I received the first batch of PCBs after several tweaks during the prototype phase. I’m very pleased with the result:

I built several boards to test the various combinations that this flexible CCS board offers:

• Top MOSFET (DN2540/IXTP08N100D) mounted close to the edge to allow the CCS to be bolted to a big heatsink. Alternative a small heatsink over the TO-220 can be used for 1-2W dissipation.
• The lower device can be multiple options depending on the target use
  • High bandwidth and low noise CCS for phono stages (up to 20mA):  
    • 2SK170 (TO-92) 
    • BF862 (SMD)
    • Alternatives to BF862 due to EOL: 2SK3557 and CPH3910
  • High bandwidth and general CCS use with jFETs:
    • J310 (<60mA) – TO-92
    • J112 (<20mA) – TO-92
  • Protection zener for low voltage devices (jFETs)
  • High-current CCS
    • DN2540 (<120mA) – TO-92
    • DN2540 (<170mA) – SOT-89 (mounted on reverse PCB side)
    • DN2540 (<500mA) – TO-220
    • BSP129 (<250mA) – SOT-223 (mounted on reverse PCB side)
    • BSP149 (<660mA) -SOT-223 (mounted on reverse PCB side)
  • Gate stoppers can be either through-hole or SMD (mounted on reverse PCB side)
  • Mu-follower output for low output impedance configuration
  • Current measurement resistor: 1 or 10Ω
  • Current setting arrangement:
    • Fixed resistor
    • Multi-turn trimpot and series resistor (optional)

Below is the diagram of the CCS board:

This is an universal cascoded FET CCS which has been used extensively in audio. It can be used for anode loads (phono, drivers), parafeed output stages, tail CCS, LTP stages, bias supplies arrangements, VR valve supplies regulation, etc.

Still need to do all the documentation, but if you are interested you can send me a request here:

This is a long overdue post which I never had the time to write about. I was hoping to get my measurement gear down to where my system is to take a final FR sweep analysis of my 4P1L PSE amplifier, but never got around to do it.  However, after the recent posts in DIY Audio, it was time I shared the measurements made and my listening impressions of the NP Acoustic Transformers.

I found them long time ago when searching around for transformers. I took the risk and purchased a pair of their 3K2 SE OPT (NP3.2SE18F100) with amorphous core and with very promising specs. As usual, specs can be very tricky in particular when the testing conditions are not close to real life and also when they are not even documented! You will find this to be very common in transformer manufacturers out there. Each transformer will behave differently depending on the driver as well as the overall implementation circuit. 

For this reason, I decided to put the transformer to the mercy of my test bench and here are the results.

Testing the 3K2 OPT transformer

First test was made on a pair of NP3.2SE18F100 shipped by NP Acoustics. I found a slight inconsistent HF frequency response between the pair due to the winding technique which was corrected by NP Acoustics who sent me another pair of OPTs. 

The first step was to measure the OPTs with the LCR:

Interestingly the HF response (-3dB) is about 50kHz. If you look at much more expensive OPTs like the Monolith Magnetics S-9 (which doesn’t have an  amorphous core), you would be surprised that the HF response isn’t much higher, however as expected due to its specs the LF is much better:

Now, that is pretty impressive for the price of the OPT I think! Below there are some listening impressions, so hang on.

I’ve been using this OPT in my latest version of the 4P1L PSE amplifier for about a year and I have to say that I’m delighted with its performance. 

Testing the 5K OPT transformer

My next adventure was on the NP5SE18F100 5KΩ:8Ω OPT.

Below are the LCR: measurements 

You can see that given the step-down characteristics of this OPT, the Cip and Lsp are much higher than the 3K2 version.  In theory again, the response should be around 5Hz to 107kHz.  

In real life, here is the response with the 4P1L valve. In this case its biased to 30mA but doesn’t make a big difference. The HF is limited to 37kHz:

Here is a comparison with the LL1623, which has a slightly better response, HF up to 42kHz:

I was slightly disappointed with the response compared to a Lundahl given the latter isn’t amorphous. However is still great value for money.

I exchanged several emails with the NP Acoustic team and unfortunately couldn’t settle a clear view on their FR testing protocol. They didn’t take on board my measurement feedback and is slightly disappointing to see the FR specs shown on their site for this 5K OPT. Despite I can’t claim myself an OPT expert, I think they have work to do to improve the winding of this transformer. 

OPT Reviews

A year ago, I shared a pair of these OPTs with my friend Andy Evans. I also asked him for a review, which is posted below, and as said earlier I struggled to find the time to write up this blog post. Obviously because I wanted to measure them on my system, which I didn’t. 

In my view, these OPTs are great. I haven’t replaced them with the Monolith Magnetics as they are reserved for my 300B amp. However, considering the price, they are great value for money. The 4P1L provides a unique clarity which in partnership with the amorphous core OPT brings a level of detail on the treble which is outstanding.  Yet, the bass is powerful and strong. For a single ended amp, I’m impressed how good it is. The 4P1L PSE and these OPTs match really well with the Alpair speakers I have.

Here is Andy Evans’ review from a year ago:

"Amorphous core transformers have the reputation of offering increased clarity and low level detail, particularly in the upper mids and treble. For some this comes at the expense of some degree of harshness or hardness on first listening, but there are many user reports of break-in after 50 hours offering a more smooth perceived response.

