other ways including high current, red plate, and general instability. When a semiconductor device operating on several hundred volts oscillates it just blows up. I had several just go away before I discovered the stopper resistor. The only ones that I heve blown since then have been due to my own stupid mistakes.
IXYS claimed that the 10M45 is a special CCS IC. Many believe that it is just a depletion mode mosfet. After successfully using it as a mosfet, I am now inclined to believe this. It is possible to use a Supertex DN2540N5 in this same circuit if the IXYS part isn't available.
Recomended output transformer
Dear 5-20 builders
I read your threads and I would warmly recomend you Sowter Output transformers. I think for your aplication you need the UO27 5K primary, 50W
output power. You can ask for UL taps, Mr. Sowter will be ready to help.
His products are not cheap. This transformer is 169 GBP per unit, but it's worth every penny.
I recently bought a pair of output transformers for Antique Sound Lab
AQ1008DT monoblocks and the result are excellent.
Regards and good luck
Tigro
Dear 5-20 builders
I read your threads and I would warmly recomend you Sowter Output transformers. I think for your aplication you need the UO27 5K primary, 50W
output power. You can ask for UL taps, Mr. Sowter will be ready to help.
His products are not cheap. This transformer is 169 GBP per unit, but it's worth every penny.
I recently bought a pair of output transformers for Antique Sound Lab
AQ1008DT monoblocks and the result are excellent.
Regards and good luck
Tigro