I’ve been playing for some weeks the marvellous UV-201a stage. The addition of the source follower brings a new dimension to this stage. However, I experienced random oscillation on this setup. I knew it was the MOSFET, but I wanted to pin it down.
I replaced the SiC version with the famous AOT1N60. That cure the issue, but I wanted to find out what was going on, of course blaming my selection of gate stoppers.
In case you were asking, how does the follower oscillation manifests? Well, simply like a random harsh noise on the background. You can tame it by restarting the HT supply, but that is a temporary workaround.
The guilty bastard was the 1KΩ carbon composition gate stopper. Here is what you get typically:
A nice crest around 5-6Mhz. And this is the most stable, try the SiC MOSFET instead of the AOT1N60, it will go wild!
So if you change the stoppers from 1KΩ to simply 100Ω, this is what you get:
And surprisingly, this is what you get with the C3M028009 SiC used before:
My FR system can do better than this at 10Mz. A flat response all the way up, I like that.
Doggy bag
Lesson learned: change the MOSFET enhancement mode stopper from 1KΩ down to 100Ω in a follower mode. Otherwise, it will go crazy!
Travelling around Europe on business is paying its toll. I’m away from home every week and pretty exhausted now. I don’t have much time free and whatever is available I spend with my family. Hence, the lack of posts recently. I hope this will change in the future.
Anyway, what’s up in the DHT world? I listened the Aa/Ba valves long time ago but never played with them. Mainly due to their higher anode resistance. With the gyrator load and the source follower output, things take a different dimension.
German precision
I have a nice stash of Aa from Valvo (globe) and Ba from Siemens. Interesting to see that curves are not easy to find, so I submit them both to the mercy of the uTracer. Nice to see the linear curves with high mu about 14 on the Ba to 30 in the Aa.
Here is an example of the Ba loadline:
With a maximum of 230V running it at about 200V/5mA and -3.4V seems reasonable to provide good headroom.
A simple but yet effective DHT stage
With such a high Ra and low anode current, the marriage with the gyrator load and the output source follower is ideal. I use this with great success on the UV-201a preamp which sounds amazing.
The circuit is dead simple as I have implemented this before again and again. If you need like 22dB of gain in your system, then look into this option. I have to implement to see how it sounds, and I guess it would be great. The Ba has a high reputation as well as its predecessor Aa.
I’ve been posting not very frequently lately. This is mainly due to lack of time and the level of business travel which reduced to nearly none the time available for DIY audio.
Nevertheless, the scarce time always pays off. It’s incredible how selective I have to be in order to prioritise which project I should work on. The list is long though.
Last time I did a quick exercise on the Ba DHT based on the curves I traced and the LTSpice simulation. Well, you always need to build and test in order to check against simulations. The result is, that you may need to adjust and learn from your practical experiences.
The Ba (like the Aa) are tricky to use. They pick up any electrostatic induced noise. You don’t need even to place your hand close, the mains noise is induced already in its plate. This force you to shield these valves if you want to use them. Am afraid, that is what it is. My friend Rob (DHTRob) warned me, thank you.
The circuit I posted here, had to be readjusted. Distortion was way too high. The operating point wasn’t good enough. You’d normally get inclined to run the valve as hot as you can, but I was wrong here with this one.
The distortion is about 0.09% for 2Vrms and 0.1% for 4Vrms. It’s about 0.15% for 10Vrms or less at this operating point. What is interesting to see below is the intermodulation of the 50Hz and 100Hz byproducts of the noise picked up on my workbench:
The 115kHz bandwidth of this stage can only be achieved thanks to the source follower.
If you’re planning to build this stage, get a nice shield screen for the valves. Either a mesh or enclose them under any sort of faraday cage. It won’t be practical otherwise.
The circuit is very easy to implement if you use a gyrator PCB and a Source Follower PCB. This reduces the complexity enormously.
Now it’s time for listening to this stage after modding the Mule….
Well, after enjoying the 2P29L as part of the Mule preamp project, I decided it was worth building it so I can continue with my adventures on the DHT land. The 2P29L sounds so good that is worth having it as a standing preamp so here is my new version in progress. Thanks DHTRob for your inspiration using the IKEA chopping boards in a slightly different way I used them so far:
Simple and neat. The 2P29L were stripped out of their aluminiun cans. Naked as when we come to life. The gyrator boards are ready and tested. Just need to get the Rod Coleman regulators and complete the wiring.
After doing all the soldering part (which I enjoy much), the preamp is now finished. It sounds as good as the original breadboard:
For the curious ones, here you have a picture of the inside:
The teflon sockets are bolted straight into the 4mm top aluminium plate. No microphonic noise this way. Rod Coleman V7 regulators set to 200mA. A pair of Russian Military NOS wire-wound resistors in parallel provides the filament bias. The gyrator PCB is set as per original circuit and each valve at 3mA. These are DC coupled to the MOSFET follower PCB set at 10mA each. The output is then taken out from a pair of FT-3 teflon caps.
Last time I wrote about the 4P1L in screen mode. It was great to see some DIYAudio member (Blitz) to post about his work on the 4P1L with screen as anode. I call it screen mode but probably is incorrect.
His post about G3 structure remind me to post this, I have tested it but never blogged about it. Yes having G3 as part of the anode structure will increase conductance and will form a nice “mesh” anode. Here it is how I implemented:
The pin 4 (G3) is now connected to 3 (G2) to form the anode. I reduced the anode voltage down to 110V to get 10mA. It could be increased, sure within the Pd limits.
The response is very good:
Here you have the distortion at 1kHz:
How does it sound?
Well, I wrote about it before. The 4P1L is one of my favourite valves. In this mode it sounds great, with a particular clear detail in the treble. I like this valve and will play it for some time to get further impressions.
Finally I managed to build a stable version of the 2P29L preamp. The wooden modular version inspired on DHTRob designs paid off. Here it is:
The result is fantastic. The 2P29L is a superb DHT, quiet and it sounds amazing. The circuit is similar to the original one but with a slightly different operating point due to the valves themselves:
I’m still running them at 20mA. This time the anode voltage has to be increased to 150V to achieve this. Consistent across the 2 valves used.
C2 was modified to fit a gift I received from my friend Vyacheslav. Instead of the usual 220nF, I fit a bigger 470nF FT-3 version which improves the LF response.
The frequency response is very good, as I published before. For the curious ones, here are some further pictures of the build:
I can’t say more than what I said before. This is a fantastic DHT, it provides a great detailed sound and is quiet. No microphonic noise and the dynamics are great. One of my favourite clearly.
If you like your VR valve glowing in the dark but you’re concerned about how quiet it will be, then you may be interested in this post.
I like the VR, I used them a lot. They are quieter than most people claim. However, they’re not the quietest if you’re looking for the lowest noise level.
Below is a very simple circuit which you can implement easily and use your lovely VR valves glowing in the dark! The VR (U1) is fed from the HT source via R1 .R1 should be sized to at least provide 10mA to the VR valve. C2 is the maximum allowed cap. R2 and C3 form the cap multiplier section. R2 isolates U1 from C3 to ensures it doesn’t oscillate. C3 to 10uF can provide about 50dB reduction at 100Hz, so great to smash out any remaining noise from the VR. Q1 provides current limitation in conjunction with R4. This will protect M1. The output is about 5V lower than the VR level. Well, that’s what you pay with a follower or cap multiplier.
The PSRR of the cap multiplier below is about 110dB @100Hz. Great performance:
The source follower PCB can be twicked easily to use for this purpose. It can also be used as an electronic choke (aka gyrator) and or a simple cap multiplier.
Q1 can be a BC547 thanks to the protection diodes D1 and D2 (15V) which will prevent from exceeding VCEO levels.
I’d use this circuit in many configurations, not just the screen supply.
Here’s an excellent build of the 01a Preamp with the gyrator PCBs and the Source Follower PCBs by Paul. I will leave to him to write down his impressions of this preamp and the build experience.
