As time is very limited these days, I’m focused in continuing my modular building approach in LEGO style. I have developed several PCB modules which are flexible to be used in multiple amplifier and pre-amplifier designs. Now, I used the power of PCBs to build some additional supporting modules to speed up my breadboarding over the IKEA boards. Not the most elegant approach, but building becomes a very fast process this way.

You will see what I’m saying when you see a few of the following additions:

The previous modules are very handy for point to point wiring as well as connecting transformers with different configurations, etc.

Now for HT and LT connection between boards (e.g. power supply and amplifier) I use extensively the Neutrik NL2 series. They come with 2 screws in diagonal array. Here is a small plate adapter:

For signal input/output I use the high-quality Jakeband RCA connectors. The plate has also a GND connection if needed:

Same PCB plate can be used with a generic RCA connector:
Speaker connector:

For sockets, here are a few examples of Octal and Loctal dual socket modules as well as 7-pin miniature socket and UX-4 from Jakeband:

Here is a handy INS-1 Neon module for HT indication. It comes with a built-in resistor and connection and holds tightly the INS-1 Nixie bulb between the 2 plates:

There are a few more to come, this is just a snippet.

I’ve got hold of a NOS quartet of UX-120 Radiotron valves made in 1925. These are 95 year old beauties. Unbelievable how well conserved they are and operate.

This valve developed by RCA was the first output valve intended to be used with dry-cells (3 No. 6 cells in series). See the data sheet extract below:

There is no much data available on this valve, if you have some please share and let me know. Here are the curves I traced from a sample which measured 114%:

I developed a Spice model as well for the ones interested in this valve:

You can download the model here: UX120-DHT-model

It should be an interesting candidate for a line stage with its low Mu. With a hybrid mu-follower using the gyrator PCB should work nicely. Could be biased around -6V/6mA which should give a very low anode voltage (aproximately below 70V). This could be implemented with a SiC filament bias array of 6 diodes. The tungsten filament requires only 125mA. Big question here is how microphonic the valve is? Only way is to test it. I will add it on the list

I made a few additions to the PCBs available. Check these out:

1. Moving Coil Transformer (1:16 or 1:32) –  (LL1943)
2. High Level Mic Transformer PCB (1:2 to 1:8 configurable) – (LL7903)

Available until the PCBs run out

In the process of rebuilding my old 300B amplifier, I decided to make a new filament power supply. It all came up around the components I had at hand, so it could’ve been improved but that meant extra cost:

I have a pair of custom-made JMS transformers with multi-taps secondaries. This helps me tweaking the right output voltage. Anyhow, any 15V transformer would do. Perhaps you want 14V to ensure you don’t dissipate too much on the filament regulators (e.g. Rod Coleman regulators)

I used my flexible LT supply PCB which allowed me to build this in less than 1 hour. I also used some existing chokes made in the UK by “Spirit” which are ok for this purpose. The Lundahls are in use, so can’t reuse them:

I used some SOT-128-2 schottky rectifiers but any other should work as well. Resistors are wirewound and the CMR choke is what I had in stock as well. A simple 15mH/3A should do fine.

The output measured well at 9V with a 6R load which dragged 1.5A. A bit more than the 300B but should be a good indication of performance. Also ripple level is good at 2.5mVrms. The rest will be cleaned up by the regulator itself

I hate saying goodbye. However, that’s life and it’s sad to see the great people jump on the next bus. I was very young when I first heard McCoy Tyner. I was learning saxophone when obviously Coltrane took me to him and his piano. Unique style, complex tempo and chords made Coltrane explore the deep end. That Coltrane Quartet was magic.

He did a lot for music indeed. More importantly he made a big difference to a few individuals on the planet, including me. Probably my love for music and thrive for innovation was due to him and Trane.

Many moons back went to see Mingus Big Band in New York. On the table next to us, McCoy sat down with Mingus’ wife. We were crazy about his presence but didn’t want to make him feel uncomfortable. We had a nice brief conversation and enjoyed the music as equal audience members. Still, his presence next to me disturbed my mind.