My experience was similar to the above. My first reaction was that the NP32 output transformers lacked the overall smoothness and sweetness of the Lundahl LL1664/70mA I have had in my system for around 6 months, and which was my preference from a bunch of other transformers. My output stage consists of two 4P1L DHT pentodes in triode mode, used in filament bias which means a small value cathode resistor with no cathode bypass cap. Output is similar to a 2a3, but gain is around mu=11 which allows me to use a simple 2-stage setup with a DHT 01A Gen 2 preamp as the input stage. This is also in filament bias, and both use choke input filament supplies with Rod Coleman regulators. Despite the 01A tubes dating from the 1920s and 1930s, this is a modern design using advanced solid state designs in the filament supply and the gyrator active load. The combination of the 2 stages offers a very high level of detail and good reproduction of the timbre of acoustic instruments. Timbre is important to me as a professional conservatoire trained musician. Input is Mac based through Audirvana+ software and an ES9023 DAC. Speakers are full-range Mark Audio Alpair 10s in infinite baffle columns.

So for the first hours of listening I found the sound lacked the sweetness I had become used to. Listening material was a lot of opera and vocal music, both classical and also jazz and popular vocals such as modern gospel and singer-songwriters. Massed vocals are quite a severe test of a system’s smoothness and resolving power. Added to which they can show up any resonance peaks or high frequency oscillations. Resonance peaks in the treble are a common downside of single unit speakers such as mine, and this offsets the gain in clarity from not using a crossover. So some degree of resonance can be laid at the door of the system, even though the Lundahl OPTs seemed to mask this.

After a week’s listening every day the strengths of these OPTs became increasingly obvious. The most obvious gain was clarity. Orchestral and vocal textures were so much clearer, words so much easier to hear, that I started to listen at lower volumes. This was quite noticeable. I was getting all the information I was used to at a lower overall volume. This also offset the greater perceived energy I was getting in the treble, which was bothering me at first listen. Whether there were effects of the much reported  break-in happening I can’t be sure – it’s a subjective thing. But for sure, I was starting to get quite absorbed in the new clarity I was getting. Vocal music was a joy with the lyrics clearly audible.   

So would I go back to my smooth and sweet Lundahl OPTs? Well - they're excellent OPTs and superior to most conventional core models, but no, I wouldn’t. In comparison the sound from them is less immediate, more recessed. The detail may still be there somewhere, but it’s less obvious and the temptation is therefore to turn up the volume to hear it better. So I’m addicted to amorphous now. I have to have the clarity, now I know it’s there in the recordings I listen to. It’s too difficult to go back knowing that some of the low level detail is going to go off the radar. These OPTs are a hard master – they show up any faults in your system and demand the highest level of detail and smoothness in the signal presented to them. What you get in return is a level of directness and transparency in the sound that is a whole new experience for users of tube equipment. Yes – it’s addictive and I’m addicted!

These transformers are well made and very attractively finished. They are very much in the tradition of Tango and Tamura, with large square cases and a circle of pins underneath for attaching the connections. Builders of high-end tube equipment now have a new name to add to the select few famous names of audiophile transformers, which have become rarer and rarer to find in recent years. NP Acoustics deserve to do well if they can make equipment of this quality at competitive prices.

Andy Evans MA, MBPsS, PCC

Director: Performance and Media Coaching UK

January 2017"

Well, I hope you find this review useful and hopefully support this great manufacture of hi-fi OPTs. They do a good product at a fantastic price point.

Now fitted on the 4P1L PSE stage:

Had to redo the top plate when new frame arrived (arghhh). Here it’s the new Preamp just rebuilt

Very close but now ready for ETF in a few weeks

Frame is made out of walnut root. A great job by Danielle Ardito.

Packing up the system for the shootout

ETF.18 has been an emotional journey. I moved house literarily when I got back from ETF so my life has been more than hectic over the past few weeks.

I promise I will do a write-up of this amazing experience. There’s a lot of people out there who would love to attend so is my duty to reflect and share as much as I can.

In the meantime, I wanted to share the lecture I gave at ETF on DHT preamps. It was a challenge on its own but went really well. This was my first ETF and without knowing the audience I had to guess the level of detail, entertainment and expectations of an unknown audience. I knew a fair bit of the ETF folks, but audience was big and wide.

I struggled to find the time to prepare this lecture I have to confess. Between moving house, house building works, my second daughter’s arrival , weekly work travel and everything else, I seemed not to find the time to get this done. Thank you Morgan Jones and Rod Coleman for proof-reading and making this an easier task.

I hope you enjoy it. There are some notes on the slides I put together for the people who didn’t attend ETF. Otherwise the slides aren’t of much use on their own.

I’m writing this blog entry whilst enjoying the lovely Bourbon that Pete Millett gave me on the way back. Thanks Pete!

ETF Lecture on DHT Preamps (with notes):

DHT-Preamps-ETF2018-final-notes

ETF Lecture on DHT Preamps (slides):

DHT-Preamps-ETF2018-final

It’s been far too long since I last posted on this blog. With the limited spare time I’ve got these days, I concentrated in setting up the new workshop and system since we moved back to our place. I’m nearly there, so now it’s time to get back to work

I made some updates to the “gyrator” PCB. I’ll stop referencing it gyrator from now on, since the name is misleading. However, it got popular that way. Nevertheless, it’s a hybrid mu-follower circuit but if only if you take the output from the anode, it behaves like a “gyrator” from a frequency response perspective. If you’re interested in this circuit in more detail, please read the lecture I gave last year at ETF.18. You can download it from here.

