Recently I was asked about whether I could write on my blog about how to design a filament bias stage. My immediate answer was:
I don’t have much time these days am afraid to write extensive articles (and sometimes to even write-up at all)
Thomas Mayer has written about it (see here). Of course, I completely forgot that Thomas never completed his intended series of posts around filament bias, so I decided to attempt explaining the practical aspects of its design in this blog.
Before you continue reading this post, I suggest you read first Thomas’ article above and get yourself acquainted with DHTs and triode amplification. I’m not going to cover any of that theory which I will give it for granted that the reader is experienced with valve circuits and in particular with the hybrid mu-follower amplification stage with gyrator load.
It’s always great to come back and revisit a great design. The 4P1L preamp performs flawlessly so I tweaked the gyrator board to see how it worked with the BF862 FET. The result is great, it sounds as good as it measures:
The 4P1L is biased to 150V/25mA which is the maximum current that the BF862 can do (IDSS max). You can see that the frequency response is flat up to 1.5MHz. The LF response of my test mule is affected by the AC coupling of the measuring gear. However it should be around 5-10Hz.
The distortion of low-level signals is really good:
THD @ 4Vrms
Predominantly H2, it’s very nice to see THD<0.015% for a 4Vrms output. The load is 100KΩ which is the typical input impedance of an amplifier (with exception of solid state gear)
This low distortion manifests across the entire audio band (ignore the THD below 20Hz which is a byproduct of my testing gear):
THD version frequency @ 4Vrms
The nice thing to see also, it’s how well the 4P1L can drive larger voltage swings:
4P1L THD @ 10Vrms
We can see H4 popping up, however odd harmonics are lower (H5 in fact is higher than H3). THD at 10Vrms is still below 0.03%!
It’s been a long time since I haven’t tweaked my speakers. After more than 9 years I decided to change the drivers after falling in love with the Alpair 10M from Mark Audio. I listened to my friend Andy’s system (4P1L PSE driving the Alpairs) and decided to get hold of them.
A simple upgrade
As I don’t have much time left for DIY audio these days, I needed a simple solution. I couldn’t build a new set of speakers despite the love I have for some horn-type designs. Bringing new speakers was out of the question, so I had to modify my existing boxes to replace the FE167E. Sadly they didn’t fit straight on, so my friend Tony made me a pair of adapter boards to fit these. Made of MDF I painted them in grey:
I always loved the 27 valve. It was one of the first line stages I built many years ago before adventuring in the DHT world. I still have a large collection of them and I was very fond of the mesh anode ones. Please check Thomas’ blog in which he wrote a very nice note about it.
With the hybrid mu-follower (a.ka. gyrator) configuration, we can build a minimalistic and great preamp stage. The 27 has a mu of 9, so in some scenarios this may be a bit too much gain, but for many cases, it’s just what we need to drive the valve amps. Someone recently asked me for help on this, so here it goes my version:
The circuit is dead simple. The 27 is biased with a battery via a grid leak resistor (R1). C1 blocks DC from input and contributes to LF response by forming a pole with R1. 150nF is good enough but if you don’t have any, use 220nF. The operating point is 6mA looking at my old notebook. The supply doesn’t need any funky regulation, and 180-200V should do. The top FET should be either DN2540 or any other depletion of your choice. The lower JFET should be either a 2SK170GR or 2SK170BL (preferably). You can use a J310 here as well (or SMD BF862).
The sound is beautiful and THD is very low driven by H2 only, as you would expect from this triode.
If you don’t want battery bias, you can add a 1K5 resistor in the cathode with its decoupling cap and remove the battery and C1. R1 should be changed to 47k then.
More than 4 years ago I ran a lovely 71a preamp which sounded amazing. I used it for some time and enjoy its sound up until I continued with my exploration around DHT preamps. Recently I was asked about how to implement this lovely valve again.
The CX371a / 71a valve is a great candidate for a line stage with its low mu and anode resistance. In my experience you have to run it above 20mA and over 100V to get the best out of this valve:
CX371a curves
The implementation of this preamp is dead simple and a few components are needed on top of the gyrator PCB:
I haven’t starved the filaments as I found this valve not to be microphonic. If you have an 01a preamp you can modify it slightly. The interesting thing is that you can run it with just 180V. Even 150V should work and you need 25mA on each channel. A J310 or BF862 lower JFET device will work fine and you will need a heatsink for the top device (e.g. DN2540). Filament resistor is anything close to 50Ω. I used some 51Ω Russian NOS wire wound resistors, but any combination will be fine.
As part of improving my bench test gear to do sweep tests and impedance measurements, I ended up building a great preamp and buffer gig based on the SSM2019 device as described previously here.
Here is the main circuit for the preamp:
SSM2019 preamp
The circuit is same as described before. I added a rotary switch to select gain from 0dB to 60dB (ignore the diagram labels). The circuit has a DC input (differential floating) and an AC input for voltages less than 60V. There is a switch for AC/DC selection and also a switch to ground the negative input for DC mode when we don’t want it floating.
The preamp AC output is not shown but is a 1uF with a 220k resistor to ground.
The buffer circuit is similar to the above but without AC circuit and no gain selection.
I have used Jakeband teflon sockets for over 3 years. They are very well made and of high quality. Luciano from Jakeband can provide you with any custom socket you may need. This time, I requested a set of sockets for my 813 transmitting valves, a pair of UX-4 for the 300B and a pair of octal sockets for the driver stage.
813 socket
813 socket
Octal
Octal
Octal
UX-4
UX-4
UX-4
Construction overview
Jakeband uses the following materials for the socket:
Tellurium Copper (CuTe) for pins
Virgin teflon made in Germany
Jakeband sockets
The socket pins are machined with CNC from a solid core copper tellurium CuTe with tolerances within +/- 0.01 mm. The contact surface which is obtained is the highest possible.Thus there is a better grip and and a lower electrical resistance and therefore better heat dissipation (e.g. on the filament pins).
The pins undergo three surface treatments:
1° electrolytic acid copper to remove any impurities
3° 24k gold thickness 3 microns to prevent oxidation and give more surface hardness
Why using CuTe pins?
The copper has a conductivity rating of at least 100% IACS (International Annealed Copper Standard). Brass has a conductivity rating of 28% IACS. Tellurium copper(CuTe) contact pin provides up to 320% greater conductivity than a standard brass pin
Pin gold plating
The gold plate is 3 micro-inches thick. It is purely there to prevent oxidation (and increases the surface hardness) however doesn’t contribute to the conductive process. In fact gold is less conductive than tellurium copper. The gold plating is direct, nickel free.
Threaded pins
The new sockets come with threaded pins. This is an interesting concept that Jakeband is introducing and will be keen to test various connectors which will simplify the wiring to the socket.
I hope to test the new sockets shortly.
Contacting Jakeband
If you want to request them, you can reach out directly to Jakeband using the following form:
I’m not a big fan of rolling valves. Perhaps it’s likely to do with the fact that I don’t have too much free time these days. However, I do look into burning in the valves. A noticeable change found on my 4P1L preamp after 350-400 hours use. The microphonic noise reduced to a minimum whilst the sound became more rounded. I added an electronic clock (LCD module) to the HT supply to monitor the exact number of hours between any changes I made to the system. It’s very handy. I’ve been running now the same pair of 4P1L valves for over 850 hours and they sound better than before. No further mechanical expansion noise is heard during warm up, something which was noticeable at the early days of use.
The second aspect I’m also in constant monitoring is the impact of the filament starvation in the 01a. Running them at 20% less current than expected is not recommended as the filaments will not operate at expected target temperature. However, after some years of joy, I’ve not seen any issues. I may remove and trace the pair of 01a in use currently to check their health.
Our previous west meets east circuit can be improve further. In fact, a compromise made with the filament bias design is that coupling between driver (FET follower) and the output stage wasn’t DC. We want DC coupling to get best performance, to ensure we can drive well the output stage and provide sufficient grid current even when not operating in A2. This can be done with filament bias, however, since we are already introducing a negative supply, I’d prefer removing the filament bias and go for proper grid bias to get best performance of output stage in terms of maximum power and linearity.
The below circuit can be easily implemented with just few modifications from previous version:
What has changed here? Not much, the coupling cap C2 is now between the gyrator and the FET follower. The gate bias resistor R6 provides high impedance to the gyrator load to ensure maximum performance of the 01a driver (minimum distortion given size of load). Not as good as previous version, but good enough. The R6 is connected to a potentiometer which sets the bias voltage. The bias voltage is derived from V2, the -50V negative supply. You can see that this circuit will put more stress into the M1 FET as now there is an additional 25V of drop across it so power burned on this device increases.
The output of the follower is directly coupled (DC) to the output stage. The filament bias resistors are removed and we use the Coleman regulators directly on the filaments of the 4P1L.
This amplifier responds better to the grid current of the output stage once the output power goes over 3.5W. At 4.5W the distortion is just above 3% (3.2%) with a 3Vpp input signal. A tad more and you can get to the 5W and a bit more into A2 operation.
The interesting combination to explore from our previous designs is to mix some western valves like 01a into the Russian parade.
The result would be quite interesting, as the sound of the 01a has proven to be amazing. Therefore 01a driving 4P1L is possible as the 4P1L doesn’t need a lot of drive. Instead of using 4P1L as a driver, we can opt for the 01a which has a similar gain. What is interesting is that the voltage swing required by 4P1L wouldn’t force the 01a outside the zone in which is highly linear, hence, with some modifications, it can work as a great driver here.
The circuit
01a into 4P1L PSE
Instead of starving the filaments of the 01a, given the voltage swing requirements for a driver, we ought to drive it at full tilt. In the circuit above, the 01a hasn’t got the stones to drive the 4P1L pair, therefore we have added a cathode follower as explained here. The M1 follower will then drive easily the output stage.