I will spin a Love Supreme and My Favourite things on vinyl tonight. Have a good trip McCoy, I will continue to enjoy your magic…

It’s been a while since I experimented with preamps again. I settled as a permanent setup with my 01a/ER801a which I love. However, I wanted to continue with my experiments so I built a new Mule, let’s call it “Mule 2” for now. It’s based out of multiple PCBs which simplifies the construction process and reduces build time:

I wanted to revisit as a first preamp, the old and beloved RE-084/A-408 DHT which I never had the chance to share. This is a great valve. I had a decent stash of Valvo pairs which sound like angels. Here was my previous incarnation:

I used filament bias and also a source follower as was driving directly the 4P1L output stage. Performance was great as it was the way it sounded:

THD only rises to 0.03% at 20Vpp showing how linear this valve is!

I also played with an unbypassed cathode resistor instead of filament bias:

As  you  can see below, the frequency response is impacted at LF with the increased reflected impedance at the anode due to the unbypassed cathode resistor. Gain is also reduced a tad. This can be sorted with the increase of the gyrator capacitor C1 which I left at 100nF:

My latest version is very similar and went back to filament bias, but with SiC diodes instead. I love the sound of these diodes. I biased the valves at lower voltage but will likely push them up to 6mA/150V as before:

I tested the new mini-gyrator PCBs based out of SMD devices. They work brilliantly!

I love the 46 DHT. To the extent that I went on the crazy idea to implement it in filament bias. It was my winter heating indeed. Stupid idea, but sounded brilliant.

I had several questions about the 46 lately. As I’m rebuilding my 300B amp, I’d love to play with the 46 again. Here is how I’d implement it today:

The gain of the stage is about 40. The 46 has a mu of 5 in triode mode. Instead of burning loads of heat in the filament bias arrangement, I use a degenerative cathode resistor (unbypassed). The input set up transformer is the brilliant LL7903 wired in 1:8. I have a PCB made for this which it’s very handy. A zobel arrangement (C1+R2) helps taming down the frequency response at HF. Running the valve at 30-33mA is ideal. 200V at the anode is ok and the HT may be adjusted depending on the swing you need here. The stage has all the bearings to drive whatever output stage. I will add a source follower prior to the 300B as I run the 300B in fixed bias mode.

Hope this helps

Ale

Time ago I made a set of gyrator boards for Chul in South Korea. His build skills are amazing and look at his work below. I love the structure and design, it’s simply beautiful.

Chul asked me also to engrave one of my cartoons on his preamp, a real honour. Look at it and judge it yourself

Hopefully Chul can chip in here and leave his impressions of the preamp. I believe he tried before other configurations and also is looking to build a 10Y soon!

From left to right: 1) Standard Rev08 PCB with full flexibility of FET and TH components. 2) Rev 1.0s board with SMD except Rmu, protection drain resistor and LED as well as space for any nice big PIO capacitor. 3) the smallest version of all, all SMD except Rmu, film cap and standard TO-220 top FET and multiturn trimmer.

Very happy with the results in the board development. It does take more time and precision (you will need a microscope) to work with the MELF resistors and the SMD components in general. However, it’s worth the trouble if you’re looking to reduce the footprint.

Long time ago I developed a CCS board which provides full flexibility in terms of FETs/MOSFETs used. It was a 2-terminal CCS, a very well known circuit.

Recently, I looked at developing a small PCB to hold a basic CCS with depletion FETs. The top device is for the IXTP08N100D or DN2540 in a TO-220 case which can be bolted or use a clip-on heatsink. The lower device is a SOT-23 SMD FET. I use from BF862 to J105 /112or J113. Current ranges from 1mA to about 40-50mA depending on the sample IDSS you get on the FET. There are a few combinations quite interesting for very low Tempco. In the ETF.18 lecture I published on the blog, there are a few points and formulae to get the best tempco configuration with these FETs.

This CCS isn’t the most stable one from a tempco point of view. There are other and more complex circuits to look into if you need so. Generally speaking, these CCS are more than stable enough for the valve audio circuits we work with. And that is more than good for me.

Having a small footprint of 3 x 4 cm on the board is pretty good.

A Brave New World

Surely you’re as tired as I’m with COVID-19. One of the best things I can do to distract my mind is to keep myself away from social media. Every stone you turn, there is COVID or a statement about it. I won’t moan as I have a job for now and a healthy family. Some members of my family were infected but nothing major. I can only say is that the world has change. And so my day to day life looking after the young family whilst working is a real challenge. Starting my fifth week of lockdown, I have to distract somehow my mind at times, otherwise will go mad.

