I got these ARC Classic 120's couple weeks ago and they had old tired Chinese KT88's in them so I ordered new Sovtek 6550 tubes and 6CG7 driver tubes and installed them and proceeded to set the bias. with the bias control at it's minimum setting the bias was at 68mv (supposed to be 65mv for 6550 tubes) and slowly creeping up so I shut it down and measured the resistor at R97 (next to the bias pot. It's the resistor you change if you want to run KT90 or 91 tubes). Turns out that my two amps had different resistors in that position, one was 100K ohms and the other was 77K ohms. Schematic says it should be 127K ohm resistor for 6550's so I changed them to 130K (couldn't find 127K). I hooked up my multimeter and fired it up expecting it to slowly climb to the 65mv range and settle out so I can set the bias. What actually happens is it waits about 30 seconds and then it rushes up to 120mv and I shut it down because I'm expecting it to blow. Is this normal? My previous amps had test points in front that were easily accessed. These have to be accessed from the bottom of the amp and the bias pot is on top. So I set it on it's side and hook up my long reach alligator clips before turning on the amps. So I have never actually watched what happens with the multimeter as I turn on the amps. Just wondering if it's normal for cold tubes to run way past the bias set point.
What is the resistance of each 6550 g1 to ground?
For fixed adjustable bias (and for fixed bias) it is supposed to be no more than 50k, g1 to ground.
Any thing higher than 50k may cause problems, especially with many 6650 tubes.
The KT88 tubes are Much stronger in that specification than a 6550 is.
For fixed adjustable bias (and for fixed bias) it is supposed to be no more than:
220k (for plate plus screen dissipation less than, or equal to 35 Watts).
100k (for plate plus screen dissipation more than 35 Watts).
You can see that no 6550 is able to duplicate the KT88 task.
For most amplifiers, the designer has to re-design the driver to drive 50k required by the 6550; instead of the easier to drive 100k or 220k required by the KT88.
Check out the DCR from g1 to ground on your ARC Classic 120.
Just my opinion (I have had an original Tungsol 6550 go into thermal run-away, and red plate, because the g1 resistance to ground was too high).
Happy Listening
For fixed adjustable bias (and for fixed bias) it is supposed to be no more than 50k, g1 to ground.
Any thing higher than 50k may cause problems, especially with many 6650 tubes.
The KT88 tubes are Much stronger in that specification than a 6550 is.
For fixed adjustable bias (and for fixed bias) it is supposed to be no more than:
220k (for plate plus screen dissipation less than, or equal to 35 Watts).
100k (for plate plus screen dissipation more than 35 Watts).
You can see that no 6550 is able to duplicate the KT88 task.
For most amplifiers, the designer has to re-design the driver to drive 50k required by the 6550; instead of the easier to drive 100k or 220k required by the KT88.
Check out the DCR from g1 to ground on your ARC Classic 120.
Just my opinion (I have had an original Tungsol 6550 go into thermal run-away, and red plate, because the g1 resistance to ground was too high).
Happy Listening
Last edited:
CL120_BiasAdjust.pdf - Audio Research
Hello, I noticed that R97 is different as stated in service manual and schematic, maybe this is why you don't get 65mV across R68 (0.25 Ohms TP1 and TP2), 65mV/0.25=260mA (/4 = 65mA). If all the tubes are tested ok or matched ok, but maybe not safe to plug all 8 tubes but one tube each side since a bad tube can give the fault you just experienced?
PS What is the bias voltage you get for -55V supply (when not output tubes are plugin ?)
Hello, I noticed that R97 is different as stated in service manual and schematic, maybe this is why you don't get 65mV across R68 (0.25 Ohms TP1 and TP2), 65mV/0.25=260mA (/4 = 65mA). If all the tubes are tested ok or matched ok, but maybe not safe to plug all 8 tubes but one tube each side since a bad tube can give the fault you just experienced?
PS What is the bias voltage you get for -55V supply (when not output tubes are plugin ?)
Attachments
Last edited:
Resistance from g1 to ground is 49.8k ohms on all tubes. People have suggested that I should have gone the other way with the resistor and used 77k which doesn’t make sense to me. I was at minimum on the adjustment and reading 68mv so I thought I would need a larger value resistor to bring that adjustment in range.
Could you post a schematic I can only find the CL60. Is it the one which adjusts the bias on one tube to match the other?
Attachments
OK it does have that strange circuit with one side balanced to the other. Mr Summer is correct really the max grid leak resistor is 47k per tube so 74k is grossly incorrect for four - so really you should use KT88's in this application. Shame cause I like 6550's. However that will not be the problem at switch on. Unplug all output tude and measure the grid voltage on each in turn. If you want to disable that bias circuit and just use matched quads remove U7 and replace R114 with 15k. I would do this to test. Check C11-C14 etc for leakage. Also not sure why so many output tubes are needed 4x6550 can generate 120W rms in UL. Triode mode I guess needs twice as many thats a lot of tubes and heater current! I think if you wanted to use 6550 I would go upgrade for servo bias boards as this means you can use much bigger grid leaks and retain the drive. Oh this is more wered its measuring screen current+plate too. Op-amp at HT nice. I would also check the screen resistors too.
Last edited:
The Classic 120 was originally built to use 6550’s. KT88’s were a later option Then a slight mod allowed the use of KT90 and 91’s.
Then I can only think they did not read the 6550 data sheet. To support fixed bias without danger of thermal runway would need R45 = 50k/4 which would be impossible to drive. The minimum I would use is 100K/4 = 25K which is 3x smaller then the 74k. Anyway to help I would check C12 and C13 for leakage. I would also disconnect the master/slave bias as shown above to make testing easier. Using the secondary for the cathode FB OK unless the ground of the 4R tap comes off.
Last edited:
Look at the bottom set of four triode wired 6550 tubes.
Each grid has individual 1k grid stoppers.
The -55V bias is regulated (a stiff DC resistance, and a stiff AC impedance).
The -55V bias is then divided by 1.96k connecting to 30k to ground.
30k / 31.96k = 0.939 divider
0.939 x -55V = -51.6V.
But the schematic shows -46V there.
That means the four 6550 grids must be drawing DC grid current, even at quiescent conditions.
Which ‘fact’ is wrong?
Which is right, -55V, -46V, 1.96k and 30k divider?
Or, the calculation result of -51.6V?
Something does not agree.
And, even with either -46V, or with -51.6V, that bias has to go through a single 71.4k resistor to the four 1k individual grid stoppers.
71.4k x 4 = 285.6k (the resistor is shared by four 6550 tubes).
When you parallel tubes, the g1 resistor is less effective, than the same 71.4k would be for a single tube.
That is a far cry from the requirements of:
One 6550 50k g1
Two 6550 25k g1
Four 6550 12.5k g1
Talk about a violation of the tube specifications.
And in this case, "one bad apple (bad 6550) Will spoil the whole darn bunch".
Note 1: the -55V bias supply has a voltage adjustment pot.
Note 2: the upper four 6550 tubes has a DC balance adjustment pot (and Vactec LED / CDS Photocell) to be able to set the Push and Pull DC quiescent currents to be equal (the output transformer requires equal DC currents to minimize early saturation of the laminations).
I am not sure I would even try using KT88 tubes, given the very large g1 grid resistance, even if it is to a stiff voltage point.
But I would bet on good KT88s versus good 6550s any day in that circuit.
Your mileage may vary (and it seems to be running away, doesn’t it?).
Some think all engineering results are professional and good.
Remember the original Tacoma Narrows Bridge?
KT88 fixed bias:
One KT88 100k
Two KT88 50k
Four KT88 25k.
Conclusion:
Can you live with less output power, I bet you can. Unless this is Woodstock III, or you have those MBL Radiostraller speakers, or you are hard of hearing.
If it were me, I would go for less power output:
Strip out all of the -55V bias supply circuitry.
