Friday, 26 February 2010

The 45 Concept

For those who have read some previous entries in this blog, you will no doubt have read my ramblings regarding my directly heater triode amp. It began as the 300B story, and thus far, ended at the Electra Print DRD45. Having gained more insight into amp design and building, as well as more reading into the various standard reference books and Internet forums, I have concluded that the 45 is the tube to stay with. Helped in no small part to my 20+ 45 tubes. And in some twisted logic, I still hold that the 45 is a good tube for those wishing to trying different NOS brands and new stock. My reasoning to this; there are good range of NOS 45 available. They are easy to find and readily available on most stores and the price is still the sub US$100 per tube. The globle versions (UX145, UX245, UX345 and UX445) are much more sought after and range in US$150 to US$200 per tube. Whilst expensive they are still readily available. In terms of new production 45, I have only come across the Emission Labs and TJ brands. The range of available 45 is great and the NOS supply is still relatively affordable. This is in contrast to the 300B, 2A3, PX25/60, AD1, 50, 211, 845 and all the other DHT power tubes.

The 45 amp in broad design will be:

  • C3g driver tube. Either Triode of Pentode mode. To be reviewed.
  • Interstage Coupling. Using 1:1 or near interstage from either Sowter, Lundahl or Tango-ISO.
  • Fixed grid bias for 45. Using Tentlabs Negative Bias Module
  • Filament heating through Tentlabs filament module for 45.
  • Output Transformer is Electra Print 5K with 8R and 16R output. May consider changing to Tamura Amorphous core when the finances allow.
  • Passive PSU with 2 choke stages. Using a Lundahl LL1673/140mA common mode choke and an Electra-Print 5532 choke for each channel. All capacitors to be ASC Oil X386 types.
  • Rectifier tube 2x6AU4GT.

Most of the parts will be recycled from the DRD45 build that has yet to be done. This will be constructed as a stereo block with no attenuator.

Operating point of the 45 would be at the standard 220V with a -40V bias at 36mA onto a 5K load. Output should be around 1.3W, and will require the driver stage to provide around 40V of swing, or 80Vp-p or 28.4Vrms. If we have a sensitivity of 2Vrms then the driver stage needs to provide a gain of around 15x. An easy task for the C3g. The mu of the C3g is 40, so it might be worthwhile to consider using a 2:1 step-down interstage transformer. This may help reduce the reflected load and increase bandwidth. Being my first self designed amp, the above operating points may be completely wrong. However, it is a start, nonetheless. The other interesting note is that with the gain requirement of the driver stage of 15, it may open up possibilities of using an alternate driver tube, even using another DHT. Nothing ground breaking here.

Below are the heater requirements.

  1. 45 Tube: 2.5V @ 1.5A = 3.75 W per tube total 7.5 W.
  2. C3g Tube: 6.3V @ 370mA = 2.33 W per tube total 4.7 W.
  3. 6AU4GT Tube: 6.3V @ 1.8A = 11.34 W per tube total 22.5 W.

Total Heater consumption = 35 W (approx). More to come later.

Wednesday, 24 February 2010

Further Adjustments on Millet

After using the amp for around 20 hours, I re-measured the tube bias. I found that it drifted to sub 10V. So in this whole process, I adjusted the B+ to 26.8V and thus biased each tube at 13.4V. This would result in around 12.6V on the heaters with the 10R heater resistor. I found that the sound improved and tighten up a lot. Over the course of the next 100 hours or so, adjusting for bias every 20 or so hours. I also took the chance to change the buffer bias to 110 mV. Regardless of these changes, the amp remains deadly silent with no hiss, hum or noise through the HD650. With my neighbourhood being very quite even during the day, it makes for a beautifully silent background.

It would be interesting that after the first 100 or so hours to play around with the tube bias and the buffer bias to see its effect on the quality of sound. It would appear from previous experiments that running the tube hotter and closer to its limits tends to produce a bigger and more dynamic sound. This amp would be an excellent platform to test such differences.

One of the delights of headphone audio is that you can really start to hear the music in a way that you cannot with speakers. Some faint spoken lyrics or background vocals becomes crystal clear. You can hear subtle noise in the background, like the creaking of piano pedals, or the pianist moving in the seat. Such minor details, though not affecting the overall enjoyment and appreciation of the music, tends to adds a new dimension to provide a much memorable experience.

Tuesday, 16 February 2010

Mini Max Millet Part 3

After a a few day's worth of soldering and assembly, the amp is finally working! As usual, the assembly work took longer than the PCB assembly itself. See some of the photos below for the progress builds. The biggest mistake I made was to solder the fuse clips on the wrong side of the board. Now I have to remove the PCB fully to be able to change the fuse. On hindsight, that may be a good thing, as a blow fuse would usually mean more of a problem down stream. So having to remove the PCB also forces me to have a quick check of the PCB itself.
The wiring for the volume control, crossfeed, etc was all done with solid silver wire and the use of tabs. Although this took a very long time, the results are well worthwhile. Using 24 AWG meant that I had to strip the teflon insulation off very carefully, otherwise any slight cut/scratch of the wire will cause it to break off when it is bent.

Biasing and setting up the amp wasn't too difficult. The B+ adjustment was straight forward, and I biased it to 25.6V, and it is now hovering around 25.58V. The diamond buffer bias is set at 100mA at present. Reading through the advice, I will leave this at this setting for a few weeks of use and then revisit after that. I want to let the various capacitors, resistors and transistors break-in. The tube bias was quite straight forward. At present, the channel imbalance doesn't seem too bad, and both channels are well balanced.

Once everything was biased and running for around a minute after that, I did the measure of the DC offset at the outputs. It was measured around 0.1mV. Not too bad at all! Everything appeared to be in good order. The heatsinks were getting warm, but only slightly so. With the diamond bias below their recommended levels, I'd assume that it isn't doing as hard work as it was designed to, and running cooler.
Once the final assembly was completed, I hooked them up with my test pear of cheap Sennheiser ear-buds. I first noticed a little hum, but it was not too loud. Connecting my test iPod up to the amp, I waited for the delay relay to kick in. Music! The sigh of relief was immense when sound was coming out of the ear buds! After letting this run in for about ten minutes, I powered off the amp to change to my usual cans, the HD650.
The sound, in short, was amazing. The hum had disappeared. Not sure whether that was from moving the temporary power cables around, or from a different headphone, or whatever, it was gone. Dead quiet, and this was late at night with very little ambient sounds. Fantastic! Powering through music, the first conclusions I made was; I did not like the crossfeed. At this time, I am not sure if the crossfeed was causing the first problem, or whether it was wrong wiring.
Using the crossfeed for the first time, I found that it muted the highs and made the sound very flat and even a little restricted. Turning it off made the sound more fuller and far more enjoyable. I was thinking that I probably should've tried the crossfeed before on another amp and would've saved many hours. Maybe, the caps need some time to open up, but at the moment, I think I will be listening without the crossfeed. I did not notice any widening of the soundstage or it moving from the off position. I think my ears/brain probably doesn't respond to the crossfeed at all.
Regardless of the issue mentioned above, the sound was amazing. The bass was awesome, very authoritative, and I have never really heard such clean and articulate bass before (headphones or speakers). It was almost scary. The mids was warm and clear and the highs crystal clear. All in all, a great sound. I think there will be no need for any bass boost. The balance at this moment, is very good to my ears.
At the end, I used both a Vitamin-Q and a Vishay MK1837 as a bypass on the output capacitors, and though only have less then 1 hour, the sound is already amazing. Looking forward to seeing them break in and get better over time. Part of me still wanted to try the Mundorf Silver-oil caps, but the leads were too thick, and I was in no mood to do any more modifications yet.
Overall, very impressed with the sound and quality. The remaining actions are:
  1. Remove Crossfeed
  2. Build legs to raise chassis off the ground to allow air flow inwards from bottom.
  3. Possibly install a few more LED to create a more glowing effect.
Enjoy some of the progress build Photos
The photos below show the construction of the PCB with the most of the components installed. It also shows all the trimmers and terminal blocks on the bottom to allow for easy access and adjustment without having to remove the entire amp.

Below are the diamond buffer devices mounted on the heatsinks, ready for installation.

Below are some shots of the completed PCB. Note the Vitamin-Q mounted on the bottom of the board.

A few shots of the assembled amp. Note all the hook-up wires from the cross-feed to the switches and the various input and out connections. It was such a shame that I did not like the cross-feed and most likely remove it eventually.

Below the completed amp ready for power-up and first use.

Thursday, 11 February 2010


After much reading in the last year or so, I have finally settled on a DAC Design to begin experimentation. The Analog Devices AD1865 family. To this end, I recently purchased from eBay the AD1865NOS-1.0 kit. As a start, I will hook it up with any modifications and let the unit run for a few months to let it break it and understand the type of sound this DAC is capable of. But the longer term plan is to modify the unit, potentially using a different I/V stage to draw more out of the chip. The long term plan would be to make a poor man's Audio Note DAC 4. This would most likely entail a CS8414 receiver, feeding into the AD1865. This would then be coupled to the tube gain stage via transformer, which the gain stage would be coupled to the world via another transformer. The intent would be similar to Ciuffoli's DAC-End, with the exception of an added transformer between the DAC and the tube. The power supplies at this stage would be Twister Pear Audio's Placid and TubeCAD's Janus.

Tuesday, 9 February 2010

Mini Max Millet Part 2

Finally, everything is ready to go, every part, socket, tube, panel, etc is on-hand. See below the timber frame ready to receive the amp. Note that I am using a XLR input (as is my new standard for all my own equipment) rather than 2 RCA jacks. Also note the use of a lock TRS jack for the headphones. You can see below a few photos of the arrived panel on the timber chassis. The Front Panel Express panel is fantastic. The quality of finish and precision is very good. I will continue to use them for all future projects.

It was a tough choice, but as you can see, I decided to use the orange text on the panel. Combined with an orange LED under the chassis, I am hoping the whole unit will have a very warm colour glow during use, particularly at night, when I will be using it the most.

You can see below the partially completed crossfeed board. For the moment, I have only use stock caps, but I might change that some time in the near future. But the Panasonic ECQ caps are pretty decent. The trouble will be finding a cap of the required size to fit.

The capacitors used are Nichicon Muse for the large signal output capacitor with a Mundorf Silve/Oil as the high quality bypass cap. The power supply capacitors are Nichicon Fine Gold. The other parts are the stock parts as recommended on the MiniMax website.

Construction of the amp will begin after a final check to make sure everything fits. I anticipate that the construction will take around a full day, and I am eagerly looking forward to its completion. Over the last few years, I have built a small collection of 12AE6/12FK6/12FM6 tubes from a range of different manufacturers. By far the best sounding has been the Brimar. I am desperately trying to source more of the Brimar 12AE6A, but to date, no avail. Since I will be using Sennheiser HD650, I'd think that the higher gain tubes would be preferred and ultimately be the permanent tube in the amp. Look for the next part of progress build photos!

Monday, 8 February 2010

Mini Max Millet Part 1

In the last few months, due to family, I have had a resurgent interest in Headphones. Having not really listened to my Sennheiser HD650 for a few months, I thought it time to dust them off and relax. During those months away from the headphones, I also took the task of upgrading my computer, during which I decided on a sound card upgrade. This was in the form of a Asus Xonar Essence STX. Whilst mainly functioning as a gaming PC, my music library is stored on this computer and shared over iTunes.

Naturally, I started to enjoy some headphone music through this setup. Initially, the sound was a little dark, but as the hours stacked up, became increasingly more open and less laid back. This combination is both easy on the ears and non-fatiguing. The sound stage for me was more forward, instead of in the middle of my head, it was at the front, near the forehead. The details that you get from headphones is so much more then through speakers. Although I find at this stage of the burn-in period, the Asus lacks some of the dynamics and 'sparkle' of my PPA headphone amp, I am sure that with time, this will improve.

