Forgot about this one as well. The LTP choice has been defined somewhat. The best choice apart from linearity, is where the peak-to-peak signal to be delivered (in this case to the KT88s) fits the load line the best. To make that clear will need repetition of basic load line theory. That is dealt with quite well in the previously mentioned Steve Bench paper, but at the risk of boring the knowledgable I can run through it again. That could best be done with the aid of a Ia-Va graph family, which I will prepare shortly.
Now I'm confused. 🙄
I thought R1 has been 470R for a while now? (I can't recall why or when that changed.) Should I change it back to 220R? And to confirm, the EF86 supply voltage is correct at 288V, yes?
rev 0.7c (with R1 = 220R) attached.
..Todd
The latest R1 should be 470R. It yields the needed 5V drop between the KT88 B+ and the LTP B+, and provides a slightly smoother voltage ripple for the LTP supply. The 220R only gave about 3V drop with the currents specified; but it would also work fine. The 470R was a slight refinement.
When did the first stage B+ change to 288V? I thought it was 240V...
hey-Hey!!!,
Jus' to make a bit ov a rant...the earlier staves( v. the powr stage ) ought to have as much voltage as the tubes can stand. Given that we're looking for maximum headroom here, too much resistance in the decoupling stages seems rather foolish. If the filtering is good, don't drop the voltage.
cheers,
Douglas
Hmm... I don't think anyone was interested in dropping the voltage, we just need clarification of what the actual values are that Johan was working with. If they turn out to be too low for your comfort (whatever they are), THEN please let us know.
..Todd
Last edited:
You may have to adjust the 470R as now there is over twice the current in the LTP compared to the old t7 design. Otherwise you will drop more voltage than you want.
That's right. 330R gets closer to 5V drop than 470R. Not sure how picky that is, but from Bandersnatch's message, it sounds like less drop is probably better. So 220 may be fine.
..Todd
..Todd
Question: The cathode resistors on the KT88's. They are currently a parallel pair of 22R resistors, and TubeMack mentioned swapping them for a single 10R. Is this a good idea?
I've seen parallel-pairs used usually to cut inductance in half, individual power handling in half (ie. use more smaller, cheaper parts), and to simplify value matching with ones on the other half. Are any of these considerations important enough here to worry about?
..Todd
I've seen parallel-pairs used usually to cut inductance in half, individual power handling in half (ie. use more smaller, cheaper parts), and to simplify value matching with ones on the other half. Are any of these considerations important enough here to worry about?
..Todd
Last edited:
Final value will be determined by the input tube. I don't have the sim anymore for this circuit, but should run it through to obtain the minimum value needed to decouple and address ripple.
A 220R resistor will have a cuttoff @ 3.3Hz and 31dB attenuation. 3.8V drop
A 330R resistor will have a cuttoff @ 2.2Hz and 35dB attenuation. 5.7V drop
A 220R resistor will have a cuttoff @ 3.3Hz and 31dB attenuation. 3.8V drop
A 330R resistor will have a cuttoff @ 2.2Hz and 35dB attenuation. 5.7V drop
Hmm.. Seems you're calculating on 17mA? I used 14mA from the CCS. Is that wrong? The results are relatively close enough for our discussion, but I am curious about how to properly calculate that, in case I am not doing it right.
[Edit] Oh, I see! Adding in 2.5 mA for the EF86 too.
..Todd
Last edited:
BTW guy's. I've asked Taj to make these last changes, and lock down the design at Ver 7B.
The amp design is Done! And it only took 3 months. 🙂
At least for me, i've decided to stay with the EF86 for the intial build, though johan has listed some options if anyones game. I don't think it matters much as the driver can always be changed to something else with minor circuit changes in the future.
A good reason to stay with a 9 pin tube for the driver, is that Sgregory has mentioned to me the prospect that one could start with the easier amp, and later with just a few changes switch over to his design. Thats sounds like a nice option. As you know these amps are very similar. I don't how the PS would translate.
A big thank you to all that have contributed. I think we all learned a little!
The amp design is Done! And it only took 3 months. 🙂
At least for me, i've decided to stay with the EF86 for the intial build, though johan has listed some options if anyones game. I don't think it matters much as the driver can always be changed to something else with minor circuit changes in the future.
A good reason to stay with a 9 pin tube for the driver, is that Sgregory has mentioned to me the prospect that one could start with the easier amp, and later with just a few changes switch over to his design. Thats sounds like a nice option. As you know these amps are very similar. I don't how the PS would translate.
A big thank you to all that have contributed. I think we all learned a little!
Last edited:
Calculating the bias current (by measuring the voltage drop across these R's) would be easier with 10R instead of 11R (parallel 22R), i.e. 75 milliamps=.75V instead of .825V with the 11 ohm (parallel 22R) R's.
Some folks like to make these resistors small to act as fuses if the current runs away, although as mentioned in a post way back, CC R's can catch fire when acting as a fuse.
Taj: You may also want to show bias test points on these R's. The test points would be on the cathodes.
Last edited:
I see, that makes good sense.
Good idea. I would like to show lots of things, especially component ratings so that I can create a fairly comprehensive BoM.
Please let me know as many component ratings as you can. I don't have enough experience with tubes to know how much tolerance to work with. I'm talking mostly about cap dielectric ratings. and resistor ratings/choices (other than 1/4w metal film).
..todd
And nearly 700 messages. 🙄
Seriously, once we get the power supply questions answered to everyone's satisfaction, I will create a version 1.0 schematic and work on a BoM.
Did you decide on a name yet? Pretoria88? BlowMonkey? SpiderFish? How about the "TubeMack88"? 😎
..Todd
Last edited:
OK
Sorry for my possibly mixing up more than one circuit diagram in referring to components. To set that straight, the diagram should be the one given in post #610 (and shortly afterwards post #612). The designations and electrode voltages there represent my suggestions. Only, recently, C6 was increased to 32µF, which was decreased in the immediately previous post to 22µF. If my subsequent posts differ, ignore and put it down to .... well whatever.
(No pressing reason for my 32µF; it is more of an available value here than 22µF. A higher value will guarantee a little more against l.f. instability and is physically quite small as said previously, but we do not know that l.f. instability will be a factor at all. Thus 22µF if that is more available over there.)
Regarding several later remarks:
The feeding h.t. to the EF86 as high as possible: The maximum signal at EF86 anode is only some 10Vpp before the power stage overloads - which is the 'end-of-the-road' signal-wise anyway. The reason for landing at about 300V supply (R2 of 68K) for the EL86 proper is the requirement of not higher than some 80V on the LTP cathodes. Yes, one could have drawn some more EF86 anode current to get down to 80V Va by adjusting screen and bias values, but the bias over R15 is already low (as Hugh has warned). Increasing Vg2 to get Va low would probably have landed Vg2 higher than Va, not a good idea. Increasing R13 would again have increased the Miller effect of the ECC88 - and so on. But none of this is set in stone - that is why I hoped others would comment - simply my suggestions.
Value of R1 (the correct one this time!) I have little to quibble with that.
Lastly just to remember my previous remark about R12-C11: Those components are there as needed for stability with gNFB. C11 comes in at about 33kHz; R12 of 4,7K might be too small. Hopefully the amplifier will not be unstable (h.f. oscillation) from the start and allow a stable square wave to be viewed. R12, C11, C13 can then be selected for optimality. Thus also the effect of C17-R23. (I presume NFB stability criteria is known - perhaps if necessary someone else can come in on that, before I suffer from over-exposure! That might be a separate thread.) Since it seems that most members are going to use the same components, particularly the OPT, an experienced member might sort these values out for others to follow.
Afterthought: It is always a good idea to have some 'limiting' resistance over the OPT output (parallel). Traditionally a 1K resistor is used; if the loudspeaker load is going to be 8 ohms, such a resistor can be lower. A 560 ohm resistor will absorb only 1,5% of the power (5 - 10W resistor!).
Again the post is becoming drawn out; I recall someone mention R34 - R37. Yes those can be only one resistor per cathode, conveniently 10 ohm. And it is hardly ever necessary to worry about the inductance of wirewound resistors in audio. The Q is simply too low to cause noticable effect, in my experience.
Sorry for my possibly mixing up more than one circuit diagram in referring to components. To set that straight, the diagram should be the one given in post #610 (and shortly afterwards post #612). The designations and electrode voltages there represent my suggestions. Only, recently, C6 was increased to 32µF, which was decreased in the immediately previous post to 22µF. If my subsequent posts differ, ignore and put it down to .... well whatever.
(No pressing reason for my 32µF; it is more of an available value here than 22µF. A higher value will guarantee a little more against l.f. instability and is physically quite small as said previously, but we do not know that l.f. instability will be a factor at all. Thus 22µF if that is more available over there.)
Regarding several later remarks:
The feeding h.t. to the EF86 as high as possible: The maximum signal at EF86 anode is only some 10Vpp before the power stage overloads - which is the 'end-of-the-road' signal-wise anyway. The reason for landing at about 300V supply (R2 of 68K) for the EL86 proper is the requirement of not higher than some 80V on the LTP cathodes. Yes, one could have drawn some more EF86 anode current to get down to 80V Va by adjusting screen and bias values, but the bias over R15 is already low (as Hugh has warned). Increasing Vg2 to get Va low would probably have landed Vg2 higher than Va, not a good idea. Increasing R13 would again have increased the Miller effect of the ECC88 - and so on. But none of this is set in stone - that is why I hoped others would comment - simply my suggestions.
Value of R1 (the correct one this time!) I have little to quibble with that.
Lastly just to remember my previous remark about R12-C11: Those components are there as needed for stability with gNFB. C11 comes in at about 33kHz; R12 of 4,7K might be too small. Hopefully the amplifier will not be unstable (h.f. oscillation) from the start and allow a stable square wave to be viewed. R12, C11, C13 can then be selected for optimality. Thus also the effect of C17-R23. (I presume NFB stability criteria is known - perhaps if necessary someone else can come in on that, before I suffer from over-exposure! That might be a separate thread.) Since it seems that most members are going to use the same components, particularly the OPT, an experienced member might sort these values out for others to follow.
Afterthought: It is always a good idea to have some 'limiting' resistance over the OPT output (parallel). Traditionally a 1K resistor is used; if the loudspeaker load is going to be 8 ohms, such a resistor can be lower. A 560 ohm resistor will absorb only 1,5% of the power (5 - 10W resistor!).
Again the post is becoming drawn out; I recall someone mention R34 - R37. Yes those can be only one resistor per cathode, conveniently 10 ohm. And it is hardly ever necessary to worry about the inductance of wirewound resistors in audio. The Q is simply too low to cause noticable effect, in my experience.
Last edited:
Another OUCH (yet again)
I have just noticed a further correction needed on the diagram. On the diagram above the EF86, it shows the correct B+ at 288V, but below on the power supply diagram it still shows the old value of 240V. Then perhaps also, the value of R10 = 200K, is a non-standard over here. I notice that a lot of 'strange' resistor values are available over your ways, but that value is uncritical and might well be 220K or 270K or some standard value, if it matters.
Then: Eli and others warned about the oscillation-prone behaviour of r.f. tubes, hence we decided to add R18, R25. With this in view, I would recommend that C17 rather be shifted to the 'R19-side' of R18, away from directly on ECC88 pin 2. Just being careful.
And: I omitted to add resistor dissipation up to now. In a final diagram that needs to be stated. Just simply because ½w resistors look small in a tube construction, one might prefer to use 1W, but ½w is sufficient for all resistors except those specially mentioned. The only dissipating resistors are R21 and R22. The diagram states 1W, but they dissipate 1,5W working, and it might get hot under-chassis. I prefer 5W wire-wounds there; over here 3W metal oxides are not easily available, which should be the lowest wattage used there. (Where I showed frequency responses some posts ago, I tested the circuit with 27K 5W wirewounds.)
Thanks
I have just noticed a further correction needed on the diagram. On the diagram above the EF86, it shows the correct B+ at 288V, but below on the power supply diagram it still shows the old value of 240V. Then perhaps also, the value of R10 = 200K, is a non-standard over here. I notice that a lot of 'strange' resistor values are available over your ways, but that value is uncritical and might well be 220K or 270K or some standard value, if it matters.
Then: Eli and others warned about the oscillation-prone behaviour of r.f. tubes, hence we decided to add R18, R25. With this in view, I would recommend that C17 rather be shifted to the 'R19-side' of R18, away from directly on ECC88 pin 2. Just being careful.
And: I omitted to add resistor dissipation up to now. In a final diagram that needs to be stated. Just simply because ½w resistors look small in a tube construction, one might prefer to use 1W, but ½w is sufficient for all resistors except those specially mentioned. The only dissipating resistors are R21 and R22. The diagram states 1W, but they dissipate 1,5W working, and it might get hot under-chassis. I prefer 5W wire-wounds there; over here 3W metal oxides are not easily available, which should be the lowest wattage used there. (Where I showed frequency responses some posts ago, I tested the circuit with 27K 5W wirewounds.)
Thanks
Both 33uF and 22uF are E12 values so fairly common here (even at 350v).
I think that we should pencil these in at the suggested value and finalize them AFTER TubeMack has built the prototype and viewed some square waves on his shiny new scope. Version 1.0 schematic after that... tried, tested and true. Umm.. Did you mean 33pF instead of 33KHz, I guess both could make sense there (without doing some arithmetic to see).
Yup, that was fixed in rev. 7c
200K is pretty common here (E24 I think), but if it's not elsewhere, then it's certainly easy and worthwhile to change, which I will do.
Yup, did that already in an earlier update.
Okay, I'll update this.
..Todd
Last edited: