Transcendent Sound Forum
Latest Topics
 
 
 


Reply
  Author   Comment  
SteveM UK
Reply with quote  #1 

Hi All,

Have finally got the Beasts back up and running after a lengthy period with them up in the attic.

They were originally scratch built some years ago, using Bruce’s basic pcb and diode kit, and whilst they showed considerable promise, I was never totally happy with them and had a number of problems, primarily with the output tubes. To keep costs down, I had decided to use the Svetlana EL509’s but poor reliability and the classic problem of trying to balance the banks to zero the offset eventually  forced me to retire them in favour of my Mullard 5-20 based amp, and a Krell KSA50.

We have recently moved house, and so the amps were duly extracted from the attic for the move. I decided that I really ought to have another go at getting them going having invested a great deal of time and money into them, so started looking at solutions to the problems I originally encountered. As mentioned, the fundamental problem was the output tubes. I couldn’t afford to re-tube with the JJ octal units, plus it would have meant totally re-building the output stage, so the first step was to buy a few more Svetlana EL509’s. The second step was to trawl through this forum to look for similar problems and a number of tips and suggestions were logged. The third (and what turned out to be the most significant) step was to build a tube tester. Whilst I have a basic knowledge of tube circuits, I am by no means an expert, so decided that replicating the output stage circuit for one tube should do the trick. My Beasts effectively use the output configuration in the T8s as transformer availability meant I am running at +/- 170V rails. Screen grid resistors are 100 ohm and I am only using 12 tubes rather than 16.

A quick rummage through the odds and sods box unearthed a number of useful parts for the tester, and a trip to the local electronics store got me the rest. Bias can be adjusted from -30 to -60V, and to help with stability, the grid is AC coupled to earth via a capacitor. (Not sure if this is necessary) Bias and current measurements are taken via hand held multimeters plugged into sockets on the tester.

Once I started testing the tubes, the reason for the original problems very quickly became apparent. Around 10% of the tubes were basically dead (or had died during my initial sessions with the amps), and many others either had very low, or very high current draw. The other issue I noted was that several of the tubes were virtually switched off at the -50V bias voltage. Having read some of the posts here where other owners had gained benefits from moving towards -40V (Which is where the T8’s are biased) I decided to look at the pcb’s and change a few resistor values. I settled on a combination which gave me -42V, and so re-tested the tubes at this level. I finally ended up with around 30 tubes, all between 20mA and 80mA current draw at this bias setting. A quick bit of maths and I had 4 banks of 6 tubes all with similar total current.

Things were starting to look positive. The next step was to check out the amps, I fired them up with the driver tubes in place but no output tubes. Voltages all seemed to be OK around the pcb, but checks on the output tube pins highlighted a couple of dead 100 ohm screen grid resistors which were duly replaced. Bias was checked and set at -42V and -212V. The only thing left to do now was fit the output tubes, fit a load resistor across the output and switch on with fingers crossed.

All went well. No bang, no smoke. Everything just sat there quite happily. It all just worked as it should. Both amps were left running for a couple of hours. Bias needed tweaking slightly, but no real issues and very easy to get them balanced. So, the final step. They were connected into the system and switched on. I use the GG as a preamp and the combination is amazingly quiet. No hum, and just the faintest bit of noise if you have your ear to the speaker. Stuck a CD into the player, sat down and was just blown away with what these things now do. Totally different to when they were originally built. They sounded a bit ragged for the first hour or so, but very quickly started to settle down and open up. You guys know how these things sound. I don’t need to go into detail, but as you can imagine, I am over the moon with the transformation.

So, one very happy bunny. It has been quite an interesting challenge. I’ve learnt a great deal from it, and would suggest that anyone who uses the old Svetlanas would benefit from having a tube tester. It makes setting up the amps so much easier. Yes, you can check the bias current with the tubes in the amp, if you are happy and able to get to the pins while the amp is running, but this assumes you can get the amp to run in a stable condition in the first place. I had a couple of tubes in which the current just kept steadily increasing, and with them in the tester, was able to see this and discard them. If they had been in the amp, they would probably have just continued to pass more and more current until they blew a fuse. I have probably discarded around 25% of the Svetlana tubes I had. There was little wonder I was having problems.

On the positive side though, I now have several spare pairs of matched tubes and the layout in the output stage is also with matched pairs where possible. If one goes down, I will simply replace the pair to maintain balance. I believe there is the circuit for a tube tester in Bruce’s latest book, but as mentioned above, it is fairly easy to build one particularly if you are only going to use it for the tubes in the Beast. For those of you who are comfortable swapping out components, I would also encourage you to have a go at moving the bias point to around -40V, especially if you again have the Svetlanas. It certainly seems to have worked for me.

Thanks to you all for your continued input into this forum. It was a great help in getting the amps back up and running. And, of course thanks to Bruce for a superb design.

Steve.

Beast 1.jpg 


glt
Reply with quote  #2 
Nice post. I enjoyed reading about your efforts and success.
SONDEK
Reply with quote  #3 
Me too.

I love a happy ending!
[biggrin]
SteveM UK
Reply with quote  #4 

Thanks Guys,

I have a couple of questions which you (or Bruce) may be able to answer for me. Apologies if the questions are dumb, have been answered before, or are explained in one of Bruce’s books.

In the T8s, the 2 ohm resistors in the output stage are connected to the cathode, whereas in the Beast they are on the anode. Is one position better than the other, and does it matter?

The second question is regarding the bias voltage circuit. In both amps, we have a separate circuit to generate the bias voltages, and this is referenced to earth. This, I assume is fine for the +ve bank, but on the –ve bank, the cathode is connected to the –ve supply. If this supply rail varies under load, it will modulate the current through the tube in the same way that the music signal does on the grid, but will actually work against it. For example, in my case, I have -212V bias and -170V rails, giving a bias voltage of -42V. If the rails are pulled down to say -165v on transients, by bias voltage is now -47V. I appreciate that the voltage drive for the signal sits on top of this, but the reduction in rail voltage will be reducing current through the tube while the signal is trying to increase it, thus the two are fighting against each other. (If my simple logic is correct)

So, this is a fairly long winded way of asking if it would be better to have a separate bias supply for the –ve bank, referenced to the –ve supply rail, so it can track the rail voltage, or is it simply not worth the extra complexity that this would involve.

I look forward to hearing your thoughts. Thank you.

Steve.

Bruce Rozenblit
Reply with quote  #5 
Very shrewd of you to figure that out.  This problem has been addressed in the Son of Beast amp.  The details are in my book.

 
SteveM UK
Reply with quote  #6 
Thanks Bruce,

Book has been ordered [smile]

Steve
SteveM UK
Reply with quote  #7 
Hi All,

Following Bruce's comments in post #5, I just had to have a go at seeing what would happen if I tied the bias for the negative bank to the -ve rail as Bruce has done in the Son of Beast amplifier.

Having looked through the book, the SoB amp uses a voltage quadrupling circuit powered from the 12v filament supply. I decided to go a slightly different route and built two simple zener diode regulated 60v supplies on a separate small circuit board. These were mounted in the chassis and connected to the rest of the amp. The output was connected across a 22k resistor and linear 22k pot in series giving an adjustment of -30 to -60V (In my amp, I have fixed the bias of the +ve bank at -42V and adjust for zero offset at the output using the -ve bank) The positive side of this board was connected to the -ve supply rail, with power being taken from pads WT1 and WT6 on the main board which Bruce had thoughtfully included in parallel with the connections to the existing on board bias transformer and left spare. The bias voltage was connected directly into the output stage wiring.

We obviously needed to retain the -ve drive signal, but disconnect the on-board bias. The simplest way to achieve this was to remove R34.

Whilst I had the amplifier upside down carrying out the mods, I noticed that R32 and R33, the two cathode resisitors on the -ve half of V3, the 12BH7, were a little discoloured and were obviously getting hot. Once I switched the amp back on to set the bias voltages, I checked out the voltages around V3 and found that the anode was at around 545V. I assumed that this high voltage was potentially causing the problem with the cathode resistors. Fortunately I had a few spares of the 62V zeners that I had used in the new bias circuit, so swapped out one of the 100V zeners in the string controlling the voltage to pin 6. This brought the anode voltage down to just over 500V and everything seemed fine.

Once I had the bias voltages adjusted, with the output re-set, I settled down to listen to some music and see what the mods had done. Unfortunately this was short lived, as a small pop through the speakers signalled the start of a frustrating hour as one channel was now seriously distorted and much quieter than the other. I assumed I had lost either the +ve or -ve part of the signal. To try and isolate the problem, I left the amp connected to the pre-amp, but wired a dummy load across the output. With a CD still playing, I was able to trace the signal through the amp with the multimeter set to AC voltage. Sure enough, the -ve signal stopped at V3, and on further investigation, yes you've guessed correctly, those two cathode resistors had finally given up the ghost. Kicking myself for not replacing them when I had the chance, I rummaged through my parts bin to find a combination of resistors that would give me the correct value, and all was well again. I'm not sure if it was co-incidence, or if altering the voltage at the anode had finally caused them to die.

So is it worth carrying out the mod? The amps haven't got too many hours on them since they were resurrected, so they are still settling in. I can't say that I noticed anything immediately obvious, but more I listen to them however, the more it is becoming apparent that they are just generally a little bit better at doing most things. As you all know, they are very detailed amps, but the amount of information that they can now dig out of a recording is simply astonishing. Transients are also now handled with ease. I have literally been startled at times by the speed of these amps and the way they can go very loud, very quickly without the slightest hint of compression. Plucked instruments take on a whole new life-like quality, and well recorded acoustic guitar is just stunning.

I have also noticed two other advantages. The offset at the output now seems to be more stable. With the original set-up, the output would fluctuate by around 20mV, but now even a digital multimeter hardly changes once everything has warmed up. The other improvement is that now, when switching the amps off, they are perfectly silent. Prior to the mod, there was always a small 'thump' from the speakers.

So, yes, I am happy with the mods. I think it is worth having a go at this, and I would be very interested to hear from anyone who has done the same thing. I am also happy to help if anyone is thinking of trying it out.

Cheers,

Steve.
desaudio
Reply with quote  #8 
Quote:
Originally Posted by SteveM UK
Hi All,

Following Bruce's comments in post #5, I just had to have a go at seeing what would happen if I tied the bias for the negative bank to the -ve rail as Bruce has done in the Son of Beast amplifier.

Having looked through the book, the SoB amp uses a voltage quadrupling circuit powered from the 12v filament supply. I decided to go a slightly different route and built two simple zener diode regulated 60v supplies on a separate small circuit board. These were mounted in the chassis and connected to the rest of the amp. The output was connected across a 22k resistor and linear 22k pot in series giving an adjustment of -30 to -60V (In my amp, I have fixed the bias of the +ve bank at -42V and adjust for zero offset at the output using the -ve bank) The positive side of this board was connected to the -ve supply rail, with power being taken from pads WT1 and WT6 on the main board which Bruce had thoughtfully included in parallel with the connections to the existing on board bias transformer and left spare. The bias voltage was connected directly into the output stage wiring.

We obviously needed to retain the -ve drive signal, but disconnect the on-board bias. The simplest way to achieve this was to remove R34.

Whilst I had the amplifier upside down carrying out the mods, I noticed that R32 and R33, the two cathode resisitors on the -ve half of V3, the 12BH7, were a little discoloured and were obviously getting hot. Once I switched the amp back on to set the bias voltages, I checked out the voltages around V3 and found that the anode was at around 545V. I assumed that this high voltage was potentially causing the problem with the cathode resistors. Fortunately I had a few spares of the 62V zeners that I had used in the new bias circuit, so swapped out one of the 100V zeners in the string controlling the voltage to pin 6. This brought the anode voltage down to just over 500V and everything seemed fine.

Once I had the bias voltages adjusted, with the output re-set, I settled down to listen to some music and see what the mods had done. Unfortunately this was short lived, as a small pop through the speakers signalled the start of a frustrating hour as one channel was now seriously distorted and much quieter than the other. I assumed I had lost either the +ve or -ve part of the signal. To try and isolate the problem, I left the amp connected to the pre-amp, but wired a dummy load across the output. With a CD still playing, I was able to trace the signal through the amp with the multimeter set to AC voltage. Sure enough, the -ve signal stopped at V3, and on further investigation, yes you've guessed correctly, those two cathode resistors had finally given up the ghost. Kicking myself for not replacing them when I had the chance, I rummaged through my parts bin to find a combination of resistors that would give me the correct value, and all was well again. I'm not sure if it was co-incidence, or if altering the voltage at the anode had finally caused them to die.

So is it worth carrying out the mod? The amps haven't got too many hours on them since they were resurrected, so they are still settling in. I can't say that I noticed anything immediately obvious, but more I listen to them however, the more it is becoming apparent that they are just generally a little bit better at doing most things. As you all know, they are very detailed amps, but the amount of information that they can now dig out of a recording is simply astonishing. Transients are also now handled with ease. I have literally been startled at times by the speed of these amps and the way they can go very loud, very quickly without the slightest hint of compression. Plucked instruments take on a whole new life-like quality, and well recorded acoustic guitar is just stunning.

I have also noticed two other advantages. The offset at the output now seems to be more stable. With the original set-up, the output would fluctuate by around 20mV, but now even a digital multimeter hardly changes once everything has warmed up. The other improvement is that now, when switching the amps off, they are perfectly silent. Prior to the mod, there was always a small 'thump' from the speakers.

So, yes, I am happy with the mods. I think it is worth having a go at this, and I would be very interested to hear from anyone who has done the same thing. I am also happy to help if anyone is thinking of trying it out.

Cheers,

Steve.


Great stuff,

I did something similar on my test bed Beast OTL. I originally bought the boards and the Manual. I built one channel to compare it to the one channel of the T8 I built out of Audio Reality.

My changes to the Beast OTL are as follows

1. I used 8 tubes.
2. My output tubes are 17KV6A tubes.
3. I did not install the original bias circuit components except for the Hammond 160F120 transformer and I used only one 60v output (had to cut 2 traces). From that I built a half wave negative supply for the bottom tubes and grounded the positive side of the supply to the negative output tube high voltage. So it does as Steve's does and swings with changes in the output negative supply. I went one further and made my top tube negative supply off of a voltage divider between the +180 and -180 output tube supply and using a 220uf cap I hand the positive side of the cap off of the center speaker output rail to get my top tube negative bias. What now happens is that both my bias supplies swing with the amps output voltage supplies top and bottom. Both supplies are adjustable.
4. Heater voltage for the 17KV6a output tubes is tubes in series on a 34vdc floating supply.

I am breaking the two different amps in and trying to decide which pair of BR's OTL's I will build in the end, The T8 or the Beast. The Beast is winning so far. The T8 is using 6LF6 output tubes.
Wolfgang
Reply with quote  #9 
Unfortunately I don’t have the schematics and it’s probably not allowed to post them although the “Beasts” are no longer sold by TS.

Despite the clear improvement with the different bias connection (plus both biases adjustable)something still bothers me: How are the tubes protected during the powering up process and how does the fact that more than one tube are biased together affect the individual current through each tube and the DC offset?

In my understanding each tube warms up a little differently and one or more will hog more current until all tubes have kind of balanced each other in a more or less equal condition depending heavily on the individual tube characteristics/matching (that’s where the fluctuations of the DC offset must come from). This can mean that for some seconds some tubes can conduct a lot more current which doesn’t heat up the plate too much but still seems to be a lot of stress for the individual tube which does that every time. The more output tubes in parallel the less of a problem I guess. If tubes are not precisely matched in the amp this could lead to tube failure over time.

To delay HV at these voltages (and in general!) is not a good idea (arcing relays), so what about having two different bias settings for each rail, one where the amp operates and one higher (can come from the already existing neg bias or a second one set for the higher voltage) that basically shuts the tube down during warm up (maybe 30-40sec)? As a positive side effect the same relay could be triggered by a window comparator(long before the fuse would kick in) to protect the tubes when the DC across the anode resistors rises above a certain value. Practically the exact value would have to be defined under operating conditions with a sine signal.
Kelvin Tyler
Reply with quote  #10 
One might reasonably think that a tube which was designed for TV line-output duty should be rather reluctant to flash-over. In this application there will generally be several kV present at the anode. The octal cap-less version of the EL509 is not suitable for line-output duty, but should be a fine and very rugged tube in OTL audio applications, as indeed it is. The problem arises from the woeful quality control that seems to be the case with the 'modern' manufacturing methods. No tube of any type should leave the factory with such a wide parameter spread as is the case with this tube. Many things are important in producing close tolerance tubes; materials quality including the glass, precision of assembly, careful gettering, and adequate 'screening' with constant monitoring of ion current. Only tubes which are visibly good and have parameters close to the nominal values should be labelled and released for sale. That is how it was done in the Mullard plant Where I worked in the 1950s. It does not seem to be the case now. Flash-over is usually due to high ion currents with slightly gassy tubes and is often induced by the large currents present near clipping or hard driving.
I have built three of Bruce's pp designs to date- SOB, Pinnacle and, very recently, an eight tube version of the Beast. The biasing arrangement for the Beast has evolved in the SOB and Pinnacle, and, indeed, I have made changes in my 8-tube Beast after trying the original scheme. The DC offset wobble of a few tens of mV is due to the lower tubes bias being the difference between the unregulated cathode voltage and the zener regulated negative bias voltage. Getting current matched tubes is a good idea if not absolutely essential. Bruce has suggested that a little imbalance can be beneficial in softening the crossover point in the AB operation.
I pretest these tubes in a simple rig with variable bias to get an idea of the Ip spread and some sense of the gain differences. This makes it easy for me to make up suitable upper and lower banks. I realize some folk may not have the facility to do this very easily so just stick to Bruce's instructions.
I built my scratch SOB about a year ago now and it has not missed a beat. The DC offset has remained close to zero in that time and the emissions have barely decreased. Not much evidence of undue startup tube stress in this case. Monitoring the DC offset with an 8 ohm load during startup shows less than a volt deviation from zero for a very short time. A zero DC offset at idle does of course indicate the upper and lower banks are passing the same current. Over a shorter period my experience with the Pinnacle is following similar lines.
One is constantly hearing of excellent tube lives with these great sounding amps; often no output tube replacements for many years. Just do not drive them too hard and certainly avoid prolonged clipping, especially during the burn in period. I do not see the need for both upper and lower biases to be variable. Just preset the lower bias voltage to determine the desired idle current.

Kelvin
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.