Back to the board, here are the few changes made:

1. Moved the trimpot P1 for easier fixing of the board with the M3 standoffs.
2. Added a gate stopper resistor (R9) to avoid oscillation at low anode current (<10mA) when using high-gm MOSFETs in the lower position J4. This was evident with devices like BSH111BNK
3. Added an LED (D4) indicator and a series resistor (R8) at the drain of M3. This enables indication of:
   1. Normal operation subject to  value of R8
   2. Source current into load (e.g. like in A2 operation) subject to value of R8.
   3. Short output to ground. Depending on duration and current limitation of power supply, this may prevent damaging M3 MOSFET. Not guaranteed, but in some scenarios will work.

Here is the circuit diagram for reference:

Of course you can continue using jFETs in the board. However, I found that the BSH111BK/BSN20BK and other high-gm perform really well in this circuit. They are still plenty and available out there, not like the BF862. Unfortunately this one is EOL (see lecture for replacement options).

Here is a sample test of the 01a preamp stage with this new board:

In this case with the use of an IXTP3N100D2 and BSH111BK operating at 3-4mA the frequency response is flat to nearly 380kHz. Same combination of FETs in a Rev07 board will oscillate at LF due to the lack of gate stopper.

With a Rev07 IXTP08N100D and BF862 board, the response only gets up to 150-180kHz depending on the valve. This is not a limitation whatsoever in audio band, let’s make this point clear.

A snapshot of a completed board:

in the following picture you will see the three new components added so you can compare to the previous Rev07 version. Also note that this is a prototype version so it doesn’t have an ENIG Gold board finish:

The new boards were tested on my 01a preamp. They sound amazing, as good as the previous ones. Interestingly enough, I made a mistake during breadboarding and the LED turn on really brightly with 100mA shunted through the output. The FETs survived this time thanks to the limiting resistor and the current protection I have on my bench power supply.

The new listening room (at least for some time) is now ready. One of the loft rooms is now purposed for guests and for music. Everything is set up. The 4P1L PSE with Monolith Magnetics S9 OPTs, the ER801a driver plus the 01a preamp are in place. Have the DAC and the Lenco turntable as sources. A Kozmo stepped attenuator and input selector. Subwoofer, record cleaning machine and the Alpair 10 in the reflex boxes. I removed the dome tweeters to experiment for a while.

My Starlight transport and discrete DAC are on the box still. I need to fit the CD towers on the wall first. Tons of CDs to rack again. Yet, TIDAL is a good alternative. Love exploring other records I don’t own.

At the moment playing the Howling Wolf record we enjoyed late at night with Jeffrey Jackson and Dave Slagle at ETF.18.

Life is good….

Long time ago I used this SLCF circuit as an output buffer with excellent results. It sounds as clean as it can get in my opinion.  Recently, I made up a PCB to hold this circuit with a varied of options:

The circuit is actually much simpler than it looks. Let’s start with the main objective of this board. The MOSFET M5 is there to keep the anode-cathode voltages as constant as possible, therefore reducing the modulation distortion produced by the anode resistance of the valve. Please read the previous posts on SLCF to understand how this circuit works. There are 2 options for biasing M5. The simplest one is by using the resistor divider formed by R4 and R6. Alternatively, you can use the CCS formed by M1 and M3 for better PSRR. The CCS will provide a tiny current (circa 300uA) to create the right bias voltage across R6. R7 and D7 provide a peak current indicator. R10 is the cathode bias resistor when you’re bootstrapping the grid via R12. There is an option to DC-couple the input (see next diagram). The tail CCS is formed by T1 and T2. Options are provided to use different transistors depending on the sink current requirements. The array of LEDs will generate the right bias for the CCS. The LED current is sourced by R5 which can be connected to +B or GND depending on the supply arrangements. The current of the CCS is fixed by R13 or a trimpot (P1).

This board provide all flexibility needed when looking to implement a bootstrapped-CF (SLCF) circuit.

If you want to connect the SLCF directly coupled to previous stage then you can do this:

Testing the board

Well, I built the previously used circuit using the beloved 6J52P/6Z52P Russian pentode.  A D3a will do a similar job here. There are multiple high-gm pentode/triodes to use otherwise.

Here is my somehow abused board:

The performance is superb, have a look below. I used the simple bias method first and a MOSFET which has the b2b zeners included in the package. Here’s the circuit:

Let’s look at the frequency response performance first:

Flat from beginning to end. Ignore the capacitors used as it’s what I have at hand. You need to tune them for a LF response. Interestingly, the stage is flat up to 4MHz.

The distortion is extremely low, here is a snapshot at 3Vrms:

Harmonics are down from 80dB so very clean profile. It sounds as good as it looks though.

I need to make some changes to the PCB (doh) but what can I say. Great device to use at the output of a DAC or to drive TVCs or low impedance SS amplifiers!

Chasing that sound

The 45 DHT is probably one of the best sounding valves out there. In fact, I have struggled to get a similar level of detail and timbre in a 300B or 4P1L output stage. Even my 814 SE Amplifier (which was class A2 and had thoriated-tungsten filaments) couldn’t replicate that sound. I posted time ago my incarnation of the 45 single-ended amplifier here.  The main challenge with this valve is that it can only put out there nearly 2W, not more. With its 10W anode dissipation, you will struggle to get more juice from it in class A at a low distortion level.

However, if we look at a push-pull amplifier with the 45, we can hopefully retain the timbre characteristic of the valve, despite it won’t  be a single-ended one. Well, I love good PP amps, so why not?

If you read carefully the datasheet, you will find that you can extract a lot of power from this valve in A2 operation. Precisely, AB2 in PP mode. Up to 19W from a pair! Wow, that’s impressive. I’d be happy with 10W on my Alpairs. Having enough headroom is good from a dynamic response point of view. You won’t probably need more than 3-5W on average.

