• 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.

Ideas input for a Interstage driven 300B SET


Account Closed
Joined 2008
D3a (or E280F) and C3g (or rare C3o) as triode are perfect VAS tube for two stage amplifiers.
Both have low distortion even at large swing, and harmonics distribution results "good tone".

View attachment 1296373
It's exactly the tone I don't like. They might be good paper and on the bench but not as good as other solutions in real use. That's my opinion, of course.

@Matthew78 4-5K with 400-420V anode voltage means 10-12W clean power. What's the point of spending all that money for getting 7W with 5% THD?

@stephe here can get 6-7W with 1% distortion using a 6SN7 cascode driver for a complete 2-stage amp. I think the OPT is 5K as well in her case. Videos on youtube....

I could get almost 13W using 4.2K and biasing the 300B at 70mA+ with 420V anode voltage. There was slightly positive grid drive but the driver stage had no problem and no significant change to the distortion pattern except the normal rise. The driver was the 46. The first prototype used LC coupling and soon after it was (coincidentally) upgraded with the NC-20!

You can do SUT+EML20B but you cannot drive it with anything. The input capacitance of the 20B will be 16x higher with 1:4 SUT or 36x with 1:6 (to match the 26 ACTUAL gain) . I am not sure the 20B has a low miller capacitance. You can buy the best SUT there is but there is no way around it: you just shift the issue upstream to the source. Actually the only way around is to drive it with a preamplifier using a step-down output transformer but then you end up with 3-stage amplifier (pre+power) with added stuff (2 transformers + another separate power supply + other components).
My favourite input transformers are all 1:1, 1:2 max or step-down.
Last edited:
  • Like
Reactions: 1 user
@45 thanks, for the OPT I hadn’t thought about power output, I’m not too worried about the power output (the current 45 plays loud enough), but thought the higher load might bring a sonic benefit.

As for the SUT, I can’t find the miller capacitance, the EML website hints at a possibility with a 1:2 input and a 1:2 interstage, but high roll off is probably the last thing I want.
Now you mention the 46, it’s a small step to a 45 driving the 300B, could that be the sweetspot? :)

You are correct.
Shishido designed amplifiers with many different topologies.
There are lots of amplifiers that are very similar to his schematics in this thread.

But I consider Shishido's non-air gapped interstage circuits to be the proof of his genius.
To me, that is his defining moment.
Both the good performance from that interstage transformer; and the ability to use A2 on tubes that were made for lots of grid current.
All wrapped up by his unique circuit.


Account Closed
Joined 2008
In reality the interstage is gapped because there is a static DC current absorbed by the grid of the power tube and the DC current of driver on the other side. However the DC current going into the power tube with signal will grow. So it's variable and the air gap is needed for that mainly + if you have some extra headroom have more choice for drivers and power tubes because don't need to satisfy strict conditions. It's still a big advantage because the air-gap can be smaller than usual or the transformer can be smaller or have less turns etc....

The Softone RC-20 is good for this application. See connection n.5:
Last edited:


Account Closed
Joined 2008
@45 thanks, for the OPT I hadn’t thought about power output, I’m not too worried about the power output (the current 45 plays loud enough), but thought the higher load might bring a sonic benefit.

As for the SUT, I can’t find the miller capacitance, the EML website hints at a possibility with a 1:2 input and a 1:2 interstage, but high roll off is probably the last thing I want.
Now you mention the 46, it’s a small step to a 45 driving the 300B, could that be the sweetspot? :)
The 46 is extremely similar to 45. Some people even prefer it. I would use the 46 because of its extra gain. It's almost 2x.....
Finally, if someone has ideas on full triode (preferably DHT) 3 stage amp, I'd be keen to hear them.
I listen to 3 stages of DHT. My first advice is to use speakers in the 100dB/1w neighborhood with SS powered subs to handle the very bottom. Sounds like you have this covered if a 45 is loud enough for you.

My preamp is dual mono on the same chassis. 300B or 6A3 with regulated CCS B+ and regulated DC on filaments.(I listen to 6A3)
Amp is CCS 3B7 with regulated DC filaments driving 2A3 parafeed.
Each stage is run at near max…I use digital input volume control.

I want to try a 10Y DC 2A3 schematic from euro21.
Also want to try a 20B/300B because I have the EML tubes.

Good luck and have fun!
  • Like
Reactions: 1 user
Hi, just an update here. I spoke to someone in real life (scary), and we brainstormed a bit. The result of this brainstorm was to up the game a bit and go for a three stage amp: 76 (or 56) -> 45 -> NC20 -> 300B.
Although it's 3 stage, and has a capacitor (perhaps initially) between the 76 and 45, at least the circuit could be simple enough for me to understand / come up with and the focus can be on listening on the effects of tuning via component choices. I'll be keeping the PSU out of sight (for increased WAF it can go in the cupboard) and as a result can probably separate the 76 and 45 power supply from the 300B power supply, either with complete separate transformers and rectifiers, and then left and right with separate chokes, or at least separate chokes.

