Millennium Tube Amplifier

 Millennium Tube Amplifier

History

The amplifier was built in 2000 which was the reason for its name.  I was then in the Chicago, US on assignment, while working for Motorola.  In 1999, my brother-in-law gave me two pieces of old US army components.  These are visible in the image above, the huge HV power transformer and a 10H 250mA choke.  When I was in the US, my brother-in-law asked to buy some vacuum tubes and Shure phono head.  I manage to locate a stores in downtown Chicago, to buy the Shure phono head and the vacuum tubes from two separate stores that were quite near each other.  Since I was alone, I had some time researching on tube amplifier design where I could put the transformer and choke to good use.  I was intrigued by the simplicity of the design.  I was particularly interested in the DTN Williamson design done in 1947.  I also decided that I can built it since I have the key components given to me and only lacking in the audio output transformer.  The most price reasonable transformer I could afford are those from Hammond.  After some research and design, I decided I would use the Hammond 1645 transformer largely because it is able to drive 30W,  I knew I could build using a EL34 design which can drive 25W so I though a 30W output transformer was a good choice so as not to drive the transformer to its capacity.  I ended up buying the vacuum tubes and audio transformer online with delivery.  I think I paid about USD190 for 2 transformers and match quad EL34 (Svetlana) and quad EL34 (JJ), 6SN7GTB and 6189 (Military version of 12AU7).  That was quite a sum given that the conversion was S$1.8 to 1USD.

The parts finally arrived at my apartment in Woodstock, Illinois.  The last piece that I need was a chassis which I order from Hammond.  However, when chassis arrived, it looks too small and thickness of the aluminium too thin to be able to hold the huge transformer, so I decided I needed a backup chassis.  I went to the supermarket near my apartment and stumble upon a rectangular korean cake tin, made with 1mm thick aluminium.  The size and weight of the tin looks big enough and at the right gauge, so I bought it.  I did not build the amplifier when I was in Woodstock, but simply gathering all the necessary components.  On the day when I left the US for Singapore, I packed the parts into my luggage and haversack.  I was on business class so I could carry quite a lot including my clothing.  By then, I was in the US for 3 months.  During that period, I had time to design the amplifier by looking at several designs which are very close copy of the original DTN Williamson design.  One motivation for me to build it was to appreciate what is the audio reproduction capability in the 1947 was.  

Original Build

The millennium amplifier was build in the months following my return.  The original design uses a direct drive to the -ve bias grid with a fix voltage.  This provide a direct DC bias to EL34 which is good since the bias is well control, but the cathode current is basically controlled by the grid bias and a cathode resistor to maintaining a cathode current of ~32mA.  The tube used were quad balanced, so the voltage grid bias works well.  With that, the cathode voltage is 16V, and cathode current is 32mA. 

The first design was using cathode bias for EL34.  The output tube were quad balanced, so there is no need to use cathode bias.  However, in view of the possibility that the tube might be replaced in future and replacement tube may not be balanced, the cathode bias use used to ensure auto current adjusted biasing for EL34.  

I was amazed by the quality of the sound.  I tested the audio bandwidth which reached about 100khz.  This is quite amazing given that Hammond 1645 only stated up to 30Khz bandwidth. I used it for about 10 years until one of the JJ EL34 tube blew.  After the tube blew, I reviewed that the design for an improvement to the biasing.  I found that I could use LM317 to use it as a current control bias for the cathode.  The basic design is taken from LM317 datasheet as shown below.

By selecting R1, the current limit can be set using ILimit = 1.2/R1.  For 32mA, R1 = 37.5 ohms, the closes being either 39 ohms.

2nd Design

The second design made some change indicated above.  To allow for more voltage to the cathode for control, the grid bias was reduced to -16V.  The new design basically removes the need for any cathode resistor since the LM317 now acts like a cathode resistor.  The cathode voltage with the grid voltage at -16V is at about 24~26V, which is within the maximum working voltage of 40V for LM317.  Notice that U5 and U6 are now added with R19 and R22 to set the constant current at 32mA.  The 4 JJ EL34 are now replaced with Svetlana type since the JJ type may not be balanced if only the spoilt tube is replaced.  

Again, the new design served me for about 12 years until one of the EL34 blew again.  I investigated, and I realized that the 3 out of 4 LM317 were damaged.  There was a problem with the earlier design change.  I calculated the power dissipation on the LM317 based on the current and voltage across it and found that it was carrying a heavy load, which could easily push the LM317 above its maximum junction temperature.  

With the cathode voltage of 24V, the power dissipated on LM317 is

Power (U5,U6) = 24*0.032 = 0.768W.  

With thermal resistance of 65°C/W, the junction temperature rise = 0.768*65 = 50°C

Together with the ambient of about 30°C, the final junction temperature = 80°C.

This is within specification for the LM317 without additional cooling.  The LM317 was mounted within the chassis without ventilation, which could push the temperature higher to reach the maximum junction of 125°C.  The LM317 has been operating at this elevated temperature without ventilation leading to early breakdown.  

The design itself is not fault safe for the output tube, because if LM317 is faulty resulting in a shorted input to output, the current will shoot up and burnt the tube instantly.  This was the reason the EL34 blew up when it was powered because of a shorted (Input to Output) LM317.  A correction to the design was needed to make it fault safe for the output tubes.

3rd Design

To make it fault safe, 2 consideration were taken;

1.  Heat dissipation to LM317 has to be reduced

2.  In case of the LM317 Input and Output shorted, the cathode current must be limited to a safe limit and should be fused to break current when this fault happen.

To do this, the heat loss to LM317 is reduce to about 40% and the rest of the heat is dissipated to a fusible resistor in series.  With Cathode at about 25V, about 15V must be across the resistor.  This means the resistor value is Rs = 15/0.032 = 468.5 ohms.  However, the highest fusible resistor was 200 ohms. so I could use 2 resistor to form 400 ohms.  To have a slightly higher voltage drop on the resistor, the cathode current is raised to 35mA, so that the voltage drop across the resistor is 400*0.035 = 14V.  This leaves about 11~12V across LM317 which is about 44% of the previous heat dissipation.

With thermal resistance of 65°C/W, the junction temperature rise = 0.035*11*65 = 25°C.  This is a reduction of 25°C from design 2.  With the ambient of about 30°C, the final junction temperature = 55°C.  This leaves 70°C for headroom before LM317 reach the junction temperature limit of 125°C.

Under normal operating condition

Power on resistor (R18+R20) or (R21+R23) = 0.035*0.035*400 = 0.49W

Fault Analysis

Case 1: Normal operation

Under operating condition, the operating point is at 1 with an anode current of 35mA.  Power dissipation on the cathode resistor is at 0.035*0.035*400 = 0.49W

Case 2: Shorted LM317

When LM317 is shorted, the grid-cathode voltage will reduce and push the anode current up which correspondingly increases the cathode voltage to increase the grid-cathode voltage.  An equilibrium is reached at about 50mA, when the cathode voltage = 400*0.05 = 20V and the grid with -16V will exhibit a grid cathode bias of about -36V which is at point 2.  This means the 400 ohm cathode resistor will limit the cathode current to 50mA.

Case 3:  Grid voltage is falls to 0V and shorted LM317

In this case, the grid voltage will drop to -24V initially and tries to raise the anode current which will correspondingly raise the cathode voltage.  This increase in cathode voltage will limit the anode current increase.  An equilibrium is reached at about 80mA and the grid volt is about 0.08*400 = 32V which is at point 3.  The power loss on the cathode resistor = 0.08*32 = 2.56W which exceeds the power rating of the resistor.  The EL34 is now dissipating 0.08*400 = 32W which also exceeds the safe limit of the tube.  I expect the fuse to break before it cause further damage to the tube.  It is actually better to have a smaller wattage for the cathode resistor of 0.5W each, which will ensure that the tube is protected in this case.  

What's Next

The next part is to document the power supply design.  The power supply is designed to start the vacuum tube with a lower anode voltage and steps up to full voltage after about 30 secs after the tubes are fully heated.