Tracing the 45

I had at hand a pair of Sylvania 45 which I submitted to the mercy of the eTracer. The one shown below measure well above 80%, so it’s a good reference. Nevertherless, I’m aiming on a fixed-bias design, so providing the pairs are matched and measure decently well, then it should be fine.

Here are the lovely curves of the 45:Creating the Spice model was easy. You can download it from here: 45 DHT Sylvania model.

The output stage

The suggested operating points from the datasheet are pretty obvious. At the end it will come up to what OPT you have at hand. Somewhere around 3.2KΩ – 4KΩ  Zaa is preferred. Below is a snapshot of how much we could get out of the valves with a 4KΩ transformer:

Of course it doesn’t match the datasheet recommended point, it’s fairly close. With 240V/28mA on the anodes you will need a bias of about -46V. The grid needs to be driven pretty hard up to close 200Vpp to achieve full power. The grid will also sink about 20mA of grid current during A2 operation.  Please ignore the THD estimate from the tool as it’s wrong.

Simulating this circuit in LTSpice with an LL1682PP (which is 8.8KΩ Zaa), can deliver 10W at 1.1% THD, mainly H3.

Below is the circuit I’ve been toying with for some time:

The circuit is very simple. The output stage has a DC filament regulators (e.g. Rod Coleman) and a mix of fixed bias with filament bias for better stability. As we need to provide the A2 grid current, a source follower is mandatory here. The grid bias is generated from a Swenson Regulator PCB (or a Rod Coleman fixed bias board) and adjusted independently per board to ensure there is balanced anode currents.

The grid voltage can peak to 60V so at least 80V are needed on the top MOSFET of the SF PCB. Likewise, the grid can go as negative as -140V so -160V is needed as the negative supply.

Building the above in separate chassis provide some flexibility in terms of the drivers used. I was going to use a differential pair of pentode valves in a hybrid mu-follower configuration. That is with a gyrator PCB and a tail CCS. very simple and yet effective. It can deliver sterling gain and linearity for 200Vpp operation. Similarly, triode-strapped pentodes like D3a, 6J52P or similar would work very well here.

Lots to work on, just wanted to share initial ideas on this design. Probably will inspire others.

Pentode drivers are very interesting. You can get excellent results out of them. Here is my version of a differential pair using pentodes. I want to try them out in the 45 PP design I wrote previously

People may get scared with the above diagram thinking: this is way too complex. In fact for me, it’s not. Using the PCBs I have it makes it easier and simple to build.

The differential pair is formed by V1 and V2. They have a tail CCS which sets the shared cathode current. Each valve has a gyrator PCB and a resistor to ground from the anode which sets the gain. The anode voltage can be adjusted by the gyrator but will not make a big difference on the anode current of the pentode (think of a flat anode curve). Remember the gain is defined by gm and Ra. However, if we vary the screen voltage we can adjust minor differences due to unmatched valves.

Regulating the screen is very good to keep distortion down. The additional complexity here is to create a resistor divider to adjust slightly (and independently) each screen voltage. The voltage reference from VR1 is isolated byR3 and R4 to avoid oscillating VR1 by adding additional filtering capacitance as C3 and C4. The SF boards are used only to provide MOSFETs driving the screen current to each valve.

You can simplify this circuit of course. I will try this and experiment a bit.

Back at the beginning of the year I purchased a pair of Emission Labs 300B for my new amplifier version (still working on it). I own several 300Bs but was keen to try these out given their reputation.

Jac from JacMusic sent me a matched pair which I traced them with the eTracer:

Lovely linear curves as expected. Here is the Spice Model I developed if you’re interested in playing with it:

EML300B-Spice-Model

You need about 140Vpp to push the 300B to full output power. Here is the operating point I worked out on my previous implementation:

Yes it’s rather hot at 35W. With a Monolith Magnetics OPT of 3K2, you should get 8W when driven with 140Vpp. You can use this excellent driver:

The D3a in triode mode is probably one of the few valves which can do 250Vpp without any issues and very low distortion. Below is a THD plot for 240Vpp which is impressive. You won’t get this level of performance if it’s not with a hybrid mu-follower. Here is the amplifier circuit:

If you are interested in less gain and adding a pre-amplifier stage (e.g. like an 01a) then you can look at implementing the 6e6p-DR (or a 6e5p) like in my original design:

I really like the sound of this valve and has a lower H3 component when driving large volts. See below:

Here is a plot of the frequency response of the driver. Less gain of course (i.e. 30dB) but with great performance:

Alternatively, if you need more gain, one of my favourite drivers is the 6SF5GT. This one can give you 40dB in sterling performance with a hybrid mu-follower (a.k.a. gyrator) topology. You need to bias the valve a bit higher than the D3a and similar triode-strapped pentodes.  Therefore the HT has to be increased to 450V. You can add a drop RC to feed into the 300B output stage (R3 and C2 below):

Here is the frequency response of the driver stage:

There are so many options to work on. Even the lovely Russian 6C45P triode (which is a wild oscillation beast) can be used at a gain of 32dB:

Amazing response as well. You need to set the anode voltage to 200V and the cathode resistor (un-bypassed) to 120Ω.

For me this time round, I will build the 300B output stage separately so I can play with different drivers. Yes, I want to experiment as usual.

A belated test of this simple, yet effective PCB. I made it as small as possible, however in order to provide flexible connections, it’s actually double the size. Still at 4 x 4 cm is small enough.

