Tubelab;
As you know I have been kicking around some ideas for both a preamp/crossover and a SE power amp for the high frequency drivers (above 150Hz) in my system. The more I think it over I am drawn to something similar to your simpleSE except that I would be using point to point wiring, no GNF and no CCS.
I am inclined toward the edcor 5K open frame with switchable UL and I like your idea of relatively high B+.
In your write up you state that you needed CCS with the 12AT7 to get enough gain. It seems like the CCS might not be needed if no FB is used as there will be no feedback losses. Do you have a feel for whether a resistively loaded AT would have adequate gain to drive EL34 or 6550?
You also mentioned that a couple of obvious choices in higher gain tubes (12AX7 and 6SL7) would not provide the frequency response that you needed because of capacitive loads of the output stage. Is that because the maximum cathode current is too low to drive the C load or because of low Gm (or is that the same thing in effect)?
Would paralleling the dual triode sections solve that problem or are we still short?
I appreciate your patience with a rookie's questions. I don't really understand how you calculated the necessary drive for a particular frequency response.
mike
As you know I have been kicking around some ideas for both a preamp/crossover and a SE power amp for the high frequency drivers (above 150Hz) in my system. The more I think it over I am drawn to something similar to your simpleSE except that I would be using point to point wiring, no GNF and no CCS.
I am inclined toward the edcor 5K open frame with switchable UL and I like your idea of relatively high B+.
In your write up you state that you needed CCS with the 12AT7 to get enough gain. It seems like the CCS might not be needed if no FB is used as there will be no feedback losses. Do you have a feel for whether a resistively loaded AT would have adequate gain to drive EL34 or 6550?
You also mentioned that a couple of obvious choices in higher gain tubes (12AX7 and 6SL7) would not provide the frequency response that you needed because of capacitive loads of the output stage. Is that because the maximum cathode current is too low to drive the C load or because of low Gm (or is that the same thing in effect)?
Would paralleling the dual triode sections solve that problem or are we still short?
I appreciate your patience with a rookie's questions. I don't really understand how you calculated the necessary drive for a particular frequency response.
mike
mashaffer said:
This part is easy. The grid circuit of the finals has capacitance. The grid doesn't see the true value of the signal until that capacitance is fully charged. This is the slew rate problem that can cause loss of detail at the high end. Lots of designers seem to think it's just a solid state problem, but it's not.
The input capacitance is: Ci= Cgk + Cmiller + Cstray. You can guesstimate Cstray to be about 30pF, and get Cgk from the spec sheet. Cmiller being the reverse transfer capacitance multiplied by (Av + 1). Then use: Xc= 1/wC to find out the impedance at 30KHz, for example. If the total Ci= 100pf, then
Xc= 1/(2pi * 30E3 * 100E-12)= 53K1.
If the input swing is +/- 20Vp, then:
I= 20 / 53.1E3= 0.38mA. This is the current that the driver must source. Here, what I do is apply the "Rule of Five" from solid state practice, so that the plate current of the driver should be at least:
Ipq= 5 * 0.38= 1.9mA or better.
Unless you have some wierd situation, a 12AT7 ought to be able to handle grid drive for just about anything, except for possibly some power triodes with large input swing requirements and large Cmiller's. Some of these triodes have an additional problem of grid current draw even before the control grid actually goes positive. In cases like that, you may need a cathode or source follower driver.
The small Edcor is an excellent choice when accurate bass below 100 Hz in not needed. The SimpleSE does not use GNFB. It can (optional) use local (cathode) feedback around the output stage. CFB is generally not needed for applications such as yours. It is a way to extend the frequency response of low cost OPT's and lower the output impedance. The Edcor goes to 40 KHz without feedback, and you don't need the low end enhancement. I would probably use triode mode unless you need the extra power of UL.
A 12AT7 with a pure resistive load does not have enough gain reserve to drive every output tube (6L6GC, EL34, KT88) to full output from some CD players under all conditions. These are the conditions I had to meet with the design since this is how most SimpleSE's are used. If your preamp/crossover has a few db of gain from the CD input to the power amp, then a resistively loaded 12AT7 will do fine.
The calculations can be done as Miles stated. It also should be noted that many tubes that are made today (and a lot of older ones) bear no relationship to the data sheets. It is not uncommon to find new Russian or Chinese tubes sold using data copied from old RCA info, or sold with no data at all. Capacitance data can be off considerably. For these reasons I tend to test every combination for frequency response and phase shift.
A 12AX7 or 6SL7 is marginal because of the typical 1 mA operating current. Paralleled sections may do fine, but I have not tested those combinations. A high gain input tube may also exhibit high frequency rolloff when driven by a high impedance source due to its own miller capacitance. This should not be an issue with a preamp or crossover source, but can be a problem if the volume control is connected to the input tube grid. In this case the volume control value should be limited to 50 K ohms or less.
A 12AT7 with a pure resistive load does not have enough gain reserve to drive every output tube (6L6GC, EL34, KT88) to full output from some CD players under all conditions. These are the conditions I had to meet with the design since this is how most SimpleSE's are used. If your preamp/crossover has a few db of gain from the CD input to the power amp, then a resistively loaded 12AT7 will do fine.
The calculations can be done as Miles stated. It also should be noted that many tubes that are made today (and a lot of older ones) bear no relationship to the data sheets. It is not uncommon to find new Russian or Chinese tubes sold using data copied from old RCA info, or sold with no data at all. Capacitance data can be off considerably. For these reasons I tend to test every combination for frequency response and phase shift.
A 12AX7 or 6SL7 is marginal because of the typical 1 mA operating current. Paralleled sections may do fine, but I have not tested those combinations. A high gain input tube may also exhibit high frequency rolloff when driven by a high impedance source due to its own miller capacitance. This should not be an issue with a preamp or crossover source, but can be a problem if the volume control is connected to the input tube grid. In this case the volume control value should be limited to 50 K ohms or less.
So let's see if I understand this. Given a tube with the following parameters...
Mu 10
Cgk 15pf
Cgp 1pf
Cstray 30pf
The Worst case input capacitance is
Cgk+Cstray+((Mu+1)*Cgp)=15+30+11=56pF
Of course actual gain will be less than Mu so the actual input capacitance would be calculated by substituting the actual gain for Mu above correct?
So given 56pF input capacitance the Zc(30kHz)=94.8kohms.
So wee need to source about 0.21mA.
BTW the numbers I chose were intended to approximate an EL34.
Does that look OK?
BTW another thought has occurred to me. We are accounting for Cgp and Cgk but what about the capacitances WRT g2? Would these too not enter into the picture? And would the effects be different when in Triode v.s. Pentode v.s. UL connection?
mike
BTW I am looking seriously at EL34, 6L6, 6550 and K88 but I will need to crunch a lot more numbers to decide on the best approach for me. The 6550 seem to pack a real wallop but a quick glance at the curves seems to indicate a pretty large voltage swing to take full advantage of it's capabilities.
Mu 10
Cgk 15pf
Cgp 1pf
Cstray 30pf
The Worst case input capacitance is
Cgk+Cstray+((Mu+1)*Cgp)=15+30+11=56pF
Of course actual gain will be less than Mu so the actual input capacitance would be calculated by substituting the actual gain for Mu above correct?
So given 56pF input capacitance the Zc(30kHz)=94.8kohms.
So wee need to source about 0.21mA.
BTW the numbers I chose were intended to approximate an EL34.
Does that look OK?
BTW another thought has occurred to me. We are accounting for Cgp and Cgk but what about the capacitances WRT g2? Would these too not enter into the picture? And would the effects be different when in Triode v.s. Pentode v.s. UL connection?
mike
BTW I am looking seriously at EL34, 6L6, 6550 and K88 but I will need to crunch a lot more numbers to decide on the best approach for me. The 6550 seem to pack a real wallop but a quick glance at the curves seems to indicate a pretty large voltage swing to take full advantage of it's capabilities.
In triode mode G2 is connected to the plate, so the G1 to plate capacitance and the G1 to G2 capacitance must be added together before being multiplied by Mu+1.
In Ultralinear mode you use the Multiply the G1 to G2 capacitance by the UL ratio before using it in the triode formula.
In pentode mode the miller effect is almost nil. You just add up the G1 to everything else capacitance and use it.
- Status
- Not open for further replies.