Curious to hear people's thoughts. I'm Dutch and therefore can take a bit of direct feedback, so please tell me if you think this doesn't make sense :)
Again thanks for all the input so far, just by considering I'm learning a fair bit in understanding how the different parameters interact.


Account Closed
Joined 2008
That 3 stages should have a bit more gain IMHO. You could try like that or with a 6J5 in place of the 76 if need more gain.

You can easily DC-couple the 76 (6J5) and the 45. You can use the modified Komuro circuit that @directdriver posted at #13. It's a nice amp and will work with one plate voltage supply for all stages.

You only need to change the cathode bias resistors. The first stage as is with the 6J5 or replace 470R with 1K with 76; the second stage with the 45 replacing the 6V6, replace 4K with 4.7K. That should work ok for start and then you can make small adjustments if necessary.

There is not much to understand about the DC-coupling of the first and second stage: the grid of the second tube will have the voltage of the plate of the first stage and so the cathode of the second stage needs to be at the grid voltage + bias. It works exactly the same way, your reference for the cathode bias of the second tube is the plate voltage of the first stage instead of 0V.
Last edited:
Thanks @45, that might be a good starting point indeed. I was thinking of separating the power supplies for "higher fidelity", perhaps building a separate supply for left and right might make more sense in that case. Also that's the most understandable explanation of DC-coupling I've read so far.

1. Shishido's genius of how he connected the interstage, not only cancelled the primary's quiescent current's magnetic field, because of the secondary's quiescent current's magnetic field . . .

2. It also was connected so that when the plate of the driver tube's current increased with signal applied, the grid current of the following state's RF tube also had more grid current. Again, a tendency to cancell the magnetic field during signal conditions.

Now you can begin to see why the air gap required was lower . . .
During quiescent current
During signal current

For the other direction (polarity of signal)
When the driver current reduced during the other polarity of signal . . .
Then the RF tube grid current also reduced during the other polarity of signal.

Also, when the driver tube draws more current, its plate impedance, rp, is reduced. That is the same time that the RF tube's grid's impedance is reducing, and demanding more current.

As I said, Genius

Shishidos other circuits are very similar to other designers circuits.
But not so, for his special interstage circuit connections. I am not aware of anyone else who has used that unique technique.
At least, I think he was the first to do so.
Last edited:


Account Closed
Joined 2008
@6A3sUMMER, I know how it works and it 's not easy to do. However "people" (99% Japanese DIYers, I would say) have found alternative solutions.

I think that the solution with DC-coupled driver of my friend Mitchi works better taking into account all aspects:

http://www6.plala.or.jp/Michi/4P1L 811A 8/Third version of the 811A amp.htm

The operative conditions of the 881a are such it never leaves positive grid until it starts clipping. Avoiding the cross-over from positive to negative is key to getting lowest distortion.
For standard negative bias, one can see that driving the tube into positive grid is ok for few volts of positive swing (depends on the current flowing into the grid) but then there is a sharp change in distortion spectrum and figures. For this kind of amplifier (i.e. negative bias but driven deep into positive grid field) THD is basically proportional to the Zout of the driver when grid current grows quickly from near zero to max and one really needs a very low Zout, very low distortion driver. If wanna do it with tubes one could use a proper MacDonald augmented CF:

Funny enough if you always stay in the positive grid filed it's "easier" to get better perfomance.....
Last edited:


Account Closed
Joined 2008
Thanks @45, that might be a good starting point indeed. I was thinking of separating the power supplies for "higher fidelity", perhaps building a separate supply for left and right might make more sense in that case. Also that's the most understandable explanation of DC-coupling I've read so far.
For the 45's DC operative conditions pick a point you like, for example 180V/31mA as in the datasheet. You will need:
100V (plate voltage of the driver) + 180V (plate voltage of the 45) + 31.5V cathode bias of the 45 + few volts dropped because of the NC20 DC resistance. That's apporx. 320V.
Regarding the cathode resistor for the 45, you can see you would need 1K if the grid were at 0V and then need to add more reistance to take into account the additional 100V: 100V/31mA=3.22K. So it's 4.2K in total for the cathode resistor. Sanity check: 4.2K x 31.5 mA = 132.3V (i.e. that's about 100V+31.5V bias). If you parallel 4.7K with 200K (shown as 100K+100K in parallel to the main 4K cathode bias resistor) you will get 4.59K. Just need to drop a bit those 100K resistors....

In the schematics the 300B works at 350V/60mA with 61V cathode bias +few volts drop on the OPT. Let's say 420V. You have 100V to spare! You can even go up in plate voltage with the 45, say 200-220V and still have plenty of voltage that can be dropped with very good filtering and separation. You could add more filtering and regulate the voltage supply of the 6J5+45 combo as is.

The 296V supply is not the right one....that was good for the 6V6.
  • Like
Reactions: 1 user

I always enjoy your many posts.

As you know, there are many different engineering designs.
The various good topologies are successful only . . . if they are implemented properly, from quality parts, quiescent volts and current, wiring layout, magnetic device angular orientation and spacing to other magnetic devices, and control of ground loops (hum, and signal loops too).