A New Concept

Last year at ETF.19 I was discussing with my friend Pete the challenge I had around finding time to build audio gear these days due to the family and business demands and also how much I hated metal work and never been good at (nor had the space for) woodworking. So historically had to rely on other people to build me a nice wood case for my amps and learnt how to design and make my own ultra high-quality top plates, which are expensive so not for a prototype at all. Anyhow, I mentioned that I could only work very fast in building gear when I used the prototype approach over the Ikea chopping boards, an idea I borrowed from my friend DHT Rob.

As a result of that conversation, Pete put me in contact with his friend Simon Mears who is brilliant at woodworking. Check his website when you have some time. Simon is an amazing character, very creative and open to help and contribute with ideas. Sadly never managed to pop around his place to listen to his horns due to the lockdown but is on my to-do-list for when this is over.

I needed to make a cabinet to fit the IKEA chopping boards and hide them from my wife and kids. The cabinet has some rear ventilation and holes for wiring mains and output cables:

The previous picture shows my initial test using the Headphone Amplifier. I can slot at least 4 boards and have some additional supports made for adding extra boards as needed. There should be plenty of space.

The 300B Amp

Ok, why 300B? I had a previous one which I loved and want to get back to it for the simple reason that I want to experiment with different output stages and will start with it.

In principle the design is as follows conceptually:

1. DHT output stage with a (huge) nano-crystalline core transformer.
   1. The transformers are massive and may push me to implement this design in mono-blocks. The are 3K3/8R or 4R. 12W output at 25Hz and can do 5 to 80kHz (-3dB) for 800R source impedance. I will take a picture and post it when I get them out of the storage are in my loft.
   2. Fixed bias with +50 and -200 to -300V supply
   3. Source follower DC coupling
   4. Flexible driver
      1. D3a
      2. 46 with input SUT
      3. Other valves to try!

As the amplifier is modular, I can reuse the power supplies and create different amplifiers easily. This is what I’ve been doing for years with the DHT preamps and allowed me to test multiple versions very quickly.

Here is a view on my first chopping board for the fixed bias supply and 300B filament supplies. I’m in COVID-19 reuse mode so I’m making this amplifier with everything I can find in the workshop. Luckily have some multi-tapped  power transformers made to order by JMS. All with split bobbing and copper screening. They also have multi-taps in the secondary for voltage adjustment and trimming:

The second board will hold an HT supply and another set of filament supplies intended for the driver. The HT supply uses my PCB design for a hybrid rectifier, film cap input and an LCLCC output. Last capacitor in the chain is a WIMA DC-Link 45uF one. Kemet electrolytic caps and a pair of 15H/200mA chokes made many moons ago by Ogonowski. The HT mains transformer is from Thomas Mayer and the 100VA split bobbin filament transformers are from JMS as well. I used them in my 45/46 SE and my 6C4C push-pull amps years ago:

I can only spend about 30min to 1 hour over the weekend when I get a break from the kids. In that time I can quickly do the drilling and get all the mounting work done. I use plastic hex standoff which I drill partially the board (i.e. only 5mm deep, not all the way through the board) with a 2.6mm drill bit. I then use an M3 tapping tool and a bit of oil to allow self tapping the plastic hex standoff without getting it damaged and ensuring a perfect locking. I then reuse whatever standoff I have at hand (brass, plated, plastic) to mount the Speakon connectors which I use for wiring power to the other boards. Also made recently some PCBs to hold octal sockets for the valve rectifier, in-rush surge protector / HT delay, etc.

Enough for now, it’s been long since I devoted a bit of time to this blog but was worth sharing some of the slow project on this amp. Surely some of you may find it inspiring of ideas and may share some of your experience.

Keeping up with the request from readers for help isn’t easy. There is a lot of people out there trying my designs and building several preamps which is very encouraging and rewarding. Happy to be able to share what I learn myself and help others.

For the ones out there who have now more time than usual, please build build and build and then listen, listen and have an audio overdose.

For everyone, keep safe.

One positive thing I can get out of this COVID-19 lockdown is that people are building more audio equipment. Here is a great story from Sridhar Ganti who has built a few DHT preamps over the years.

VT-25 with SMPS filament supply

Sridhar strived to get the VT-25 preamp working with SMPS supply for the heaters. As everyone knows, SMPS are challenging to implement due to their wide-band ripple noise.  They are ok to implement in an output stage where the step-down effect of the output transformer and the SNR is significant to make this noise very low at the output . However, at a preamp level is another story. The signal levels are tenfold or more lower so the noise injected at the cathode is crucial. Think about the DHT as a differential amplifier. Is not just what gets at the grid, it’s also about the cathode. What the triode amplifies is the signal difference between the grid and the cathode.

