I would certainly give it time to break-in! Also the circuit can be quite revealing; a cheap coupling cap can easily cause issues- one example is the brown Zicon polypropylene. I don't know why; I've heard them do this consistently but not actually measured them to see why.
Your B+ seems a bit low. If your filament voltage is low as well, that can really shut down the transconductance in the output stage!
BTW 24 hours was probably insufficient time to precondition the 6080s; don't be surprised if you find that they arc over quite easily (and I would test them right now to see if they are alright; I've seen up to 60% of them fail within 5 minutes of service when not given enough preconditioning time)!
I have been drooling over a tube OTL recently, so I guess I might just go ahead and try building one. I like the circlotron output stage for its symmetry, and I have a bunch of 6C33C tubes coming to me. So I need a suitable input + driver stage. I could just follow the M-60 circuit in post 25, which certainly is a good starting point.
However, I like to cook my own stuff... Before getting into the details of the circuit, I need to wrap my head around the amount of gain I need from the input+driver stage. Looking at the Sofia curves of the 6C33C, I don't need more than +/-60 V (peak) to drive the output stage. With my DAC ouput of +/-3.2 V (peak), I'd need a gain of about 20. However, adding the global negative feedback (GNF) into the equation, I need a lot more gain.
The question is: how much GNF do I need with a circlotron ouput, if I have 4 x 6C33C tubes per channel? My simple calculation is this:
- Output impedance without global feedback: Zout = Rp / (2+mu), where Rp is the plate resistance of the paralleled 6C33C (Rp = 130 Ohm / 2 = 65 Ohm) and mu = 2.4 (from eyeballing the Sofia curves). Therefore: Zout is approximately 15 Ohm without any GNF.
- If I am aiming for a damping factor of 10 with an "8 Ohm" speaker, I'd have to aim for 0.8 Ohm output impedance, which would require GNF = 20*log10(15/0.8) = 26 dB.
With these numbers, I'd need an additional 26 dB of gain from the input+driver stage: gain = 20*15/0.8 = 375.
Do these numbers seem about right or is this complete nonsense?
However, I like to cook my own stuff... Before getting into the details of the circuit, I need to wrap my head around the amount of gain I need from the input+driver stage. Looking at the Sofia curves of the 6C33C, I don't need more than +/-60 V (peak) to drive the output stage. With my DAC ouput of +/-3.2 V (peak), I'd need a gain of about 20. However, adding the global negative feedback (GNF) into the equation, I need a lot more gain.
The question is: how much GNF do I need with a circlotron ouput, if I have 4 x 6C33C tubes per channel? My simple calculation is this:
- Output impedance without global feedback: Zout = Rp / (2+mu), where Rp is the plate resistance of the paralleled 6C33C (Rp = 130 Ohm / 2 = 65 Ohm) and mu = 2.4 (from eyeballing the Sofia curves). Therefore: Zout is approximately 15 Ohm without any GNF.
- If I am aiming for a damping factor of 10 with an "8 Ohm" speaker, I'd have to aim for 0.8 Ohm output impedance, which would require GNF = 20*log10(15/0.8) = 26 dB.
With these numbers, I'd need an additional 26 dB of gain from the input+driver stage: gain = 20*15/0.8 = 375.
Do these numbers seem about right or is this complete nonsense?
I run my 4x6C33 with a M60 input without any feedback at all and I have fullrange speakers connected and the bass really good. Adding feedback to this circuit does not, in my ears, make the amplifier better on the contrary.
I say that in comparison with amplifiers like a Goldmund Telos 350 Hypex N-core Pass constructions and so on. I fear you will loose some of the sonic characteristic of the OTL by using so much feedback.
You will probably also run into stability problems.
I know of successful OTL builds with 6C33C output tubes that utilise John Broskie's 6SN7 Aikido Cascade for the input/driver stages.
Aikido Cascade & 300B Power Boosted
It is better to make an amplifier with a 45dB open loop gain and then with a global NFB of 20 dB ( no more than 20dB ) you will have an amp with normal gain and a fair low output impedance ! .
No global negative feedback at all? The M-60 schematic posted earlier in this thread does show a bit of NFB (100k + 2M resistors). I would think that the high gain of the M-60 input stage would make this amp a bit unpractical, because you'd have to turn the volume control to a very low setting for normal listening, no?
What is the impedance of your speakers?
Interesting. However, the Aikido board you linked to is a SE unit. A circlotron needs a driver with balanced outputs. Also, I prefer balanced inputs.
Aha, ok. Why is 20 dB NFB better than the 25 dB I guesstimated in my previous post?
Yes, of course, something of a brain fart - sorry.
I'm actually working on a SE-OTL project using the Aikido cascade to drive a single 13E1 output tube. I'll only get about 1W or so but that's fine for my high sensitivity 15 ohm drive units in back-loaded horns.
I said 20dB NFB because with higher amount you may have oscillations in low frequencies , known as Howling , due to coupling capacitors and poles in the signal path of the amp .
Last edited:
No high gain at all , only one amplifying stage with the 6SN7 tube , which mean low gain , I suggest to use 6SL7 or 12ΑΧ7 to have more gain for the use of more NFB ! .
Ah, ok, I see what you mean. The trick is to spread the poles of the C-R links between the input / driver / output stages in order to avoid instability problems. Been there, done that.
At the moment I am just thinking about voltage gain required from the input+driver stage on a generic level. Implementing these requirements is for later.
The M-60 has four triode units in the input stage (=two 6SN7 tubes). They are arranged as a long-tailed pair (LTP) with a cascode in each leg. Each cascode consists of two triode units. The 6SN7 has mu = 20, so the cascode gain = 20 * 20 = 400 (approximately). Without NFB, this would mean that a 0.1 Vrms input signal would drive the amp into clipping (+/- 60 Vpk at the 6C33C grids). This does not seem right.
Since you are saying you have only one "6SN7 amplifying stage", I'd tend to believe that you actually do not use the M-60 circuit.
Can you show your circuit?
I have been there too ( many times ) , you have to do that with a smart way , but the big problem is when you have NFB higher than 20dB , you have to be very careful and very patient to achieve a good bass and low distortions ! .
As for the cascode you making a mistake , the gain of this stage is aproximately the same as a common cathode stage , but with lower Miller capacitances and of course better high frequency response ! .
And yes in my OTL I have completely different design , two gain stages and a driver to drive the output tubes ,12 PL36 in each channel ! .
Last edited:
The M-60 has about 1 db of negative feedback. My speakers is nominally 8 Ohms.
I don't think so. Look here: The Valve Wizard
Or here (Morgan Jones book): Valve Amplifiers, Jones, Morgan, eBook - Amazon.com
For whatever reason, your calculations are wrong , as the data for the M-60 says (and my own measurements of my amp):
http://www.atma-sphere.com/Products/M-60
This is a quote from one of your links:
The bias point of the lower triode must lie on the vertical red line. For centre bias, it will be half-way up of course. Wherever you decide to put it, read off the gm (transconductance). In this case it is about 1.4mA/V. Because the lower triode generates all the current swing, its gm decides the gm of the whole cascode, and this allows us to immediately predict the gain, which is the same as for a pentode:
A = gmRa
A = 0.0014 x 47000
= 66. That's about the same as a single ECC83 triode, but we're getting it from an ECC82, and it should sound a bit like a pentode! In practice the gain may be a little lower, but not much.
The mu of an ECC82 is 17 so in your calculation the gain of the cascode would be 17x17 = 289 but it actually is 66 per this quote (without load)
Whoops! I looked at the wrong equation when I calculated the gain of the cascode input stage... This is why I ask people if my numbers make sense. Thanks!
New attempt at dissecting the voltage gains in the M-60:
Input stage (6SN7 cascode):
gm = 2.7 mA/V
load resistor value Ra = 100 kOhm:
gain = gm x Ra = 270 (not 400 as in my previous post)
Driver stage (6SN7 cathode followers):
mu = 20
gain = mu / (mu+1) = 0.95
Output stage (6C33C "cathode followers" in the circlotron):
mu = 2.6
gain = mu / (mu+1) = 0.7
Total gain without feedback:
270 x 0.95 x 0.7 = 180 = +45 dB
According to the AtmaSphere website, the M-60 (with 6SA7 tubes) has a total gain of +20 dB. Ignoring the mu difference between the stock 6SA7 and your 6C33C tubes, this means there must be about -25 dB feedback in the M-60.
New attempt at dissecting the voltage gains in the M-60:
Input stage (6SN7 cascode):
gm = 2.7 mA/V
load resistor value Ra = 100 kOhm:
gain = gm x Ra = 270 (not 400 as in my previous post)
Driver stage (6SN7 cathode followers):
mu = 20
gain = mu / (mu+1) = 0.95
Output stage (6C33C "cathode followers" in the circlotron):
mu = 2.6
gain = mu / (mu+1) = 0.7
Total gain without feedback:
270 x 0.95 x 0.7 = 180 = +45 dB
According to the AtmaSphere website, the M-60 (with 6SA7 tubes) has a total gain of +20 dB. Ignoring the mu difference between the stock 6SA7 and your 6C33C tubes, this means there must be about -25 dB feedback in the M-60.
Last edited: