• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Tubelab SPP - Bias of 12AT7 Phase Splitter?

DYI Audio folks

I have built a point-to-point version of Tubelab's SPP amp and had some difficulty understanding how the bias of the phase splitter section of the 12AT7 works. When first built, using the specified value 75K resistor as R106, the grid voltage on pin 2 was a couple of volts higher than the Cathode voltage on pin 3. For example, 72.2v on pin 2 and 70.6v on pin 3, when B+ is around 280v. I changed R106 to 150K ohms, which nicely turned this difference around. This approach was a suggestion by rknize on the SPP thread in the Vendor's forum. FWIW, I'm using new reissue Mullard 12AT7s

Is there a simple explanation for how this splitter is biased that anyone can offer, or refer me to, that would explain how the tube is biased and why changing the 75K resistor to 150K worked, and whether that's the best resolution of my bias issue? Thanks in advance for any assistance on this.


Tubelab SPP.jpg
Joined 2021
Paid Member
Hi Simon,
Could you show the voltages in the 12AT7 pins (heater not needed)? It is a bit strange indeed, because the stage is calculated using the average 12AT7 curves, and I would expect any re-issue to behave similarly. Having to increase R106 that much is not normal. I have seen mistakes with resistor values due to marking colours not being clear enough, double check just in case. Also, can you change try another tube?
There is a good explanation of cathodine calculations at valvewizard site: https://www.valvewizard.co.uk/cathodyne.html
The first section of the 12AT7 is a typical voltage amplifier stage. Its plate voltage is determined by the plate load resistor (R106) and the cathode voltage which is determined by the total drop across the combination of R102, R103 and R104 which is grounded through the OPT. This should give you about 1/3 of the total B+ voltage on the plate (pin 6) which provides the bias for the grid of the second stage and eliminates a coupling cap in the signal path.

The SPP design was developed in 2006 / 2007 before Mullard reissue tubes were available. All of the initial testing a tweaking was done with various NOS and used 12AT7's. 6201'a and other 12AT7 variants. It was also done at a B+ voltage in the 300 to 350 volt range where many users ran their 6BQ5 / EL84 tubes.

The Mullard and Tung Sol reissue tubes are Mike Mathews brands that may be based on a typical Russian tube which may not be representative of a 50+ year old 12AT7.

I could see the need to tweak the values in the first stage to get a good plate voltage, but I would not expect the need for that big of a change. First make sure that the cathode resistors are the correct value, and the bypass cap is installed with the proper polarity. If all looks good and you like the way the amp sounds, you could leave it alone.

Old US made 12AT7's tend to like a bit more current so you could experiment with lower values of plate resistor like 100 or 75 K and bring the plate voltage down by lowering R102. I have never tried the reissue tubes so I can't offer a good starting point.
  • Like
Reactions: 1 user
I built this amp, point to point soldering. Great sounding amp, right from the start. I did not change any values on the resistors. I have built some el 84 pp amps, and some of them did not sound any good, but with this amp sounded great from the moment I turned it on. This is the best curcuit if you are thinking el 84 pp!
  • Like
Reactions: 1 users
Lots of these have been built on PCB's and point to point according to the schematic. I do remember some odd stuff with a particular tube and member rknize was involved but it had to do with some JJ tubes that were not exactly like a typical 12AT7.

If this was my amp and I wanted to use the Mullard reissues, I would put a pot across R102 and tune for minimum distortion. The TSE and TSE-II boards have a pot with a series resistor (so you can't turn it down to zero ohms) across the cathode resistor because WE417's and 5842's are highly variable even among several identical samples. There is no single optimum resistor value for those tubes.
  • Like
Reactions: 1 user
Thank you George, for weighing in. I have another pair of 12AT7s to try (but, ugggh, their new Psvanes). And I'll look at trying the variable R102 idea, thanks.

Thanks for the sim, baudouin0. Yes, they're definitely 12AT7/ECC81 tubes.

Thanks jcalvarez for the ValveWizard reference - and I see that there's a more in-depth and technical treatment at https://www.valvewizard.co.uk/cathodyne.pdf. I checked, double-checked and re-checked the resistor values, but baudouin0's LTSpice sim does not reflect the voltages I seem to get. See table below. So I'll have to check one more time.