You can solder the 2mm PCB connector or alternatively you can use the 2-pole block connectors. It provides the full functionality for a cathode follower, however you can bypass some elements when using it as a tail CCS for a differential pair for example:

You can use either R2 or P1 to set the CCS current. T1 is any high hFE NPN you have at hand. I use BC547C because I have plenty. For the top device you have 2 options in order to accommodate different power requirements. I use the 2SC3503 as it has low capacitance and works up to 300V with 7W of PD. T2 is for TO-126 devices and T3 is for TO-92 ones. You can use a plethora of different NPNs here depending your voltage requirements.

If you want to self-bias the cathode follower you have R3 and R4. If not required, R3 should be replaced with a wire jumper and R4 ignored.

R1 set the current for the LED voltage reference array for both T2/3 and T1.

R11 is a build out resistor. Again if not needed, can be replaced with a wire jumper.

Cathode Follower Test

As per my previous test for the SLCF board, I used the 6J52P/6Z52P high-gm pentode in triode mode. Here is what I breadboarded quickly this morning:

Being lazy I didn’t play much with the operating point. I left the CCS set to 15mA, not 20mA as per previous example. The results of this stage are impressive. Here is the neat frequency response:

The stage is flat up to 4.2Mhz. Also the distortion is extremely low. Here is an example of the harmonics at 3Vrms:

You can see the level of harmonics is very low with a THD of 0.003%. It’s just down there.

Here’s a picture of the rat nest!

A bit less HF response due to the lack of bootstrapping in the anode and similar distortion performance to the SLCF in terms of distortion. Both stages are operating at different bias levels so can’t compare apples to apples.

PCB prototype was a success…

This design is inspired on the great work from Gary Pimm. The 47 is a sterling directly heated pentode and worth using it. I have several on my stash, including some good globe versions.

The following topology is well known.  The push-pull output stage has local feedback (a.k.a. “A la Schade”) between anode and grid. Obviously the drivers are pentodes. Subtle difference is that I added a source follower in between grids to provide negative bias and grid current drive.

The drivers are based out of the beloved 6J49P-DR. Need to finalise it. The stage can do 8W at less than 0.6% THD with an OPT with Zaa 8KΩ.

I prototyped a new PCB with a simple circuit to provide Screen voltage regulation based on a VR. Quite handy in a circuit like this. Let’s see how it turns out.

Looking at my notes and archive from 4-5 years ago, here are the “Schaded” anode curves:

Quite a lot of FB applied here to get the anode resistance down to 800-900Ω.

This will make a nice sounding amplifier!

A unique sound

If you have a small room and high-efficiency speakers, then keep reading. If not, you can still enjoy reading about probably the best sounding output valve in my view. The 45. I wrote about it few times and have to say, it’s still one of the best. Better than a 300B but unfortunately can do only 2W. You can use it in Push-Pull and is superb. Have a look at this.

Lately I had a few requests from different people about this circuit which was my main amp for few years. If I were to build it again, I’d do some changes like the following ones.

For the minimalists, you can implement the 7193 driver with SiC diodes in the cathode. Actually is a hybrid setup between 4 SiC diodes and a cathode resistor, unbypassed of course. You don’t need the SIC PCB, you can hard-wire the diodes, however the PCB simplifies the build. Each diode drops about 1V, less with small current like in this driver. The cathode resistor R3 provides the additional bias. It’s a mu-follower/SRPP circuit so no need to bypass R3:

7193 driving the 45 with fixed-bias

The operating point was found years ago upon testing and listening. The gyrator PCB is straight forward with a BF862 (or a BSH111BK with Rev08), The pair R4 and C3 provides the voltage drop for the output stage. C3 is an oil cap or film – MKP (DC-Link). You don’t want an electrolytic here as is the final cap.

The NP accoustic transformer was reviewed before. It’s a great OPT and the value for money is fantastic. You won’t get an amorphous core OPT sounding like this one at the price point.

The 7193 drives the 45 grid directly. I use Rod Coleman’s fixed regulator and the output valve has an NP Acoustic amorphous-core transformer which is great value for money.

You don’t need necessarily to heat the driver valve with DC. I do it, but isn’t mandatory if you pay careful attention to the heating wiring.

The output stage topology was discussed before at length. I’m running the 45 rather hot but works well within specs.

You need 120Vpp to drive the 45 at full tilt in A1. This is a great SE amp and is as simple as it gets. You won’t regret it and you will get hopefully the best of the DHT sound transparency and level of detail.

Improving the basic circuit

Ok, you want to improve this, don’t you? The best way to avoid blocking distortion and to improve the performance of the output stage is by adding a Source Follower to drive the 45 grid. A very simple step here. You can reduce the coupling cap (C2) value as the grid resistor here can be as high as 1Meg (or more). No issue with the grid current as the MOSFET in the source follower doesn’t grab any at DC. The 7193 driver can source/sink enough gate current to the MOSFET. The source follower provides enough current (source and sink) with the 20mA idle current at very low impedance to the 45 grid.

The negative swing can go as low as -120V so -150V is preferable to provide enough headroom. The positive swing can be 0V in A1 or we can allow some headroom for A2 operation with a +50V supply. You can get about 3.5W maximum in A2 operation if you drive the grid positively to +20V:

The 7193 has a gain of 20. This may prove to be a tad short. If you have a pre-amp then is fine. I used to run this amplifier with a 26 DHT preamp stage. However, if you want more gain, you can replace the 7193 with a higher gain driver like the D3a. Also the C3g, 6J52P, E810F, 6e5p/6e6p, 6C45p and others works really well here. Similar circuit. Only change is to dial the right anode voltage in the gyrator PCB as well as setting effectively Rk.

It’s a circuit you can play with and swap different drivers with minimum changes (just dialling anode voltage and changing Rk). You can experiment a lot and listen to the differences. I’d add to the chassis 2 sockets (one octal and one 9-pin noval). That way you have all the choices needed.

You can of course experiment as well with a DHT driver. Plenty of options. Likely you’d like to use a SUT before the DHT. Best drivers in my opinion are the 4P1L and the 2P29L. Both with low mu of about 8. You will need the SUT to deliver a gain of 5 to 10.

You won’t regret in building this amplifier. You can get the new Emission Labs 45 series. You can try the EML45B (I haven’t had the chance yet) if you need more power. It can deliver twice the output power allegedly with the 45 sound. If you have tried it, I’m keen to hear your comments about it.

Sometimes less is more. The 45 valve is a great example of that mantra. I remember enjoying several late hours of music. Playing Coltrane and Miles Davis through my Fostex FE167e was a real pleasure. The 45 is magic, please try a bit of its addictive tonic, you won’t regret it.

I’ve got asked the question on how an EML20 (or EML30) can drive an EML45B. Answer is straight forward and hopefully this brief blog entry can shed some light to this request.

I recently looked back again on the EML20A. With a gain of 20, it’s a good candidate but it hasn’t got just enough gain to drive the output stage to full power:

As you can see on the curves above, the 45B needs 176Vpp to produce about 4.6W. You can push it up to  6W in class A2 with 200Vpp input signal. I’m not sure how the Emission Labs grid will work with positive grid current so I wouldn’t try this without checking with the manufacturer.

With a 2Vrms input signal you will get about 113Vpp at the output of the EML20A. Not enough for full power and will give you 2W. You will need to add an extra input stage or a simple 1:2 SUT at the input.

The EML20A has to be biased at higher voltage. Here is my suggested operating point:

The EML20A has to be run at 300V at least. The EML30 is an alternative. I don’t know how linear and good this one is as I haven’t tried it myself. However looking at the curves you will need to run it at 400V at least. With the voltages involved and swing required, probably better to run it with choke load.

As I’ve done it in the EML20A preamp circuit, the valve is biased with filament bias. The SiC diodes get hot, but can handle it. They drop close to 1.5V when running through large currents like in this circuit. So we will need 3 diodes to get about 4.5V. In this way we can allow 180Vpp swing. We may want to increase this to 5V bias and give some further headroom for 200Vpp or more. The HT is 410V and you notice as well that the negative raw bias supply has to be around -200V as the 45B needs -88V at its grid and you need at least 100V swing voltage (in both directions).

I’m sure this would make a great sounding DHT amp with these 2 nice valves!

A friend from Canada ordered a pair of boards and I used the opportunity to test the newly arrived hybrid mu-follower (aka as gyrator) Rev08 PCBs.  He will be using these boards in the 300B design posted here.

Here is the driver circuit diagram:

As you can see, I increased the anode voltage to get about 14mA of anode current. This was due to the DUT I had at hand. The protection resistor R8 is set to 150Ω so you just get a slight dim light on the LED under normal operation. It will lit completely under an output short (which likely kill the LED) or under A2 operation (e.g. grid current)

The actual load resistor is 100KΩ which when connected to the audio test interface reflects a total input impedance of 50KΩ.

Here is a test of the frequency response:

And here is the harmonic profile under 200Vpp swing:

Nice results, very happy.

The 6F12P is used (by many) in Europe (mainly) in RIAA circuit. You would wonder why? Well, it’s a great Russian frame-pentode which has high-mu and high-gm therefore driving larger currents at low distortion.

Recently, Anatoliy Lisovskiy from Wavebourn posted this great summary review on Russian valves which made me immediately connect and look on my files. I had traced, experimented a lot with this valve before using it on some RIAA designs:

The history of tubes:
1. 6S19P triodes were designed for military power supplies as pass regulators. They have low internal resistance, and because they are long for power dissipation, they also are very linear. Grids are relatively far from cathodes, so they need high voltage swing to drive them, and pretty good current due to capacitances. But they are happy in A1 mode, don’t need A2 drive like other triodes like 300B, for example.
2. Gu-17 tube was copied after QQE 03/12 to work in UHF transmitters. It contain 2 tetrodes per bulb. Electrodes are long, so it can produce high linear swing when used as a LTP.
3. 6F12P is a frame grid pentode-triode. It has both high mu and high gm, unlike 12AX7 that has low gm, or 12AU7 that has low mu. As the result, amplification stages with such tube have amplification like 12AX7, but faster, like 12AU7, and can provide higher load current with lower distortions. Frame grid tubes were designed at the end of a vacuum tube era, in the quest for better linearity.
Military QC diamond stamps look ugly, though…

Anatoliy Lisovskiy – Wavebourn.com

Triode Section

1. Mu is about 100. Gain should be maximum 40dB
2. total input capacitance measured is 418pF @gain of 40.5dB (x106)
3. From data sheet Cgk=4.6pF.  Given Cin=Cgk+(mu+1)*Cag -> Cag = (Cin-Cgk)/(mu+1) = 3.86pF

**** 6FF12P_TRIODE1 ******************************************
* Created on 06/12/2016 11:08 using paint_kit.jar 2.9
* www.dmitrynizh.com/tubeparams_image.htm
* Plate Curves image file: 6FF12P-triode1.png
* Data source link:
*
* Traced and model developed by Ale Moglia valves@bartola.co.uk
* www.bartola.co.uk/valves
*
* 1. total input capacitance measured is 418pF @gain of 40.5dB (x106)
* 2. From data sheet Cgk=4.6pF.  Given Cin=Cgk+(mu+1)*Cag -> Cag = (Cin-Cgk)/(mu+1) = 3.86pF
*----------------------------------------------------------------------------------

.SUBCKT TRIODE_6FF12P-1 1 2 3 ; Plate Grid Cathode
+ PARAMS: CCG=4.6P  CGP=3.86P CCP=0.26P RGI=2000
+ MU=100 KG1=165 KP=768 KVB=66 VCT=0.188 EX=1.44
* Vp_MAX=200 Ip_MAX=30 Vg_step=0.2 Vg_start=0 Vg_count=10
* Rp=4000 Vg_ac=55 P_max=40 Vg_qui=-48 Vp_qui=300
* X_MIN=75 Y_MIN=51 X_SIZE=371 Y_SIZE=497 FSZ_X=1072 FSZ_Y=683 XYGrid=false
* showLoadLine=n showIp=y isDHT=n isPP=n isAsymPP=n showDissipLimit=y
* showIg1=n gridLevel2=n isInputSnapped=n  
* XYProjections=n harmonicPlot=n harmonics=y
*----------------------------------------------------------------------------------
E1 7 0 VALUE={V(1,3)/KP*LOG(1+EXP(KP*(1/MU+(VCT+V(2,3))/SQRT(KVB+V(1,3)*V(1,3)))))}
RE1 7 0 1G  ; TO AVOID FLOATING NODES
G1 1 3 VALUE={(PWR(V(7),EX)+PWRS(V(7),EX))/KG1}
RCP 1 3 1G   ; TO AVOID FLOATING NODES
C1 2 3 {CCG} ; CATHODE-GRID
C2 2 1 {CGP} ; GRID=PLATE
C3 1 3 {CCP} ; CATHODE-PLATE
D3 5 3 DX ; POSITIVE GRID CURRENT
R1 2 5 {RGI} ; POSITIVE GRID CURRENT
.MODEL DX D(IS=1N RS=1 CJO=10PF TT=1N)
.ENDS
*$

The pentode triode-strapped LTSpice 6F12P triode2 model below:

**** 6F12P TRIODE2 pentode section triode strapped ******************************************
* Created on 06/12/2016 11:21 using paint_kit.jar 2.9
* www.dmitrynizh.com/tubeparams_image.htm
* Plate Curves image file: 6F12P triode2.png
* Data source link:
* PENTODE Section (triode strapped)
* Traced and model developed by Ale Moglia valves@bartola.co.uk
* www.bartola.co.uk/valves
*
* 1. total input capacitance measured is 605pF @gain of 39dB (x89.1)
* 2. From data sheet Cgk=6.6pF.  Given Cin=Cgk+(mu+1)*Cag -> Cag = (Cin-Cgk)/(mu+1)  =6.64pF
*
*----------------------------------------------------------------------------------
.SUBCKT TRIODE_6F12P-2 1 2 3 ; Plate Grid Cathode
+ PARAMS: CCG=6.6P  CGP=6.64P CCP=1.9P RGI=2000
+ MU=86.9 KG1=120 KP=512 KVB=435 VCT=0.228 EX=1.2
* Vp_MAX=200 Ip_MAX=40 Vg_step=0.2 Vg_start=0 Vg_count=10
* Rp=4000 Vg_ac=55 P_max=40 Vg_qui=-48 Vp_qui=300
* X_MIN=76 Y_MIN=50 X_SIZE=630 Y_SIZE=516 FSZ_X=1333 FSZ_Y=664 XYGrid=false
* showLoadLine=n showIp=y isDHT=n isPP=n isAsymPP=n showDissipLimit=y
* showIg1=n gridLevel2=n isInputSnapped=n  
* XYProjections=n harmonicPlot=n harmonics=y
*----------------------------------------------------------------------------------
E1 7 0 VALUE={V(1,3)/KP*LOG(1+EXP(KP*(1/MU+(VCT+V(2,3))/SQRT(KVB+V(1,3)*V(1,3)))))}
RE1 7 0 1G  ; TO AVOID FLOATING NODES
G1 1 3 VALUE={(PWR(V(7),EX)+PWRS(V(7),EX))/KG1}
RCP 1 3 1G   ; TO AVOID FLOATING NODES
C1 2 3 {CCG} ; CATHODE-GRID
C2 2 1 {CGP} ; GRID=PLATE
C3 1 3 {CCP} ; CATHODE-PLATE
D3 5 3 DX ; POSITIVE GRID CURRENT
R1 2 5 {RGI} ; POSITIVE GRID CURRENT
.MODEL DX D(IS=1N RS=1 CJO=10PF TT=1N)
.ENDS
*$

Pentode Section (triode strapped) performance

Just be careful given the transconductance of this valve, if you don’t put the screen and gate stoppers it oscillates widely.

Here is the response of the pentode section, again 39-40dB of gain expected:

Triode Section performance

The interesting aspect of this valve is that you can still have a wide BW (up to 100kHz) without the source follower. If this valve is going to be used as a gain stage in a RIAA, then is better to avoid the SF from a noise point of view unless this is the last stage of the RIAA and you need to drive longer cables. In that scenario, the above configuration is the best solution.

In  summary, a great valve which is perfect for a RIAA stage. Very linear (as seen on the above measurements) and with sufficient gain to work well in a pre-amplifier topology.

Over the last few months life has been very hectic for me. Between work travel and family, I’ve struggled to find time to work on any audio related stuff. You probably noticed this as have reduced significantly the amount of posts over the last 4-6 months.

This is sad, as have many things piled up for sharing. Either way, my focus has diverted slightly over the last few months. I decided to get back to music performance after so many years. This is a great decision, which I’m very happy about. When I thought about which instrument to play again, I immediately looked into my tenor saxophone – which I played for so many years. Unfortunately due to family commitments, I can only play during night time, so saxophone was ruled out. So, what should I do? Of course, my mind then got stuck on to my next big instrument preference: the Chapman Stick. When I was at university, I discovered this amazing instrument. I believe it was Tony Levin from King Crimson the first one I heard playing it. Anyhow, as I couldn’t either afford one and also there were probably none available in Argentina at that time, I decided I would build one myself. Luckily my dad was very experienced in woodworking and crafts in general. With his help we designed and build one, which sounded really good.

Fast forward more than 20 years, I ended up buying a new Chapman stick myself. An amazing instrument which is extremely flexible in terms of what you can create, tone, expression and range. I’m yet a beginner, but a happy one. I’m struggling to find the time though to practice, which is what you need. I will get there eventually.

As you can imagine, the return to music performance ignited a few projects on the diy audio side. First one was the construction of a Marshall clone PP amplifier. I will write about it in due course.

Second project turned into diy distortion stomp boxes. Who’d have thought so? I spent many years in my quest to reduce distortions and now I’m going on the reverse path!

I spent quite a lot of my spare time to try different Germanium PNP transistor in different fuzz and/or distortion stages. Amazing results as they do sound very nice. I made some PCBS and am yet to complete some final stomp boxes for them.

With the Locky curve tracer I traced tons of Russian germanium transistors. When I get the time I will publish the curves as am sure there are quite a few folks out there who would be interested in grabbing their hands on them. There aren’t many available so this is gold dust!

Let’s now return to DIY hi-fi audio now…

You’d probably ask yourself what is this whole thing of “Sand and Iron”. Well, as you know I play a lot with sand with valves. But I love iron, specially where it performs at it best. A year ago I concluded a series of experiments (which I believe I shared on DIYAudio) around headphone amps. I posted here as well on this topic. The outcome of my experiments was the 2P29L headphone amp which is shown below. An amazing circuit which I highly recommend you build if you like HP amps:

2P29L Headphone Amplifier

The circuit is very simple, yet there are some subtle things about it. The 2P29L was covered extensively here, but there are 3 things about this valve which took me down the path of selecting it for the HP amplifier:

1. Tone: this valve has an unique tone and detail as a DHT.
2. Lack of microphonic noise: this is paramount if you’re looking to implement a HP amp
3. Filament current is very low – just 140mA.

I breadboarded this amp after so many different tests and found it to sound amazing.

Worth spending the money in the Sowter 8665 transformer. It’s a marvellous piece of transformer designed for headphones. The resulting circuit performance is really good with very low distortion and nice harmonic profile. I love this HP amp. I yet have to build this in a nice case.

Surely you’d ask yourself, can I build this without the output transformer? You can, but you will need to modify the hybrid mu-follower to balance current into the 32 ohm load. Also the 2P29L will struggle to provide the current needed. Even for just 100mW you will need about 60mA. You would need to have a PSE arrangement. Also distortion will be unavoidably higher. Yet, it’s possible.

What I ended up doing was to replace the cathode filament resistor for an arrangement of a SiC diodes and a small resistor for filament bias. Why? Simply because it sounds better. My SiC diode PCB holds only 6 diodes and to get 7V, I added a pair of 15R in parallel to deliver the extra volt needed:

The value of C2 is critical to provide a smooth frequency response and will depend on the transformer used and wiring configuration. If you have a 300 ohm headphones you will have to adjust the value of C2. I used Russian PIO ones which sound great. Any other boutique capacitor of your preference would work as well here.

For the 10Y/VT-25 fanatics you can easily turn your preamp stage into a similar HP amp:

The VT-25 can be microphonic enough for a HP amp. You should take enough measurements to tame this noise. The filament bias array made up of the SiC diodes will get pretty hot. They can withstand without heatsink with about 1.2W on each diode. You will need to allow for good ventilation. Otherwise you can bolt them to the chassis as the plastic TO220 case is very convenient.

Recently I have some queries about using the VT25/10Y preamp stage to drive an output stage and ways of eliminating a previous DHT stage. I did this long time ago with my 814SE Amplifier which has the crazy idea of a 46 driver with filament bias (yes it was my heating system) and an input step up transformer.

Here are two ways in which you can use the 801a/10Y/VT-25 in a driver configuration with a SUT. First example below shows a SUT made up with an interstate transformer:

The SUT gain may be limited here and you need the right OPT which can deliver lots of volts at very low distortion. With the parafeed arrangement though, the OPT can be smaller and easier to build to provide very good frequency response. The VT25 will carry the heavy load of the output stage grid capacitance which is multiplied by the SUT configuration. This arrangement also facilitates the negative grid bias of the output stage, which is quite practical as it has been done for ages. Limitation here will be on A2 performance though.

What I’d do instead is this second example. The SUT is located at the input stage. In this way the source will be challenged by the Miller capacitance of the VT-25 valve multiplied by the SUT effect. You should take this into consideration or otherwise you will risk of loosing some treble if the source + volume control cannot handle this:

The SUT can be easily 1:5. There are good Lundahl examples as well as the ones made up by my friend Dorin. I posted some measurements of his transformers before.

In this way, the output stage can operate to its best given the help provided by the source follower. If you are planning to use a transmitting valve like 845, 211, etc. This is mandatory.

Hope you find these examples useful.