I’ve been prototyping a flexible CCS PCB. The intent is to provide a cascoded FET pair with some interesting features:
The lower FET can be multiple devices depending on the choice of reverse capacitance and transconductance. These include jFETs and depletion MOSFETs like the 2SK170, J310, BF862 and of course DN2540. For this purpose several pads are provided for SMD devices as well as TO-92 ones, just like the gyrator PCB. A protection Zener diode between drain and source can be soldered when using low VDSS devices.
There is either a string of trimpot plus a resistor to set the CCS current manually during test given the variance in the FET parameters. There is also an option to put a fixed resistor.
There is a mu-output connection provided.
The board is very flexible and can be used for multiple purposes:
shunt regulators (including VR valves)
Anode load for phono preamps, drivers, LTPs, etc.
LTP tail CCSs
I’ve been running some tests with excellent results.
If there is interest, I will run a batch of PCB to offer to the DIY community.
Since my early days of valves and DIY audio, I developed an obsession around testing and tracing valves. This led me to design and build my analogue curve tracer which I used for many years successfully until I build my uTracer, which was a great innovation in curve tracing. I do have many valve testers (some which I made myself) so why building another one?
Well, Chris Chang from Essues Technologies developed a fantastic new digital curve tracer for valves, the eTracer. There are a few things which will grab anyone’s attention on this curve tracer. Firstly, the power supplies can accommodate a large range of valves which the uTracer can’t. HT can go as high as 750V @ 300mA and the grid supply down to -170V! This is exactly what you need to test your transmitting valves or even a 300B. Secondly, the tracing speed is surprisingly fast. This is a nice feature, specially when you want to trace pentodes at various screen voltages to develop a Spice model for example.
Build process
The tracer can be purchased from Essues and thanks to my feedback now Chris offers pricing broken down between the Hardware and Software. This helps to minimise the costs associated with the hardware shipping and import duties. I purchased the full kit for assembly and I believe you can buy ready units at a premium price of course. The PCB is very well made and is fully assembled, tested and calibrated. The SMPS supply can also be purchased from Essues. There are some additional parts which you need to procure yourself:
Low noise FAN – optional.
4mm stackable plugs
HT PTFE cable
Additional ferrite beads (I added more)
I had some challenges with my shipping as there were a couple of parts missing. Chris sent me everything at no extra cost. I think it would be good to have a detailed BOM of what is supplied and what you need to buy yourself. Also some extra header pins would be nice in case you have an issue when crimping/soldering them.
The build process is simple, you only need to take special care to the polarity of the supply as it’s not well marked on the PCB and you can make a real mess if you get it wrong. There’re not many pictures in the build guide and the description is high level and doesn’t focus on the practical aspects of building the kit. I guess if you are an experienced builder you won’t have an issue. However, if you are not, you may struggle a bit to get through it so beware.
Chris has done a great job with the chassis. Sockets are provided and these are low cost Chinese ones. Some work fine, but others are really poor. This has been my struggling experience with the local socket over the years. I replaced the Octal, Loctal and B9a Noval with NOS ceramic Russian ones. The loctal doesn’t fit on the existing hole so you need to mount it on top:
The socket wiring is very tedious job. You have up to 10 pins and 10 sockets. Chris provides you with ferrite beads, but if you want to allocate one bead on each socket connection (like I did) you have to add several more. I had some NOS Russian ferrite rings at hand which I used them:
Here is a side view of the tester. You can see the SMPS supply on the right and the marks I added to make sure I remembered well the polarity of the cables into the PCB connector. Also I had to use a different pin in the GND connection of the 6-pin header as I was short of the original ones. I used a circular one which did the trick:
The result is a fantastic finished product. Below is my initial test of a 4P1L once I hacked the top plate to place the NOS socket:
The clever design of the chassis allows you to swap in the future the socket plate and cater for different sockets. This is a nice feature. I’d love to add the transmitting tube sockets (e.g. 813/GM-70) as well as a varied set for Compactron (B12C – 12 pin) and others.
The Software
The software is very easy to use and intuitive. I managed to find some bugs which helped Chris to refine the beta releases. There is a major release planned which will incorporate features missing. There are some drawbacks in my opinion which some of them are being addressed by Chris in the new version:
Heater voltage compensation on the 0.1Ω sensing resistor. This is needed for larger heater currents
Curve graphics are basic, you can’t zoom/re-size without loosing your current plot.
