If we think they really don't care maybe there is nothing to care about. They don't have peak music power to worry about?
I am lost in translation here by my limited english.
Here is my question reworded again:
What is the threshold being used when Mullard 5-20 designers claim their amplifiers as 20W amp? I think it is the output power when the output tubes reach Vgk=0.
In other words, they don't design for the driver to clip before the output tubes (the driver is always designed to be able to clip the output tubes)
Here is my question reworded again:
What is the threshold being used when Mullard 5-20 designers claim their amplifiers as 20W amp? I think it is the output power when the output tubes reach Vgk=0.
In other words, they don't design for the driver to clip before the output tubes (the driver is always designed to be able to clip the output tubes)
My answer need not be amp specific: Auto bias adjustment, there is a patent for it. I think Cary Amp 300 use this method. You actual get more output this way. The bottom line is just reduce clipping by increasing power.
Nothing wrong with your English, it is much better than our Bahasa...🙂 You are correct to say that most tube amplifier's output rating are specified when the output tubes are driven to Vgk=0V or (at the onset of clipping whichever comes first). But also note that most of the popular tube amps such as the Mullard 5-20 uses "audio power tubes" with well known characteristics.
In this case, there is no data given for the 6C33C above say 600mA at Vgk=0V, so we really have no clue as to how the tube behaves at the estimated peak current of 2A. I have not found any test result done at full power to except for very short duration.
So even if there is no grid current, we are still not sure how long the amp can last under continuous use at its rated output power because that is what the FTC rating requires, and how you make an apple-to-apple comparison.
But don't let the above stop you from trying new ideas, I think you are more interested in the overall topology than getting into power race anyway...😀
You desparately need to get your facts right, and stop trying to twist facts to justify your rotten design.
There were several versions of the Mullard 5-10 published by Mullard branches around the world. My comments below apply to the version shown in Mullard Circuits for Audio Amplifiers, 2nd ed, Mullard UK 1960.
Mullard rated the amplifier as a 10W amp, specifically 10W at <0.3% distortion. 10W is the point at which SMPTE intermod reaches 1%, belived at the time to be the point at which distortion is just perceptable. It was common to rate amplifiers that way at the time.
It has 2 x EL84/6BQ5 operating with Ik 2 x 36 mA and HT 318V. The clipping point, determined by the point at which the tubes go out of their linear range in trying to pull their anodes down (+ve peaks), coincident with going into cutoff (-ve peaks), is 14 watts.
By modern standards, which rate amplifiers by their onset of clipping, it's a 14W amplifier.
The cathode bias is 9.7 V. The grid swing to get 14W is 9V Pk.
The phase splitter drives the output tubes and was impemented with high mu twin triode in see-saw (LTP) configuration, cathode voltage 64V, anode voltage 200V. The phase splitter is thus well and truely able to drive the output stage into grid current.
But before you conclude that Mullard engineers could not care about grid current and blocking, consider these important points:-
1. It was designed in the late 1950's, when the only source of recoderd music was vinyl with compression applied during mastering. Liberties that could be taken back then are not so easily justified now that music is available on CD's, which offer deeper bass and larger peaks.
2. Clipping oocurs at a grid drive of 9V pk, and grid current begins at 9.7 V pk. Grid current thus occurs at an output of 16.3 W, if the amp could do that much. Thus grid current does not occur unless the volume is advanced high enough to produce objectional distortion from clipping, so the user will back off the volume. This aspect is less valid now we have CD's.
3. The output grid coupling time constant is rather high by the standards of the day, at 84 millisecs. This did attract some criticism from profesional audio circuit engineers (eg Thiele) at the time on that basis - if the amp is driven into grid current, recovery artifacts may be audible.
The circuit was never intended to be of the highest quality. It was meant to be good quality at a low price - or if you like, compromised to keep cost down.
A genuine Mullard 5-10 per the original design is a pentode output amplifier. Later, in respnse to enquiries, Mullard published an ultra-linear version. Bias was increased to 12V (Ik 24 mA). The result is that cliiping occurs at 11W due to cut-off instead of 14W, but the distortion at 10W and lower is less than in the pentode version. The margin between clipping and grid current onset is slightly improved.
Last edited:
It seems to me that if ballpencil is wanting to make an amplifier that can achieve, let's say, 20W output power, and if it turns out that Vgk is always negative in order to achieve that goal, then it is not fair to criticise him on the grounds that if he did crank up the volume beyond that point, he would drive the tubes into the Vgk > 0 regime.
We seemingly have no very precise information available as to what is the maximum anode current that a 6C33C can deliver before Vgk goes positive (or before the grid current becomes non-negligible). In the relevant plate voltage regime it must be somewhere between about 1.5A and about 3A, and it may vary quite a bit from one 6C33C tube to another.
My own experience would seem to suggest that it is probably just about of order 2.5A in the particular examples of 6C33C tubes that I have in my OTLs. My reason for thinking that is because I typically find that when driving into an 8 ohm resistive load, the output waveform for a sinewave looks pretty clean, and measures to have a respectably low distortion, until just about 25W (i.e about 14V rms; 20V peak, and 2.5A peak current). If pushed beyond that, the peaks of the sinewave get flattened off, as if a ceiling and a floor have been hit. I never investigated exactly what was causing the cut-off. Maybe it is because 2.5A represents the maximum anode current before grid current becomes appreciable, and since the driver in my OTL is not capable of supplying significant grid current it therefore chops the peaks? I am really not familiar with what might happen when getting into that sort of regime. But before clipping, i.e. power up to but not exceeding about 25W, the distortion is pretty low.
Anyway, it seems to me that what one can say about ballpencil's design is that it is class A up to some power level X; class AB for power outputs between X and some power Y; class B for power levels between Y and some power Z, and that the maximum power Z is effectively defined by something of an untoward nature happening, either clipping in the op-amp, onset of grid current, or whatever. The power X at which it transitions out of class A is rather easily calculable. The power Y at which it "transitions" from AB to B seems to be subject to rather vague and not well defined notions of what "appreciably" means, and hence becomes (as we saw earlier in the thread) a suitable topic for heated and inconclusive debate. But as long as Z, the maximum power before something undesirable happens, is big enough to meet his aims and goals, he can surely be reasonably happy?
Chris
It is fair - and my objective is not to knock him for having a go, but to point out errors and get him to do it better.
His design may do 16 W, but will struggle to do 20W. He now seems to think he can keep the volume control down so it desn't get to even 16W.
It's a fact - it has been established that grid blocking will occur at less than 16W. It's a fact that his circuit cannot achieve full output (set by op-amp clipping, assuming the tubes hold up beyond their ratings) without Vgk positive on peaks.
If he made the few simple changes I've suggested, his amp is still not very good by qulaity audio standards, but it will be considerably better than before. As that is the case, criticism is indeed fair.
Last edited:
Indeed it is. In a long ago post I calcuated it to be 0.63W.
Whether he is happy is entirely up to him. Even if distortion is 50% at all times (I'm not suggesting it will be that bad), if he's happy, he's happy.
However, his amp is prone to grid blocking, and there is no cost or technical reason why it has to be that way. Only his inability to take on board that his design is not optimal.
As I said before, it is equally important to get clean clipping and no recovery artifacts as it is to keep steady tone distortion down.
Thanks for putting it in the best way i can describe my goals, chris.
When i started OTL 6C33C project, i do have a target output power and that is 25W since based on my research on the internet, most people claim this is what you can expect out of it.
Then i go about and design the amp and it turns out the model for my 6C33C gives 2A at Vgk=0. Hence, i am saying at the start of the thread, it is 16W amp.
If it turns out that actually 6C33C puts out only, say, 1.5A at Vgk=0 for output power of 9W.. then so be it. At least I have done my best to research OTL designs and most of them say a pair of 6C33C is good for 25W.
When i started OTL 6C33C project, i do have a target output power and that is 25W since based on my research on the internet, most people claim this is what you can expect out of it.
Then i go about and design the amp and it turns out the model for my 6C33C gives 2A at Vgk=0. Hence, i am saying at the start of the thread, it is 16W amp.
If it turns out that actually 6C33C puts out only, say, 1.5A at Vgk=0 for output power of 9W.. then so be it. At least I have done my best to research OTL designs and most of them say a pair of 6C33C is good for 25W.
Sure a pair of 6C33C can put out 25W. In fact they can do around 40W. In Class A without any grid current. Without exceeding ratings. In OTL.
But not in your circuit.
I can understand your rationale for 150V HT rails, but why are you cemented to grid capacitors and insufficient bias?
But not in your circuit.
I can understand your rationale for 150V HT rails, but why are you cemented to grid capacitors and insufficient bias?
Last edited:
Keit,
As stated a few posts earlier, i have constraints/restrictions to follow.. Main reason is economy and this breaks down to the list i mentioned a few posts earlier. For example, grid capacitors. I will be happy to remove them if i can maintain the biasing scheme, which is DC servo for the upper tube and voltage multiplier (adjustable by VR) for the lower tube. In fact, i know it will be beneficial for stability reason alone to remove them, if not for avoiding blocking distortion as you mentioned. Especially since i don't know if the amp is stable or not (yes you have mentioned it is not since the LF poles are similar, but i think the solution for that is easy: re-adjust one of the pole's frequency.. what's left is whether that xfrmr becomes the non-rectifiable instability source).
Insufficient bias, how is this insufficient? Can you please explain? 200mA at Vak 150V equals half of the maximum dissipation allowed. Considering the original plan is that these tubes will output peaks of 2.5A, i figured this is the best compromise between low crossover distortion and tube reliability.
As stated a few posts earlier, i have constraints/restrictions to follow.. Main reason is economy and this breaks down to the list i mentioned a few posts earlier. For example, grid capacitors. I will be happy to remove them if i can maintain the biasing scheme, which is DC servo for the upper tube and voltage multiplier (adjustable by VR) for the lower tube. In fact, i know it will be beneficial for stability reason alone to remove them, if not for avoiding blocking distortion as you mentioned. Especially since i don't know if the amp is stable or not (yes you have mentioned it is not since the LF poles are similar, but i think the solution for that is easy: re-adjust one of the pole's frequency.. what's left is whether that xfrmr becomes the non-rectifiable instability source).
Insufficient bias, how is this insufficient? Can you please explain? 200mA at Vak 150V equals half of the maximum dissipation allowed. Considering the original plan is that these tubes will output peaks of 2.5A, i figured this is the best compromise between low crossover distortion and tube reliability.
40 watts (OTL) in class A into 8 ohm load indicates Iq=~1600 mA. That’s quite a lot..
@jane, hint: he didn't mention the load impedance 🙂
i think it's the 600R speakers he mentioned earlier.
i think it's the 600R speakers he mentioned earlier.
"Insufficient bias, how is this insufficient? Can you please explain? 200mA at Vak 150V equals half of the maximum dissipation allowed."
I don't understand why you have to insist 30W idling instead of 10W or less. Grid bias varies from -45V to -70V for 200mA idling, you may be get only 3 pairs out 20 tubes of your specs of -47V 200mA. So you get a pair when exactly the specs you need and this other way out say -70V 200mA, their gain and power would be different, + the waste...
I don't understand why you have to insist 30W idling instead of 10W or less. Grid bias varies from -45V to -70V for 200mA idling, you may be get only 3 pairs out 20 tubes of your specs of -47V 200mA. So you get a pair when exactly the specs you need and this other way out say -70V 200mA, their gain and power would be different, + the waste...
So the question perhaps becomes: why are you wedded to an op-amp servo?
You could just have simple resistor divider between 0V and the -150V rail.
Of course if you got rid of that input triode with feedback to its anode, the bias shift problem would go away. As I and other have pointed out, feedback on the anode of a triode makes some sense if the following gain block is a high quality gain block (to get benign triode sound) or another triode (to cancel distortion), but this is not the case with your design. You are just adding some distortion to more distortion.
You could of course power the servo system from a separate small transformer, and/or include an emitter or source follower. That would allow you to have a low impedance servo output, and you can then connect the transformer secondary return to it instead of the cathode.
I wouldn't bother though - I'd just go for simple resistor divider without a servo.
Yes, readjusting the poles to separate them will work, as will removing a pole - the tube grid capacitors.
As to whether the transformer is a stability issue at LF depends on the drive capability of the op-amp. I'm too lazy to look up the op-amp specs but I don't think the transformer is a genuine LF stability issue - and if it is, that's easily fixed too.
Leakage reactance at HF may, in conjunction with tube internal capacities and roll-off due to the op-amp internal compensation cause HF instability.
However, a) I expect the op-amp HF pol to be dominant, and b) if it isn't, the probalm is easily fixed with a CR network.
You need to get the bias up to near cutoff. About -70V.
As Koonw pointed out sometime ago, doing this reduces crossover distortion. There is an optimum bias level with Class B. Too low a tube current and you get notch-type crossover distortion. Too high a tube current and you get gain-doubling type cross-over distortion.
That's one of several reasons why it is important to understand your amp is Class B amp and thus you need to do things that Class B needs...
It gets even better....
If you crank the bias up to near cutoff, you reduce tube dissipation. And on signal peaks, you get the same power output with a lower anode peak current
It gets better still...
If you crank the bias up to near cutoff (ie -70V), obviously you need a MUCH bigger grid signal drive to get the grids positive on peaks.
That's why Koonw's version has only small amounts of grid current at large signal levels.
Last edited:
No it doesn't. It indicates a higher loudspeaker impedance. That's what is normally done with OTL. Philip made 400 ohm and 800 ohm speakers for this purpose. And you can buy 32 ohm drivers and series them up as required.
Did I not say my OTL will be for 400 ohm speakers?
100 V p-p = 70V RMS, 70*70/(2*400) about ~6W rms, hum we're out to tea talk on this one. 32 Ohms speaker is ok, depend on OTL topology, if higher, not much to be gain due to limited higher drive level.
Last edited:
Did I say the output was only 70 V RMS? No, I didn't.
The design speaker load is 400 ohm - I did mention that before. Hence the required output is 126V RMS. Taking into account tube practicalities, the required HT is in the vicinity of 450 volts.
And in case some dippy thinks I'm using 6C33C, no I'm not. This is a paid job with a warrantee, so using some dodgy uncommon communist tube would be quite unethical. I'm using common television line output pentodes, 2 in parallel + 2 in parallel.
However, if you want to use 2 x 6C33C in OTL, 450 V total HT is allowable. But the higher allowed cathode current enables 40W into 150 ohms (77V RMS) without exceeding ratings or going outsidee the published curves. 150 ohm speakers are very doable with series strings and matching the speaker cone resonances.
Last edited:
Some OTL authors reported in professional peer reviewed journal achieving 0.02% percent distortion with certain tube OTL topologies incorporating feedforward error correction. That's measured, in a professional lab, not some dodgy SPICE sim. With beautiful clean clipping.
Few if any transformer output tube amps could be that good. So there is something to gain.
- Status
- Not open for further replies.