@Tubelab_com may I ask you why the same 12AT7 works at higher current on your SSE project compared to the SPP? Is it due to the capacities it has to drive, or something more?
Thanks
Roberto
Thanks
Roberto
To first approximation, the first stage anode voltage = phase splitter grid voltage ~= phase splitter cathode voltage = Vk. The current through the phase splitter I is then calculated by ohms law as Vk/R where R = phase splitter load resistance. The phase splitter anode voltage Va then = B+ minus I x R, again by ohms law. The voltage across the phase splitter triode (Va - Vk) can be calculated as B+ minus 2Vk. So its practically constant and independent of the load resistance, whereas the current reduces as load resistance increases. If you then go and look at the 12at7 curves and draw a vertical line representing the constant voltage, you will see that as current decreases (by increasing the load resistance) you require more negative bias.
So say first stage anode is 75V, Vk is close to 75V and B+ is 300V, the Va-Vk is 300 - 2x75 = about 150V. With a 24K load resistor, current is 75/24000 = 3.125 mA. From the curves, bias required for 150V and ~3 mA is about -1.95V (although your actual tubes don't follow the curves well). If you doubled resistance to 48K, current is 75/48000 = ~1.6 mA. Va - Vk is still about 150V. Bias from the curves is then about -2.5V = more negative.
There are no "actual voltages", even with tubes of the same make. You can only say "with this particular tube I get...".
By principle a cathodyne PI is self-biasing via cathode degeneration/NFB, no matter if it's DC coupled or if the grid voltage is povided by a voltage divider.
Means that the cathode voltage always self-adjusts to the grid voltage + the positive value of the Vgk necessary for the plate current.
According to tube specs a real 12AT7 should have a plate current around 8mA when Vgk is positive.
The cathode voltage measured by the OP indicates a plate current of only 4mA.
In other words the tube used doesn't conform to the 12AT7 datasheet and shouldn't be marketed as a 12AT7.
Means that the cathode voltage always self-adjusts to the grid voltage + the positive value of the Vgk necessary for the plate current.
According to tube specs a real 12AT7 should have a plate current around 8mA when Vgk is positive.
The cathode voltage measured by the OP indicates a plate current of only 4mA.
In other words the tube used doesn't conform to the 12AT7 datasheet and shouldn't be marketed as a 12AT7.
Last edited:
tikiroo - thank you so much for this!! This explains it in simple terms and provides a theoretical basis for some experimentation now. To reflect it back, when we increase the load resistors R107 & R108, the voltage across the tube (Va - Vk) does not change significantly (I'd wondered about that) and with the reduced current, the tube now operates around the lower of the red dots on the Plate Characteristics as curve pictured below, rather than the higher red dot as before:
Simon
Hi again folks - I may be in need of some knowledgeable guidance again - to help address what might have been weird behaviour?
With the Psvanes 12AT7s in, and having raised B+ to 320V (giving 310V on the back of R109) I seemed to have good bias on both channels, around -1.5 to 1.6V. I wanted to try the Mullards again and found I had some funny business with voltages.
Mullard 12AT7s
With a digital multimeter probe measuring pin 3 Cathode voltage in the right channel, connecting/disconnecting anything to Grid pin 2 changed the Cathode voltage. For example, with digital multimeters on both Cathode and Grid, the Cathode was at 92.9V (and Grid at 92.3V). Replacing the Grid probe with a 60 cm wire dropped the Cathode to 92.7V. A short lead on the Grid dropped the Cathode to 90.4V (and the music got louder), and a screwdriver, 89.1V, and then 88V with nothing on the Grid and the music distorted. The left channel did not display this behaviour and the bias seemed to stay good.
Psvane 12AT7s
Back to Psvane 12AT7s and I observed this weird behaviour continue. Probing both the Cathode and Grid in the right channel, the bias seemed consistently good - about 83.4 and 81.8 = -1.5V. Swapping the Grid probe for a short wire reduced the Cathode voltage to 83.8V and then to 83V with nothing connected to the Grid. This didn't happen in the left channel where I had 83.4V on Cathode and 81.3V on Grid and the Cathode seemed to go up 0.1V when the Grid probe was removed.
Other symptoms in Right Channel
The right channel seems much more susceptible to scratchiness and hum from touching the grid or cathode with a probe, or insulated screwdriver.
