Hi there everybody,
I have a newbie tube question.
I have a system made of a simple, single-ended 5687 line "control" preamp feeding a two-stage push-pull 6H30Pi to 2A3 power amp.
My preamp has its volume pot on the output. There's a 50k Noble pot in there now. Ignoring the issue of whether the pot should be on the input or the output of the line level stage, or if I should be using a stepped attenuator or whatever...
I've been experimenting with different values of grid leak resistors, and I've found that changing the value in the preamp changes the sound pretty drastically.
So the question is:
Since the value of grid leak resistor can vary over a fairly wide range, how does one arrive at an "optimal" value? I'm running a home-brew 5687 preamp at Ep = 135V, Ek = 3.6V using two red LEDs, Ip = 17.5mA, Rload of 15k, Ebb is about 400V...
The preceding stages are a phono amp with 7119/ECC182 output (so an output Z of about 3k, I reckon), a stock Pioneer Elite SACD player and a Kenwood solid-state tuner (both probably with output Z of < or = 1000 ohms, right?).
1) I started out with a value of 47k for the grid leak resistors. I had no grid stopper on the 5687. I liked the sound for a long time. Not at all harsh, highs very smooth. But I thought the bass didn't go all that deep, and sounded a little "loose."
2) I finally wore out the 5687, so I decided to go through the preamp to see if anything could be "improved." I saw that 47k grid leak resistor without a grid stopper and thought that could be improved upon. Looking up the 5687 datasheet, I found that the max grid leak resistance for a cathode-biased 5687 is 1M ohm. I decided to try a 470k grid leak resistor (1/4 watt metal film) with a 680 ohm grid stopper (carbon comp).
Sure enough, the bass seemed tighter. Really nice and seemed to go lower/deeper. That part I liked. However, I found the sound on the high end to be noticeably "harder" (for lack of a better term). Very much more intense, almost shrieking. What did I do wrong?
3) I decided that the 470k grid leak must be too high a value. I replaced with 100k carbon comp, but forgot to put in a grid stopper (there's no grid stopper there now). The sound is more relaxed, but still a bit harder/harsher than I remember. Or is that just my audiophilia nervosa at work?
I was afraid that 47k would load down the output from the phono amp, so just now, I decided to employ some math...
Phono amp output stage:
7119 with Rload of 15k
Output cap is 2.2uF
output load resistor of 499k
I figure the output impedance of the final 7119 stage is about 3k, so that would put 3k in parallel with the 47k grid leak resistor = ~42k. The 7119's output Z would be working into the grid leak resistor and the load presented by the 5687, plus whatever strays from the input switch (totaling about 80pF or so, I reckon).
So on the low frequency end, 2.2uF into ~42k load = -3dB at under 2Hz. I guess 47k should be fine then.
On the high frequency end, if I choose 100k instead of 47k for the 5687 grid leak, and with the 5687's input C of about 80pF, the -3dB point is ~20kHz! 😱
Hmmm... Maybe I really do want to keep the grid leak resistor value as small as possible. Using a 47k grid leak resistor and with the 80pF of input capacitance to be expected from a 5687 including strays, the -3dB point for high frequencies is still low, at 42kHz. Can I (or should I) use an even lower value grid leak R? What about 33k?
Did I figure that correctly, or am I totally off-base?
Help me, please.
--
I have a newbie tube question.
I have a system made of a simple, single-ended 5687 line "control" preamp feeding a two-stage push-pull 6H30Pi to 2A3 power amp.
My preamp has its volume pot on the output. There's a 50k Noble pot in there now. Ignoring the issue of whether the pot should be on the input or the output of the line level stage, or if I should be using a stepped attenuator or whatever...
I've been experimenting with different values of grid leak resistors, and I've found that changing the value in the preamp changes the sound pretty drastically.
So the question is:
Since the value of grid leak resistor can vary over a fairly wide range, how does one arrive at an "optimal" value? I'm running a home-brew 5687 preamp at Ep = 135V, Ek = 3.6V using two red LEDs, Ip = 17.5mA, Rload of 15k, Ebb is about 400V...
The preceding stages are a phono amp with 7119/ECC182 output (so an output Z of about 3k, I reckon), a stock Pioneer Elite SACD player and a Kenwood solid-state tuner (both probably with output Z of < or = 1000 ohms, right?).
1) I started out with a value of 47k for the grid leak resistors. I had no grid stopper on the 5687. I liked the sound for a long time. Not at all harsh, highs very smooth. But I thought the bass didn't go all that deep, and sounded a little "loose."
2) I finally wore out the 5687, so I decided to go through the preamp to see if anything could be "improved." I saw that 47k grid leak resistor without a grid stopper and thought that could be improved upon. Looking up the 5687 datasheet, I found that the max grid leak resistance for a cathode-biased 5687 is 1M ohm. I decided to try a 470k grid leak resistor (1/4 watt metal film) with a 680 ohm grid stopper (carbon comp).
Sure enough, the bass seemed tighter. Really nice and seemed to go lower/deeper. That part I liked. However, I found the sound on the high end to be noticeably "harder" (for lack of a better term). Very much more intense, almost shrieking. What did I do wrong?
3) I decided that the 470k grid leak must be too high a value. I replaced with 100k carbon comp, but forgot to put in a grid stopper (there's no grid stopper there now). The sound is more relaxed, but still a bit harder/harsher than I remember. Or is that just my audiophilia nervosa at work?
I was afraid that 47k would load down the output from the phono amp, so just now, I decided to employ some math...
Phono amp output stage:
7119 with Rload of 15k
Output cap is 2.2uF
output load resistor of 499k
I figure the output impedance of the final 7119 stage is about 3k, so that would put 3k in parallel with the 47k grid leak resistor = ~42k. The 7119's output Z would be working into the grid leak resistor and the load presented by the 5687, plus whatever strays from the input switch (totaling about 80pF or so, I reckon).
So on the low frequency end, 2.2uF into ~42k load = -3dB at under 2Hz. I guess 47k should be fine then.
On the high frequency end, if I choose 100k instead of 47k for the 5687 grid leak, and with the 5687's input C of about 80pF, the -3dB point is ~20kHz! 😱
Hmmm... Maybe I really do want to keep the grid leak resistor value as small as possible. Using a 47k grid leak resistor and with the 80pF of input capacitance to be expected from a 5687 including strays, the -3dB point for high frequencies is still low, at 42kHz. Can I (or should I) use an even lower value grid leak R? What about 33k?
Did I figure that correctly, or am I totally off-base?
Help me, please.
--
The maximum grid leak given in the data sheet for cathode bias assumes cathode bias via a resistor. This can take up the slack caused by any bias shifts due to grid current. If you are using LED bias than this is more or less the same as fixed bias so needs a smaller grid resistor. Using the larger grid leak would have shifted the bias point, which may have been audible.
Your LF calculation is not quite right. The 3k preamp output impedance has to be added to the 47K, so you have 2.2uF with 50k (-3dB at 1.45Hz). This is already too low for a phono source, so you will be getting lots of subsonic signals from record warps unless your preamp has a good filter. You won't hear these signals directly, but they may cause intermodulation because your amp still has to process them.
At HF the relevant resistance to use is preamp output impedance in parallel with grid leak, then in series with the grid stopper. So 3K output in parallel with 47K is 2.82K, plus 680 ohm stopper gives 3.5K. With 80pF this is -3dB at 568kHz.
The other effect of the grid leak is to load the anode of the previous stage. So 47K in parallel with 15K anode resistor gives an effective AC load of about 11K. This may increase distortion in the previous stage. The general rule with grid leaks is to put them towards the upper end of the range specified in the data sheet, for the relevant bias method. The 5687 data sheet does not give a value for fixed bias, but as a wild guess I would go for 20% of the cathode bias value - so 200K. I would therefore use something like 150K.
Your LF calculation is not quite right. The 3k preamp output impedance has to be added to the 47K, so you have 2.2uF with 50k (-3dB at 1.45Hz). This is already too low for a phono source, so you will be getting lots of subsonic signals from record warps unless your preamp has a good filter. You won't hear these signals directly, but they may cause intermodulation because your amp still has to process them.
At HF the relevant resistance to use is preamp output impedance in parallel with grid leak, then in series with the grid stopper. So 3K output in parallel with 47K is 2.82K, plus 680 ohm stopper gives 3.5K. With 80pF this is -3dB at 568kHz.
The other effect of the grid leak is to load the anode of the previous stage. So 47K in parallel with 15K anode resistor gives an effective AC load of about 11K. This may increase distortion in the previous stage. The general rule with grid leaks is to put them towards the upper end of the range specified in the data sheet, for the relevant bias method. The 5687 data sheet does not give a value for fixed bias, but as a wild guess I would go for 20% of the cathode bias value - so 200K. I would therefore use something like 150K.
Thanks for helping me out.
Hmm, I wonder if I was enjoying a little extra harmonic distortion from my phono preamp... It seems a lot of people find a little "tubey" distortion pleasing. Fortunately, an 11k load on a 7119 would probably be just OK (rp of about 3k). Better with a 5687 (rp of about 2.5k).
In a previous thread, it was pointed out to me that my preamp includes a built-in LF rolloff between stages. http://www.diyaudio.com/forums/tubes-valves/147717-cathode-bias-questions-9.html. I hope that rolloff takes care of the record warp problem. I could also reduce the value of the output coupling cap. I have some really good-sounding Roederstein 220nF parts I could use (f3 = 15.4 Hz), maybe two in parallel for 440nF (f3 = 7.7 Hz)...
I have 100k in the grid leak of the line preamp now. That's probably good enough. Now for the power amp, which has 220k grid leak resistors. The power amp's input stage is a 6H30Pi long-tail pair w/ LM317 ccs in the tail. Does the LTP count as fixed bias or cathode bias? I'd figure cathode bias, because the impedance seen by the cathodes is very high. But it does have an active load...
--
Hmm, I wonder if I was enjoying a little extra harmonic distortion from my phono preamp... It seems a lot of people find a little "tubey" distortion pleasing. Fortunately, an 11k load on a 7119 would probably be just OK (rp of about 3k). Better with a 5687 (rp of about 2.5k).
In a previous thread, it was pointed out to me that my preamp includes a built-in LF rolloff between stages. http://www.diyaudio.com/forums/tubes-valves/147717-cathode-bias-questions-9.html. I hope that rolloff takes care of the record warp problem. I could also reduce the value of the output coupling cap. I have some really good-sounding Roederstein 220nF parts I could use (f3 = 15.4 Hz), maybe two in parallel for 440nF (f3 = 7.7 Hz)...
I have 100k in the grid leak of the line preamp now. That's probably good enough. Now for the power amp, which has 220k grid leak resistors. The power amp's input stage is a 6H30Pi long-tail pair w/ LM317 ccs in the tail. Does the LTP count as fixed bias or cathode bias? I'd figure cathode bias, because the impedance seen by the cathodes is very high. But it does have an active load...
--
Lets cover the technical "requirements" fro the grid leak resistor.
At NORMAL negative bias we have (hopefully small) NEGATIVE or REVERSE grid current flowing. This negative current is the sum of:
- Gas Ionization Current
- Leakage Current
- Grid Emission Current
In general the last 2 will be negligible an so we are primarily concerned with gas ionization current. When the electron stream accelerates "up" the tube from cathode to anode some electrons can "crash" into any residual gas atoms in the tube. This collision is energetic enough to strip outer orbit electrons effectively turning the gas atom into a positively charged ion which of course then accelerates back toward the cathode AND the grid which is likely to be at the most negative potential of any element in the tube. Ions hitting the grid will then cause NEGATIVE grid current and the grid will take on a negative voltage from this current developing a potential across the grid to ground (Grid Leak or Rg1) resistor.
It is obvious from this that the problem becomes worse as tube current increases as there are more electrons to smash into residual gas atoms (so it is worse in power tubes). It is also obvious that the problem becomes worse as the tubes vacuum decreases (more gas atoms to smash into).
The 3rd thing that is obvious is that for a given level of negative grid current the problem gets worse as the grid leak resistor is increased in value.
What is perhaps less obvious is that this problem is worse in high mu tubes compared to low mu tubes.
The negative voltage developed at the grid ADDS to bias and disturbs the operating point. Using cathode bias can give some "self adjustment" of this and that is why you see a max Rg1 value specified as a certain value for fixed bias and a larger value for cathode bias.
RDH contains some useful guide lines.
For high mu tubes, if using fixed bias, The grid leak resistor Rg1 should be no larger than 2 x the value of the anode load resistor. If using cathode bias then Rg1 can be 1.5 times larger again or 3 times the value of the anode load resistor.
Sticking to these guidelines will generally guarantee that the grid will sit at not more than -0.3V and more typically -0.15V. Some adjustment of the Cathode Bias resistor (downward) or the fixed bias voltage (downward) can made to compensate for that.
LED bias is fixed bias (regardless of the fact that the LED is in the cathode).
For low and medium mu tubes the above maybe relaxed a little.
Cheers,
Ian
At NORMAL negative bias we have (hopefully small) NEGATIVE or REVERSE grid current flowing. This negative current is the sum of:
- Gas Ionization Current
- Leakage Current
- Grid Emission Current
In general the last 2 will be negligible an so we are primarily concerned with gas ionization current. When the electron stream accelerates "up" the tube from cathode to anode some electrons can "crash" into any residual gas atoms in the tube. This collision is energetic enough to strip outer orbit electrons effectively turning the gas atom into a positively charged ion which of course then accelerates back toward the cathode AND the grid which is likely to be at the most negative potential of any element in the tube. Ions hitting the grid will then cause NEGATIVE grid current and the grid will take on a negative voltage from this current developing a potential across the grid to ground (Grid Leak or Rg1) resistor.
It is obvious from this that the problem becomes worse as tube current increases as there are more electrons to smash into residual gas atoms (so it is worse in power tubes). It is also obvious that the problem becomes worse as the tubes vacuum decreases (more gas atoms to smash into).
The 3rd thing that is obvious is that for a given level of negative grid current the problem gets worse as the grid leak resistor is increased in value.
What is perhaps less obvious is that this problem is worse in high mu tubes compared to low mu tubes.
The negative voltage developed at the grid ADDS to bias and disturbs the operating point. Using cathode bias can give some "self adjustment" of this and that is why you see a max Rg1 value specified as a certain value for fixed bias and a larger value for cathode bias.
RDH contains some useful guide lines.
For high mu tubes, if using fixed bias, The grid leak resistor Rg1 should be no larger than 2 x the value of the anode load resistor. If using cathode bias then Rg1 can be 1.5 times larger again or 3 times the value of the anode load resistor.
Sticking to these guidelines will generally guarantee that the grid will sit at not more than -0.3V and more typically -0.15V. Some adjustment of the Cathode Bias resistor (downward) or the fixed bias voltage (downward) can made to compensate for that.
LED bias is fixed bias (regardless of the fact that the LED is in the cathode).
For low and medium mu tubes the above maybe relaxed a little.
Cheers,
Ian
Gas ionization current makes the grid positive, which is why you can get thermal runaway in an output valve which has gone a bit gassy. The ions are positive, which is why they are attracted to the grid/cathode. There is also a small electronic grid current due to the thermoelectric effect. For small negative grid voltages this usually dominates, which is why it is possible to use grid leak bias (10M resistor) on input valves. There is great scope for confusion here. For grid leak bias (cathode grounded) you need a high grid leak in order to generate sufficient negative voltage from the small electron current. For conventional cathode bias you need a lowish grid leak to stop the grid from moving in a positive direction. I have seen a graph which shows this, but I can't find it now.
A high anode load resistor will limit the possible anode current, so you can get away with a slightly higher grid leak resistor.
A CCS in the cathode fixes the current so a high grid leak is fine. LTP with resistor is similar.
A high anode load resistor will limit the possible anode current, so you can get away with a slightly higher grid leak resistor.
A CCS in the cathode fixes the current so a high grid leak is fine. LTP with resistor is similar.
Thanks for this discussion. I learn a lot when you guys get talking to each other.
Re: RDH
OK, with LED bias being fixed bias, and treating the 5687 as a high-mu tube (which it's not, but)...
- If anode load resistor = 15k, then Rg1 should be no larger than 2x that, or 30k ohms (33k would be close enough, yes?).
Being that the 5687 is a medium-mu tube and I'm using it with fixed bias, then would it be safe to say that an Rg1 value of 3x or 4x the value of the anode load would be close? That would be 45k to 60k. 150k would be 10x the anode load.
While the RDH seems to mention mu as the determining factor, how does gm play into this? The 5687 and 6H30Pi tubes I'm using have pretty high gm. 5687 is also advertised as "high perveance," which might also affect things? If it conducts readily, might it not be prone to grid current? I'm untrained, so only going on intuition here...
When I get home, I'll measure the voltage at the grid of the 5687 to see how far it is from 0V ground. If there is no grid current flowing, I should see 0V, correct?
So a LTP with CCS in the cathode allows a higher value grid leak R with even less likelihood of problems?
--
Re: RDH
OK, with LED bias being fixed bias, and treating the 5687 as a high-mu tube (which it's not, but)...
- If anode load resistor = 15k, then Rg1 should be no larger than 2x that, or 30k ohms (33k would be close enough, yes?).
Being that the 5687 is a medium-mu tube and I'm using it with fixed bias, then would it be safe to say that an Rg1 value of 3x or 4x the value of the anode load would be close? That would be 45k to 60k. 150k would be 10x the anode load.
While the RDH seems to mention mu as the determining factor, how does gm play into this? The 5687 and 6H30Pi tubes I'm using have pretty high gm. 5687 is also advertised as "high perveance," which might also affect things? If it conducts readily, might it not be prone to grid current? I'm untrained, so only going on intuition here...
When I get home, I'll measure the voltage at the grid of the 5687 to see how far it is from 0V ground. If there is no grid current flowing, I should see 0V, correct?
So a LTP with CCS in the cathode allows a higher value grid leak R with even less likelihood of problems?
--
I think the thing to bear in mind is that there is not a 'correct' grid leak value, just a range of acceptable values. If it is too small then the bias will be stable, but it will load the previous stage - RDH says at least 4 times the anode resistor of the previous stage. If it is too big then the bias shifts. The data sheet gives the maximum, usually for cathode bias. For fixed bias RDH suggests a ratio of 1.2-2 (e.g. 1.5 close enough?) smaller - my figure of 5 was a wild guess, so please ignore it.
Considering all that, I would say that the original 47K was a bit too small and the 470K was possibly a bit too high. I would go for something between 68K and 330K. Measuring the grid voltage will only tell you what happens with that particular valve - another 5687 could be quite different.
Bear in mind that RDH regards mu=20 as being fairly high!
Considering all that, I would say that the original 47K was a bit too small and the 470K was possibly a bit too high. I would go for something between 68K and 330K. Measuring the grid voltage will only tell you what happens with that particular valve - another 5687 could be quite different.
Bear in mind that RDH regards mu=20 as being fairly high!
- Status
- Not open for further replies.