Display of measurement (e.g. current at specific voltage) with the mouse dragged over the curves.
Variety of tracing types (like the uTracer does) for Screen drive, transfer curves, etc.
If you don’t set properly the axis type with pentodes you will get the wrong measurements. The eTracer doesn’t enforce or check your setting for pentode mode.
The speed is very good, here is a video with an example of the 4P1L pentode curve tracing:
4P1L pentode tracing
A quick test on the 10Y/VT-25:
VT-25/10Y tracing300B tracing
Also Chris has developed a conversion tool to use the output files from eTracer to match the input requirements of Derk Reefman’s ExtractModel tool to create pentode Spice models. I haven’t tried this yet, is on my list
Conclusion
The eTracer is an amazing piece of technology. I’d highly recommend you get one if you work with valves often. As every new product, there is further enhancements which will come over time as it can be developed and refined via Software. The hardware is solid and provides an ample voltage range.
I still have lots of tests to do with this little unit, but didn’t want to miss the opportunity to write up an initial review. I look forward to the new Software release.
I’ve been using my choke-input DHT filament supplies for many years. I’ve got many of them as you’d expect. Something I do really hate is to build power supplies. It’s dull and boring, but they’re a necessary evil. I’m afraid I have to admit.
Recently I experienced the dreaded smoke of capacitors blowing up. It’s actually funny in hindsight, however at the moment of the fireworks you don’t laugh. I killed one Coleman regulator as well with my experiments on a flexible DHT supply. That made me revise the design and the stress put on the components, in particular when you’re using filament bias in anger, as I do.
Anyhow, over the last 3 months I’ve been playing with SMPS supplies to try to get the commercial available ones quiet enough to be used with a DHT preamp. I started with classic filter stages like CLCLC, morphed to gyrator filtering (which wasted a lot of heat) and then resisted using LDO regulators which I knew it would do the trick. However, getting rid of the HF noise is a daunting task.
Actually, I have a variable HT SMPS supply built which sounds really good and is extremely quiet. It can deliver 2 channels of 600V/100mA. For a filament supply, the SMPS challenge is of a major league game. You can get the noise to about 1mVrms but the harmonics spread well over the mid and high range. Big chokes have also a big leakage capacitance which makes the choke not that effective at filtering off this HF noise. 1mVrms in 600V is fine, however 1mV in a 16V supply which si feeding the filament bias resistor is a problem.
After giving up, my patience these days isn’t the best I have to admit. In particular when time available for DIY audio is very limited. So I said to myself: “sod the SMPS, I will get a nice pair of custom made transformers with multiple taps and job done”. And that is what I precisely did. JMS transformers in the UK provide an amazing service. I’ve mentioned them in the past several times. I get all my power transformers from them these days. I ordered a pair of split bobbin, with outer copper screens and multiple taps to cater for all the voltage ranges needed for my output stages. From a 4P1L to a 300B.
As I always use choke input supplies for filaments, I used the LL2733s I have in stock and carelessly wired it on 400mH (series bobbins) which provides a huge voltage peak output when the minimum choke input current isn’t in place. This happens at start up, the voltage will raise to the level of cap input supply when the filament is just starting up thanks to the gentle raise of the Coleman regulator. The result is the high voltage peak output which can damage the capacitors.
The solution was to wind the input choke with the two bobbins in parallel to get inductance down to 100mH. Also the resistance is reduced significantly which avoid wasting too much voltage drop on the choke. Also adding a tuning cap (C3) to make the supply to operate in a hybrid mode between choke and cap input is great to dial in the output voltage.
Here is my supply which I use for the VT-25 and the 4P1L preamps:
I’m not going to go through this basic (quasi) choke-input supply because it has been covered extensively by others. Rod Coleman in fact, provides a nice set of recommendations worth following when building LT power supplies. I have some components at hand which I tend to reuse on my supplies. R1 and C2 are the snubber network which I trimmed manually. C3 is the voltage running cap, a polymer aluminium one to ensure low impedance. I use the EPCOS 22mF caps all over my filtering stages and in between I add a CMR choke. C6 and C7 are to improve HF filtering. It could also be a stack pair with the centre connected to earth. I haven’t had any issues with any HF noise on these power supplies, so never had the need to improve them.
Bear in mind that most of them are mounted in a piece of wood
Here is the PSUD2 simulation to demonstrate the initial peak of the choke input supply when the regulator is slowly coming up. You can see that it goes up to 26-28V. With a higher L1, it goes higher and can damage the caps as it will get close to the Cap input maximum voltage. Also the reverse voltage can put some pressure on the diode bridge, so you want to avoid this.
So here is my final version of the “flexible” filament supply. I have the transformers tapped at 0.5, 1, 2.5, 5, 10 and 20V. This allows me to adjust and trim the output voltage to avoid wasting power on the Rod Coleman regs for example. The filtering stage is slightly different (it has an HF choke at the output L4) which is inheritated from my SMPS experiments. I reaped off the filtering section of the SMPS and reused it here. You can make your own tweaks:
This supply will be used in a varied set of DHT output stages with fixed bias.
Here is a view of the final build using some nice chopping boards I got as a present
Recently I’ve spent some time designing some PCBs for my own use and why not to share them as well to the DIY audio community. Currently I have a flexible CCS board for anode loads and a tail CCS, which is the one I will write about on this post.
I came up with a 3cm x 3cm PCB board which provides the maximum flexibility for a tail CCS circuit. Below is the diagram:
There are 2 options to go here: BJT or MOSFET. There are some minor differences, but the same board can accommodate both circuits, which is pretty handy.
Let’s start with the MOSFET version. This one is the classic ring of two but with a MOSFET (M1). The board can accommodate TO-220 as well as TO-92 and SMD versions. The protection zeners. Z1 and Z2 will ensure you can use a HV supply at +B. R2 is the gate stopper and can be either through hole or SMD. You can set the CCS current with the resistor R3 or if variable current or precise adjustment is needed you can use instead a trimpot (P1) and R4 in series. This CCS is very tight and has a good temp-co as well as stability considering changes to +B. Given the MOSFETs have lower transconductance compared to the BJT, the output impedance won’t be that high, but will be high enough for most purposes. I considered implementing a cascode stage, but wanted to keep the board small and simple. The temp-co is negative, which means the current will drop when temperature rises. MOSFETs on the other hand are available at higher voltages, so this stage may be useful to implement when higher negative voltages across the CCS are required (e.g. 100V or more)
The BJT option is an improved design. The NPN can be any of your choice: TO-220 or TO-92 and can accommodate low voltage to even high voltage devices (e.g. 400V) . In most of the applications you will probably use voltages below 50V at the tail CCS. The stability (both on temperature as well as on variations of the power supply) is improved thanks to the shunt voltage reference. U1 can be either the LT1009 or the ZR431. Both TO-92 or SMD options are available on the board.
Well, I still need to do some further tests on the prototype PCBs but so far so good. I think this tiny PCB is very flexible and can be used in many circuits like cathode follower stages as well as LTPs, etc.
If there is some interest out there, I may make the PCB available.
There is one DHT which attracted me from Emission Labs which is the EML20. If you’re looking for a mid-mu DHT valve these days, this one is the way to go. In order to meet with the two key requirements of linearity and low microphonic noise, EML made a great effort in producing this valve.
For me it’s a great candidate as a driver or for a Spud amp for headphones. I will likely use them in several places, but will start with a preamp stage, as you would expect from me anyway.
First step: tracing the EML20AM
Using the eTracer, I managed to plot the nice set of curves up to 650V:
EML20AM curves
The next step was to produce a nice set of curves to make it easier the process of developing the Spice model:
The model adjusts very well. Unfortunately there’s no capacitance data on the EML data sheet. I have to say that this is an area where they fall short. These days you would expect a detailed data sheet. That is not the case so far, I hope they improve on this in the future.
**** EML20AM TRIODE Composite DHT *****************************************
* Created on 01/21/2018 12:10 using paint_kit.jar 3.0
* www.dmitrynizh.com/tubeparams_image.htm
* Plate Curves image file: EML20AM triode.png
* Data source link: www.bartola.co.uk/valves
*
* Model and curve traced by Ale Moglia
* (c) Bartola Ltd. UK, London, UK
* valves@bartola.co.uk
*
* Curved traced with eTracer
*
* Notes: missing capacitance data, not available in datasheet
*
*----------------------------------------------------------------------------------
.SUBCKT DHT_EML20AM 1 2 3 4 ; P G K1 K2
+ PARAMS: CCG=1P CGP=1P CCP=1P RFIL=3.33
+ MU=20.7 KG1=1230 KP=300 KVB=60 VCT=-3.68 EX=1.4 RGI=2000
* Vp_MAX=600 Ip_MAX=60 Vg_step=2 Vg_start=0 Vg_count=14
* Rp=4000 Vg_ac=55 P_max=9 Vg_qui=-48 Vp_qui=300
* X_MIN=75 Y_MIN=51 X_SIZE=509 Y_SIZE=556 FSZ_X=1178 FSZ_Y=705 XYGrid=false
* showLoadLine=n showIp=y isDHT=y isPP=n isAsymPP=n showDissipLimit=y
* showIg1=n gridLevel2=n isInputSnapped=n
* XYProjections=n harmonicPlot=n harmonics=y
*----------------------------------------------------------------------------------
RFIL_LEFT 3 31 {RFIL/4}
RFIL_RIGHT 4 41 {RFIL/4}
RFIL_MIDDLE1 31 34 {RFIL/4}
RFIL_MIDDLE2 34 41 {RFIL/4}
E11 32 0 VALUE={V(1,31)/KP*LOG(1+EXP(KP*(1/MU+V(2,31)/SQRT(KVB+V(1,31)*V(1,31)))))}
E12 42 0 VALUE={V(1,41)/KP*LOG(1+EXP(KP*(1/MU+V(2,41)/SQRT(KVB+V(1,41)*V(1,41)))))}
RE11 32 0 1G
RE12 42 0 1G
G11 1 31 VALUE={(PWR(V(32),EX)+PWRS(V(32),EX))/(2*KG1)}
G12 1 41 VALUE={(PWR(V(42),EX)+PWRS(V(42),EX))/(2*KG1)}
RCP1 1 34 1G
C1 2 34 {CCG} ; CATHODE-GRID
C2 2 1 {CGP} ; GRID=PLATE
C3 1 34 {CCP} ; CATHODE-PLATE
D3 5 3 DX ; FOR GRID CURRENT
D4 6 4 DX ; FOR GRID CURRENT
RG1 2 5 {2*RGI} ; FOR GRID CURRENT
RG2 2 6 {2*RGI} ; FOR GRID CURRENT
.MODEL DX D(IS=1N RS=1 CJO=10PF TT=1N)
.ENDS
*$
The Spice Model works really well. You can download it here: EML20AM spice model
Still need to undertake the usual THD measurements for various operating points. It looks like it’s best to run it around 300V. My current DHT preamp HT supply will not fit the optimal operating point, but playing in LTSpice, I can implement this one easily with filament bias.
So the developing is progressing. This time I tested the first PCB prototype of my enhanced “Swenson” regulator which I baptised Swenson+. It’s a great circuit and performs really well. Although component selection and track cree page and clearance is for about 1kV, I think it should safely operate at least to 800V. Not really thinking to use it at those voltage levels, but rather as phono stage supply instead, which requires much lower voltages.
As usual the focus of the PCB is to provide as much flexibility in component selection (SMD and TH) as well as use cases.
Still some layout refining and further testing but so far so good! Here is a taster which shows at least 60dB rejection of the incoming power supply noise which wipes out all rectified-related harmonics (100Hz etc). The 100dB 50Hz and byproducts is likely to come from my workbench:
Many times I get emails from DIY Audio builders who embark on building a DHT preamp when they don’t need gain, but instead what they need is a simple line stage to drive their amplifiers and interconnect cables effectively. Then they come back asking: “can I reduce the gain of the 01a or 4P1L preamps?”
For those who don’t need the gain, here is an interesting idea which brings together several design decisions which makes the DHT sound to its best. The challenge with many of the best sounding DHTs of low-mu is that is very hard to implement with filament bias. I’ve done a driver with a 46 in filament bias which was a crazy idea. I could turn of the heating with the amp running! It was a nice experiment though. With exception of the 71a and some other few DHTs, if you’re looking for good anode current and low ra, you’re in trouble. The 300B, 45/46, 50 and some other variations can’t be used in filament bias.
Subject to your religious beliefs in audio, you may not want to add a capacitor in the cathode, like me. I won’t dive into this discussion which is pointless as is a personal decision. If you continue reading this is simply because you value the sound difference in the DHT without a capacitor bypass in the cathode. Keep reading then…
A simple and yet effective way of creating a bias similar to the filament bias arrangement is shown below. When the current requirements are low, you can derive it from the HT and you end up with the simplest arrangement. Unfortunately when you need +20V bias levels, you will not be able to do this. If you reduce the bias current over RFIL1, you need a higher value of RFIL1. This translates into a higher resistance reflected at the anode and a degenerated stage which reduces further the gain. A compromise point can be achieved between filament bias currents (>1A) and a high bias resistor.
The above circuit is very simple. A classic DHT stage in hybrid mu-follower mode with a gyrator PCB. the 300B is run very cool at 25mA which is fine for a line stage. The 27V bias is generated by a third power supply which feeds a CCS set at 100mA which in turns generates the 27V bias requirement. The filament is floating and run at DC regulated with a Rod Coleman reg. It can also be a Tentlabs reg. The same 60V/200mA supply can be shared as the crosstalk between channels is highly minimised thanks that the CCS PSR.
Thanks to the low anode resistance of the 300B, the bandwidth of this stage is amazing. You can get a flat response up to 2MHz. This forces us to be cautious with the necessary stoppers etc as the stage could oscillate with stray inductance and capacitances at stake.
I’m sure many builders out there will entertain this idea and adventure on building one.
My test rig for DHT/IHT stages (and even Pentodes) has evolved over the years. Lately I settled with some nice modifications to allow testing the majority of valves I have. I use a modular socket system, nothing fancy and can add/remove a source follower stage at the output. There is also a screen regulator in case pentodes are submitted under the mercy of the jig.
Here is the simplified diagram. I added a nice fixed bias supply formed by a SMPS PCB board which delivers up to 400V, however the output is dial to about 100V. Then I use a Swenson Regulator to knock down the noise by about 100dB. A simple pot provides the voltage needed between 0 and -100V. It can be tweaked for whatever range you need. The pot is 20T wirewound so allows a fine adjustment on the bias:
For the curious builders, here is the rig mounted on a piece of floor board:
After tweaking my test rig to fit a variable supply for fixed bias, I proceeded with the tests with the EML20AM valve.
After playing a bit today with the valve, found that it picks up hum like the Ba or any mid-mu valve due to the large grid which acts like a proper antenna. If you screen the valve is dead quiet. I used the same rig test with a CX301a to compare and definitely the 20A picks up more induced noise from the workbench.
The H3 levels are higher than I’d have expect, about 7-8dB difference with H2. The harmonics have a nice waterfall shape which is good and THD shows how linear the valve is, very nice. You get only 0.08% for 40Vrms. Didn’t test further but I expect it will perform really well:
As a driver is outstanding. You won’t find many valve out there this linear capable of swinging 40-50Vrms with just THD below 0.10%.
If you need the gain as your system is only one output stage (like mine) then this valve is a good option. With a step up transformer of 1:2 or 1:4 can make a very nice DHT driver for a 300B/50/211, etc. For larger swing volts I’d probably look at biasing it above 220V, albeit it can work very well there. The anode current above 15mA makes the stage able to manage the output valve’s capacitance without any slew rate distortions.
Here is a quick snapshot of the frequency response using the gyrator load:
A nice crest around 5-6Mhz. And this is the most stable, try the SiC MOSFET instead of the AOT1N60, it will go wild! 