Measuring the bias directly with the probes of a single digital multimeter on Grid and Cathode, seemed to give quite different readings compared to using two different multimeters measuring to Cathode and Grid voltages simultaneously and doing the subtraction, in the right channel - but apparently no so in the left.
I suspect something like this was going on all the time, and affected some of the measuring I was previously doing.
Guidance Requested
I think this is what I need to know: What is best practice for measuring bias? And what might going on with the PI bias being badly affected by connections to the Grid, seemingly, only in the right channel (I suppose a wiring error, out-of-spec component or dry solder joint - but where is it likely to be)?
Simon
With the Psvanes 12AT7s in, and having raised B+ to 320V (giving 310V on the back of R109) I seemed to have good bias on both channels, around -1.5 to 1.6V. I wanted to try the Mullards again and found I had some funny business with voltages.
Mullard 12AT7s
With a digital multimeter probe measuring pin 3 Cathode voltage in the right channel, connecting/disconnecting anything to Grid pin 2 changed the Cathode voltage. For example, with digital multimeters on both Cathode and Grid, the Cathode was at 92.9V (and Grid at 92.3V). Replacing the Grid probe with a 60 cm wire dropped the Cathode to 92.7V. A short lead on the Grid dropped the Cathode to 90.4V (and the music got louder), and a screwdriver, 89.1V, and then 88V with nothing on the Grid and the music distorted. The left channel did not display this behaviour and the bias seemed to stay good.
Psvane 12AT7s
Back to Psvane 12AT7s and I observed this weird behaviour continue. Probing both the Cathode and Grid in the right channel, the bias seemed consistently good - about 83.4 and 81.8 = -1.5V. Swapping the Grid probe for a short wire reduced the Cathode voltage to 83.8V and then to 83V with nothing connected to the Grid. This didn't happen in the left channel where I had 83.4V on Cathode and 81.3V on Grid and the Cathode seemed to go up 0.1V when the Grid probe was removed.
Other symptoms in Right Channel
The right channel seems much more susceptible to scratchiness and hum from touching the grid or cathode with a probe, or insulated screwdriver.
Measuring the bias directly with the probes of a single digital multimeter on Grid and Cathode, seemed to give quite different readings compared to using two different multimeters measuring to Cathode and Grid voltages simultaneously and doing the subtraction, in the right channel - but apparently no so in the left.
I suspect something like this was going on all the time, and affected some of the measuring I was previously doing.
Guidance Requested
I think this is what I need to know: What is best practice for measuring bias? And what might going on with the PI bias being badly affected by connections to the Grid, seemingly, only in the right channel (I suppose a wiring error, out-of-spec component or dry solder joint - but where is it likely to be)?
Simon
RE THE ABOVE:
At ease, gentlefolk - you may stand down!
I went back and checked the circuit for the umpteenth time, checked all component values, and re-soldered every joint in the driver/splitter areas. And now the music has been playing sweetly for the last half hour and we have rock stable bias in the right channel at -0.1.5V - just swaying gently with the rise and fall the jazz! Aahh - Sidewinder by Lee Morgan has just come on. What glorious and now logical bliss! Long may it continue.
For a while I'd thought the gremlins might win.
🙂
Simon
At ease, gentlefolk - you may stand down!
I went back and checked the circuit for the umpteenth time, checked all component values, and re-soldered every joint in the driver/splitter areas. And now the music has been playing sweetly for the last half hour and we have rock stable bias in the right channel at -0.1.5V - just swaying gently with the rise and fall the jazz! Aahh - Sidewinder by Lee Morgan has just come on. What glorious and now logical bliss! Long may it continue.
For a while I'd thought the gremlins might win.
🙂
Simon
So there may have been cold solder joints? Hey, that can happen so easily, and causes all sorts of craziness.
Now that the circuit is working better, do you still see those nearly-zero bias voltages with the Russian 12AT7?
The bias voltages you reported from the Psvane 12AT7 seem about right.
I'm curious to see if the 'Mullard' reissue 12AT7 is really that far off spec. I hope not.
Sorry I led you to spend more money on those JAN 12AT7WCs, but now with three sets of 12AT7s you should be set for life! 🙂
Now that the circuit is working better, do you still see those nearly-zero bias voltages with the Russian 12AT7?
