| Saurav |
Option 1:
| code: |
o B+
|
|
|
3||E----o
3||E
3||E
3||E
3||E
3||E----o
|
|
_____
/ \
o---------| 2A3 |
| \_____/
| |
| |____
\ \ |
/ / |
\ R R \ = C
/ / |
| | |
|__________|___|
|
|
O CCS
O
|
|
V
|
Please forgive the ASCII art, I don't have a good drawing program here. Anyway, that's an idea for using a CCS in a 2A3 (or any other) output stage. I left the cathode bias resistor in there because I don't have any other way of making a 45V bias supply. The grid leak should probably be referenced to the bottom of the cathode bias resistor, otherwise the bias will include the drop across the CCS, which we probably don't want, right (unless we calculate for it). I'm not sure if the bypass cap should bypass only the cathode resistor, or should go straight to ground - do we want the AC going through the CCS? Probably not.
Anyway... those IXYS chips should be able to handle 60mA easily, so is there any reason why this is a bad idea? Has anyone tried anything like this? |
|
|
| Saurav |
Option 2 would be to move the CCS up above the OPT, i.e. between the OPT and the B+ point. The PS cap is still connected to the B+ point (i.e. before the CCS). Putting the cap after the CCS makes the CCS part of the power supply, and the cap will feed the 2A3 so it won't really be a CCS any more (but it'll be a nicely regulated PS).
| code: |
o B+
|______________
| |
O CCS |
O = PSU C
| |
3||E----o V
3||E
3||E
3||E
3||E
3||E----o
|
|
_____
/ \
----------| 2A3 |
| \_____/
| |
| |____
\ \ |
/ / |
\ R R \ = C
/ / |
| | |
|__________|___|
|
|
|
V
|
|
|
|
| Sch3mat1c |
| quote: | Originally posted by Saurav
Option 1:
so is there any reason why this is a bad idea? Has anyone tried anything like this? |
For one, the CCS will try to maintain constant cathode current. This means that cathode voltage will exactly follow grid voltage changes. (It won't allow any change in the G-K bias voltage, because that would cause a change in current based on the Gm of the tube.) And also, since Ip = Ik, and delta I = 0, you will get exactly zero power into your load resistance.
And finally, the tube's plate resistance is multiplied by the CCS, making it a very good CCS for the OPT. But that of course is useless...
In the second case you have the same effect, though not as significant because the tube isn't acting as an error amplifier. I'm sure the CCS is still quite good though, so you'll have the same net result.
From your circuits, I don't see what your goal was. If you did bypass the CCS with a cap, you would attain constant bias no matter what tube is in place. (You can omit the bias resistor because the CCS pulls the current to - 60mA you said? - it adjusts its terminal voltage until whatever around it puts 60mA through.)
If stable bias is the case, then it will be mu+1 (I think) times more effective in the cathode than the plate circuit, it will provide the necessary bias voltage, and will not waste plate voltage.
If your goal was to reduce distortion, well distortion is a fact of life and a side effect of producing power. Only way around it is to cancel it, as with NFB or a PP stage.
Tim |
|
|
| janneman |
The second option would work if there was a way for the varying anode current back to the psu. If you remove the transformer (replace by short) and take the signal off the anode (via a cap if necessary) with a load resistor to gnd or B+, that would work quite well.
Jan Didden |
|
|
| Saurav |
| quote: | | you will get exactly zero power into your load resistance |
OK, that would be slightly counter-productive :)
So does no one use CCS's under cathodes? And how does a CCS work as a plate load for an input/driver tube? Even there it would try to hold a constant current through the tube.... OK, so the current changes due to the signal show up across the load (i.e. grid leak of the following stage), so it works because the CCS is in parallel with the load.
If that is correct, then I need to find a way to put the CCS in parallel with the OPT. Which would probably end up looking like a parafeed circuit, with the CCS replacing the choke, and the OPT connected through a cap? |
|
|
| Saurav |
| quote: | | From your circuits, I don't see what your goal was. |
I'm not sure, actually. It was just an idea that popped into my head, so I thought I'd throw it out for comments. That, and a vague idea of trying to take the advantages that a CCS brings to an input/driver stage and apply them to an output stage. And my amp has a single ended 2A3, so that's what I was thinking about.
| quote: | | If you remove the transformer (replace by short) and take the signal off the anode (via a cap if necessary) with a load resistor to gnd or B+, that would work quite well. |
But then I won't be able to drive an 8 ohm speaker, will I? So it comes back to putting the transformer in parallel with the CCS, and connected through a cap like you said.
Or am I still missing something? |
|
|
| dhaen |
| quote: | Originally posted by Saurav
OK, that would be slightly counter-productive :)
So does no one use CCS's under cathodes? ..snip... |
7N7 used one in his "Instock" design on this thread":
http://www.diyaudio.com/forums/show...0103&highlight=
It's on a Long Tailed Pair, so maybe it doesn't count.
Cheers, |
|
|
| Sch3mat1c |
Yeah, LTP is a balanced system. You can draw actual power from that because one side's current can go up, the other's just has to go down to compensate. The purpose of a CCS here is just that, to balance it.
You can use a CCS in a PP class A amp.
Saurav: yes you'd have to go parafeed. The disadvantage is, you have to run approx. 1/2 B+ across the CCS, so that the ideal maximum voltage range will be 0V to B+, peak to peak. With an inductor, its flyback effect allows peaks from 0V to 2B+, or even more if the duty cycle is odd (say, 0V for 90% of the time would balance with like 20B+ on the flyback peak! :bigeyes: ). So you only lose something by not going through the transformer.
Tim |
|
|
| Saurav |
| quote: | | The disadvantage is, you have to run approx. 1/2 B+ across the CCS |
Yup, I realized that after I typed out that post. So that's roughly equal dissipation in the tube and in the CCS, i.e. ~ 15W for a 2A3. Which is quite a bit of heat to dissipate in an IC.
OK, here's another idea. Assuming you had the B+ to spare, you could replace the parafeed choke with a CCS + R. The R would be in series with the CCS so wouldn't hurt the high impedance, and would reduce the amount of power dissipated in the CCS.
This is starting to look like it's more trouble than it's worth though, like a brute force way of adding a CCS to the circuit just for the heck of it. Oh well, it was fun to think about :) |
|
|
| G |
Why don't you just use the active load on the plate and use a coupling cap to connect it to an output transformer? In other words just Parafeed it. You could get away with using a PP output transformer since there would be no DC in the primary. Just forget about the center tap a wire it up.
G |
|
|
| Saurav |
| Because of the power dissipation required in the CCS to do that. As Tim explained, you'd pretty much need to double your B+ supply in order to allow the plate to swing the same voltage that it could swing with the choke/transformer load (I think). Since you still want roughly B+ at the plate, that means you need to drop B+'s worth of voltage across the CCS, which combined with the current of the output stage, comes to quite a bit. In fact, the CCS dissipation will be roughly equal to the tube's own dissipation, I think. For a 2A3 at 'classic' operating points, that's 15W, which is a lot for a little IC. |
|
|
| Sch3mat1c |
I had assumed your CCS was something a bit beefier, like a TO-220 power transistor/MOSFET? If it's a generic DIP package, that wouldn't seem like it could handle much over 100V or 200mA...
Tim |
|
|
| Saurav |
Well, it probably is in a TO-220 package, but 15W would still require some serious heatsinks, wouldn't it?
OK, it is a TO-220, and I was wrong, it looks like it can handle a 2A3's worth of power.
Here's the only person I know of who's actually tested this device:
http://www.pmillett.addr.com/current_source.htm
Fdegrove has mentioned them here in the past, I think, but I don't know if he's tried them.
OK, so to take this idea further. I'd need a high quality cap for the parafeed, obviously. And the whole thing is a little pointless unless I use a good parafeed OPT. I currently have a pretty decent (for the price) air-gapped SE OPT, the One Electron UBT-3. Would this benefit from the removal of DC across it? Or does an air-gapped OPT perform worse when you remove the DC across it? |
|
|
| fdegrove |
Hi,
| quote: | | Would this benefit from the removal of DC across it? Or does an air-gapped OPT perform worse when you remove the DC across it? |
No. The OPT would perform just fine used like that, it doesn't depend on any DC being present.
Cheers,;) |
|
|
| G |
I could be wrong( I have been a time or two ) but I'm a little sceptical about these so called "Parafeed output transformers" that look like they came out of an old magnavox console and sell for $300 - $700 a pair. I can understand the need for an exceptional choke for parafeed but for $300 I would like to see something that looks a little more finished. Mr. Sowter across the pond makes parafeed chokes and output transformers that look nice and are still less expensive than that other manufacturers even with the duties and shipping thrown in.
G |
|
|
| dhaen |
| quote: | Originally posted by G
snip.. Mr. Sowter across the pond makes parafeed chokes and output transformers that look nice and are still less expensive than that other manufacturers even with the duties and shipping thrown in.
G |
If you are happy with the price of Brian Sowter's transformers, then look no further. They are very good indeed.
Remember that he only has parafeed transformers up to 5W. I presume (but don't know for sure) that they are useable up to full power without too much falloff of performance. |
|
|
| Saurav |
Agreed, I don't think I'd spend money on the "big dogs" for my little amp. Probably something along the lines of Lundahl or Sowter, I've heard good things about both.
Does Sowter have a US distributor? |
|
|
| janneman |
| quote: | Originally posted by Saurav
OK, that would be slightly counter-productive :)
So does no one use CCS's under cathodes? And how does a CCS work as a plate load for an input/driver tube? Even there it would try to hold a constant current through the tube.... OK, so the current changes due to the signal show up across the load (i.e. grid leak of the following stage), so it works because the CCS is in parallel with the load.
If that is correct, then I need to find a way to put the CCS in parallel with the OPT. Which would probably end up looking like a parafeed circuit, with the CCS replacing the choke, and the OPT connected through a cap? |
It may help the insight if you think of the CCS as a very high plate resistor (say 1GigaOhms) connected to a very high B+ (say 10kV). Then it is just any other circuit and, in my view, much easier to analyse.
Jan Didden |
|
|
| Saurav |
Thanks, I should have checked the website before I asked. I'll probably have to write to IAG, as parts of their website seem to be "under construction", and have been for a while.
Anyway... I doubt I'll be spending money on new OPTs any time soon. If I try this CCS parafeed thing, I'll probably try it with my current OPTs first.
Jan, thanks for that insight, that does help visualise what's going on. |
|
|
| DrG |
To get back to the original circuit query: there IS an advantage to placing a CCS in the cathode circuit. Tube mA bias can be forced. In a PP design balance can be forced, eliminating all possibility of imbalance and DC in o/p transformer.
Naturally the CCS requires capacitative by-passing as with a regular cathode bias resistor... |
|
|
| Saurav |
Resurrecting this from a long time ago:
| quote: | | Naturally the CCS requires capacitative by-passing as with a regular cathode bias resistor... |
So if I replaced the cathode bias resistor with a CCS and kept the bypass cap as it is (so it bypasses the whole CCS), would that work? Can the circuit now be viewed as a DC component which goes through the CCS and an AC component which goes through the cap? Or will the CCS still hold Vg-k to be constant and cause zero power output?
So the circuit would now look like this:
| code: |
o B+
|
|
|
3||E----o
3||E
3||E
3||E
3||E
3||E----o
|
|
_____
/ \
o---------| 2A3 |
| \_____/
| |
| |____
\ | |
/ | |
\ R CCS O = C
/ O |
| | |
|__________|___|
|
|
V
|
|
|
|
| Sch3mat1c |
The CCS will keep current constant (duh), restricting it to class A. If the tube tries turning on or off more it will counter the effect. With the cap bypass, this does not apply to AC signals and you get full gain, and power output.
Tim |
|
|
| Saurav |
Thanks. I was hoping I got it right this time.
So does that mean it'll clip harder when pushed to the limits of class A operation, or something like that? It's single ended, so it shouldn't ever go out of class A, right, or I'm in pretty high distortion land.
One more question - the CCS should look like a very high impedance looking in from either direction, right? So that should change the cathode bypass capacitor requirements, meaning you should be able to get by with a smaller capacitor and still get the same cutoff frequency?
Tubecad seems to say that I only consider Rk when figuring the capacitor value, but I'm probably reading that wrong, shouldn't rp come into the picture somehow? Because if it's only Rk, and the CCS has a few Mohm resistance, I would only need a really small cap, and that doesn't sound right. |
|
|
| Sch3mat1c |
The actual factor is the tube's cathode-follower-style cathode impedance, which should be very nearly the same as the tube at the same operating point without a plate load, in parallel with the cathode resistor (which being a CCS is very high to AC), and can be ignored).
So let's see. IIRC, Zo ~= 1/Gm, or something like 200 ohms. So the cap's reactance should equal 200 ohms at the lowest frequency of interest. This is something like 40uF for 20Hz, so anything above 100uF will work just fine.
BTW, a quirk of this system is there is *no* gain at DC - it all goes into the CCS - so the pole frequency above will indeed be the -3dB cutoff frequency of this stage. (A normal resistor-biased stage retains some grid-plate gain at F's below that of the bypass's cutoff.) Another good reason to use 220 or 470uF instead.
Tim |
|
|
| Saurav |
Hmm... I was trying to find a way to use smaller caps so I could try films instead of electrolytes :) My speakers start to roll off around 60Hz though, so I could use that to justify smaller caps, or use this cap as a filter to shape the low end response of my amp. Obviously I'd take it out if I were to give the amp to someone else, but as long as it's in my system, I think I could implement a high-pass filter in this way.
Thanks a lot for the help.
Saurav |
|
|
| fdegrove |
Hi,
| quote: | | My speakers start to roll off around 60Hz though, so I could use that to justify smaller caps, or use this cap as a filter to shape the low end response of my amp. Obviously I'd take it out if I were to give the amp to someone else, but as long as it's in my system, I think I could implement a high-pass filter in this way. |
Don't use that as an excuse or pretty son you won't have any bass left.
Try a BG for a decoupling cap and use the largest cap you can afford.
You may be surprised with what your speakers put out.
Cheers,;) |
|
|
| Sch3mat1c |
Errr, you mean he should use like ten ultracaps? WTF?
Between 100uF and 1000uF will be more than sufficient. Just ignore Frank :p :rolleyes: :rolleyes: :o :D
Tim |
|
|
| fdegrove |
Hi,
Thanks Tim...
I'll remember that, especially since it seems you don't know what it is you're talking about.
Ignorance is not a sin. I know,:rolleyes: |
|
|
| Saurav |
| quote: | Don't use that as an excuse or pretty son you won't have any bass left.
...
You may be surprised with what your speakers put out. |
Well... I do have a subwoofer. And so far I've been having a hard time trying to get it to integrate smoothly if I let my mains run full range. My room has a huge peak at around 35Hz which even my prior small bookshelf monitors excited, even though you couldn't hear a thing from 55Hz down to 40Hz. So I've finally decided to try the other option, which is to put a high pass on the mains and see if that helps. And I'm looking for the least harmful ways in which to implement that filter, i.e. no extra caps if possible.
Though I am curious about trying out a 3-way active XO (digital or analog, tube or SS) and going fully multi-amped. Use my 2A3 SET on top, use SY's "ST70 on steroids" for the woofers (until he asks for it back, and then I'll have to build something, maybe a gainclone for kicks?), and keep my sub with its 100W cheapo plate amp.
What do you think... is that an idea worth spending time on?
OK :) |
|
|
| fdegrove |
Hi,
Good to know and much less work for me...thank you, Tim.
Cheers,;) |
|
|
| Saurav |
| That was meant to be a joke. My apologies. |
|
|
| Sch3mat1c |
Ditto, although much more striking. Couldn't think of a better way to show my sarcasm...
But really Frank, wtf is "biggest you can afford"? There are relatively cheap caps these days that will literally take HOURS to stabilize on this circuit.
Tim |
|
|
| fdegrove |
Hi,
| quote: | | That was meant to be a joke. My apologies. |
You won't see me laughing for years...unless it's at both your and Tim's expense.
Apologies accepted.
I'm sure our cocky benjamin didn't mean it as a joke?
Cheers and good luck,;) |
|
|
| Saurav |
| quote: | | Cheers and good luck |
Thank you. |
|
|
| fdegrove |
Hi,
| quote: | | But really Frank, wtf is "biggest you can afford"? There are relatively cheap caps these days that will literally take HOURS to stabilize on this circuit. |
The biggest cap one can afford I had in mind was a BG cap...size matters or so women tell me....:angel:
What I meant is, better use 470µF cap than a 100µF one here.
HOURS to stablise?
Never noticed that, even with thousands of µF as a bypass cap.
Surely YMMV,;) |
|
|
| Sch3mat1c |
Hours indeed. An *electrically* large cap, big as you can afford, could be quite a few hundred, even thousand farads - although getting one in 50V sizes appropriate for this project would be difficult. At this point you obviously would need to series-parallel the ultracap modules.
Tim |
|
|
|