I was not trying to promote Shishido's unique design as the best design, just a very successful one.

In photography, look at the latest statement from Sigmas owner and CEO, about the development of a full frame Foveon sensor (lots of tradeoffs, but a unique look too).

Opinions are a dime a dozen, but lots of people are enjoying their dimes.

I am not sure that all diyAudio designers are willing to give up on a design, just because it is difficult (or very expen$ive as well).

Some others like myself will not use that difficult of a design, and will not use very expen$ive parts.
Someday, I would like to purchase a pair of Monolith Magnetics push pull output transformers. I can not justify the expen$e, but probably will break down and order a pair anyway.
Hi, as my thinking progresses and am starting to lean to the all triode 76/56 DC coupled 45 -> NC20 -> 300b path… I’d like to pick your brain on the WAF of all of this and practicalities of how to build this.

Initially was thinking of a PSU with only DC going to the Amplifier section. The PSU would be hidden in a cupboard out of size.

I have two chassis like I used for my simple 45, but I’m thinking that putting 6 valves an attenuator and 4 big transformers in, might be a bit of a squeeze. Which lead me to think splitting the pre stage from the driver stage.
I could then “hide” the driver stage for increased WAF.

Would doing this after the interstage be an option? I lot am i setting myself up to introduce more problems than I’m solving? The leads to the hidden box would have to be roughly 2 meters, I could then also experiment with different end drivers more easily?
As always thanks for all the inputs!

Ps here’s a picture of the chassis I’m referring to


  • IMG_0146.jpeg
    314.6 KB · Views: 35


Account Closed
Joined 2008
The more you split it up the more messy it will be. WAF will decrease, IMHO.

Amplifier with separate PSU is what I would do if had no choice. But you can make it all in one. Something like in the picture below. If it can be done with 4 x 211 (2 are output tubes and the other 2 are used as rectifiers!) can also do it with the 300B and 0 or 1 tube rectifier.

Other things you might consider: Using a 6SN7 as input tube will save you quite some space. Can also find space below the plate to install inductors inside. Etc...


  • 211SE.PNG
    211.9 KB · Views: 54
  • Like
Reactions: 1 users
That’s a tight fit, and nicely done. It would probably still fall in the “bulky” category as my wife called the Simple 45.

I’m coming from the 47 labs Shigaraki, which obviously is “petite” compared to anything with proper DHT valves :)

That said even though I would need to cram in some rectifiers, input selection and an attenuator, that symmetric layout is quite neat, also puts the interstage transformers in a natural position. (So far I thought the 300b might end up hidden behind the ITs).
In any case you have me considering the single box solution again :)

As for WAF I guess I was thinking of a small pre amp in sight and the power amp out of sight, but before we know it we’re back to the eternal integrated vs. pre-power discussion :) or did someone reach a final verdict there?
Some more progress in my thinking and the first components start to come in (two 76 tubes on the bench, and 45s on the way).

I’m keen to forego maximum wattage for now and run the 300b with lower voltage and amps, also with an idea in the back of my mind to keep the door open for the EML Mesh version (which should be run at lower dissipation). Therefore I’m thinking an anode voltage of 350V, 63mA and the grid at -74. This, if my thinking is correct, should result in a 425V B+.

Would this imply I could run the 45 at 250v 34mA and -50v grid as per the datasheet or go a a bit higher at 275v, 36mA and -56 at the grid?
Intuitively I’d rather run on a lower voltage as I suspect distortion is lower, but I also see that the output impedance goes from 3900 to 4600. Would this have a positive effect on sound quality? Or would a lower impedance be better? Should I consider 180V and go for more filtering?

If the 76 is run at 100v anode, then 100+250+50+ a few volts through the NC-20, would conveniently be rather close to the 425 B+ required for the 300b.

Any inputs much appreciated!
Joined 2004
Paid Member
"lower voltage as I suspect distortion is lower"

Higher current of tube == lower distortion and lower output impedance.
Lower output impedance == higher driving ability == lower distortion with problematic (for example high Cp) load.

If you have the opportunity to increase the current (within the reasonable limits), must be done.
Hi @euro21 thanks for the response, I didn't realise about the 30mA for the NC20, nice catch, not worth risking these expensive ITs.

The schematic above, did you build that in Spice? again impressive how quickly you turned that around. I was planning to do some simulations to see if I could calculate all resistor values etc... correctly. Although seeing the above, I can perhaps skip that and use your values :):)

As for the distortion, perhaps I'm not using the right technical term, I thought by not running the valve at max power, one could stay in a more linear / symmetric region of the load lines, but that's just an unscientific assumption, underpinned only by the fact that a lot of classic schemes in case of the 300B seem to bias only at 320-350 with +/-60 mA.

I'm not looking for the highest output wattage, my current 45 SET goes plenty loud, nor for the best measuring amp, but looking for that fluid natural Single Ended Triode sound. Again thanks!