Saturday, January 23, 2016

The heart attack special - 6S17K-V powerdrive GK71 SE amplifier

This amp operates at very high voltages that can kill you. Do not attempt to build this amplifier unless you have already built several other amplifiers, preferably kits so that you can learn good techniques. I would highly recommend the Bottlehead kits. The exposed pins on top of the GK71s are a definite safety hazard. Do not use this tube if you have small children around, or even curious adults. Curious cats are also in danger.

For those of us who build amplifiers and have a few sitting around, there needs to be a good reason for building yet another amplifier. In this case, the reason was a mild heart attack which lead to a triple bypass operation on New Years Eve. Fortunately all went well and I have recovered quickly, probably because I do a three mile walk around my community about five times a week.

It's not a procedure that anyone would undertake just for the fun of it. Even with pain killing drugs, it's not entirely comfortable and lying in a hospital bed for most of the day gets boring. To help pass the time, I thought about building an amplifier. It's good mental exercise. My wife loaned me her Kindle Fire tablet and I was able to do some research.

I had a trip planned to India for five weeks in February which I realized would need to be postponed. The guidelines for recovery from bypass surgery include not carrying anything heavier than 7 to 8 pounds (4 kg) for quite a few weeks. Luggage becomes a real problem.

Moving boat anchor tube amplifiers also becomes a problem. So my thoughts turned to possibly building a reasonably powerful ( 10 watts ) amplifier that might weigh under ten pounds (5 kg) or so. Before all the health problems I had been thinking about building an amp based on a GM70, 813 or GK71 as the output tube. It looked like the GK71 was easier to drive than the others and could be operated at a lower plate voltage to achieve about 10 watts which would be adequate for my Tannoy 15" red speakers. Here are links to the GK71 datasheet and the GK71 triode curves.

To make this amplifier all directly heated, I decided to use the Russian 6S17K-V planar tube that I had enjoyed in a line stage I built a few years ago. Here is a link to the datasheet. This tube is actually indirectly heated but the cathode is connected to the heater which is somewhat unusual. In a topology sense, you could say it is semi-directly heated. It self biases with 6.3 vdc. It does not suffer at all from micro-phonics and gives a very clear sound.

If you decide to use this tube, read about the cooling requirements of this type of tube.

The tiny 6S17K-V tube would not have enough oomph to drive a GK71 but I reckoned that it might work if I included Tubelab's Powerdrive. I had tried Powerdrive in my 6V6 amp and liked it. Although one of the reasons for implementing Powerdrive is to provide sufficient drive for the output tube to achieve class A2 operation, the other advantages are the elimination of the cathode bias capacitor and the ability to adjust the bias with a potentiometer. 

So here is the schematic for the amplifier.

I had a couple of Edcore 8K:8 GXSE10-8-8K transformers from yet another failed project and crucially, the specs say that each can handle 100 ma but I will try to run them at 90-95 ma. If you are used to the concept of Powerdrive, there is nothing too unusual about this circuit other than I omitted the CCS to load the 6S17K-V because it is such a high mu tube to begin with and I wanted the filament bias to set the current, not a CCS.

So that is the easy part. The hard and usually heavy part of a direct heated transmission tube type amplifier is the B+ power supply and the heaters. The primary object of this amplifier is not to be the best 10 watt amplifier on the planet, but to be light weight. Great sound is allowed to be compromised in favour of reduced weight.

Let's look at B+. No doubt some of you have come across the Greenvalve GM70 website where Erick Bates uses a stack of Cisco 48v SMPS to get a B+ of 816 volts and a B- of -48 volts. Since my circuit called for about 500 - 600 volts and a B- of -200 volts, I figured I could use 12 @ 48 to get a B+ of 576 volts and 4 @ 48 to get a B- of -192 volts. Close enough. A vendor on Ebay was offering 16 of them for $99 including shipping so I ordered them.

Looking at the triode curves for the GK71, I am hoping for a bias point of 576 volts, 90 ma and about -80 volts bias. Hopefully these settings on the bias pots will be close enough to start with.

Let's look at the GK71 heaters. 20 volts at 3 amps each and directly heated. This time I was inspired by Alex Kitic's use of electronic transformers for halogen bulbs in his 813 amplifier. He gave me some advice on the versions I should use for 110 volts and so I ordered a couple for about $20. This high frequency AC heating solution should weigh a whole lot less than a hefty transformer and the usual methods of implementing DC heating such as CLC or even the Rod Coleman solution.

In my original 6S17K-V linestage, I used a simple LM317 to get the 6.3 VDC for the heater. During my hospital research, I came across a DIY group in the UK called Audio Talk that created a group buy for a 6.3 VDC indirectly heated pcb. They published a schematic and that is what I used. They also had a thread for a similar direct heated pcb, but kept the schematic secret which is quite understandable. I figured the indirect version would probably do fine for the 0.3 amps required for the 6S17K-V tubes.

By the way, about ten years ago, I met Paul Barker ( he is a contributor to the group) at his house to listen to what I remember was an 833 amp. It was basically a bread board with clip leads everywhere along with an enormous power transformer. Unfortunately there was something wrong with the amp that day, but I was more than impressed by the outrageousness of it all. This hobby is meant to be fun and a bit over the top.

Incidentally, the Audio Talk had a thread on building a GK71 amp which was a mine of information from which I decided to use 5K transformers and the operating point.

Normally I would just use some strip board to do the Powerdrive and 6.3vdc circuits. Something changes after a triple bypass operation. You look at life a bit differently and I wanted to learn and experience more so I decided to lash out and create my first PCB using Express PCB. It seemed easy enough to do the Powerdrive circuit and I discovered I had enough room to include most of the 6.3 vdc circuit as well.

