Hello,
I just wired my EL34 Simple SE amplifier in Pentode (from Triode) mode and it brought my FE127E bi-poles to life! but I notice two things
1. The sound is most defiantly less Linear
2. There is more distortion
So I wanted to add a little (3-9db) of NFB to smooth everything out, except, I have no clue how to properly add such feedback, all I know is that It involves putting a resistor from the secondary on the OPT to the Cathode of the driver stage, but I don't know any of the rules!
please help, because I'm fairly sure adding NFB is stupid easy once you figure it out.
Thanks
Moose
I just wired my EL34 Simple SE amplifier in Pentode (from Triode) mode and it brought my FE127E bi-poles to life! but I notice two things
1. The sound is most defiantly less Linear
2. There is more distortion
So I wanted to add a little (3-9db) of NFB to smooth everything out, except, I have no clue how to properly add such feedback, all I know is that It involves putting a resistor from the secondary on the OPT to the Cathode of the driver stage, but I don't know any of the rules!
please help, because I'm fairly sure adding NFB is stupid easy once you figure it out.
Thanks
Moose
alexmoose said:
Yep, that's gonna happen. I never did anything with EL34s, and don't know exactly what you're getting there. With pentodes, you will get more h3 and higher order harmonics that'll make anything from an excessively "edgey" or "aggressive" sound to something really annoying. You will definitely be needing some gNFB to tame that.
Figuring it really is NBD:
Avcl= Avol / (1 + BAvol)
You first need to know what your open loop gain is. That would be from the input to the output at the secondary of the OPT. If you wanted 6db(v) of gNFB, then:
Avcl= 20log(Avol) - 6 (In db(v) )
Convert that back to actual gain:
Avcl= 10^(Avcl(dbv) / 20)
Then:
Avol / Avcl= 1 + BAvol
And solve for B, which will be the "gain" of the voltage divider formed from the Rf resistor connected between the OPT secondary and the cathode resistor.
When connecting this and testing for the first time, make sure you use a resistive dummy load in case Mr. Murphy wired it for positive feedback and makes the amp into a high powered square wave oscillator. Best not to use the speeks for this since it'll make onehelluva racket. If that's the case, simply switch the primary OPT leads to reverse the phasing.
Once you've gotten that straightened out, make sure you're not getting any RF or ultrasonic oscillations that'll require additional compensation.
Alrighty,
Avol = open loop gain = 36.6666667 =~37 (Measured) Does this make sense for an EL34 in pentode driven by an ECC85? I can re-measure with more imput signal if necessary
Avcl = (20 Log (37)) - 6 = 25.36
Avcl= 10^(25.36/20) = 18.54
B = 1/avcl - 1/avol
B = (1/18.54) - (1/37)= .0269???
Is this right?
where do I go from here?
I form a voltage divider between the OPT secondary, and this resistor?
Avol = open loop gain = 36.6666667 =~37 (Measured) Does this make sense for an EL34 in pentode driven by an ECC85? I can re-measure with more imput signal if necessary
Avcl = (20 Log (37)) - 6 = 25.36
Avcl= 10^(25.36/20) = 18.54
B = 1/avcl - 1/avol
B = (1/18.54) - (1/37)= .0269???
Is this right?
where do I go from here?
I form a voltage divider between the OPT secondary, and this resistor?
alexmoose said:
Yes. You lose a helluvalot of voltage across the OPT, which is working as a voltage step-down xfmr.
Yup.
Now you make a voltage divider with the input end at the OPT secondary, and the output end at the cathode resistor, which must be unypassed for this to work. Then:
0.0269= (1 X Rk) / (Rf + Rk)
Solve for Rf.
okay, I found the time and I crunched the numbers
Rf = (Rk /.0269) - Rk
Rk = 475 ohms (taken from schematic)
Rf = (475/.0269)-475 = 17.18299 Kohms or 17.2k (for 6db)
Rf = (475/.0111)-475 = 42.31779 Kohms or 42.3 k (for 3db)
How beefy of a resistor does this have to be? I have some 15k 1/4watts, and a ton of 2.2ks at any wattage you can think of.
Rf = (Rk /.0269) - Rk
Rk = 475 ohms (taken from schematic)
Rf = (475/.0269)-475 = 17.18299 Kohms or 17.2k (for 6db)
Rf = (475/.0111)-475 = 42.31779 Kohms or 42.3 k (for 3db)
How beefy of a resistor does this have to be? I have some 15k 1/4watts, and a ton of 2.2ks at any wattage you can think of.
alexmoose said:
Yes, adding 6db(v) of gNFB will cut the open loop gain in half, just like you calculated before.
The reason for paralleling the feedback resistor with a capacitor is to enhance high frequency stability by rolling off the high end gain. So far, I haven't found this necessary since I'm using RF coax to lead the feedback from the output back to the input, and its inherent capacitance may be doing the compensating. This is one thing you have to determine by o'scoping to see if there's any hint of high frequency oscillation or damped oscillations.
Usually, you see this where the feedback factors are a good deal higher.
Calculate the power dissipated in the resistor with the usual formula: P= V^2 / R and get a resistor at least twice that power rating.
I wired it up for 3db, checked all my wiring, and gave it a go. Due to the legitimate concern that it would convert from an amplifier to a "high efficiency" square wave oscillator, I wired it up to my old junky speakers I found in a garage. The amplifier warmed up, and on the first try delivered some Pentode tunes with lower distortion, and improved linearity (qualitative observations made on a $10 pair of speakers). I checked the operating point of the tube, and everything seemed okay.
When I plugged the pentodes into my Fostex bi-poles, I noticed that the sound was just plain terrible, distortion was still high, sibilants sounded like sandpaper.
I decicded to remove the crossover cap I use two identical drivers in series (FE127) with a capacitor in between them to shunt high frequencies to ground, this way, all of the high frequencies goes into the front driver, while the baffle step correcting low frequencies get spit out both drivers. This makes the speaker itself behave somewhere between an 8 and 16 ohm load.
I removed the capacitor, and wired the drivers in true series with each other, and powered up. I hear music for a few seconds, then a putrid oscillation somewhere between 15 and 20khz. What a vile noise that makes!
so I figure that the primary impedance is too high (5k) making the amplifier oscillate, when I wire one half of my bi-poles (8 ohms) there is no oscillation, however there is no baffle-step correction either!
do I dare wire the drivers on each speaker in parallel (4 ohms) dropping the primary impedance to 1250 ohms? or is that too low?
I'm guessing I'm just going to re-wire it for 2500ohms on a 16ohm load. will that work?
-Moose
When I plugged the pentodes into my Fostex bi-poles, I noticed that the sound was just plain terrible, distortion was still high, sibilants sounded like sandpaper.
I decicded to remove the crossover cap I use two identical drivers in series (FE127) with a capacitor in between them to shunt high frequencies to ground, this way, all of the high frequencies goes into the front driver, while the baffle step correcting low frequencies get spit out both drivers. This makes the speaker itself behave somewhere between an 8 and 16 ohm load.
I removed the capacitor, and wired the drivers in true series with each other, and powered up. I hear music for a few seconds, then a putrid oscillation somewhere between 15 and 20khz. What a vile noise that makes!
so I figure that the primary impedance is too high (5k) making the amplifier oscillate, when I wire one half of my bi-poles (8 ohms) there is no oscillation, however there is no baffle-step correction either!
do I dare wire the drivers on each speaker in parallel (4 ohms) dropping the primary impedance to 1250 ohms? or is that too low?
I'm guessing I'm just going to re-wire it for 2500ohms on a 16ohm load. will that work?
-Moose
^^^^
Those Fostex drivers seem to be doing something squirrelly with the phase stability. In cases like this, time to get out the o'scope and signal generator and find out what the problem frequency is. That 15 -- 20KHz probably isn't it, and that is most likely a heterodyne of a much higher frequency. Ultrasonic oscillation probably also explains why the thing sounded "off" in the first place.
The other thing to consider is that the feedback be taken from the same OPT tap that the speeks are connected to. Frequently, it's connected to the highest tap even if it isn't being used, and that causes all sorts of problems.
You could tap off a smaller plate load at the expense of worse linearity to load down the parasitic circuit so that it can't oscillate, but that's more of a ghetto method. The real problem isn't really solved.
Those Fostex drivers seem to be doing something squirrelly with the phase stability. In cases like this, time to get out the o'scope and signal generator and find out what the problem frequency is. That 15 -- 20KHz probably isn't it, and that is most likely a heterodyne of a much higher frequency. Ultrasonic oscillation probably also explains why the thing sounded "off" in the first place.
The other thing to consider is that the feedback be taken from the same OPT tap that the speeks are connected to. Frequently, it's connected to the highest tap even if it isn't being used, and that causes all sorts of problems.
You could tap off a smaller plate load at the expense of worse linearity to load down the parasitic circuit so that it can't oscillate, but that's more of a ghetto method. The real problem isn't really solved.
That makes sense, I took a measurement, when the amplifier was working "fine" (junky speakers) and Measured a Gain of 100, nothing screams "oscillation" like a gain of 100 from a pentode amplifier with 3db of Negative feedback....
Now that I think about it, the capacitor gave the amp an 8ohm load way up there (the oscillation frequency) which kept it out of the audible range, when the load changed to 16ohms, the oscillation frequency got pushed down into the audible range.
I have no o'scope, but I have a frequency Gen, do I see what freq makes the amplifier loose stability? or do I have to see what frequency the amp is producing when it looses stability?. In any event, I'll start poking around, and see if I can find some other culprit. now that I know its not the 5k on the tube, I'll poke around.
-Moose
Now that I think about it, the capacitor gave the amp an 8ohm load way up there (the oscillation frequency) which kept it out of the audible range, when the load changed to 16ohms, the oscillation frequency got pushed down into the audible range.
I have no o'scope, but I have a frequency Gen, do I see what freq makes the amplifier loose stability? or do I have to see what frequency the amp is producing when it looses stability?. In any event, I'll start poking around, and see if I can find some other culprit. now that I know its not the 5k on the tube, I'll poke around.
-Moose
This sounds suspiciously like a case needing a Zobel network.
Probably something like 22 ohms and .47uf in series... that will probably take care of any impedance rise from the speakers. Wire it across the amp secondary...
You can experiment with other values... just keep the resistance at or above the impedance of the speaker load, and I would think the appropriate cap value would be for equal Z to the resistor, at about 10KHz-20KHz or so (you can use a first-order crossover calculator to find this- just plug in the resistor value into where it asks for "tweeter impedance", and the frequency you want (somewhere in the range above, or maybe higher or lower, depending on what works best)...
Regards,
Gordon.
Probably something like 22 ohms and .47uf in series... that will probably take care of any impedance rise from the speakers. Wire it across the amp secondary...
You can experiment with other values... just keep the resistance at or above the impedance of the speaker load, and I would think the appropriate cap value would be for equal Z to the resistor, at about 10KHz-20KHz or so (you can use a first-order crossover calculator to find this- just plug in the resistor value into where it asks for "tweeter impedance", and the frequency you want (somewhere in the range above, or maybe higher or lower, depending on what works best)...
Regards,
Gordon.
alexmoose said:
I don't think so. Over at DiY Tube someone was having the same problem. This time with a commercial amp and Aspen full range speeks. For whatever reason, those FR drivers seem to be a very difficult load for any VT amp that incorporates gNFB. Alas, no solutions over there either.
You might try the Zobel on the primary side of the OPT. Make R equal to the primary load, and Xc= R at 20KHz or so and see what that does in terms of stability and sonics.
If Global NFB isn't making it happy, Perhaps I should attempt Local? maybe some Plate to grid feedback? i've been dying to try it after the high praise George (Tubelab) gave it, I'm just afraid to go it alone on implementing it.
Perhaps Cathode feedback would help? I used to use it on my JBL e60s to keep the bass under control (not a problem when the amp is crossed at 100hz), but perhaps it will allow help to tame the highs as well (the curse of the Midrange shout).
That Oscillation noise haunts my dreams!
-Moose
Perhaps Cathode feedback would help? I used to use it on my JBL e60s to keep the bass under control (not a problem when the amp is crossed at 100hz), but perhaps it will allow help to tame the highs as well (the curse of the Midrange shout).
That Oscillation noise haunts my dreams!
-Moose
can you use direct coupled plate to plate feedback as in the RH amplifiers?. Should be plenty easy, but check with Mr TubeLab first.
http://www.tubeaudio.8m.com/RH84/rh84.html
Make sure to use as short a path as possible
http://www.tubeaudio.8m.com/RH84/rh84.html
Make sure to use as short a path as possible
If your speaking in regards to Pentodes high output impedance. I have actually had really good results with Pentodes driving Fostex. I used the Pentode, Pentode wired with no feedback. It behaves like a current source amplifier. You can then tweak the frequency response by altering the impedance plot. This is done by shunting the driver with various passive components.
The driver I used was FE167E. The end result sounded better than any of my voltage source amplifiers pushing the same speakers. Since your speakers are Bi-Poles the bass boosting characteristic may or may/not be beneficial to you. Shortly after experimenting with current source amplifiers I found that Nelson Pass did some similar testing
http://www.passdiy.com/pdf/cs-amps-speakers.pdf
My approach was slightly different. I can explain if anyone wants to experiment with it.
The driver I used was FE167E. The end result sounded better than any of my voltage source amplifiers pushing the same speakers. Since your speakers are Bi-Poles the bass boosting characteristic may or may/not be beneficial to you. Shortly after experimenting with current source amplifiers I found that Nelson Pass did some similar testing
http://www.passdiy.com/pdf/cs-amps-speakers.pdf
My approach was slightly different. I can explain if anyone wants to experiment with it.
hey Jeb-D,
Sounds like we have much to discuss! What kind of enclosure were you using to drive the 167e? and how did that affect your end result
I would love to hear how your methods differed from Mr. Pass's, as well, what kind of Pentode amplifier did you you use? push-pull? or Single ended?
Sounds like we have much to discuss! What kind of enclosure were you using to drive the 167e? and how did that affect your end result
I would love to hear how your methods differed from Mr. Pass's, as well, what kind of Pentode amplifier did you you use? push-pull? or Single ended?
It is a little push-pull amp, but the principal should work well with single ended too. The FE167E's are in a simple bass reflex.
We all know that full rangers dominate in the mid-range. The main idea of adapting the driver for a current source amplifier is to flatten the driver impedance, but not completely. Since the drivers impedance increases in the low and high frequencies (the two areas full rangers fall short). We can get the amplifier to pump out extra juice in those regions.
I didn't find the notch filter that Nelson used, necessary for the FE167E. I used a Zobel instead to flatten out the impedance rise in the upper frequencies. For the Zobel capacitor, 6.8uF seemed to work well for this driver. You can adjust the "treble" by adjusting the resistor value in the Zobel. Don't use a resistor that is smaller in value than that of the drivers nominal impedance. You can go higher in value, but not lower. This Zobel made the largest difference out of all the circuits tried. It sounded horribly harsh before this was added.
The circuit I used to flatten the drivers bass resonance was the same as Nelson's, except with a capacitor in series to bring the impedance of the network back up below the resonant frequency. I have not got to do any with/without cap comparisons yet. The "Bass" can be adjusted with the resistor value.
(This part not recommended for bipoles)
I also tried adding 3dB of Baffle-step compensation (put the BSC circuit before the amp), and tuned the driver impedance so the bass was +3dB on top of that (+6dB total). So, the bass was getting full-space compensation, and the lower mid was getting somewhere in between full and half space compensation. I found this approach to give the better sonics at some listening levels and speaker placement, but worse in some. If the speakers are a few feet from the wall, I'd recommend this(but not for bipoles), but if it's fairly close to the wall then I think this circuit won't be worth the trouble. Just tweaking with the two resistors of the circuits listed above will produce good results.
The good thing about these type of circuits is that they are all in parallel with the driver, so the speakers can still be driven just fine by a voltage source amp
We all know that full rangers dominate in the mid-range. The main idea of adapting the driver for a current source amplifier is to flatten the driver impedance, but not completely. Since the drivers impedance increases in the low and high frequencies (the two areas full rangers fall short). We can get the amplifier to pump out extra juice in those regions.
I didn't find the notch filter that Nelson used, necessary for the FE167E. I used a Zobel instead to flatten out the impedance rise in the upper frequencies. For the Zobel capacitor, 6.8uF seemed to work well for this driver. You can adjust the "treble" by adjusting the resistor value in the Zobel. Don't use a resistor that is smaller in value than that of the drivers nominal impedance. You can go higher in value, but not lower. This Zobel made the largest difference out of all the circuits tried. It sounded horribly harsh before this was added.
The circuit I used to flatten the drivers bass resonance was the same as Nelson's, except with a capacitor in series to bring the impedance of the network back up below the resonant frequency. I have not got to do any with/without cap comparisons yet. The "Bass" can be adjusted with the resistor value.
(This part not recommended for bipoles)
I also tried adding 3dB of Baffle-step compensation (put the BSC circuit before the amp), and tuned the driver impedance so the bass was +3dB on top of that (+6dB total). So, the bass was getting full-space compensation, and the lower mid was getting somewhere in between full and half space compensation. I found this approach to give the better sonics at some listening levels and speaker placement, but worse in some. If the speakers are a few feet from the wall, I'd recommend this(but not for bipoles), but if it's fairly close to the wall then I think this circuit won't be worth the trouble. Just tweaking with the two resistors of the circuits listed above will produce good results.
The good thing about these type of circuits is that they are all in parallel with the driver, so the speakers can still be driven just fine by a voltage source amp
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.