Interesting thought, Tony. I will give it a try.
Currently, with 4 x 6SN7, the sound has way too much hardness in the tops, but it is possible that this is just due to one pair of the 6SN7 being brand new (NOS).
With this in mind, I will also revert to the "normal" configuration (2 x 6SL7 / 2 x 6SN7) using the new tubes in the 6SN7 position. If the hardness in the tops prevails, this will tell me if it the hardness problem is down to the new tubes - or the configuration.
I will try the 4 x 6SL7 experiment thereafter.
My gut tells me that the current 4 x 6SN7 configuration is not getting a fair chance presently, hampered by the (now) wrong resistor values in R2, R3, R4 and R5.
Currently, with 4 x 6SN7, the sound has way too much hardness in the tops, but it is possible that this is just due to one pair of the 6SN7 being brand new (NOS).
With this in mind, I will also revert to the "normal" configuration (2 x 6SL7 / 2 x 6SN7) using the new tubes in the 6SN7 position. If the hardness in the tops prevails, this will tell me if it the hardness problem is down to the new tubes - or the configuration.
I will try the 4 x 6SL7 experiment thereafter.
My gut tells me that the current 4 x 6SN7 configuration is not getting a fair chance presently, hampered by the (now) wrong resistor values in R2, R3, R4 and R5.
i can not see why the cathode degeneration resistors at first hole is 1.2k, even 100 ohms is big enough imo..why degrade tube transconductance of a tube not known to have high gm to begin with...?
highs are affected when not enough drive current is available form the driver tube to swing the grid of the 6550,
so increasing the current to say 10ma from 5ma will double the current available to swing the grid with its grid cap and stray caps at high frequencies...
but then thing go messy quickly as the 6sn7 will have lost plate voltage in this way, perhaps the 10k dropping resistor is changed to 1k, and voltage drop at this point is 10 volts instead of 50 volts previously...
just thinking out loud.....
Tony,
If I decide to try the current 4 x 6SN7 configuration with with zero NFB, can I simply lift one leg of the NFB resistor to break the circuit?
Or do I need to remove the resistor completely and then bypass it with a jumper, in order to maintain the circuit?
(Still learning…)
I completed a bit more testing today, with interesting results.
Again, my "normal" configuration for this amp is 2 x 6SL7GT tubes in the first hole and 2 x 6SN7GT tubes in the second hole. The output tubes are 4 x 6550.
TEST 3: I simply ran 6SL7GT tubes in all 4 preamp positions and had a good listen.
The sound was quite good with good levels of detail, but again sounded rather homogenised. Dynamics were also squashed - which surprised me. When volume increased, bass became a bit unwieldly and unrealistic. So, not terrible sound, just less good and less enjoyable than the normal configuration.
So, after three different tests, to my ears the "normal" BEWITCH configuration sounds best - by far. It is probably not surprising given that the BEWITCH circuitry remains optimised for this configuration.
Can I get better sound by reconfiguring the first tube circuitry for 6SN7? Truthfully, I do not know. Given that I am not totally convinced of the best particular values for those ten resistors that control tube hole 1, I feel it would be nothing more than a gamble for me to try it.
Given also that in the early days - documented earlier in this thread - I completed three different listening tests with three different NFB values (each reducing NFB) and found that the "factory" value still sounded best - I will hold-off on further NFB value testing.
My next step is to retube the amp completely and record live test values at each important circuit juncture - much like saltonm73 has already done with his configuration schematic. Once those values are recorded on my schematic, I will report back and ask for comments and feedback.
After all, as one wag inferred earlier in this thread, "discussing circuit design where there are no measurement given values, is utterly pointless..."
I have now learned enough to agree with him.
Again, my "normal" configuration for this amp is 2 x 6SL7GT tubes in the first hole and 2 x 6SN7GT tubes in the second hole. The output tubes are 4 x 6550.
TEST 3: I simply ran 6SL7GT tubes in all 4 preamp positions and had a good listen.
The sound was quite good with good levels of detail, but again sounded rather homogenised. Dynamics were also squashed - which surprised me. When volume increased, bass became a bit unwieldly and unrealistic. So, not terrible sound, just less good and less enjoyable than the normal configuration.
So, after three different tests, to my ears the "normal" BEWITCH configuration sounds best - by far. It is probably not surprising given that the BEWITCH circuitry remains optimised for this configuration.
Can I get better sound by reconfiguring the first tube circuitry for 6SN7? Truthfully, I do not know. Given that I am not totally convinced of the best particular values for those ten resistors that control tube hole 1, I feel it would be nothing more than a gamble for me to try it.
Given also that in the early days - documented earlier in this thread - I completed three different listening tests with three different NFB values (each reducing NFB) and found that the "factory" value still sounded best - I will hold-off on further NFB value testing.
My next step is to retube the amp completely and record live test values at each important circuit juncture - much like saltonm73 has already done with his configuration schematic. Once those values are recorded on my schematic, I will report back and ask for comments and feedback.
After all, as one wag inferred earlier in this thread, "discussing circuit design where there are no measurement given values, is utterly pointless..."
I have now learned enough to agree with him.
As an interesting aside to the above, I have had quite a time finding affordable 6550 tubes - even here in China.
I had hoped to continue with (NOS) Soviet 6550 tubes as I have found them both sonorous and long-lived. The current Russia/Ukraine conflict has scuttled that idea.
Instead, I have ordered new-make Chinese tubes - and this time, they will be the KT88 variety, which are not identical to 6550, but close enough for it not to matter. Reportedly, BEWITCH transformers have no problem dealing with the slightly higher plate demands of the KT88.
What the KT88 will sound like - compared with my present Soviet NOS 6550 - remains an unknown.
Let's see...
SHUGUANG Electro Tube KT88-98
A matched quad of KT88 - including hand-delivery to my doorstep within 72-hours - all for $US 90.00. AND including a 1-week right of return, something that I hope will not be necessary.
(China is VERY affordable, for most things...)
Evidently, Shuguang is one of the oldest tube manufacturers in the world and is said to make the lion-share of all new-make vacuum tubes, including other brands.
I shall report!
I had hoped to continue with (NOS) Soviet 6550 tubes as I have found them both sonorous and long-lived. The current Russia/Ukraine conflict has scuttled that idea.
Instead, I have ordered new-make Chinese tubes - and this time, they will be the KT88 variety, which are not identical to 6550, but close enough for it not to matter. Reportedly, BEWITCH transformers have no problem dealing with the slightly higher plate demands of the KT88.
What the KT88 will sound like - compared with my present Soviet NOS 6550 - remains an unknown.
Let's see...
SHUGUANG Electro Tube KT88-98
A matched quad of KT88 - including hand-delivery to my doorstep within 72-hours - all for $US 90.00. AND including a 1-week right of return, something that I hope will not be necessary.
(China is VERY affordable, for most things...)
Evidently, Shuguang is one of the oldest tube manufacturers in the world and is said to make the lion-share of all new-make vacuum tubes, including other brands.
I shall report!
Looks like the designer is a solid state guy. You can get away with small tail resistors since you can drive the gm of a transistor much higher than that of any VT, so the balance can still be good even with such short tails. Seems he forgot that detail here. A hollow state LTP needs a large tail resistor, and preferably an active tail load if you're to have good balance.
This isn't the first time I've seen this.
I understand that the BEWITCH 6550 circuit is based upon a 1950s design that was granted a French Patent. The designer was one PIERRE LOYEZ.
Attached is his (alleged) original schematic.
Based upon previous investigations by others, it seems there is far more going on in this circuit than meets the eye... certainly far more than I can understand.
For anyone wanting to follow the rabbit-hole, here is a previous discussion on this forum.
Understanding the complexities of the BEWITCH 6550 circuit...
Attached is his (alleged) original schematic.
Based upon previous investigations by others, it seems there is far more going on in this circuit than meets the eye... certainly far more than I can understand.
For anyone wanting to follow the rabbit-hole, here is a previous discussion on this forum.
Understanding the complexities of the BEWITCH 6550 circuit...
yes, although lifting the wire from the speaker terminal jacks seem to be easier...
replying to post #189, indeed, there are a lot of things going on in the feedback department with that circuit, OShade, then cross grid feedback and then global negative feedback, way too much complexity for my taste..
the Mullard 5-20, even the Radford 25 are not so overly complicated and yet delivered the goods as far as i am concerned...
i love the psu part of the below schematic though for its elegance and simplicity...
the Mullard 5-20, even the Radford 25 are not so overly complicated and yet delivered the goods as far as i am concerned...
i love the psu part of the below schematic though for its elegance and simplicity...
If you or other readers don't like the circuit variant of the Bewitch 6550 and the Lampizator modification, there are the mods by Andrey Tatukov.
http://checkcfg.narod.ru/audio/tubes/index.htm
I can report that I have had a matched quartet of SHUGUANG ELECTRO TUBE KT88-98 output tubes in play for just over a week.
After initially sounding quite thin and bright, they quickly settled down to deliver really impressive sound in my BEWITCH 6550.
As expected, plate voltage has increased from ~405V with the 6550, to 411V with the KT88. The power transformer (cover) runs hot-to-the-touch, but subsequent inspections of the interior of the amp indicate that everything is happy.
Following some swapping around and experimentation, I have happily returned to the "standard" 6SL7/6SN7 Phase Splitter/Gain Stages, as prescribed by the factory and continue to optimise the circuit - to my ears. I'm extremely happy with the improvements.
Whilst the stock circuit sounded okay, this amp is capable of sublime music quality, with thoughtful tweaks.
My next step is to reduce the present 375-Ohm KT88 (shared) cathode resistor, down to 250-Ohms. This will bring each pair of KT88 bias point from ~45%% of max. dis. to somewhere around 70%.
I will report.
After initially sounding quite thin and bright, they quickly settled down to deliver really impressive sound in my BEWITCH 6550.
As expected, plate voltage has increased from ~405V with the 6550, to 411V with the KT88. The power transformer (cover) runs hot-to-the-touch, but subsequent inspections of the interior of the amp indicate that everything is happy.
Following some swapping around and experimentation, I have happily returned to the "standard" 6SL7/6SN7 Phase Splitter/Gain Stages, as prescribed by the factory and continue to optimise the circuit - to my ears. I'm extremely happy with the improvements.
Whilst the stock circuit sounded okay, this amp is capable of sublime music quality, with thoughtful tweaks.
My next step is to reduce the present 375-Ohm KT88 (shared) cathode resistor, down to 250-Ohms. This will bring each pair of KT88 bias point from ~45%% of max. dis. to somewhere around 70%.
I will report.
Last edited:
with plate voltage in the 400's vdc, a 6 volt change is nothing, well, if you are not to picky on plate swing, increasing pate voltage can only mean that your tube is drawing lesser plate current...
will increase cathode currents and lower plate voltages some more...
i use separate cathode resistors for biasing to prevent "current hogging" that can happen with not so good matched tube..
My objective for reducing the value of the (shared) cathode resistor, is to simply listen to how increased bias changes the sound of the amplifier - if at all.
I am not really sure what to expect, but understand that increased bias - WITHOUT over-biasing the tube - tends to create a more linear response, helping reduce distortion.
If the increased bias - from reducing the cathode resistor by 100-Ohms - does NOT change or improve the BEWITCH sound, I will revert to the higher cathode resistor value, in an effort to promote long tube life.
I will report.
I am not really sure what to expect, but understand that increased bias - WITHOUT over-biasing the tube - tends to create a more linear response, helping reduce distortion.
If the increased bias - from reducing the cathode resistor by 100-Ohms - does NOT change or improve the BEWITCH sound, I will revert to the higher cathode resistor value, in an effort to promote long tube life.
I will report.
As an aside, I am struggling to understand how and why higher tube bias shortens tube life.
Perhaps this is an urban myth?
The tube heaters run at a given voltage (heat), no matter what the bias, so an increased tube bias CANNOT make the tube run any "hotter" can it?
[Please understand that in making these comments, I am EXCLUDING instances where a tube is actually over-biased!]
If I increase the tube bias, so that the tube is running at it a more linear operation point, I would have expected this optimal bias to extend tube life - NOT reduce it.
QUESTION: Will running an output tube at 45% of max. dis. really extend the tube's life? If so, why?
What do others think?
Perhaps this is an urban myth?
The tube heaters run at a given voltage (heat), no matter what the bias, so an increased tube bias CANNOT make the tube run any "hotter" can it?
[Please understand that in making these comments, I am EXCLUDING instances where a tube is actually over-biased!]
If I increase the tube bias, so that the tube is running at it a more linear operation point, I would have expected this optimal bias to extend tube life - NOT reduce it.
QUESTION: Will running an output tube at 45% of max. dis. really extend the tube's life? If so, why?
What do others think?
ideas will have consequences if you try them, how subtle or pronounced? you will find that out..
yes it will, but also the filament voltages needs careful consideration, try not to stray away too much....
plate biasing percentage of pate watt ratings... https://www.amplifiedparts.com/tech-articles/tube-amplifier-bias-calculator
Thanks @TonyTecson.
The site I have been using for biasing simulations is Rob Robinette - Tube Bias Calculator. It's the same sort of bias information but offers a lot more detail and "how to".
But one thing that seems to be tripping me up.
The original factory values of the shared cathode resistor bias the KT88 tubes at 51mV delivering ~42% of max. plate dissipation.
[Shuguang Electron Tube KT88-98 confirmed by factory as 42W max. dissipation...]
I calculated my new shared cathode bias resistor value in the RR simulator above and reduced it from 375R - with a voltage drop of 40V - and with 370V on the plate (measured plate to cathode) to 250R.
Once I had installed the new shared cathode resistor value of 250R - the voltage drop reduced to 35V - with an also reduced new reading of 349V on the plate (measured plate to cathode).
So, these other changing parameters have messed with the results.
With the 250R cathode resistor, the tube bias has increased to ~57mV, but I am nowhere near my 70% of max. plate dissipation target I had hoped for.
I guess I just have to keep trying lower cathode resistor values until the tube bias value reaches the right point, unless there is a fancy calculation that can save me some time and effort.
There is no mention of such a fancy calculation on any of the tube bias sites that I have been reading - so I have more trial and error ahead of me.
The site I have been using for biasing simulations is Rob Robinette - Tube Bias Calculator. It's the same sort of bias information but offers a lot more detail and "how to".
But one thing that seems to be tripping me up.
The original factory values of the shared cathode resistor bias the KT88 tubes at 51mV delivering ~42% of max. plate dissipation.
[Shuguang Electron Tube KT88-98 confirmed by factory as 42W max. dissipation...]
I calculated my new shared cathode bias resistor value in the RR simulator above and reduced it from 375R - with a voltage drop of 40V - and with 370V on the plate (measured plate to cathode) to 250R.
Once I had installed the new shared cathode resistor value of 250R - the voltage drop reduced to 35V - with an also reduced new reading of 349V on the plate (measured plate to cathode).
So, these other changing parameters have messed with the results.
With the 250R cathode resistor, the tube bias has increased to ~57mV, but I am nowhere near my 70% of max. plate dissipation target I had hoped for.
I guess I just have to keep trying lower cathode resistor values until the tube bias value reaches the right point, unless there is a fancy calculation that can save me some time and effort.
There is no mention of such a fancy calculation on any of the tube bias sites that I have been reading - so I have more trial and error ahead of me.