I tried and failed with the VT-25. I couldn’t get rid of the midband noise. I added a series choke and it wasn’t good enough. In the end I gave up as I needed to add such a filtering stage which wasn’t worth the trouble.

On the other hand, Sridhar went through this. Despite my discourage to avoid the SMPS, he went through the experimental journey and succeeded.

Here is Sridhar’s write-up on his preamp:

The 6 diodes give a drop of approx 7.8V. The 801’s are shock mounted on a sub-chassis using 6.32 sandwich mounts from McMaster Carr, link given here : https://www.mcmaster.com/catalog/126/1548. This really helped controlling the micro-phonics. I also mounted the power transformer on small rubber grommets to further isolate power transformer vibrations.

I used Meanwell SMPS supplies with the filtering circuit provided my friend and mentor John Levrault. Getting the preamp right was a little bit of a journey, I had messed up the filament connections, on one tube I connected the +ve to Pin1 and -ve to Pin4, and on the other tube it was reversed. The results was, when I powered up, there was power supply white noise. Since I had not used meanwells in the linestage before, I suspected that to be the culprit. Needless to to say after 1 week of tinkering around, John pointed out my mistake, and when I connected the pins with the right polarity as given in the 801A datasheet, the preamp powered up correctly, and it was dead quite. The initial Meanwell I had went up to max 14V, and with the 7.8V drop across the Sic, it was cutting it fine to get the 7.5V across the filaments. While trying to fix the initial hum problem, I ordered a pair of 24V meanwells, I used those with about 6.2 Ohms of dropping resistors to get the voltage to 15.5V The dropping resistors provide further filtration anyways.

Now coming to the sound, but before that, a little word on my old preamp — it was a 10Y, using LCRC-Coleman filament supply and cathode biased and had a 4:1 step down transformer from Dave Slagle — Intact Audio. So it was a very good benchmark. However, the gyrator load with duelund caps and meanwells filament supply took the sound to another level – more blacker background, more details, the three dimensionality with the music improved in all directions. This is undoubtedly the best linestage I have ever built.

Attached are some pictures of the line stage, its internals with the external filament supply, and picture of my system. The amps are Class A2 805 monoblocks using 211 as the driver with a source follower driving the grid of 805.

Ale – thank you for the experiments you share with the community, and also being very patient in working with novices like myself :-). Owe you a nice dinner next time I am in your part of the world.

Cheers

Sridhar

A VT-25 from South Korea

Slowly making progress during the lockdown period. Over the past few weeks I managed to achieve quite more than I was expecting. The fixed bias and 300B filament supply board (board 1) needs to be wired. On the other hand, the driver filament supply plus the HT supply board is completed (board 2):

I tested the HT supply and can deliver over 430V. It has more voltage capacity, however the filtering capacitors are 450VDC.  The DHT filament supplies can do 8V to 15V to accommodate different filament requirements (including some level of filament bias). It works beautifully!

Last night I did a quick job to build a pair of source follower boards. They will drive the 300B grid with DC coupling. The whole fixed bias arrangement will be as follows:

1. Raw supplies for -200V and +25/50V for positive grid current supply and correct operation of the pass-MOSFET in the follower
2. Swenson+ regulator for bias
3. Channel trimpot and coupling capacitor for independent channel bias setting. This will be in one PCB
4. Each channel has a source follower biased at 20mA

Here is a snapshot of the boards:

As I ran out of these PCBs, I decided to make a minor upgrade on the boards before requesting a new batch. I added a series resistor and parallel LED for either A2 operation or normal operation indicator. Also place underneath the board to solder a film/PIO 100nF decoupling capacitor as needed.

Playing with the layout a bit, here is a view of potential layout of the amplifier board. The PCBs for mounting sockets, turrets and switches are very handy. It accelerates the build process and provides full flexibility.

You may not like the open look and feel, but who cares! I love the aesthetics and those OPTs are enormous!

Wiring job is done. The fixed bias supply delivers from +50V to -300V. It has more voltage capability as am planning to use this same supply for future builds (e.g. 845 SE). The filament supply set to minimise power dissipation on the filament regulators for the 300B. All working fine, so am happy with this board. Filament noise is 0.3mV.

Last year I developed a voltage reference using an HV LED. Unfortunately these devices from OSRAM seem to be discontinued. I managed to buy enough parts for my own use though. What is interesting from these LEDs is that the dynamic resistance is low. About 150Ω with good current, or between 300Ω to 500Ω.  Tempco is very low and with such a low dynamic resistance, they are great for creating a voltage reference with a stable CCS:

The LEDs are extremely bright and found that with a 1mA current are dim enough whilst retaining the stability needed. I have a cap multiplier arrangement and the LED array is fed by a stable CCS. Jumpers on the board allow bypassing LEDs and there is also the option to use a trimmer for variable voltage adjustment. Very handy for screen grid supply and phono stages. The reference voltage is extremely quiet with more than 100dB PSR.

My latest version of the 300B amplifier will have fixed bias. Yes, adds complexity, but for me that’s not a problem. Also I was keen to DC couple the 300B to a source follower so that already gives me the voltage supply to implement fixed bias easily.

To make my life simpler here, I designed a few PCBs to support the build process. First one is the HV LED voltage reference which I already covered. It delivers a 150V fixed voltage reference and enough current thanks to the MOSFET pass device. I built a pair of source follower PCBs which you can see in my previous post here. Each SF is biased at 20mA. A bit of an overkill, but enough headroom to discharge the 300B grid capacitance as well as the SF MOSFET Cds one. Also will provide grid current and potential excursion to A2 thanks to the positive supply which can be wired to +25 or +50V. I tested the SF boards and with large finned clip-on heatsinks work perfectly fine.

The second PCB is the fixed bias splitter and AC coupling circuit to the driver. To ensure I provide a constant impedance and keep the RC constant with any bias setting variation, I added a simple emitter follower (T1/3 – PMBTA42) with a tail CCS (T1 – BSS126). All SMD devices to minimise footprint. The board has 2 voltage dividers (R9/11, P1/2 and R10/12) with a trimer for fine voltage setting. There is a couple of connectors for external multi-turn potentiometers. The board has space for a nice and large film/PIO capacitor (C1/2) and bias provided via R4/8:

Here is a snapshot of the bench test:

This worked superbly and is rock solid. Left it for an hour to get warm and couldn’t see any bias variation. This looks promising. I will mount the 4 boards (2 SF boards needed in stereo) in a mounting PCB. Love this modular building

And the fixed bias PCB is completed. All individual PCBs mounted over a ground plane PCB. It will be a stacked build. On top of this PCB, another one will hold the driver. Firstly the D3a in a hybrid mu-follower configuration:

Tested and bandwidth of these source followers is nearly 10MHz with plenty of current drive at 20mA idle.

It didn’t take long (or at least as long as I thought it would) to finish the driver board. It has a D3a hybrid mu-follower with SiC cathode bias arrangement:

The board is mounted on top of this previous board.

I usually get a lot of emails asking for more details and pictures of the building process. Here you are for the ones who asked for this. Hopefully this inspires you to build or use some of the same layout/techniques I use:

Did an initial testing which was really successful. As I used a pair of hybrid mu-follower boards (a.k.a gyrators) which weren’t designed for 250V bias point, I could only dial 230V maximum. It was fine as did a lovely test of 150Vpp anyway.

The modular approach is very handy. The small PCBs I made for turrets finally paid off. I can connect the boards easily with 2mm banana plugs and remove and/or change things as needed.

I made a set of useful PCBs. They are intended to mount large (big really big) film capacitors: WIMA DC Link ones!

I use the cost-effective 45μF/600V (MF Part No. DCP4I054507ID2KYSD) in many of my boards as the last capacitor in the filtering network. This is a 2 pin device, however when you go larger like the 80μF/900V (MF Part No. DCP4N058009JD4KYSD), this one has 4 pins and bigger size. The PCB for the later can also accommodate the smaller DC Link of 45μF/600V. The boards have turret or 2mm banana plug connections and an INS-1 Nixie indicator with its associated resistor. Finally a bleeder 3-5W resistor can be added.

The smaller board has the size of the Source Follower PCB. It can be mounted below it or can be used independently. Can fit a variety of PIO/Film capacitors for decoupling or for AC interstage coupling.

Speaking about the Source Follower PCB, I made also a new batch of PCBs as run out of the original ones. I made a minor modification and improvement by adding an LED indicator before the top MOSFET drain. This works in the same way as the gyrator Rev08 PCB. Can be used for normal operation or for A2 current source indicator. Also added an extra PIO/Film 100nF decoupling cap to be mounted under the PCB to decouple the high impedance node to the power supply: