Siberian Gen 3 (4P1L pre-amplifier)




Building a new version of the venerable 4P1L “Siberian” was very encouraging. This belated project finally came to life after some recent work on a new set of power supplies. So why 4P1L again? I always found the 4P1L sound to be unique. Great detail, overall tone and fantastic treble. What it makes it well suited for pre-amplifiers is not just its linearity (probably being the most linear valve out there) but the fact that it has a low anode resistance and current capability to ensure any challenging load can be handled effectively without any sound degradation. This can be heard particularly on the treble where the input capacitance of the amplifier is more evident and it is translated into treble loss. Other DHTs like 26, 01A, 30sp can only handle a few milliamperes of anode current and is not enough to charge and discharge the  parasitic capacitance at high frequencies. More importantly, the 4P1L has filaments which aren’t demanding. This is a unique feature amongst DHTs that is rare and very useful. Having low-current filaments that can be either configured at 325mA or 650mA, low grid voltages and high transconductance in a valve is very useful. This mean that filament bias can be easily implemented without burning unnecessary power by swinging many volts to perform the desired level of amplification.

However, nothing comes for free. The 4P1L is a dancing lady. You have to tame her singing (i.e. microphonic noise). This may discourage you and has done so to many people before. However, it can be done. Many ways have been discussed out there, but what worked well for me is the use of silent blocks and silicone rubber.  Also starving the filaments is a very effective way of reducing microphonic noise. I parallel the filaments to increase linearity and also to increase the filament current which helps keeping the filament resistor value low which doesn’t impact the output impedance much. Remember that any resistance in the cathode (e.g. the filament bias resistor) is reflected int the anode at (μ+1) times.


4P1Ls can be found in superbly NOS condition (probably not for long) and can be matched very easily. It’s surprising how close they are all to spec or at least amongst other ones in a specific batch.


When I started planning my third 4P1L pre-amplifier, I thought about what I wanted to achieve that I felt could be improved from my previous designs. I came up with the following three things:

  1. Reduce the output impedance: I don’t need further gain, just increased load drive
  2. Increase current drive capability to handle capacitive loads. This is a must if you’re using step-up transformers as an input in your amplifier.
  3. Reduce overall noise level (microphonic and residual hum). After building so many circuits using DHTs, this can be done effectively.

The first two objectives can be accomplished by the use of an LL1660 gapped at 40mA and in Alt Q mode or 4.5:1 step-down. This output transformer will help reducing the output impedance of the 4P1L whilst running the 4P1L at a very high current without compromising the frequency response given the relatively high inductance (i.e. 50H) for the gapped core it has. This feature married with the Ra of the 4P1L will ensure good low-frequency response. The transformer will also help to remove ground loop issues as it provides galvanic isolation between input and output.

The 4P1L likes it hot :) You want to run the 4P1L higher than 20mA if you want to get the most out of this valve. Triode-wired can withstand 9W or even more at 250V maximum anode. As an output stage you can stress the 4P1L to get every juicy bit out of it, but is not what we want in a preamp. We want it to be quiet and very linear. I traced 4P1L in triode mode multiple times throughout the past 3 years and looked at the THD and sound impact of starved filaments in the search of the sweet spot (if there is really only 4P1L Siberian Gen3 Pointone).

I aimed at running the 4P1L around 30-35mA as it was providing the lowest THD on my tests. A good operating point to start here was 163V and -8.7V of bias. I looked at filament bias given it sounds the best to my ears compared to battery or any sort of other fixed bias. I looked at an operating point which will lead me to a nice resistor value that could be formed out of high-quality wirewound resistors and 32mA seemed to be a great choice. The filament resistor then should be Rf=8.7V/(550mA+32mA)= 15Ω. This can be implemented with three 10Ω resistors or if you have a 5Ω at hand then better. The higher power handling capacity the better as it will help keeping the thermal noise down. The quality of this wirewound resistor is fundamental. I use NOS Dale ones. You should pay special attention to this point if you want to get the best sound of this preamplifier.

The 4P1L wired in triode will have a μ=10 and an Ra=1.6kΩ. Transconductance is high at about 6mA/V. Simulation showed that the preamplifier should provide a very low distortion about 0.02%.  Will see how this performs in reality.

The final circuit design is here:
4P1L Siberian Gen3.Nothing new for a classic DHT pre-amp. The devil is in the detail, though. The HT supply is regulated by a Salas SSHV shunt regulator which allows adjustment of the operating point but also is very quiet and proved to be reliable when delivering large currents as well. I used it on my 46 DHT driver with great results. I added a 10μF ANSAR Kelvin capacitor to improve input filtering and HF response. The regulator is set to 210V and 75mA. A 10Ω sensing resistor is added on each anode to help measuring the anode current.

Simulations show that the maximum capacitive load that can be handled by this pre-amplifier is in the region of 18,000pF – 20,000pF before the -3dB point is below 20kHz. This handling capacity should be more than what most of you need in your systems. However this is good enough to drive my step-up input transformer (1:8) and a Zobel network which included a 220k + 390pF.

With the LL1660 wired in 4.5:1 the overall gain is about 6.7dB and output impedance around 425Ω. I ended up wiring the 4P1L with G3 and G2 tied up to the anode via a 300Ω resistor. I usually connected G3 to the cathode filament end which end up with a lower anode current. I therefore ended up with a 34mA anode current.  Not a major difference, but couldn’t be bother to rewire the G3 as I’m fine with a bit more current anyway.


P1 is a stepped attenuator in my case and you can get rid of it if you have a TVC or other volume control. In that case R3 should be changed to something like 100kΩ. I don’t use grid stopper, instead I use a ferrite bead located close to the grid pin. Never had an oscillation problem with the 4P1L so far.


The most critical part of this pre-amplifier is the filament supply. You can’t underinvest here. I designed and built a choke-input filament supply which delivers 16-17V at full load to the Rod Coleman filament regulators. If you haven’t tried proper DC regulation in DHTs, I suggest you read this article. The difference in sound is overwhelming. The slow rise in filament current helps in extending the valve’s filament life.

First tests and listening impressions

It took me half-day to upgrade the 26 preamp and convert it to 4P1L. I had to replace the transformer block (now with cover), the filament regulators, the SSHV2 with a bigger heat-sink, the 4P1L loctal socket board and the filament resistors.

The initial tests on my workbench where really good.  The key highlight measures are:

  1. Bandwidth (-3dB points): 10Hz – +100kHz
  2. Noise level: -87dBV (45μV)
  3. THD: 0.014% @ 1Vrms or 0.0188% @ 2Vrms output

Here are some quick screenshots of my preliminary tests. Load in this case was the Pete Millett’s interface (100kΩ + 250pF or more):

4P1L Siberian Gen3 Freq Response

Sweeping the input to see how much the stage can handle reveals a nice performance. I have to optimise the automated sweep measurement at LF, but either way you can see that up to 8Vrms the THD stays below 0.04% which is really good:

THD 4P1L Siberian Gen3 v1

What is important to highlight is the fact that since this is a DHP in reality and is triode-wired, the H2 and H3 levels are not strictly speaking performing as a pure triode. You get about same levels for large voltage swings. For example at 2Vrms I measured the H2@ -62dB whilst the H3 at -64.8dB.


Sound-wise, I’m surprised with this preamplifier  It’s great.  The microphonic noise is very low, only you can hear it when you tap the valve or when it heats up you will get some metal expansion dings here and there. Once warm (it can get to 105°C measured) it provides the most delicate DHT sound. The dynamics and tone are superb. I can now get some high-treble detail that I lost with my 26 preamplifier. I simply love it. I’m burning in a matched pair of Svetlana NOS 1973 production which my friend Vyacheslav sent me from Ukraine. These are fantastic.


I’m still looking forward to listen this pre-amp for many hours before making further observations. I will also take further measurements when taken back to the workshop.

Now, I will enjoy the 4P1L stage for some time…



32 Replies to “Siberian Gen 3 (4P1L pre-amplifier)”

  1. Good morning,

    congratulations to all its projects, have become for me a real reference.I’m Italian and I follow for a long time, now I have decided to build the wonderful preampthat uses 4p1l.

    I ask you, please, a piece of information, I already have a pair of lundahl choke mod. 1673 by 10Henry 200ma for Anodic power supply, but I’ve seen that use choke from 20Henry.

    Can be fine, too, from 10Henry, I can fit the schema?

    I can write them again if I need help?

    Thanks again,



    1. Ciao Andrea,
      It can be used, however I suggest you play around with PSUD2 to tweak the LC circuit and also add additional primary resistance (R1 and R2) which dampens the response of the supply. If you’re not familiarised with PSUD2 then the risk is that not only you will have higher ripple in the output (L is lower so the filtering effect is impacted) but also the transient response (or response to changes in the music reflected in the anode current of the 4P1L) will be impacted.
      Unfortunately don’t have the time to play with PSUD2 for you, but hopefully you can sort this out yourself.

      All the best and keep us posted with your progress

  2. Thanks for the speedy reply and for all the advice,

    I prefer at this point, having to buy the power supply transformer, take even the inductance by 20Henry lundahl.

    Excuse me if I take his kindness, for capacitors I have two motor run from 450vca 45mfd, compared with 40mfd, the difference can create some imbalance?

    One last question, are well under way for power supply to the valve filaments,making use of the rod coleman regulator(bias).

    To simulate the resistance of the filament I simply applied ohm’s law, 2, 1V/0, 55A > 3.81 ohm, this to avoid damaging the filament of the valve for some error.

    Thanks for your patience,



    1. 45uF should be fine. You can use a 3R3 resistor to test your regulator. Don’t forget to add it in series to the filament resistor. Once you’re sure that the filament current is about right, you can test it with the 4P1L. At that point you can trim the current to the right value

  3. Hello Ale and thanks,
    A latest info, then we update, maybe,, to the preamplifier switched on…….an indication, kindly, where to find the capacitors Kelvin ansar? I found something but are too high, speak of condendatore from 10mfd to be placed at the entrance of the pre.

    Thanks for your precious help and your availability,



  4. Hi, Moglia,
    Thanks for sharing your work. Read this article and I got the buzz to try your pre. I have a few 4P1L around, a pair of LL1660 and a pair of Rod’s fillament regulators. It should be easy to build, but the LL1660 is a 18mA. I could use the 4P1L at 16mA and the LL1660 at 4.5:1 or use the 4P1L at 32mA and the LL1660 at 2.25:1 (or 2.25:1+1, if I need balanced output).
    What do you think would work better? any other suggestions?


    1. Hi Joao,
      If the LL1660 is gapped at 18mA then you should run the 4P1L at 18mA. The 4P1L work much better above 20mA given the proximity to the curve kink, however you can still get good results in a preamp as the voltage swing is relatively low. The step down required is up to your system and the level of gain needed. I’d recommend 4.5:1 to achieve the lowest output impedance. That would be halved when taking the output in differential mode.

      Good luck with the build!

      1. Hi again, seing your reply I checket wjhat else I could use to get the 4P1L to work where itr feels the best (30mA?) I found a pair of LL1677/60 I could use 8:1+1 and estimate It won’t saturate before the 30-32mA. Ever tryed this? Any comments?

  5. Hi Ale,
    finished a copy of your siberian some tima ago. Used a Tamura transformer and it works quite well. Just one very strange fenomena: had some humm and that wouldn’t desapear nomather what I would do, even shunting the 4P1L grid would not reduce the humm. Untill I placed a cap (1500uF) in the Rod regulator output. The noise is gone and all works fine, but a cap in that place will deny the benefits of the regulator. What did I do wrong? How can I correct it?


    1. Hi Joao.
      The hum noise could be due to many things. Hard to give any advice without a proper schematic and pictures of your build. Specially the grounding scheme. I don’t know what raw supply you’re using for the Rod Coleman regulators or wether you tested them before placing them on the preamp.
      You can send me your email through the “About me” contact form

  6. Hello Ale!

    Have you measured the output impedance or just calculated it?
    If the internal resistance of 4P1L is around 1600R then the ouput impedance should be around 80R to 100R (1600R divided by the squared tranformer ratio which is 4,5^2=20,25).

    Best regards,

      1. Hi Ale,
        Thanks for explanation. Seems I was a little bit too optimistic in my calculations. Good point is I am still learning.

  7. Hello Ale,
    I completed the Siberian preamplifier and I am really pleased.
    The only thing is the gain a little lower, than my final (a Single ended 211 tube).
    I wanted to ask you, politely, if I could configure the output transformer in order to have a higher gain, although the output impedance will tend to rise.
    Can you suggest a configuration for lundahl transformer 40mA 1660s? Or add an input transformer and which model eventually. …
    Thanks in advance for the time that you will grant me and for all of your valuable advice,


    1. Hi Andrea,

      I’m glad you liked the 4P1L preamp. It’s a great preamp, probably one of the bests out there in my opinion

      The LL1660S is different from the LL1660. If you can figure out the winding connections you can try replicating the Alt S which is 4.5:4. You will get the gain you may need that way albeit output impedance will rise as you said. I don’t know what your first stage of your amp is but I’d suspect if this is a high gain triode stage then the miller capacitance will potentially make it a tough load to the 4P1L in the treble connected this way. Alternatively the LL1671 (1:1 Alt S) is a better choice for an IT
      Good luck!

  8. Hi Ale,

    Just curious, I own Lundahls LL1660 configured Alt Q but data sheets says 18mA, how do you do to have 40mA?

    Do you have gyrator boards for sale?

    Thanks & have a nice day.


    1. Hi Felipe, I have an LL1660 gapped at 40mA. You need 30mA at least to get best sound
      Boards are not ready for sale yet as an refining the prototype and will let know when ready

  9. I’m using mine LL1660 18mA bought to Thomas Mayer for #26 preamp, What about the hum & microphonics using only SSHV2 + your gyrator CCS load in place of the LL1660?

  10. Hi Ale,

    I have checked Big Bear Audio for LL1660/40ma but they asked me to approach Lundahl instead. Where did you get yours?


        1. Hi Gary, I’d not do that as bass will be impacted due to reduced primary inductance and distortion due to core saturation. You can run the 4P1L at 25mA by changing the operating point instead.
          It won’t be optimal but it will work fine

  11. Hello from Bonn in Germany,
    I am a new reader of your gorgeous web pages (many many thanks for this formidable effort!!) and wanted to ask you which raw power supply schematics did you use before the Salas regulator?

    Many thanks and best regards,

  12. Hi Alex, I need balanced out, my friend has only XLR in, on the power amp. what is the best option, increment a tube as phase spitter or interstage ( witch LL )

  13. Hi,

    Thanks for all of your informative posts!

    I’m confused and perhaps you can clarify for me issues regarding the Lundahl IT.

    The LL1660 data sheet says that the primary for Alt Q (4.5:1 ratio) is 100H, but you have 55H in your schematic. So you are using an inductance lower 55H vs 100H on the data sheet. I thought that this would also mean the 2° inductance is x/(4.5^2) which is 2.716H if 1° is 55H, or 4.94H if 1° is 100H). I’m getting the 2° calculation, but not how you arrived at 55H for the 1° inductance.

    I also am confused how Rs is 277.5 ohms. I thought the resistance windings are simply paralleled with Rs = 1/((2/315)+(2/240)) = 68 ohms.

    On another schematic, the CX301a-preamp-based-on-LL2745, with the LL2745 (used similarly as the LL1660 above, also Alt Q) at 3.5:1 ratio, the data sheet says the 1° is 95H, but you listed 180H on the schematic. In this example, you got a 1° inductance increased over the data sheet (180H vs 95H). [The 2° again works out OK when using 180/(3.5^2).] I just don’t understand how you arrived at 180H instead of 95H on the data sheet.

    Basically, I’m trying to learn how to simulate Lundahl’s IT in LTSpice, and I’m confused as to how the primary values of inductance (and Rs) are being calculated.

    Regards, Robert

    1. Hi Robert,
      Primary inductance is dependant on the current gap. It’s linear to current in the region so Lp shown is based on gapped transformer I have (LL1660 @ 40mA) as well as using the LL2746 at 3mA instead of 18mA.

  14. Thanks so much for the reply. There are more variables than I realized affecting the inductance.

    I’ve read about the gapping and various current ratings but did not understand that this in turn affected the inductance. (It would be helpful is Lundahl published tables of the primary inductances for optional ratings for each model.)

    Thanks again. I truly enjoy reading your web site.

    Regards, Robert

  15. Per Lundahl does give you all the info you need, and typically values for current and inductance. From one inductance/ current you can calculate any other, as it varies linearly and inversely to each other. Per can manufacture almost any you’d like, just ask him

Leave a Reply