Considering a differential stage in a push-pull amp consisting of two triodes, is there anything problematic about using current sources as plate loads? I realize that there is huge gain by using one current sink under the cathodes, but I'm running out of room for a negative voltage supply, so I figured this is about the best I can get. I plan to implement something close to Pete Millet's generic driver schematic, but have a regulated supply for each stage with a current source on all the plates (250v regulated for the first stage, 350 for the second).
This setup will be driving basically the equivalent of four EL84's in DPP.
TIA,
-Paul
This setup will be driving basically the equivalent of four EL84's in DPP.
TIA,
-Paul
Paul,
If you use Mullard style topology, you can put an IXYS 10M45S under the LTP's tail, without having to use a negative rail.
EL84s don't need much, in the way of drive voltage. So, a 5687 as the LTP and a 7AF7 section as the voltage gain device rate to be satisfactory. High gm small signal types provide resistance against loop NFB induced slew limiting.
If you use Mullard style topology, you can put an IXYS 10M45S under the LTP's tail, without having to use a negative rail.
EL84s don't need much, in the way of drive voltage. So, a 5687 as the LTP and a 7AF7 section as the voltage gain device rate to be satisfactory. High gm small signal types provide resistance against loop NFB induced slew limiting.
As I undestand it you won't convert a push pull stage to a differential stage by using plate load CCS's. You need to have the cathodes tied to one current source to force the tubes to run at the same current.
You do know that you can't run other than full Class A1 like this? I suspect you may get fairly hard clipping if using a differential output stage. That said Allen Wright's differential power amp designs get unanimously great press.
Regards, Rob.
You do know that you can't run other than full Class A1 like this? I suspect you may get fairly hard clipping if using a differential output stage. That said Allen Wright's differential power amp designs get unanimously great press.
Regards, Rob.
Paul,
I was investigating the same thing you are. I recently got Morgan Jones book, which is a great reference, and the only way to have a "practical" CCS in a differential pair plate circuit is to use a current mirror CCS. If you don't use this type of CCS you will be constantly fiddling and matching the currents between the two individual plate circuit CCS.
I'm in the thinking stage of a PP amp with a differential pair driving 4 EL84s and 2 6V6s. I pretty much decided decided against putting the current mirror CCS in the plate stage of the differential pair and instead going for just the single cathode CCS. The previous advice about not needing much drive voltage for EL84 is right and if you need more than what you are currently getting just go for a "single" voltage amplification stage that uses a tube with an MU of 30 or better. It is better design to get the right tube for the job and eliminate needless stages.
If you do decide to add a separate stage it should be a cathode follower and not a needless additional voltage gain stage for an EL84. A cathode follower will eliminate capacitor blocking distortion between driver/voltage gain stage and EL84s, ala Tubelab dot com.
By the way, the reason I'm adding a 6V6 to mine is because I had an amp that had that topology that sounded like a little slice of heaven. This will be an upgraded version of that one with Mosfet cathode followers for each power tube. I also plan on having selectable pentode/triode for each tube pair for ultimate fine tuning of the sound.
I was investigating the same thing you are. I recently got Morgan Jones book, which is a great reference, and the only way to have a "practical" CCS in a differential pair plate circuit is to use a current mirror CCS. If you don't use this type of CCS you will be constantly fiddling and matching the currents between the two individual plate circuit CCS.
I'm in the thinking stage of a PP amp with a differential pair driving 4 EL84s and 2 6V6s. I pretty much decided decided against putting the current mirror CCS in the plate stage of the differential pair and instead going for just the single cathode CCS. The previous advice about not needing much drive voltage for EL84 is right and if you need more than what you are currently getting just go for a "single" voltage amplification stage that uses a tube with an MU of 30 or better. It is better design to get the right tube for the job and eliminate needless stages.
If you do decide to add a separate stage it should be a cathode follower and not a needless additional voltage gain stage for an EL84. A cathode follower will eliminate capacitor blocking distortion between driver/voltage gain stage and EL84s, ala Tubelab dot com.
By the way, the reason I'm adding a 6V6 to mine is because I had an amp that had that topology that sounded like a little slice of heaven. This will be an upgraded version of that one with Mosfet cathode followers for each power tube. I also plan on having selectable pentode/triode for each tube pair for ultimate fine tuning of the sound.
This may or may not have some information you are looking for.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=115756&highlight=
http://www.diyaudio.com/forums/showthread.php?s=&threadid=115756&highlight=
I think that answers it. The crude vs. strong current source discussion lends itself to what I'm doing. If I set up a common cathode 12AU7 diff stage with a resistor under the cathodes and current sources on each plate, the current sources should be happy enough. I suppose I could always split the cathodes apart too...
audiowize said:
Paul,
If you use a Schmitt circuit with both CG1 and CG2 connected to the I/P trafo, replacing RK2 by a CCS achieves your purpose. Select the value of RK1 to yield the appropriate grid bias voltage.
It's a full blown differential gain block. 350 V. should be enough to run a number of tube types. The 5965, with IB = 3.5 mA./section and 27 KOhm load resistors, looks good to me. Under 100 V. get dropped in the load resistors. So, there is LOTS of compliance for both the tube and the CCS, as approx. 30 V. are consumed by some popular cascoded CCS circuits.
Paul,
I think you may be overobsessing on the merits of a flat load line with a CCS in the plate circuit. Yes, those merits are there, plus you also get a value of gain that is much closer to intrinic gain (mu) of the driver tube with the CCS in the plate circuits. I think this could be important if you needed a high input voltage and/or current requirement of the grid of the power tubes after it. But the fact of the matter is the EL84 is normally biased at -11V and you need only 22 volts for full output. And because they are pentodes, even if there are two of them to drive, it doesn't require a lot from the driver.
Being very conservative and not using plate CCS means a differential pair will provide probably 2/3 of the intrinsic MU of the driver tube, rather than 9/10 with plate CCS loads. A normal amp input using CD source voltage will be 2.8 V pk-pk per tube on each driver input. If you want to have 6db of surplus gain driving the output requires 44 volts pk-pk at the output of the tubes. Using the conservative value of only 2/3 efficiency using cathode CCS requires a tube with an mu of 23.5. To get a little more reserve you can find any number of fine tubes, like the ECC88, 5842, 6s45p etc, each of which have over 30 mu, that have very low distortion with excellent GM and low output impedance. Believe me, you won't have a problem with high distortion with any of these driver tubes, even without plate load CCS. And if you are still concerned you will have enough reserve driving power to tame it with feedback, local or otherwise. The low drive requirements for an EL84 are a blessing so don't obsess too much about it.
I think you may be overobsessing on the merits of a flat load line with a CCS in the plate circuit. Yes, those merits are there, plus you also get a value of gain that is much closer to intrinic gain (mu) of the driver tube with the CCS in the plate circuits. I think this could be important if you needed a high input voltage and/or current requirement of the grid of the power tubes after it. But the fact of the matter is the EL84 is normally biased at -11V and you need only 22 volts for full output. And because they are pentodes, even if there are two of them to drive, it doesn't require a lot from the driver.
Being very conservative and not using plate CCS means a differential pair will provide probably 2/3 of the intrinsic MU of the driver tube, rather than 9/10 with plate CCS loads. A normal amp input using CD source voltage will be 2.8 V pk-pk per tube on each driver input. If you want to have 6db of surplus gain driving the output requires 44 volts pk-pk at the output of the tubes. Using the conservative value of only 2/3 efficiency using cathode CCS requires a tube with an mu of 23.5. To get a little more reserve you can find any number of fine tubes, like the ECC88, 5842, 6s45p etc, each of which have over 30 mu, that have very low distortion with excellent GM and low output impedance. Believe me, you won't have a problem with high distortion with any of these driver tubes, even without plate load CCS. And if you are still concerned you will have enough reserve driving power to tame it with feedback, local or otherwise. The low drive requirements for an EL84 are a blessing so don't obsess too much about it.
This theme was spoken about quite often here at diyaudio and I have seen positive comments from people as 316a, 7N7, EC8010 maybe.
The trick is to make the plate CCS's of a much lower impedance than the tail CCS, for example plain simple MJE350 at the plates and a cascoded CCS at the tail. 316a used 2M2 resistors parallel to the the MJE350. I imagine that single DN2540's or 10M45S at the plates would work as well, but fiddlier to adjust for same current.
Are you referring to page 136? That is not a current mirror... that schematic of page 136 led to some discussion here. SY sent an e-mail to MJ, who replied and SY posted the answer in one of those threads as well. 7N7 applied the technique to a PP driver stage and measured excellent performance, so it could be a keeper! At least it is something I want to try out, once...
Erik
The trick is to make the plate CCS's of a much lower impedance than the tail CCS, for example plain simple MJE350 at the plates and a cascoded CCS at the tail. 316a used 2M2 resistors parallel to the the MJE350. I imagine that single DN2540's or 10M45S at the plates would work as well, but fiddlier to adjust for same current.
Are you referring to page 136? That is not a current mirror... that schematic of page 136 led to some discussion here. SY sent an e-mail to MJ, who replied and SY posted the answer in one of those threads as well. 7N7 applied the technique to a PP driver stage and measured excellent performance, so it could be a keeper! At least it is something I want to try out, once...
Erik
Hi Wavebourn
I remember this active load appearing at earlier discussions. It was suggested that it could be used quite simply in a differential pair, with a cascoded CCS on the tail, something like adjusting the tail current and the plate currents would set automatically at the desired plate voltages. I spent some time thinkering about it, but eventually gave up, I could not grasp how that one would work. Unfortunately I am not able to find that thread, so the above may all sound vague, but if it does ring a bell, could you explain it a bit further, as I am very interested.
Erik
I remember this active load appearing at earlier discussions. It was suggested that it could be used quite simply in a differential pair, with a cascoded CCS on the tail, something like adjusting the tail current and the plate currents would set automatically at the desired plate voltages. I spent some time thinkering about it, but eventually gave up, I could not grasp how that one would work. Unfortunately I am not able to find that thread, so the above may all sound vague, but if it does ring a bell, could you explain it a bit further, as I am very interested.
Erik
ErikdeBest said:
Hi Erik,
Yeah, that was the page. I stand corrected. I'd be interested in more information also and which thread Sy talked about it. The only thing I wish MJ had done differently was to use FETs in all his implementations because I think they are preferable to BJTs. Unfortunately I don't have enough expertise to make a direct translation to FETs. For this particular case the Ixcy CCS won't work, at least I don't see how it could be made to work that wouldn't be more trouble than its worth. Anyone have any ideas how to implement the page 136 picture with FETs?
One other thing - the discussion of the complexities of having a CCS at both ends is kind of unsettling to me - I'm just getting into this. Personally I prefer having a CCS(s) on one end only. However, if you decide to have them on the plate side it seems to me you would want to use a low output resistance tube like a 5842 for best results. The lower the resistance of the plate the better able the current matched CCSs would be able to divide the current equally and eliminate the need for the cathode CCS. Am I thinking about that wrong? Also is there an equivalent quality tube to the 5842, i.e. low output impedance, but with MU only around 20-30? The gain for my needs would just be too much with the 5842 with plate CCS driving EL84/6v6.
Hi Exeric
In my view that is not a current mirror, MJ is only using the same reference string for both MJE350. If you change one of the 16k resistors for lower/higher value you will more/less current in that leg, therefore imbalance. They will not track currents as a current mirror does.
All threads I could find referring to the theme
http://www.diyaudio.com/forums/showthread.php?s=&threadid=111099
http://www.diyaudio.com/forums/showthread.php?s=&threadid=115756&highlight=
I could not find the thread I referred to earlier, but both threads above may help you a lot as well
Erik
In my view that is not a current mirror, MJ is only using the same reference string for both MJE350. If you change one of the 16k resistors for lower/higher value you will more/less current in that leg, therefore imbalance. They will not track currents as a current mirror does.
All threads I could find referring to the theme
http://www.diyaudio.com/forums/showthread.php?s=&threadid=111099
http://www.diyaudio.com/forums/showthread.php?s=&threadid=115756&highlight=
I could not find the thread I referred to earlier, but both threads above may help you a lot as well
Erik
ErikdeBest said:
Hi Erik;
The lower is the voltage on anode, the higher is current through R2 and R3, the higher is voltage drop on R2, and as soon as it is enough to break-down PN junction B-E of the upper transistor, it starts conducting C-E current. The seconde (cascoded) transistor repeats thius current increasing dynamic resistance.
So, a voltage on anode will be always defined by R2/(R2+R3) ratio. while current will be adjusted accordingly.
It looks like an ideal (in terms of distortions) choke (Gyrator, i.e. transformer of a capacitive impedance to an inductive one)
Currently I use JFET instead of an upper transistor, so R2 and R3 may have a huge value.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.