Just now, I've run voltages again on both 12AT7s, with R106 set to its original-spec 75K ohms on the Right Channel and the altered 150K ohms on the Left Channel. Then I reversed the situation, so R106 was 150K on the Right Channel and the original 75k ohms on the Left Channel. I tried to keep B+ close to 300v as possible. Results shown below. Although previously, the 150K substitution seems to fix the grid bias in the splitter, it clearly is not doing so now - particularly in the Right Channel, which is where I've had some variability before.

I also get fairly different voltages between right and left channels.

I also ran voltages with both channels using stock 75K R106s, and with both channels using 150K resistors. This further suggests I may have a slight intermittent fault somewhere in the right channel splitter circuit

So I think I need to go back and triple check for dry joints and correct component values .... again. Thank you very much everyone for interest and comments so far. Back to the building board! More later.

Interestingly, the amp sounds fine in my cheapo test speakers in all these
configurations (except when it motorboated badly with 75K R in place, when the the multimeter probe was connected to .


Test 1.JPG
Test 2.JPG


  • 75K vs 150K Ohms.JPG
    75K vs 150K Ohms.JPG
    75.2 KB · Views: 33
Thank you huggygood. The house mains voltage was around 235v - 240v, for all tests, but used a variac to adjust the AC on the PT primary to keep B+ close to 300v for the tests as noted on the results.

I can easily swap the 12AT7s around and repeat the test - perhaps tomorrow. I could even try some other (new) 12AT7s I have, if folks thought that would be diagnostic.

The RC values in both left and right FB loops are 5K1 and using 390pf as a starting point. Feedback uses the 16 ohm secondaries but speakers are connected to the 8 ohm taps.

Is there a simple and straightforward way to test/examine for inverted phasing of the output transformers? I would be very pleased to do that. I have an oscilloscope. The OTs are Hammond 1650FAs. There is no audible screeching or feedback and the sound quality is and was apparently very good throughout all these tests.

I know NFB will keep the signal level low to the input stage 12AT7, but those are very low grid bias voltages (Vg-k). Looking at the measured voltages in post #7, it looks like they're down near zero bias. 12AT7 is said to be more prone to drawing grid current than 12AX7 (for example), so it's likely your 12AT7s are indeed drawing grid current. I aim for at least -1.2V, to try to keep the 12AT7 out of grid current territory, but perhaps that's not absolutely necessary.

How much of a problem is it to have the input 12AT7 running with -0.5V grid bias?

EDIT TO ADD: I believe George nailed it. It looks like the Russian 12AT7s bias up differently.
Using Adrian Immler's model of an Electro-Harmonix 12AT7 in LTspice, it looks like even with +320V B+ the 2nd (cathodyne) stage 12AT7 will be running with a little positive bias, guaranteeing grid current.
Substituting the Ayumi N. 12AT7 model, grid bias on the 2nd (cathodyne) stage is around -1V, which is probably OK.
Substituting Adrian I.'s RFT ECC81 model, grid bias on the 2nd (cathodyne) stage is -0.7V, which is better than positive bias!

It looks like that confirms the New Sensor/Russian 12AT7s bias up differently than vintage US and Euro 12AT7s, and that could be what is causing the zero bias condition here, especially with the lower than designed B+ of only +280V. Perhaps reduce the value of R109 to 820 ohms and increase the value of C102 to 100uF?
Last edited:
  • Thank You
Reactions: 1 user
Looking at the voltage table for "Right Channel with R106 =75k" I see a pin 6 voltage of 66V and a pin 2 voltage of 104V.
That's not possible, rather those voltages should be essentially the same.

I suspect there's ultrasonic oscillation involved.
Oscillation is sometimes triggered by connecting a probe.
With oscillation most DC voltages might be off and meter readings are not reliable.
Oscillation in the PI circuit should stop when you pull the EL84s or when you disconnect the NFB.
Use your scope.
  • Like
Reactions: 1 user
Thank you jcalvarez for the full sim - yes, I see the Cathode staying about 0.8v higher than the grid.

Thank you Rikaro. You have sharp eyes. I made a transcription error from my notes - I'm sorry. With "Right Channel with R106 =75k" pin 6 voltage was not 66V as I wrote - it was actually 104V, the same as the pin 2 voltage. I have disconnected the NFB lines to take that out of the picture; the amp plays noticeably louder now but I didn't really notice any loss of quality.

For the next measurements at 300V B+ with R106 at 75K, I swapped the 12AT7s around : pins 2 and 3 in the left channel both close to 105V now (grid bias Vkg about -0.01V) and on the right both pins 2 and 3 were around 98V (Vkg about -0.06V) - so the slightly higher/lower voltages changed channels with the 12AT7s. So ... this negative grid bias is still very small using 75K R106s.

I also tried R106 at 150K with the swapped 12AT7s and got much better bias of about 1.6 - 1.7v, which is nice.

But I'd like to go back to 75K R106s if I can.

Which leads me to thank rongon for the suggestion to reduce R109 a little and double up C102 to see what that does - easy to clip on an additional resistor and capacitor to each. I'll try that next.

I very much appreciate the support and guidance I am receiving from you more experienced and knowledgeable folks.

So I decided to change out the Mullard reissue 12AT7s and put in a pair of Psvane 12AT7s that I bought as a precaution on Aussie fleabay for A$35 each (US22.50) after I started getting wobbly voltages. Well ... much better!

Going back to the original 75K ohm R106s, at B+ of 300V, I got on right / left:
Pin 1 Plate 190V / 191V
Pin2 Grid 96V / 95V
Pin3 Cathode 97V / 96V - with a grid bias of -0.4V / -0.7V
Pin6 Plate 96V / 95V
Pin7 Grid 0V / 0V
Pin8 Cathode 0.83V / 0.83V

Turning up the B+ to 310V, I got grid bias of -0.5V / -0.8V.

While there's still some imbalance in the bias, I'm much happier with these numbers. I'll try swapping these left to right to see their bias the other way round, and also try to fine tune them further of course but I think the new Mullards may have "issues"? Also, I can probably re-connect the FB lines and tweak the FB caps.

Thanks again for all your comments and suggestions!


  • Like
Reactions: 1 user
Is there a simple and straightforward way to test/examine for inverted phasing of the output transformers? I would be very pleased to do that. I have an oscilloscope. The OTs are Hammond 1650FAs. There is no audible screeching or feedback and the sound quality is and was apparently very good throughout all these tests.

Put a test signal through the amp and measure the output voltage (across the speaker or load). The actual level is not important as long as it's high enough to notice a change. Do this with and without negative feedback. Do not change the input level between tests. The level at the output should drop noticeably when the negative feedback is connected. If the level increases when the feedback is connected you need to swap the primary wires.
  • Thank You
Reactions: 1 user
I mucked around in LTspice for a few minutes with the stock SPP circuit using 320V B+ and Adrian Immler's (Russian) Electro-Harmonix 12AT7. It looks to me like this particular 12AT7 requires higher plate-cathode voltage to get its grid bias away from zero volts. With EH 12AT7 I see -0.8V grid bias on the first stage, but slightly positive bias on the second stage. If I then substitute cogsncogs's (Ayumi method) GE 12AT7 model, the difference is dramatic -- the first stage biases at -0.9V while the second stage biases at -1.3V. That looks like what was intended.

George mentioned earlier that he designed the circuit using vintage US 12AT7s, and it looks like Russian and Chinese 12AT7s are different enough that they don't work well in this circuit. Perhaps find yourself a pair of decent American or European 12AT7 or ECC81 and try them out? That could be cheaper in the long run and a lot less work than re-designing the SPP to suit these Russian 12AT7-oids.

In my experience, all the American 12AT7 versions are similar to each other; RCA. GE, Sylvania, Philips-ECG (1980s Sylvania), Westinghouse, Philco, etc. Perhaps in Australia the British and Euro ECC81s are no more expensive.

@Tubelab_com - Did you ever try 12AV7 or 5965 as the cathodyne driver in an SPP? It looks like those would work too. But I don't want to claim that they will based only on simulations. They would need to be tested in real life.
  • Thank You
  • Like
Reactions: 1 users