Maybe I should have split these questions into two threads, but it's all the same amp, so here goes:
Question 1
In the attached design, the first two stages use a CCS in the cathode circuit to bias the differential pairs. The negative rails come up immediately on power up, being SS rectified. The B+ (not shown) is delayed through damper diodes, so by the time B+ comes up, the tubes are sufficiently heated to begin conduction. What happens prior to B+ ? With no load on the negative supply, the rail is close to -100V, which the CCS has no problem with. I'm trying to figure out if the grid-cathode is zero or 100V prior to conduction, and if this is a problem. With zero gate voltage, the DN2540 tries to turn fully on, which I would think would force Vds to zero, thereby applying -100V to the cathode. No current flow yet, but still a very large positive grid voltage ??
Question 2
Using the 0.047uF input capacitor and 121K resistor to produce a single pole HPF, but the numbers don't seem to work out very well. Actual turns ratio is 1.95, so by the numbers the -3dB point should be about 106Hz. Actual measured data is just under 80Hz, which is a far cry. Passive values have been checked with 1% or better accuracy, and the source impedance of the function generator is 50 ohms. Any ideas? I can't imagine the transformer DCR would be contributing, and for the tests, no tubes were installed, this is measured right at the xfmr output at no load.
I have seen some resonance (peaking just above -3dB) when using capacitors as low as 0.01uF. Increasing values seemed to eliminate the resonance, even at 0.022 uF. Values as high as 0.1uF worked too, but then the resistive component starts to drastically decrease my input impedance. So I met halfway with the 0.047uF value.
Question 1
In the attached design, the first two stages use a CCS in the cathode circuit to bias the differential pairs. The negative rails come up immediately on power up, being SS rectified. The B+ (not shown) is delayed through damper diodes, so by the time B+ comes up, the tubes are sufficiently heated to begin conduction. What happens prior to B+ ? With no load on the negative supply, the rail is close to -100V, which the CCS has no problem with. I'm trying to figure out if the grid-cathode is zero or 100V prior to conduction, and if this is a problem. With zero gate voltage, the DN2540 tries to turn fully on, which I would think would force Vds to zero, thereby applying -100V to the cathode. No current flow yet, but still a very large positive grid voltage ??
Question 2
Using the 0.047uF input capacitor and 121K resistor to produce a single pole HPF, but the numbers don't seem to work out very well. Actual turns ratio is 1.95, so by the numbers the -3dB point should be about 106Hz. Actual measured data is just under 80Hz, which is a far cry. Passive values have been checked with 1% or better accuracy, and the source impedance of the function generator is 50 ohms. Any ideas? I can't imagine the transformer DCR would be contributing, and for the tests, no tubes were installed, this is measured right at the xfmr output at no load.
I have seen some resonance (peaking just above -3dB) when using capacitors as low as 0.01uF. Increasing values seemed to eliminate the resonance, even at 0.022 uF. Values as high as 0.1uF worked too, but then the resistive component starts to drastically decrease my input impedance. So I met halfway with the 0.047uF value.
Q1) A quick solution to this is to use a couple 1N4148s from grid to cathode on the LTPs. Anode of diodes go to grid, cathode goes to cathode. Then when there is no current through the LTP, the all current goes through the diodes and avoids high Vg-k voltages. When the tube is hot and B+ is there, cahtodes positive wrt ground the 1N4148s are reversed bias and out of the circuit.
Q2) You can try putting the resistor on the primary after the capacitor and see if the rolloff correct. Hard to say what's going on there, did you measure at the primary too?
Q2) You can try putting the resistor on the primary after the capacitor and see if the rolloff correct. Hard to say what's going on there, did you measure at the primary too?
There is no need at all for CCSs as this is a balanced configuration.
A good reference:
http://www.nutshellhifi.com/triode2.html
A good reference:
http://www.nutshellhifi.com/triode2.html
That's where I fail to get it. First you say "no current" then you say "all current". It's not both. I can have voltage drop without current flow, so is running the grid positive (with no current) a problem? If the tube begins to warm before B+ comes up, do I get copius grid current? That would be bad, I imagine.Boris_The_Blade said:
A fine idea. I'll try tonite and report.
I'm probably not the best person to ask that question, but offhand I would guess 'not at the voltages required'. Besides, where's the fun in thatdsavitsk said:
You have a point, but then again, Gary Pimm has used them in his cathodes extensively, and Morgan Jones has many good things to say about this application. IIRC, Gary mentions that installing the CCS's under the cathodes allowed him to remove the WE bypass caps with no change in sound. I LIKE the idea of as few caps as possible. Also forces my tubes to a specific operating point, which a resistor would not.revintage said:
hey-Hey!!!,
The diode from ground-cathode(s) sets the voltage at 0.7 V. The set current flows, and the added diode sets the voltage. I have done this when ever I have a LTP input stage and 'lytic B+ caps with delay in application of B+; the LTP will pull them negative and the added diode stips that. With film B+ caps I couldn't care less what the B+ rail is going to....and thus delete one more part. I have used some other Zeners, 18V/500 mW that I also use for gate-source clamps.
cheers,
Douglas
The diode from ground-cathode(s) sets the voltage at 0.7 V. The set current flows, and the added diode sets the voltage. I have done this when ever I have a LTP input stage and 'lytic B+ caps with delay in application of B+; the LTP will pull them negative and the added diode stips that. With film B+ caps I couldn't care less what the B+ rail is going to....and thus delete one more part. I have used some other Zeners, 18V/500 mW that I also use for gate-source clamps.
cheers,
Douglas
Bandersnatch said:
Okay, that makes some sense to avoid reverse biasing any lytics. My power supply is exclusively poly in oil, so no worries there.
Are you then saying that when the IDHT heats sufficiently with no B+, that the grid will in fact pass current to the cathode? I haven't gotten an answer to this question yet, at least not directly stated. If so, does it do damage to the triode, or is it merely a concern for other parts of the system (i.e. electrolytics)?
revintage said:
I would like more information on this, even though it's not what my original question was about. I see the CCS in the anode forming a shunt regulator with the gas tube. All the triode sees is a transformer fed by a voltage source, nothing else. It has no clue there is a CCS upstream.
This being the case, a CCS in the common cathodes makes sense, and should work just fine. Granted, it is easy enough to try both ways, and I may do that, but I do believe it will work.
Couldn't get around to checking HPF action last night; hopefully tonite.
It will take significant plate current too( as evidenced by the reverse polarity PS caps). Grid current will roughly be limited by the grid resistor( or volume pot; turned to Zero would allow current I think ). it'll pass some g1 current, but not in a way that would pose me any worry if the rest of the system were up to the task.
cheers,
Douglas
cheers,
Douglas
So did some testing tonite, came up with somewhat questionable results. All measurements taken with calibrated Fluke 87 and uncalibrated Fluke 8060A. Source is an unimpressive sine/tri/square frequency generator; you can see the glitch on the sine wave, indicating they generate the sine wave from the triangular.
47nF cap with 99.5K resistor on secondary of 1.95 ratio xfmr = 108Hz cutoff. Being the capacitance measurement is probably the least accurate, the numbers would indicate an actual capacitance of 56nF. Quite a difference, with a 10% tolerance cap and a more accurate multimeter.
Next test was same cap, 9.857K resistor, no other load, no xfmr. Just a simple RC HPF. Measured -3dB is 309Hz. Still results in a calculated capacitance of 52nF, outside normal tolerances.
Tried moving to a 33nF 10% tolerance capacitor, measured by Fluke 87 as 32.7nF.
With transformer in circuit, 129.3K resistor on secondary, -3dB = 118Hz, resulting in calculated capacitance of 39.7nF, still outside tolerance.
Simple RC network with no xfmr, same capacitor, 9.86K resistor, -3dB = 410Hz, results in calculated capacitance of 39.4nF. Pretty close to previous test, but well outside direct measured capacitance and nameplate tolerance.
Is it possible the quality of the sine wave is messing up the measurements that bad? I know the measured resistances are very accurate, as is the frequency and voltage measurements. Confirmed with a scope, the frequency generator magnitude is constant over the swept frequency range.
Maybe a valid question would be does anyone know of a Mac app that produces fixed frequency clean sine waves ?
47nF cap with 99.5K resistor on secondary of 1.95 ratio xfmr = 108Hz cutoff. Being the capacitance measurement is probably the least accurate, the numbers would indicate an actual capacitance of 56nF. Quite a difference, with a 10% tolerance cap and a more accurate multimeter.
Next test was same cap, 9.857K resistor, no other load, no xfmr. Just a simple RC HPF. Measured -3dB is 309Hz. Still results in a calculated capacitance of 52nF, outside normal tolerances.
Tried moving to a 33nF 10% tolerance capacitor, measured by Fluke 87 as 32.7nF.
With transformer in circuit, 129.3K resistor on secondary, -3dB = 118Hz, resulting in calculated capacitance of 39.7nF, still outside tolerance.
Simple RC network with no xfmr, same capacitor, 9.86K resistor, -3dB = 410Hz, results in calculated capacitance of 39.4nF. Pretty close to previous test, but well outside direct measured capacitance and nameplate tolerance.
Is it possible the quality of the sine wave is messing up the measurements that bad? I know the measured resistances are very accurate, as is the frequency and voltage measurements. Confirmed with a scope, the frequency generator magnitude is constant over the swept frequency range.
Maybe a valid question would be does anyone know of a Mac app that produces fixed frequency clean sine waves ?
Well, after some further testing, there is a combination of two issues. One, the purity of the sine wave drastically affects the results, even with a simple RC HPF with no transformer involved. Downloaded a function generator, using the sound card of the Mac. Scope verified a very pure sine wave, much cleaner than the previous function generator used. Numbers turned out right on the money.
Then, inserted the transformer and repeated the tests. Still off the numbers, so the transformer is definitely affecting the ideal response. Essentially, the transformer makes the overall response appear as though the input capacitor is about 15% higher than actual. Empirical testing is really the only reliable way to get the proper behavior.
End result is theory with ideal transformer required 47nF cap with 110K. Actual test results show 33nF cap with 130K for same response.
Fortunately, there are no peaks or odd behavior going on with these values, all the way down to 20Hz. Looks like from a test equipment standpoint this methodology should work; we'll see how it sounds.
At least I have obtained a good clean sine wave generator out of this fiasco...
Then, inserted the transformer and repeated the tests. Still off the numbers, so the transformer is definitely affecting the ideal response. Essentially, the transformer makes the overall response appear as though the input capacitor is about 15% higher than actual. Empirical testing is really the only reliable way to get the proper behavior.
End result is theory with ideal transformer required 47nF cap with 110K. Actual test results show 33nF cap with 130K for same response.
Fortunately, there are no peaks or odd behavior going on with these values, all the way down to 20Hz. Looks like from a test equipment standpoint this methodology should work; we'll see how it sounds.
At least I have obtained a good clean sine wave generator out of this fiasco...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.