So all light weight components except for the two output transformers and a small 12 volt transformer to provide AC for the 6.3 vdc PCB.

A Cisco 48 v @ 0.38 amps.

Case removed.

Backside. Each module is about 3.5" by 2".

IEC connector and ground wire removed. As I understand it, each module needs to float electrically so the ground connection should not be used.

Bank of four attached to a poly-carbonate sheet. The poly is just an 8 x 10 Lexan sheet from Home Depot cut in two. The poly should electrically insulate each bank of four modules from the next. Eventually I realized that the modules needed to be separated a bit more which was easy to do.

There was no convenient way to securely mount the modules so I drilled a hole in the metal heatsink and secured the module to the Lexan with a nut and bolt. Eventually I will make sure the nut does not loosen by applying a dab of nail polish.

Initially I will leave the wires long so that I can undo any mistakes. Later I can shorten them if the amp is successful.

The heavy items installed. One of the filament modules in the back to the right of the blue Edcore transformer. So far it weighs less than 12 pounds including the plywood. 

Modified filament transformer with two extra turns (yellow) of 20 gauge wire which I added to the existing 12 turns. Two turns was not enough and I finished with 5 extra turns giving about 21 volts. I tried an extra 6 turns which resulted in 24 volts.

You could also remove the existing 12 turns and replace with 17 turns. The five extra turns took about 15 inches of wire.

This is a multimeter ( Extech EX205T) that can measure RMS AC. Whether it is accurate is another matter and eventually I will manage to get hold of one of the Flukes that can measure RMS AC or a scope which is the most accurate method. Incidentally, my Fluke 75 failed dismally recording only 9 volts but it cannot measure RMS AC. 

I figure that if the AC voltage is in the 18 - 22 range, I will probably be ok. And a GK71 tube is very inexpensive at under $20.

The finished product. The capacitor is 560 uf 200 v and is an essential part of getting rid of any AC 200 hz buzz. Read more about it at Alex's web page. You will notice that for the 240 volt Europe you would use a 270 to 330 uf 400 v cap. I attached the cap to the plywood with a glue called Goop. I have used it extensively in the past, even with motor run caps. I also used it to attach the 0.1uf KSG silver mica coupling caps. The bond can be broken with some difficulty.

Alex left his versions inside the cases they came in which would be easy to mount. I left them outside to make it easier to get rid of any heat. I mounted them by screwing the T220 cases to the plywood.

As usual, I use the two DPDT switch method of delaying HT. The transformer on the left provides the filament power for the 6S17K-V tubes. Each 6S17K-V uses 0.3 amps.

Erick Bates used 4K 2watt resistors as a load for each module. I used 4.7K 2 watt resistors that were a lot cheaper. It was not too difficult to daisy chain the leads to the modules but it does get a bit fiddly and boring. I left the leads long at this stage and if the amp is successful, I will shorten them to fit better.

Be careful to ensure that you observe correct polarity as you daisy-chain the output wires together. On the modules I used, the negative wire was closest to the edge of the module but you should check this yourself. The positive wire had white printing.

In the next version of the amp I will raise the strip up a few inches to make it more accessible.

The PCB turned out good enough though I made one mistake that was easily bypassed. The DC regulator section is cramped but I was greedy. I did not have enough room for all of the components mentioned in the UK Audiotalk article so some of the larger capacitors are off the board. It was a fun exercise to do this and not all that difficult. Initially, the board produced about 6.9 vdc. Changing the 56 ohm resistor to 28 ohm reduced the voltage to 6.3.

Closeup of how I mount the tiny 6S17K-V planar tubes. The green wires are 12 gauge copper. No solder is used. The holes in the plywood behind each tube allow the heat to escape. Since they are metal they don't light up. However the advantage of the metal is very low micro-phonics.

Somebody asked me in my blog on the line-stage whether it is possible to solder wires to the metal parts of the tube. The answer is yes, but I would not be too confident about the joints.

The GK71 filament produces 60 watts ( 3A @ 20 v) which along with the plate dissipation of 55 watts is a lot of heat to get rid of. Even worse than my 6C33C amp. So I mounted the socket on the top side of the plywood and fed the connecting wires through a hole cut in the plywood.

The socket is seven pin similar to those used for 829B tubes but the GK71 pins are much thicker so a special socket is needed. The pin at the top of the tube is also thicker than those of an 829B.

Yes, it is a rats nest but I may do a second version to clean it up a bit. It will then look like there is only one rat in the nest instead of two.

The dual power switch arrangement makes testing easier. First establish that the 6.3 vdc and 20 vac boards are working ok. Before adjusting the Powerdrive potentiometer to establish the desired bias point ( -80 vdc approx), do not connect the cap to the GK71. Finally, connect the cap to the GK71 and  make the final adjustment to the potentiometer to fine tune the current through the GK71. The dcr across the Edcore output transformer is 207 ohms so I aimed for somewhere between 18.6 volts for 90 ma and 19.6 for 95 ma. I finished up with 19.1 volts for 92 ma. Bias derived from the Powerdrive is -77 volts with a B+ of 577.

A 3 amp fuse is a bit marginal and I used a 6 amp that I had on hand.

I can't hear any hum so far.

