845/75TL Single Ended Amplifier

I had been contemplating a "from scratch" power amp design for some time. I considered a number of output tubes including the 300B, SV811, SV572 and parallel 6B4G's. However, the glowing reports about the sound of the 845 not to mention it's seductive looks always kept this tube at the top of my list. This project began when I found a set of transformers and chokes suitable for an 845 amp at bargain prices (just over $1 per pound!) at a surplus electronics store.

In 2006 these amplifiers were converted to use an EIMAC 75TL output tube. The circuitry was revised by dropping the the 46 driver and cap coupling the 27 directly to the 75TL. The power output was reduced to 3.6 watts, more than sufficient for horn speakers. The 75TL tube is a remarkably transparent and musical tube that unfortunately has become nearly un-obtianable. The 75TL is a significant upgrade over the already great 845.

• Power Supply Design
• Power Supply Schematic.
• Power Sequencer Schematic

  Amplifier Design

  Amplifier Schematic

  Chassis Design

  Construction (pictures)

  Listening Evaluation

  Measurements

  They glow in the dark!

For the power supply I was shooting for a B+ of 1000 to 1100 volts preferably using choke input. I was also looking for a fast, low impedance power supply as suggested by Herb Reichert in his "flesh and blood" design. Fortunately the power transformers I found had a number of primary and secondary taps that made it possible to produce a wide range of voltages. To aid in the power supply design I used Microsim's Pspice. Simulation predicted that I could get 1030 volts for choke input and and 1057 for cap input with my available taps. The actual voltages were 1035v for choke input and 1080v for cap input. I ended up using cap input primarily because of lower ripple but I also liked having the extra 45 volts.

For rectification I decided to use 866-A mercury vapor tubes. These tubes have to be pre-warmed before the HV is applied. They can handle lots of voltage (10kv), are low impedance and have a really cool blue glow when operating.

Initially I had planned to use separate rectifiers to supply power to the output tube and the driver/input tubes. However, Pspice simulations showed that this was perhaps less important. By using two chokes in a parallel arrangement I found that the power supply isolation was better than 137db at 1khz. Using simulation I found that the parallel arrangement of chokes and dropping resistors offered considerably better isolation than the more traditional series or cascade topology. Click here to see the simulation results for isolation using parallel and series connections. Using simulation proved to be an invaluable aid in the power supply design.

For the power supply filter caps I wanted to stay away from electrolytics. I considered using Elna Cerfines but with voltage ratings of 500 volt I would have had to stack them three deep to get a rating of 1500 volts. This ended up being a very expensive solution. Luckily I found a bunch of surplus 13uf 1200v paper in oil caps at a good price.

One of the concerns that I have had about using fixed bias is the damage that can be wrought if a bias supply fails. Since I also needed some facility to preheat the 866 rectifiers I designed a "smart" power sequencer.

The power sequencer uses a AT89C4051 micro-controller and a couple solid state relays. The AT89C4051 has a built in voltage comparator that I used to sense the bias voltage. The sequencer is operated via three pushbutton switches (off, standby and on) and status is displayed by three LEDs (fault, standby and on). For power up the sequencer switches on the filaments and the bias supply. Only after the filaments have been on for at least 20 seconds and the bias supply has been at the correct voltage for 4 seconds is the high voltage switched on. If the bias voltage drops below 90% of it's nominal voltage then the high voltage is immediately switched off and the fault LED turns on. On power down the filaments are left on for an additional 90 seconds to provide more gradual cool down and also ensure that the high voltage caps are fully drained. In standby mode the filaments and bias supply are powered up, keeping everything warm. If the amp is left in standby mode for more than an hour it automaticly shuts off.

To keep the weight manageable I decided to build separate chassis for the amp and power supplies. The power supplies still ended up weighing in at 55 pounds each. The thing I like least about these amps is their size and weight. At 85 lbs. per channel I am reluctant to haul them to my workbench for tweaking and measurements.

For the input tube I had Initially planned to use a 6J5 primarily because it looked like a mu of 20 or more would be needed to get enough gain. However, after working the numbers I found that the highly regarded 26 or 27 would have just enough gain if used with the right driver tube. After studying both tubes I settled on the 27 mainly because it is indirectly heated. Since both the driver and output tubes are directly heated I figured that three DHT's may be too much of a good thing. Also the 27 has a slightly higher mu than the 26 (9.0 vs. 8.3). I started with ST shaped 27's but later replaced them with NOS blue glass globe's. They not only look cool but sound better too.

For the driver I chose a 46. This tube is a bit of a sleeper. It is highly regarded by those who have used it and generally preferred over the 45. Since is is not as well known it is also a real bargain. I also created an alternate design using the 10-Y but eventually selected the 46 because was a lot less expensive. The 845 can be a difficult tube to drive, needing lots of voltage swing and a fair amount of current. The 46 is up to the task. Operating at 22ma the 46 has a voltage swing of just over 300v (peak to peak). At 22ma the slew rate should be adequate for up to 60khz. Like the 27's I started with ST shaped 46's and later upgraded to globe's. The globe 46's have a warmer more lifelike sound. More expensive but worth it.

I am using unusually high power supply voltages for both the input and driver tubes (750 and 960 volts respectively). This allows the use of larger value plate resistors resulting in lower distortion, more voltage swing and higher gain. However, the plate resistors end up dissipating a lot of power (3 watts for the 27 and 10 watts for the 46) limiting resistor choices. I also figure there is a greater risk of fireworks if something should go wrong.

Initially I chose to use cathode bias for the 845. This seems to be a matter of taste and I had generally preferred the sound of cathode bias. Besides, cathode bias was cheaper and easier to implement.

After about a year a friend convinced me to experiment with fixed bias. For the experiment I built a quick and dirty tube rectified bias supply. As expected fixed bias made a notable difference in the lower registers. The bass was more powerful and detailed. However, what I did not expect was subtle improvements in both the midrange and treble. Fixed bias added a little more presence and detail to the sound. I expect that the differences in the mid and upper ranges is not due to fixed bias but rather that using fixed bias increased the voltage across the output tube by 145v. Regardless of the reasons, fixed bias was a clear win. This is not meant to be a blanket endorsement for fixed bias. But with the Verus amps it is clearly the best choice. Oh yes, and fixed bias makes for a great excuse to mount a vintage meter on the front of the amp.

Once I was sure that I wanted to stay with fixed bias I installed a small bias supply into the base of each of the power supply chassis. For the bias supplies I used a 6X4 rectifier with a 9hy choke and a combination of polyproplyene and paper in oil caps. The bias supplies fit nicely into the power supply chassis, but it goes against my principles to hide any tube where it can not be seen. In my opinion tubes should always be proudly displayed.

There are lots of different opinions about the best operating points for 845s. So far I have tried 700v @ 61ma, 940 @ 81ma and 1080 @ 80ma. I found that each time increased the voltage it sounded better. I wish that my power supply capacitors would let me go even higher.

One of the questions I had from the outset of this project was if DC would be required for the 845 filaments. I only wanted to use DC for the 845 heater as a last resort. When the amps were first brought up I measured 30mv of hum, too much for my 92db efficient speakers. Not wanting to give up on AC filaments I dug around and found that most of the hum was due to power supply ripple. The ripple was Initially about 4 volts and by reconfiguring the power supply as cap input the ripple dropped to 250mv. To my surprise this reduced the hum to about 6mv. Quiet enough that you have to be within about 3 feet of the speaker to hear the hum. I would not have expected that 4 volts of ripple on a 1030v power supply would have been so significant.

I was resigned to using DC for the 46 filament since it is used as a driver. I started with a simple unregulated DC supply with Schottky diodes, big caps and some small value chokes. After living with the amps for a few months I decided to try using a current source for the filament. I used a simple two transistor regulator that admittedly is not high performance but should be adequate. This change made a small but worthwhile sonic improvement. The sound was a bit warmer and more 3 dimensional. But most important this change added a subtle but very enjoyable lush character to vocals.

As for parts selection I used good quality throughout All of the cathode resistors are Caddock and AB carbon comps were used for grid stoppers and grid resistors. The high voltage and high dissipation of the plate resistors limited my choice to Mills 12 watters. The driver to 845 coupling cap is a Jensen copper foil in oil cap. I have used both the Jensen aluminum and copper foil caps before and very much like the way they sound. The aluminum is good but the copper is definitely worth the extra money. For cathode bypasses I ended up with a mix of oils, polypropylene and Elna Cerafines. The output transformers are Bartolucci double C core model 18's with a primary impedance of 7.5k. Just the ticket for a medium powered 845 amp. Wire for the signal path is 26ga XLO and the power supply connections are solid core silver in Teflon tubing.

To aid in the chassis design and layout I used the Povray ray tracing program to create images of the amplifier and power supply. Though this was a time consuming venture it has been a valuable aid in fine tuning both the physical appearance and the parts layout. This computer design stuff is great! The finished amps almost identical to the ray traced pictures. Here are a couple of ray traced samples. Click on the pictures for a larger view.

The transformers are painted with a glossy "Hammerite" paint that gives the surface an interesting and appealing texture. The top plates were machined from 3/8" aluminum plate. All of the components are mounted from the bottom side of the plate using blind holes. This way no screw heads are visible. The top plate is finished with polished black lacquer. The wood frames are made from black walnut. The wood in these renderings looks more like rose-wood but it was the closest I could get to walnut.

It's tough for a designer / builder to be objective about his own work. Perhaps I can get of my audiophile friends to write up a more objective review. However, I can say that the sonics of Verus amps have exceeded my expectations. There is a sense of realism that I have not heard before. The sound is big, warm and powerful. But what I like most is the delicacy and the way subtle nuances are reproduced. It seemed paradoxical that such a physically massive and powerful sounding amp could also excel at delicacy. Sometimes you can have your cake and eat it too.

• Frequency Response
• Power Bandwidth
• Harmonic Distortion

Here are a couple of low light shots that give you an idea of how they look in the dark. The blue glow of the 866's and the warm yellow of the 845 has a stunning effect.