The bias voltages you reported from the Psvane 12AT7 seem about right.
I'm curious to see if the 'Mullard' reissue 12AT7 is really that far off spec. I hope not.
Sorry I led you to spend more money on those JAN 12AT7WCs, but now with three sets of 12AT7s you should be set for life! 🙂
With the Psvanes
As context, I've upped R106 from 75KR to 99KR and increased the four screen resistors from 100R to 1K2R. This latter change was because with my Hammond 1650FA OTs, the Screens on the EL84s were running hotter than the Plates by about 1.5 volts. The Plates are now around 313 - 314V and the Screens nicely lower at about 309 - 310V.
As for the EL84 diss, the Plates are at 314V and the Cathodes at 12.6V through 330R. So current is 12.6/330 = 38mA and the power about (316-12.6) x 0.038 = 11.6W - so this looks OK.
Back to Mullards
These 'might' be OK but ...
The amp still displays its weird behaviour, especially in the right channel. When measuring the Cathode voltage on the 12AT7, placing a probe (or even a longish wire) on the Grid lowers the Cathode voltage by about 2V, from 86.9 to 84.9 and this wipes out the -1.0V bias that exists when I'm not measuring Pin 2. Something similar happens in the left channel but it's only a 0.2V drop when something contacts the Grid, so the bias is not completely wiped out by the Cathode drop in this channel. These particular Mullards seem susceptible to this.
Psvanes not so Weird
With the Psvanes back in, I see that something similar happens in the right channel, but not so much. It's only a 0.4V drop when contact is made with the Grid, so the bias of 1.2V is not completely overcome. In the left channel, with the Psvanes, the Cathode voltage actually goes up slightly when the Grid is touched.
Maybe it's All Alright
I'm hoping that with both sets of valves, the bias remains good when no measuring is being done - because, the Grid voltages are unaffected and the Cathodes only seem to drop when they're contacted.
I wish I knew why this happens. Or if it's weird at all - or expected? If someone can suggest a component measurement or adjustment (change a cap slightly?), I'd be grateful.
Simon
PS The forthcoming NOS 12AT7s will be good to have.
As context, I've upped R106 from 75KR to 99KR and increased the four screen resistors from 100R to 1K2R. This latter change was because with my Hammond 1650FA OTs, the Screens on the EL84s were running hotter than the Plates by about 1.5 volts. The Plates are now around 313 - 314V and the Screens nicely lower at about 309 - 310V.
As for the EL84 diss, the Plates are at 314V and the Cathodes at 12.6V through 330R. So current is 12.6/330 = 38mA and the power about (316-12.6) x 0.038 = 11.6W - so this looks OK.
Back to Mullards
These 'might' be OK but ...
The amp still displays its weird behaviour, especially in the right channel. When measuring the Cathode voltage on the 12AT7, placing a probe (or even a longish wire) on the Grid lowers the Cathode voltage by about 2V, from 86.9 to 84.9 and this wipes out the -1.0V bias that exists when I'm not measuring Pin 2. Something similar happens in the left channel but it's only a 0.2V drop when something contacts the Grid, so the bias is not completely wiped out by the Cathode drop in this channel. These particular Mullards seem susceptible to this.
Psvanes not so Weird
With the Psvanes back in, I see that something similar happens in the right channel, but not so much. It's only a 0.4V drop when contact is made with the Grid, so the bias of 1.2V is not completely overcome. In the left channel, with the Psvanes, the Cathode voltage actually goes up slightly when the Grid is touched.
Maybe it's All Alright
I'm hoping that with both sets of valves, the bias remains good when no measuring is being done - because, the Grid voltages are unaffected and the Cathodes only seem to drop when they're contacted.
I wish I knew why this happens. Or if it's weird at all - or expected? If someone can suggest a component measurement or adjustment (change a cap slightly?), I'd be grateful.
Simon
PS The forthcoming NOS 12AT7s will be good to have.
Possible Explanation of the Wierdness?
I'm out of my depth here, but I'm wondering if it's just the high input impedance (and sensitivity?) of the tube (see reference below) that results in a wire touching the Grid in the PI section to affect the Cathode voltage like this. Each tube would likely have different characteristics in this regard. I'm also thinking that when the Grid is not being touched, the bias is OK, so maybe leave well enough alone.
reference: Valve Wizard section 12.1.3 "Input Impedance" at page 436
I'm out of my depth here, but I'm wondering if it's just the high input impedance (and sensitivity?) of the tube (see reference below) that results in a wire touching the Grid in the PI section to affect the Cathode voltage like this. Each tube would likely have different characteristics in this regard. I'm also thinking that when the Grid is not being touched, the bias is OK, so maybe leave well enough alone.
reference: Valve Wizard section 12.1.3 "Input Impedance" at page 436
It is normal for an ultralinear setup to have g2 at slighly higher voltage than the anode, because g2 is connected to a primary tap, therefore it wont see the full DCR, resulting in less voltage drop.
Akso, using high resistor values for g2 is not recommended, it may result in lower output power and increased distortion.
But, at the end, it's all about how the amp sounds to you. If you like it, then that's it 👍
Thank you. I've now read numerous articles that say exactly that. I lowered the screen voltages based on Stephe's video on her SPP monoblocks. I presently have my screen grids 0.5V below the Plates at idle, using 560R resistors. I may well go back to 100R resistors and let the screen grid have its head again, but I'll do some listening first.
Simon
EPILOGUE
So I ought to bring this one to a close ...
I finally found the cause of the weird behaviour, especially in the right channel. It was the negative feedback causing oscillation, under certain conditions, namely full or zero and low settings of the 21 step volume control. There was a significant resonance at 3.37MHz - plain as day once the scope found it, slightly stronger in the right channel. Probably causing havoc locally in the HF band! I expect that I didn't catch it on the scope originally because (probably foolishly) I liked to play music softly while I checked the voltages!! Obviously, touching various components around the driver/splitter could de-tune or stop the resonance altogether. The oscillation was (1) altering the splitter bias when an attempt was made to measure it; (2) rendering components sensitive to contact by metal objects; and (3) causing a residual AC voltage on the speaker terminals when there was no input, especially on low and full volume settings and (4) driving me nuts! I discovered this while tuning the FB capacitors - I'd used 390pf to begin with but found this resonance. 200pf gave me the best 1kHz and 3kHz square waves and also stopped the oscillation. Whoopee!
So Rikaro - you were right all along.
Thanks again to everyone who gave me moral support.
Simon
PS. I tried the Mullard 12AT7s again and they still didn't have a good negative bias. No doubt I could address that by the means suggested but the PSVANEs are doing fine and I have some US-made NOS tubes coming.
.
.
So I ought to bring this one to a close ...
I finally found the cause of the weird behaviour, especially in the right channel. It was the negative feedback causing oscillation, under certain conditions, namely full or zero and low settings of the 21 step volume control. There was a significant resonance at 3.37MHz - plain as day once the scope found it, slightly stronger in the right channel. Probably causing havoc locally in the HF band! I expect that I didn't catch it on the scope originally because (probably foolishly) I liked to play music softly while I checked the voltages!! Obviously, touching various components around the driver/splitter could de-tune or stop the resonance altogether. The oscillation was (1) altering the splitter bias when an attempt was made to measure it; (2) rendering components sensitive to contact by metal objects; and (3) causing a residual AC voltage on the speaker terminals when there was no input, especially on low and full volume settings and (4) driving me nuts! I discovered this while tuning the FB capacitors - I'd used 390pf to begin with but found this resonance. 200pf gave me the best 1kHz and 3kHz square waves and also stopped the oscillation. Whoopee!
So Rikaro - you were right all along.
Thanks again to everyone who gave me moral support.
Simon
PS. I tried the Mullard 12AT7s again and they still didn't have a good negative bias. No doubt I could address that by the means suggested but the PSVANEs are doing fine and I have some US-made NOS tubes coming.
.
.
Last edited:
A lesson (for me) to remember!
Final Update - I rebuilt the now-beautifully-performing Tubelab SPP into a more compact format. Still no detectable hum or bad behaviour, it easily drives any reasonable speakers, and it makes my F-18 open baffles absolutely rock - seen here listening to Jay McShann Rockin' in Rhythm 🙂.
Thanks for all the support to get me here!
Simon
Thanks for all the support to get me here!
Simon