My estimate of power output is about 10 watts based on the discussion by the UK Audio group about building a GK71 amp. It would be quite simple to drastically increase the power by adding another bank of 4 modules (16 x 48 = 768 volts) and substituting output transformers that can cope with higher current than the 100 ma of the Edcore transformers. The GK71 has a maximum dissipation of 125 watts which is much more than the 55 watts of the current design. Throw in the ability to operate in class A2 and this amp has the potential to be a real beast at lets say 768 volts at 150 ma per channel. I don't need that kind of power,but I might consider using more exotic output transformers at some stage.

The cost of building this amp has been quite low, for me less than $300. 

16 x Cisco SMPS modules - $100
2 x Edcore output transformers - $60
2 x GK71 tubes - $40
2 x GK71 sockets - $16
2 x 6S17K-V tubes - $4
1 x 12 volt transformer - $5
sundry resistors, semiconductors, wire, switches, IEC socket, RCA sockets, binding terminals

Of course, shipping adds to the cost, but $300 is a reasonable approximation. Three of the PCB boards cost about $50.

Better output transformers would add to the cost but there is not much else in parts that would drastically improve the sound quality of the amp. If I rebuild the amp, I would probably improve the grounding by using 12 gauge wire or a strip of copper as a bus bar and re-positioning the module boards so that the ground tap lies next to the 12 gauge wire.

The finished product. My wife is a very good quilter but she recognizes that despite her considerable abilities, she cannot attain the 'technically physically perfection' that others can achieve. Obviously the same applies to me. What you can't see in this photo are the washers that I install between the top piece of plywood and the wooden sides. They provide a narrow slit around the top and presumably increase ventilation.

The 6S17K-V tubes are mounted under the one inch holes just to the right of the RCA jacks. The holes allow for ventilation.

It would be relatively simple to substitute a GM70 for the GK71. The heater requirements are the same however you would need to provide a higher B+. Erick Bates used 720 vdc and a 6K output transformer. You could possibly use an 813 but the heater requirements are different and I would advise you to look at Alex Kitic's 813 project. The advantage of the GK71 is the lower B+. 

I have no idea how safe these stacked modules are. Use them at your own risk.

Cat proof.

My speakers are 15" Tannoy reds that I use on an open baffle. The low bass is handled by 18" Goldwood 1858s powered by a digital amp. A PLLXO handles the crossover. Since the Tannoys are 15 ohm, I will eventually replace the output transformers with units with 16 and 8 ohm taps. The transformers can be smaller since they don't need to handle anything below 80 hz. I would probably look for transformers that could handle a bit more current.

As a side note, the Tannoys sound much better to my ears on an open baffle than in any box I have heard. The detail is incredible and I presume this is because the 15" cone is not having to do so much work handling the low frequencies and pushing against the air in a box. I inherited them from my father and never had the room for boxes that would be suitable. Now I will never put them in a box and I won't bother looking for other drivers.

Getting back to the amplifier and the object of the exercise, the weight of the various parts without the box turned out to be about 10 lb 8 oz (5kg). The plywood top weighs about 3 pounds and the sides about 4 lb 8 oz for a total of about 18 lb. That is pretty light weight for a direct heated transmitting tube stereo amp. The next challenge is to determine how to make the box less heavy. Perhaps I could cut large holes in the sides or use some lighter weight material. Anyhow, it has been a fun project.

No doubt, you want to know how it sounds. I used inexpensive speakers for a while before I tried the amp on the Tannoys. So far the results are impressive. The amp is quiet, no hum or buzz. This stacked SMPS and AC heating works. I am currently listening to Beethoven's Cello Sonata #2 with Steven and Carol Honinburg and the cello has a wonderful sound. The sound seems to be even more detailed than the 6C33C amp and the sound stage is deeper. Initially I thought the new amp was less dynamic but now it is just as dynamic as the 6C33C. More later as the amp breaks in but this amp is a definite keeper. Many thanks to all who provided the various ideas.

By the way, I have recovered very well from the surgery. I am back to doing my regular 5k walk around my community and I have a trip to Rome planned in late April. It was a very interesting trip with some wonderful gelati.

A followup. The amp is a bit too bright for my taste even though I hear more 'stuff' in familiar recordings than I do on other amps. I remembered that I had a pair of old 2.5 K Audio Note transformers each weighing over 20 pounds and I have arranged to run them outboard from the main amp. I use the 4 ohm taps on 8 ohm speakers and the 8 ohm taps on my 15 ohm Tannoys. And yes, they are better than the Edcores.

Monday, January 19, 2015

Cat Vomit Special - a 6E5P / 6C33C parafeed amp

When my mother-in-law died several years ago, we inherited her small ginger cat. We are primarily dog lovers, particularly Airedale Terriers. However, cats tend to like me even though I am relatively indifferent to them. Probably that is why they like me. Catty-watty is getting older and has a tendency to vomit. Since she likes to sit in warm places, amplifiers are attractive to her and she already has ruined the power supply of one old receiver with an appropriately placed dribble. While my wife would not be too upset if the cat disappeared in a puff of smoke, I am just an old softy. Consequently I have stopped using some of my tube amps, particularly those with top caps.

So here it is in all its glory, a tube amp designed to be cat vomit proof. The case is simply plywood, roughly 16" x 21" x 6". The cage on top is three wire cooling racks tied together with some wire to make a triangular shape.

The large tubes are the 6C33C output tubes. Those of you who are older might remember a Russian pilot defected to Japan in 1976 in a MIG fighter jet. The Japanese let the Americans examine the plane and they were puzzled that the Russians were still using tubes / valves in their planes. It turned out that the tubes were not effected by the electro magnetic pulse generated by a nuclear explosion. Ordinary solid state components would be effected or destroyed by the pulse. These are the same tubes used in the jet fighter. They cost about $16 each when you buy them from Russia or the former Soviet bloc countries. You might notice two large black resistors between the output tubes. These are 100 watt 1K ohm resistors that form the plate load for each tube in this parafeed design. The input / driver 6E5P tubes are close to the RCA input jacks.

The large aluminium bars are heat sinks for the 400 ohm bias resistors. They also serve as a base for the wire triangle.

These tubes and the resistors are putting out a combined 180 watts of heat approximately. This is too much heat to try to dissipate inside the chassis so I placed these items on top. The high voltage connections are covered with electricians tape, 

Here is the schematic of the amplifier and except for the parafeed topology is fairly typical of 6E5P - 6C33C amplifiers. Most parafeed amps have a plate choke instead of a large resistor at R5. It's much less expensive to run at a higher B+ voltage and use a resistor instead. I got the idea from the Steve Bench web site. (see the third example, RC coupled)

I have used this topology quite a few times and I like the sound of it. I've even used light bulbs as the resistor and I reckon they sound 'better' than a straight resistor even though they are not as linear as a resistor. Think of them as a seasoning in cooking. I may yet try the amp with a couple of 50 watt Australian bulbs. Each bulb has an approximate resistance of 100 ohms.  

P = V I
V = I R
R = V / I
R = V * V / P

250 x 250 / 50 = 1000

 I've used Australian 50 watt bulbs with my 829B amp and also with my 6V6 Lumiere amp. You have a to fine tune a bit with a resistor in series. Generally the bulbs are a bit over 1K. The 'bulb sound' is present even with a normal resistor in series with the bulb. I suspect the 'bulb sound' is some sort of resonance that happens to appeal to my ears. It may not appeal to you.

The circuit is very basic otherwise with auto bias on the out put tubes. Some builders state that this output tube sounds better with fixed bias. They may well be correct but this tube is reputed to be finicky and liable to 'run away' and self destruct with fixed bias. Since this is the first time I have used this tube, I chose the safer route.

The CCS comes from Gary Pimm and is easily set to 15 ma. Other builders using the 6E5P have used 13 or 17 ma. I chose to use the halve the difference. Another possibility is a CCS using 4 DN2540 depletion mosfets.

The ceramic sockets for the 6C33C tubes are mounted with approximately one cm clearance to allow for any heat from inside the chassis to escape easily. I have not used any fans.

Since I was not sure how the 6C33C tube was going to behave, I installed meters to monitor the bias voltage which for this amp was supposed to be -80 volts. The operating point is 200 volts, 200 ma and -80 volts which reputedly produces about 14 watts per channel. On the other side of the amp is another similar meter that monitors the B+ at 480 volts.

Each meter is a 1 ma meter

For the bias meters I hook them up with a 100K resistor between the meter and the cathode of the 6C33C and the other side of the meter connects to ground. For the B+ I use the same meter but with a string of resistors that add up to 1 meg. I allow 100 to 150 volts for each resistor so in this case I just have five 200k resistors. The B+ meter reads approximately 48.

B+ is 480.
The drop over the 1K resistor is 200.
The drop over the plate is 200.
The bias is 80.

The power supply is the contraversial LSES supply that has been extensively argued over in the TubeDIY forum. I could have used the typical large 5 or 10 henry Hammond choke from my stash but the largest chokes can only take 300 ma. Because of space limitations in my listening area, I can't fit mono-blocks so I use a single B+ power supply that feeds both channels. Since two channels would need 400 ma, my Hammond chokes would not be usable. The C-56U and C-40X chokes from Allied Electronics easily handle 400 ma. The 6E5P run at 15 ma each.

I would strongly advise you never to go to that forum and mention this power supply. You would be stirring up a hornets nest yet again which does not need to be poked. FWIW several years ago when the arguments were flying thick and fast, I built such a power supply and tried it out. As part of the process, I installed a couple of switches where I could substitute one of my 10 henry Hammond chokes for one of the C-40X chokes and switch back again relatively quickly. The small C-40X choke sounded much better than using the Hammond which seemed to put a blanket over the sound. The effect was even more pronounced when I replaced both C-40X chokes with 10 henry Hammonds. Large 5H Hammond chokes were not much better. Feel free to differ on this subject. For this amplifier, it was very convenient to use and it produced the 480 volts that the circuit called for.

The power transformer is a 400 watt Antek toroid AN- 4T450 (450 x 2 and 6.3 v x 2). I use the 6.3 volt windings to power the two 6E5P tubes. I use two separate 12.6 volt 5 amp transformers to light the filaments of the 6C33C tubes.

The capacitors are motor run which I attach to the underside of the top using Goop. It suspends the weight of these capacitors quite easily and the bond is difficult but not impossible to break. I use a star grounding system. The diodes are 2 x UF4007 in a bridge arrangement.

While planning this amp, I got a lot of information from Romy the Cat's website and his Melquaides 6C33C amp. In particular I found his 6C33C survival guide to be very useful. In particular I noted the one hour breaking period for the filaments. The dual DPDT switch arrangement I have been using for years was perfect for turning on only the filament voltage and waiting the appropriate amount of time to switch on the high voltage. Each time I fire up the amp now, I wait 10 minutes after turning on the filaments before I turn on the high voltage.

I also use a CL-90 on the 115 side of the toroid to limit the inrush current to the toroid.

I was amazed to find that the voltages throughout the amp were very close to what I had predicted, within a few volts. Pigs do fly after all.

The output transformers come from a couple of Tannoy CVS 6 ceiling speakers that I won off Ebay several years back. Included with each unit was a very well made 60 watt THP-60 line transformer. I used the speaker units but did not need transformers but realized that these transformers might be useful if used as a parafeed transformer for the 6C33C tube. Note that you cannot use them like a standard SE transformer since they do not have an air gap and cannot tolerate much DC.

I have found the following formula to be useful for line transformers.

resistance = (voltage x voltage) / power

For example on the high voltage side connecting the blue 70v 7.5 watt wire to the plate via a parafeed capacitor and the black wire to ground

(70 x 70) / 7.5 = 653 ohms 

or the green 15 watt wire instead of the blue wire

(70 x 70) / 15 = 326 ohms

I have used the blue wire 653 ohm load connection so far. Perhaps one day I might get around to trying the 326 ohm load.

The coupling capacitor C2 is two 0.1uf Russian KSG-2 500v silver mica capacitors in parallel. Let's just say I prefer them to everything else I have tried and they are quite inexpensive at about $4 each. I have read somewhere that they can leak DC but I am not too concerned.

The parafeed caps, C4 are two 2uf MGBO-2 PIO caps in parallel. I am using the 300 volt versions which is a bit too close to the 280 volts of the circuit, but I figure there is a bit of headroom and I haven't had problems so far. Next time I order Russian caps, I will get some higher voltage rated versions. I bypass each caps with a 6800 pf KSO silver mica cap. If you click on the links you can see they are relatively inexpensive.

I have not played around with varying the parafeed caps.

So no doubt you want to know if it sounds any good. I reckon it is pretty decent and it appears to be very good with details (resolution). For many years my resolution test has been a CD by the Tallis Scholars of choral music by Josquin Des Pres. The first track, Pange Lingua, has a very echoing acoustic recorded in Merton College, Oxford. Usually one person is singing at any given time but occasionally there are two or more and with the echos, it is difficult to tell the number of singers. This is the first amp I have built or listened to where I can tell which is which.

I started to total how much this amp cost and I reckon it will be somewhere about $400 - $450 depending on how much the output transformers really cost. I got both Tannoy speakers and transformers for about $50 from Ebay. Normally they cost about $170 each new.  Tannoy does not produce junk and these transformers appear to be well made.

It might be possible to use a 115:24 volt toroid as a substitute. (115 + 115) * (115 + 115) / (24 * 24) * 8 = 734 ohms). The 115 volt windings are in series and the 24 volt windings are in parallel. I used 9 volt toroids in my parafeed 6V6 amp as output transformers and after a long breaking in period, they sounded better than I expected.

Under the hood.  The Tannoy transformers are at the top two corners. The two transformers on the right are the 12.6 volt 5 amp transformers for the 6C33C tubes.  The large toroid is the power transformer. The blue wires are the 6.3 volts powering the 6E5P tubes. The high voltage power supply is lower left. I have tried to implement a star ground (green wire). The Gary Pimm CCS is between the two ceramic 6C33C sockets. It uses a small heat sink of aluminium angle. The brown objects are the Russian coupling and parafeed capacitors.

The motor run caps are attached to the plywood using Goop. The bond is quite strong. The other items are screwed into the plywood.

Back to the cat vomit. You can see the holes I cut to go around the outsides of the 6C33C tubes. The wires at the side prevent to cat from venturing inside.

So far the triangle has worked.

So here is Catty-watty (aka Honey) wandering past the amp. The first time, she looked at it and you could see her wondering 'What the hell is this?'

So the true test of the design is whether it worked. The cat vomited but not in the direction of the high voltage and has lived to meow another day.

Sunday, April 27, 2014

HV delay using two DPDT switches

Years ago Glass Audio published an article that described a simple way to delay the high voltage when turning on your tube amp. I've used it many times and it works well.

When you turn on either of the switches, the filament transformer comes on first. After a delay of your choosing, turning on the other switch turns on the high voltage. Turning off either switch turns off the high voltage first.

There are two disadvantages. First, the circuit requires two separate transformers. Second, if you lose power and then it comes back on before you can turn off the switches there is no delay. In practice, the tubes are usually still warm if the power outage is short.

The ordinary DPDT switches from Radio Shack work fine and you are only putting 115 vac through them so this circuit is good for higher voltage amplifiers. When testing your amp for the first time, it can be helpful to check that your filaments are lit and at the correct voltage before applying high voltage.

Finally, most people can't turn on the amplifier unless they are given instructions. 

This diagram may help you visualize how to wire the switches.

Wednesday, February 8, 2012

Tube Taster Linestage

Some of you who look at the Bottlehead internet Forum might remember a post by Doc B back in January of 2001 extolling the virtues of Dalwhinnie single malt whiskey. It provoked a flood of messages where the boozers amongst us extolled the virtues of Lagluvin, Laphroaig, Balvenie etc. My contribution follows:

A couple of years back, spouse gave me a twelve pack of sample size bottles. It was very instructive to go through them and see what I actually liked. My favorites were Dalwhinnie and Lagluvin.
What I need is a twelve pack of various tubes to taste. Now that is an idea. Designing a generic input tube tester should be possible for 7, 9 and octal.
The idea lingered, longer than the Lagluvin, and I decided to design and build a Tube Taster. Something where I could try out these various input tubes such as 6N1P, 5965, 76, 6SN7 etc and get an idea of the sonic flavor in a standard setup. Like many of you, I have a stash of tubes that I was going to do something with one day, a couple of 76’s, 6N1P, 5965, 5867, 12AU7, 6080 and so on. I decided that I would just take a standard 12” x 12” piece of aluminum, put in a bunch of 9 pin and octal sockets plus a couple of five pin sockets for the 76’s and see what could be done. I also wanted to do it inexpensively and simply.

I really enjoyed John Day’s Valve article on the Paraline with the output transformers. Looking at the schematic, I realized that I could use the guts of it and replicate it a number of times using switches. At this point, it probably would help to see a schematic.

Everything inside the box is repeated six times. For example, there are six R3 resistors, six R4 resistors, six R2 resistors and six C1 capacitors per channel. However, there is only one R1 resistor and one C2 capacitor and one output line transformer per channel. The key to it all is the two pole, six position switch for each channel.

As you can see, I use a two pole, six position switch for each channel. One pole is used to connect to the input resistor, R1. The other pole is connected to the capacitor C2 connecting to the output transformer. Since there are six positions, I can accommodate a maximum of six different tube types. Let’s look at wiring tube 1. Connect position 1 to the plate side of resistor R3 that connects to the plate of tube 1. Connect position 7 to the grid of tube 1.

It does not matter about the position of the switch, for each tube, current will flow from B+ thru each R3 to each tube thru the appropriate R4 to ground. The grid resistor for each tube ensures that the bias is set correctly for each tube. Since the topology is parafeed and no DC current is passing thru C2 or the output transformer, each tube operating point is stable and predictable irrespective of switch position. In fact it would be possible to listen to a 76 through the left channel and a 6N1P though the right channel.

If we desire to actually use tube 1 and the switch is positioned to contact positions 1 and 7, then the signal will flow R1 - pole A – position 7 – grid 1 – plate tube 1 – position 1 – pole B – C2.

If we desire to actually use tube 4 and the switch is positioned to contact positions 4 and 10, then the signal will flow R1 - pole A – position 10 – grid 4 – plate tube 4 – position 4 – pole B – C2.

I finished up with two five-pin sockets for the 76’s, three ninepin sockets and two octals.  To make things simple, I tried to use a standard operating point of 200 volts and 10 ma since that point is pretty easy to pick out on most plate curve graphs. Incidentally, the 76s run at 6ma.

The six position switches are break-before-make Lorlin 10WA155 from Mouser at the time of build $2.57 each. They are a nylon design rated to 1000 volts. The design calls for 200 volts to pass thru the switch, so to reduce the risk of electrocution, take care in the quality and design of the switches you use in this position. In addition, use non-metallic knobs. There are a lot of wires in this linestage and color-coding is useful if not essential. It also helps to number the tubes 1 thru 6 to avoid confusion, particularly when you use non-duals such as 76s.

The output transformers are Edcor WSM10k/600 at a whopping $8.24 each (at the time, currently $10.23). They arrived nicely packed between two pieces of plywood and appear to be built to a higher standard than the line transformers I get from Radio Shack, and the core was wrapped in a fetching shade of yellow tape. On their website, they use M6 for core laminations. These transformers are wonderful for playing with. No doubt the more expensive transformers are better, but you can go a long way with these at very little cost.

The power transformer is a simple global use dual primary 115-230, dual secondary 115-230 43va unit. Typically, these cost less than $20. Mine actually came from Allied, but Mouser and Digikey also sell them. The 43va rating is a bit low if you plan to run all of the tubes at once. I usually run with half the sockets empty. You might consider lashing out and buying the 80va model for a few extra bucks. I simply had a 43va unit on hand, left over from another incomplete (failed) project. It’s pretty easy to get 300 – 320 volts with the usual ultrafast diodes and a simple CRC or CLC smoothing circuit. There is nothing special here, though I did some of the Foreplay mods to the power supply. I’ll leave them up to you.

What is special in the power supply is the turn on circuit that I filched from an old Glass Audio article that credits an even older document. I sometimes wonder if there is really anything new in this field. It requires that the 6 volt transformer be separate from the high voltage transformer, but they are cheap. Simply turn on either one of the DPDT switches and the heater voltage is applied. Wait whatever period you think is necessary for the heaters to get warmed up and then turn on the other switch and the high voltage is applied. It does not matter which of the switches you turn on first. Turn off is just as simple. Turn off either one, which turns off the high voltage, and then turn off the other to remove the 6 volts from the heaters. I use this arrangement on my power amps as well. I used a Radio Shack 12.6VAC CT and wired 3 tubes to one 6.3VAC half and the remaining tubes to the other 6.3VAC half. The DPDT switches from Radio Shack work fine.
Be aware that this arrangement does not work if there is a short power outage. Where I live we get a lot of outages and when they occur I simply turn off the switches if I can before the power returns. If the power is out for a short time I figure the filaments are still hot.

When changing from one tube type to another on the preamp, I first remove the high voltage first from the amp, and then remove the high voltage from the preamp. I then turn the two rotary Lorlin switches to the different input tube and then turn on the high voltage for the preamp and then the high voltage for the power amp. It takes but a few seconds with the heaters still in operation the whole time. Turning the Lorin rotary switches without following this routine results in a pop at the loudspeakers.

So the big question is whether you can actually hear differences between the various tubes. The answer is yes, which is not surprising. When I first started to get music to appear at the loudspeakers I quickly shuffled between the various tubes and immediately noticed differences. Gain levels vary, of course, so I needed to adjust the volume to do real comparisons. Actually, I found it works best to resist the temptation to flick around between the tube types. Certain tubes seemed to suit certain music and I seemed to be consistent in my tastes. If I put this in whisky terms, if it’s really cold, wet and miserable outside, I prefer the peaty Islay types like Lagluvin and Laphroag. Otherwise if it’s just plain wet, then Dalwhinnie works for me. If it’s dry, then I prefer Ezra Brooks bourbon. If it’s after dinner, then Pierre Ferrand cognac is the tipple of choice. Variation according to your needs is wonderful.