Working on the premise that most of my audio enjoyment will probably come from headphones for the near future, I started looking into building a high quality headphone amp. Naturally my criteria was a tube based design. My first DIY Audio was Pete Millet's hybrid headphone amp. That was a rewarding build and a nice little amp. It combines a tube voltage gain and a IC buffer stage. Some would say combining the strengths of both technologies. However, having built that many years ago, the I decided it was time to see how that design has progressed.

Surely enough, there is a thriving community of Millet amp builders and has spawned several evolutionary designs. Looking through the forums and the net, it seems that most of the improvements lie in the output stage (employing a diamond buffer) and the biasing of the tube (i.e., using a CCS). Having the fore-sight to ensure that whatever I build be as reasonably compact as possible, I settled on the MiniMax Millet. Some of the progressions.

  1. Original Millet
  2. Max Millet
  3. Mini Max Millet
  4. Head-fi Iteration

Having ordered the various parts over the last month or so, the build is ready to begin. This time around, the chassis design will be a standard DIY Tube affair. 4 timber side walls with a metal top plate. The timber will be home made with a beeswax finish. The top metal plate has been laid out and ordered through Front Panel Express. The design allows for ventilation slots above the heatsinks with the tubes projecting past the panel. Further, most of the control are also on the top panel. In this design, I have also integrated the Tangentsoft's Modified Linkwitz Crossfeed.

Sunday, 7 February 2010

Aikido Line Stage Build Begins!

Finally got around the designing and building the chassis for the Aikido line stage. Have a look at some of the photos during the planning phase. The chassis will be the usual timber sides and metal plate on top with the tubes and regulator heat sinks sticking through the top plate. The knobs and connectors are all on the front face.

I have decided to use XLR connectors in lieu of the 'Audiophile' type uber-expensive RCA sockets. It was actually the Eichmann plugs that gave me the inspiration. They were talking about the eddy currents, etc that happens when there is a sleeve type connection. And well, that does kind of make sense. But why not go one step further and remove ANY sleeve or ring type connection in the signal chain? The XLR are pin connection only, and being a professional audio product, designed for high numbers of inserts and removals. It is also silver plated connectors, with cable locks. All this had to around $10 each. Further, with the 3 pin standard connection, I can run stereo RCA, with the grounds tied together (which is what happens anyway inside the amp), and thus half the number of cables in the system.

You can see the generic layout of the chassis, with plugs all front access. Transformers and power stage at the back. 3 inputs and 4 outputs. The outputs will be arranged so that 1 will have a High Pass filter (around 150 Hz) and the other 3 will be full range with the RC f-3db point set around 1 Hz.

Close up of the DACT Stepped Attenuator. There will be enough room around for a potential upgrade to the Acoustic Dimension 41 step ladder attenuator.

Close up of the DACT Selector. It will switch both channels and ground. Since it is a 5 position switch and you can see I only have three inputs, it will be arranged that the two unused positions will act as mute function between each active position.

Some additional parts ordered to replace the standard kit components. Tantalum resistors in the important areas and Takman resistors in others. Note also the large Elna Tonrex caps for the heaters. Moulded Teflon sockets for all tubes and Obbligato film capacitors. The choice of output stage capacitors will be Mundorf Tin foil, with a potential for one of the output to have some kind of oil cap (Teflon, Copper, Silver/oil or Beeswax).

The power transformers, one for heaters (Hammond 185D12 6.3/12,6V 43VA) and one for B+ (Hammond 185E230 115Vx2/115Vx2 80VA). The IEC inlet has both dual fuse and EMI line filter. The heater transformer is slightly over-rated. With 5687 and 12AU7 as the tubes, the current draw should be around (5687 @ 12.6V = 0.45A and 12AU7 @12.6V = 0.15A) 1.2A total draw. With an available 3.4A supply from the 185D12, there should be enough head room.


This is what happens when you leave speaker without protection from young kids! My no longer very nice Fostex FE127e dipole! Well, it may be just the excuse to try to phase plug replacement. And yes, it was done to both speakers, but luckily only to the front speakers, and not to the rear facing pair.