Strip out all of the VacTek circuitry.
Strip out all of the rest of the fixed bias circuitry.
Then . . .
You have enough room to install 8 self bias resistors, and 8 bypass capacitors.
Individual Self bias will match your Individual tube currents much much better than what you are doing now.
And . . .
Then you can use your favorite 6550s, instead of the KT88s that you do not like.
Let us know what you come up with.
Each grid has individual 1k grid stoppers.
The -55V bias is regulated (a stiff DC resistance, and a stiff AC impedance).
The -55V bias is then divided by 1.96k connecting to 30k to ground.
30k / 31.96k = 0.939 divider
0.939 x -55V = -51.6V.
But the schematic shows -46V there.
That means the four 6550 grids must be drawing DC grid current, even at quiescent conditions.
Which ‘fact’ is wrong?
Which is right, -55V, -46V, 1.96k and 30k divider?
Or, the calculation result of -51.6V?
Something does not agree.
And, even with either -46V, or with -51.6V, that bias has to go through a single 71.4k resistor to the four 1k individual grid stoppers.
71.4k x 4 = 285.6k (the resistor is shared by four 6550 tubes).
When you parallel tubes, the g1 resistor is less effective, than the same 71.4k would be for a single tube.
That is a far cry from the requirements of:
One 6550 50k g1
Two 6550 25k g1
Four 6550 12.5k g1
Talk about a violation of the tube specifications.
And in this case, "one bad apple (bad 6550) Will spoil the whole darn bunch".
Note 1: the -55V bias supply has a voltage adjustment pot.
Note 2: the upper four 6550 tubes has a DC balance adjustment pot (and Vactec LED / CDS Photocell) to be able to set the Push and Pull DC quiescent currents to be equal (the output transformer requires equal DC currents to minimize early saturation of the laminations).
I am not sure I would even try using KT88 tubes, given the very large g1 grid resistance, even if it is to a stiff voltage point.
But I would bet on good KT88s versus good 6550s any day in that circuit.
Your mileage may vary (and it seems to be running away, doesn’t it?).
Some think all engineering results are professional and good.
Remember the original Tacoma Narrows Bridge?
KT88 fixed bias:
One KT88 100k
Two KT88 50k
Four KT88 25k.
Conclusion:
Can you live with less output power, I bet you can. Unless this is Woodstock III, or you have those MBL Radiostraller speakers, or you are hard of hearing.
If it were me, I would go for less power output:
Strip out all of the -55V bias supply circuitry.
Strip out all of the VacTek circuitry.
Strip out all of the rest of the fixed bias circuitry.
Then . . .
You have enough room to install 8 self bias resistors, and 8 bypass capacitors.
Individual Self bias will match your Individual tube currents much much better than what you are doing now.
And . . .
Then you can use your favorite 6550s, instead of the KT88s that you do not like.
Let us know what you come up with.
Last edited:
Its odd that its so grossly wrong - maybe it started life as the CL30 and somebody just kept adding tubes.
Thanks guys! Turns out it was a bad tube. Caught a glimpse of it glowing a strange pinkish blue as I was trying to figure this out. Finding a bad tube in a jungle of eight while watching the multimeter and keeping a finger on the power switch so it doesn't blow is difficult. Swapped that tube out and it biased up fine. Now I'm short a tube until I get a replacement.
Audio Research CL30, CL60, CL120 are pretty much the same basic triode design.
Audio Research CL30, CL60, CL120 are pretty much the same basic triode design.
The KT88 is also 50k max. (It is a copy-cat tube in a different bottle.)
It is not necessarily bad engineering. The 50k number seems to cover "all" working conditions, even on-the-edge hot. Amps which are biased well below their dissipation rations, and using high-quality tubes, are often "OK" with larger grid resistors. A half million Fenders use 220k for two "100k max" 6L6 per side. And often idled at 50%-65% of Pdiss(max), so even if they drift a bit they don't melt. (And also, amp makers rarely warranty the tubes.)
KT88 are meant to have additional radiator fins over 6550. However I am looking at EH KT88 and EH 6550. They look identical. Maybe nowadays they are one and the same thing would make sense from a production point of view. You only likely to get significant grid current if you run the plates very hot as PRR says. You can always measure the grid current across the grid stoppers or the 74K. Anyway the amp will work with 3 each side. Glad you found it. A lot of amps do use 100k each side per tube. I think 74k for four tubes is pushing it though.
Last edited:
The attached is sim of grid bias supply regulation, it can't drawn too much grid current when the bias voltage is increasingly negative when more grid current is drawn. OP amp is arcing, it maybe not be protected this way, IMHO. Some one interested can persuade more about it. Similar regulation with HT supply but I haven't sim it yet.
Attachments
Last edited:
PRR,
Agreed.
Tubes run more reliably when they are run at lower than maximum dissipation.
The 6550 and KT88 filaments dissipate 10.08 Watts.
That is as much as a type 45 Triode plate can take.
Then the 6550 and KT88 plate and screen start conducting.
All the KT88 data sheets I have seen, for fixed bias specify max g1 return resistance of 100k, from 35 Watts and up to the limit of Plate + Screen Dissipation.
Most of the data sheets I have seen, for fixed bias, and less than 35 Watts plate + screen dissipation, say 220k maximum.
All the 6550 data sheets I have seen, for fixed bias specify max g1 return resistance of 50k, that is for 35 Watts or more dissipation.
What data sheets are you looking at that lists 50k max for the KT88?
Of course, there probably are some modern production 6550 and some modern production KT88 tubes that may not last long at maximum ratings.
And having 4 tubes in parallel at the same fixed bias voltage does not always work out very good.
The amplifier in this thread was fixed when the one bad tube was discovered.
The US navy destroyer I was on could do 28 Knots (4 boilers, 2 steam turbines, 2 shafts, and 2 propellers; but not all day long.
Forget about the bearings freezing up, the unbalanced props would tear the ship apart first.
Agreed.
Tubes run more reliably when they are run at lower than maximum dissipation.
The 6550 and KT88 filaments dissipate 10.08 Watts.
That is as much as a type 45 Triode plate can take.
Then the 6550 and KT88 plate and screen start conducting.
All the KT88 data sheets I have seen, for fixed bias specify max g1 return resistance of 100k, from 35 Watts and up to the limit of Plate + Screen Dissipation.
Most of the data sheets I have seen, for fixed bias, and less than 35 Watts plate + screen dissipation, say 220k maximum.
All the 6550 data sheets I have seen, for fixed bias specify max g1 return resistance of 50k, that is for 35 Watts or more dissipation.
What data sheets are you looking at that lists 50k max for the KT88?
Of course, there probably are some modern production 6550 and some modern production KT88 tubes that may not last long at maximum ratings.
And having 4 tubes in parallel at the same fixed bias voltage does not always work out very good.
The amplifier in this thread was fixed when the one bad tube was discovered.
The US navy destroyer I was on could do 28 Knots (4 boilers, 2 steam turbines, 2 shafts, and 2 propellers; but not all day long.
Forget about the bearings freezing up, the unbalanced props would tear the ship apart first.
Last edited:
SED Tubes - Vacuum Tubes Manufactured in St Petersburg, Russia
https://drtube.com/datasheets/6550c-sed2001.pdf
Not all 6550 have same 50k for fixed bias:
Svetlana 2001 6550c has this in the datasheet:
Control grid resistance, fixed bias 200k ohms
Also in Ge datasheet, the grid1 current is 1mA @-40V, other don't mentioned at all, who has the grid current curve for 6550 please send it.
https://drtube.com/datasheets/6550c-sed2001.pdf
Not all 6550 have same 50k for fixed bias:
Svetlana 2001 6550c has this in the datasheet:
Control grid resistance, fixed bias 200k ohms
Also in Ge datasheet, the grid1 current is 1mA @-40V, other don't mentioned at all, who has the grid current curve for 6550 please send it.