You will notice I have not said which tubes I prefer. The reason is simple, I don’t want to influence your taste. Build it yourself and then you will know what you yourself really prefer.

Construction. The unit is designed to be turned over easily and worked on. I took the unit to a Bottlehead meeting in Elkton, Maryland, and its looks caused a sensation. Takes the whole business of fit and finish to a whole new level. Perhaps "gawdawful looking" seemed to be the most appropriate description.  Some of you may know that the old fashioned artisans used to put in a deliberate flaw in their projects to show that only God was perfect.  I decided to leave the half torn off yellow sticker from McMaster on the top aluminium plate as my deliberate flaw. Just to make sure that God noticed, I did the same thing on the acrylic bottom plate.

The Edcore transformers.

Wiring this linestage was a challenge. I used wire from IBM Type 1 cable. It's got some sort of non flammable foam insulation and I like using it. If it came from IBM in those days you could be sure it was top quality.

Sound quality. Let’s just say it sounds a whole lot better than it looks. Definitely not duct tape sound. There is something quite special about the parafeed topology, even in a preamp.

829B lightbulb amplifier

A few years ago, a friend I worked with mentioned to me that the old gentleman who lived across the road from him was getting rid of some tubes since he was moving to a smaller house and would I be interested in taking them. I duly finished up with three show boxes of assorted tubes. There were two peculiar tubes that looked like they came from 'My Favourite Martian' and so I did some research about them. They were 829B transmitting tubes.

I looked them up in my copy of the RCA Transmitting Tube manual and also poked around on the web. Almost immediately I found Pete Millet's web site and his 829B single ended amp. The seed was planted in my brain and eventually I decided that I had to do something with them.

Here is a link to Tom Schlangen's plate curves for the 829B wired as a triode,

I tend to prefer the parafeed version of single ended and had built an STC amp and an 807 amp using the cheap Radio Shack line transformers. I had been surprised by how good these little cheap transformers can sound provided you don't need the amp to play very low frequencies. I typically use line arrays with 12 of the famous Parts Express 99 cent 4" full range drivers (269-469) so the line transformers would be ok because the arrays don't go below 100 hz. I have been building various subwoofers over the years and eventually settled on a pair of the Bill Fitzmaurice Tuba 18 horns that I built which I really like. I use an old Aragon amp to drive them and an electronic crossover.

I also had some 6C45Pi tubes waiting to be used for something and I decided to try a two stage amp rather than the three stages of Pete's design. I didn't have any suitable plate chokes however and I was pondering whether I would lash out and finally buy some good ones when I did one of my annual trips back home to Australia. As usual, the trip gave me time to think and ponder and somewhere along the line, I remembered Steve Bench's web site where he describes different versions of single ended topology. In particular, I found his comments about the resistor loaded version (#3) intriguing. "I would like to add that I find something magical about the sound of the resulting amplifier, in spite of its disadvantages". Then I also remembered the Nelson Pass Zen amplifier where he used light bulbs as resistor loads.

So what light bulb do you use? Based on Pete's amp, I figured I would probably be passing about 80ma through the bulbs and that I would need a load of about 4k ohms. I would be dropping 4000 x .08 volts = 320 volts and dissipating 320 x .08 = 25.6 watts. That's a fair bit of heat. The light bulb idea was looking better and better.
So what size light bulb should I use. Nelson Pass points out that the resistance of a light bulb changes with the voltage applied, but I figured I would start off with the simple P = VI and V=IR equations.

100115 0.87 132 
50115 0.43 264 
25115 0.22 529 
12 1150.10 1102 
This wasn't working too well. Since I was in Australia where the household voltage is 240 volts, I tried that.

100240 0.42 576 
50240 0.21 1152 
25240 0.10 2304
122400.05 4608 
That 25 watt bulb looked like a real possibility. I could use two in series for approximately 4600 ohms and a combined power capacity of 50 watts. Maybe it would work and maybe not, so on the way to the airport I stopped off and bought four of the bayonet sockets commonly used in Australia and six 25 watt bulbs. Somehow they let me carry the bulbs on board the plane in my hand luggage and they all arrived home safely.

Unfortunately, when you do the calculations, 115 volt bulbs don't produce resistances that are as useful. Also note that UK and Aussie lightbulbs have a bayonet mount.
It's about time you saw the schematic. But first, the warning.

This amp operates at very high voltages that can kill you. Do not attempt to build this amplifier unless you have already built several other amplifiers, preferably kits so that you can learn good techniques. I would highly recommend the Bottlehead kits. The exposed pins on top of the 829B are a definite safety hazard. Do not use this tube if you have small children around, or even curious adults.

I have not included the power supply. I will leave that up to you. This power supply will need to produce approximately 600 volts so be very careful. I would recommend that you look at other power supplies and the PSUD design tool. There is always considerable debate about what makes the best power supply. I typically do better with UF4007 rectification rather than tube, and I like using motor run caps.  I'm currently messing around with low DCR power supplies in my latest creation.

Since I typically use UF4007 diodes for rectification, I also use the two DPDT switch solution that allows you to turn on the 6 volt stuff first and then to throw the other switch for the high voltage. It doesn't matter which switch you throw first and the switches only have to handle wall voltage. It works very well and the only disadvantage I know about is that it doesn't handle a power outage where the power comes back on before you can turn the amp off. That doesn't happen too often. The circuit appeared in a Glass Audio but the author in the article got it from an old manual.

When you look at the pictures of the amp, you will see that I have a board that I got out of an old IBM computer that was to be thrown away. I figured it desrved to be kept in use. I don't know whether it helps or not.
The toroid come from Plitron.

The Hammond chokes and the 30uf 600VAC motor run capacitors came from a failed project where I realized I had bitten off more than I should chew. In retrospect, I was lucky I stopped when I did.

For the heaters, I use a Radio Shack 12 volt 3 amp transformer. It gets a bit warm with use. For the 6 volt wiring, I use some IBM type 1 plenum cable (solid copper 22#, FEP, twisted pair, sheilding galore). I ground the shielding. I like using this stuff for signal wiring as well.

I had read about being very careful with the 6C45Pi tubes with grip stoppers and cutting off the unused solder lugs to prevent oscillation, so I did as well as I could. So far, I haven't had any oscillation problems.

I used the euro-style terminal block that Pete suggests for connecting the antenna style pins on the 829B tubes. I have noticed that over time that they tend to loosen a bit and need to be tightened. You need to think about how much additional insulation you need beyond the insulation of the wires carring the high voltage to the plate pins.

I wired the output transformers as autoformers. To my ears, they sound better that way. I've since built another creation where I can switch from normal configuration to autoformer configuration and back with little interruption and I still prefer the autoformer way. You may differ in your opinion, they are your ears. Radio Shack has stopped selling the line transformers, but there are plenty of other places to buy the equivalents. Be aware that these Radio Shack transformers are not supposed to do too well below 100 hz. It's not a problem for me since I use a separate subwoofer.

I realized that with this amp, there were a number of things that might not work too well, so I tried to build the amp in modules so I could replace a module if it was unsuccessful.  I also like using polycarbonate instead of aluminium. Yes, I know I should use metal, but I hate using it since I always manage to cut myself on the holes etc. I tried polycarbonate and haven't gone back. I have since started to do what Pete Millet does and use single sided copper pcb and turrets for the main circuit. That works well for me and I still use polycarbonate for the rest of the amp.

I also had an IKEA Ivar side unit kicking around and decided to use it to support the modules. The power supply is at one end and the input 6C45Pi tubes are at the other. There is a reasonably logical progression from one end to the other.

Fortunately, the amp worked immediately, though I had some failures when I transported it to some Bottlehead meets in Maryland. I am gradually improving in my soldering abilities, but I'm still not too good at it.
I really like this amp and many others who have heard it at the Bottlehead meets have liked it as well. It certainly gets your attention though some have mentioned to me that their wives would never allow it in their living room. I'm not particularly good at describing sounds and would never make an audio reviewer, but Steve Bench's 'magical' description works for me.

Incidently, when the music gets loud, the light bulbs flicker in time to the music. Normally they put off a reaonable amount of light, but not a lot.

More on the line transformers.
Line transformers are used in public address systems to distribute output from an amplifier to multiple sets of speakers. From what I see, line transformers do not have an airgap. If you go to the Edcor web site you will see that they come for use in 25, 70 and 100 volt systems, and that they have input taps rated in watts and output taps of usually 4, 8 or 16 ohms. This all confused me initially until I looked at the Mouser website and a pdf file that was included with the description of their 42KB001 Audio Line Matching Transformer. This sheet lists the number of input and output turns for the various taps. Suddenly it all made sense. I put together a spreadsheet for the 42KB001 with the 8-ohm tap.

8 ohm
/ power
10665 857.82 61.21   490490 
5 94085 11.06 122.30 978 980 
2.5 133085 15.65 244.83 1959 1960 
1.251880  85 22.12489.19 3914 3920 
0.6252660 85 31.29 979.32 7835  7840

 If the 4 ohm tap is used, the spreadsheet becomes:

4 ohm
/ power
10665 6011.08122.84 491490 
5 94060 15.67245.44982980 
2.5 133060 22.17491.363 1965 1960 
1.251880 60 31.33981.783927 3920 
0.6252660 60 44.331965.447862  7840

As you can see the load results are much the same.

The final column is simply resistance = (voltage x voltage) / power. Using this formula, we can produce the following table for the various voltage and power combinations.

watts 25
60 1082  167
3021  163 333
15 42 327667 
1063 490 1000 
5125 980 2000 
2.5250 1960 4000 
1.25500 3920 8000 
0.6251000 7840 16000 

As you can see, there are several potentially interesting combinations.

I decided to try out one of the low DCR, LSES powersupplies that was talked about on tubediy. I built on a 12" x 24" sheet of polycarbonate with a switch that enable me to have a B+ of 300v or 600v. I also can optionally lower the voltages by switching in some resistors on the 115v side of the industrial transformer. I keep track of this with some DC voltage and currrent meters. The voltage meters are actually a 1ma current meter in series with five 200k resistors (effectively 1m). Next to the meters I have some terminal blocks that I use to attach the removable amplifier modules. The meters are useful to make sure that the voltages are where they should be and that they are at zero volts when I change amp modules. While in use, I cover the terminal blocks with a strip of polycarbonate to keep prying fingers away. I would not recommend this method if you have small children around.
The 829B amp was rebuilt on a 12" x 12" sheet of polycarbonate as an amplifier module. There were no changes to the circuit. It sounded a little brighter if anything compared to the first version. Curiously, the light bulbs do not flicker as much now, except at volumes.

I then decided to try a Gary Pimm self bias ccs load on the 6C45PI tubes. I have them set at 15ma and after some months of listening, I think there is an improvement. It seems to be a greater sense of control and dynamics.