I have been meaning to build an el34 or maybe 2a3 drd/free lunch amp for a while. I have most of the parts, but I really have just been too cheap to buy the plate loading chokes. So, I started thinking that I might be able to swap out the chokes for a CCS that would work about the same and cost $3 instead if $200 making it a cheap free lunch, which makes no sense. Moreover, the CCS will drop however many volts it needs, so it seems that one could do away with the resistor divider in the output tube's cathode and just let the CCS set the voltage.
So, here is a preliminary schematic awaiting comments.
So, here is a preliminary schematic awaiting comments.
An externally hosted image should be here but it was not working when we last tested it.
Don't forget the resistive divider helps set the bias for the output stage. Also, a choke load allows the plate voltage on the input tube to swing above B+; the CCS needs to have enough voltage drop across it so it doesn't drop out when the plate swings more positive. If you can get it to work it will cost more in terms of higher B+ required.
jeff mai said:
Here, the current through the output tube is set by the voltage drop across the CCS, so the resistive divider is not necessary.
mach1 said:
This doesn't strike me as a huge deal. I am more interested in the DC coupling part than the extra power part that DRD provides. It is a SET, which is hugely inefficient, so making it slightly more inefficient isn't a concern. I never run amps anywhere near their max. But, the EL34 might not be ideal. I probably need something that is biased a bit higher.
To expand on what Jeff said, at DC a choke's inductance equals zero. It's the driver's static curent through the choke's residual resistance which is used to set the output tube's bias. A CCS maintains megaohms impedance at DC, leaving no way to properly bias the output.
Edit: Bad timing. How is the voltage across a CCS controlled?
Edit: Bad timing. How is the voltage across a CCS controlled?
rdf said:
The constant current across the driver tube forces the CCS to drop the right amount of volts to maintain the voltage on the driver's plate. I'll say that a different way ... the operating point of the driver is determined by the convergence of 3 things -- the current, the bias, and the plate voltage. Usually, you think of specifying the voltage and the bias, and the current falls into line. But, there is no reason not to do it the other way. By specifying the current and the bias (which is quite specific due to the LED), the tube will set its own plate voltage, and so long as the voltage on the of of the ccs is higher than this, the ccs will drop the appropriate number of volts. This, then, sets the bias on the output tube. I think it is just a matter of finding the right combination of tubes, which I have not really done yet.
jeff mai said:
No doubt
I just thought of this while in the shower today, so I can still be convinced it is unworkable.
Here's a simple example that might be useful for pushing some headphones ...
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
You're on fire. I see what you're saying, if a little mixed up between which tube's terminals your explanation referred. As long as the voltage across the CCS doesn't fall below the minimum required to operate though it doesn't matter where its positive terminal is connected, the 5687's plate voltage is completely independent of the voltage on the other end of the CCS and therefore the output tube's operating condition. Given any tube/LED pair the driver's plate voltage will be set entirely by the CCS current. The DC interaction between driver and output cathode voltage that acts to stabilize a free-lunch is lost.
I wonder if replacing the 30K with a bypassed adjustable CCS as well would get some stability back. The 12B4's cathode voltage would then seem to be free to 'float' to a stable op point. A second cap from 12b4 cathode to ground will also be required to maintain the Ultrapath PS rejection. However the only thing then to be gained from tying the driver CCS to the output cathode is saving power supply draw.
I see what you're saying, if a little mixed up between which tube's terminals your explanation referred. As long as the voltage across the CCS doesn't fall below the minimum required to operate though it doesn't matter where its positive terminal is connected, the 5687's plate voltage is completely independent of the voltage on the other end of the CCS and therefore the output tube's operating condition. Given any tube/LED pair the driver's plate voltage will be set entirely by the CCS current. The DC interaction between driver and output cathode voltage that acts to stabilize a free-lunch is lost. I wonder if replacing the 30K with a bypassed adjustable CCS as well would get some stability back. The 12B4's cathode voltage would then seem to be free to 'float' to a stable op point. A second cap from 12b4 cathode to ground will also be required to maintain the Ultrapath PS rejection. However the only thing then to be gained from tying the driver CCS to the output cathode is saving power supply draw.
rdf said:You're on fire.
I hope this doesn't mean that building this is going to cause me to be so ...
rdf said:
I am going to need to keep studying to understand this.
rdf said:
I also thought about making the output stage parafeed with a plate CCS of it's own and a WE cap connection.
rdf said:
No small thing in the above example -- 40mA v. 70mA seems significant enough to think about (at least in as much as one of them could be built with stuff in my closet, and the other would require ordering yet another power transformer). But, the real idea was to get rid of a coupling cap. Maybe it isn't that important ...
rdf said:
I'm sensing encouragement ...
After selling off about 4 I only have 5 headphone amps left, so I could always use another
Hammond makes a few models of very high Hy chokes. The 156C is 150H at 8mA with a DCR of 3700 ohms. Obviously this will require lowering the current to the driver a bit, and probably make the LED impractical (though 8mA might be right on the border), but is there a reason not to try these as plate loading chokes? Is the DCR too high? They are only about $10.
As I think about it, in the 5687/12b4 schematic, these chokes would drop about 30V at 8mA, which would suggests that they could be used without the resistor divider more or less in place of the CCS to set the bias of the 12b4. The bias on the driver would need to be adjusted a bit too.
The data I have (limited to be fair) shows otherwise. A few nights ago, I did some measurements of a prospective driver tube for my new amp using a moderately sized plate load resistor and a series string of two LEDs. At the required 80V of swing, 2nd harmonic was by far predominant at about 0.6%; third was something like 0.03%, nothing else visible. Not too poopy, I think.
SY said:
And far better than tiny resistors shunted by very large electrolytic capacitors unless you can spring for a couple of differently sized and expensive BG. This is based on experience with D3A in a phono stage where my design called for bias voltages below that attainable by LED. The cathode bias resistor was 62 ohms, and the cap 1000uF, ultimately shunted with 10uF + 2.2uF film to deal with the really poor high frequency distortion performance of the original panasonic 1000uF el cap. (It sounded awful with any percussive sound.) The BG are much better, but much more expensive. In the future I'd change the design to allow the use of led based cathode bias over small resistors/big caps or go with fixed bias instead where practical.
I recently replaced the 1000uF cathode bypass caps on my C3G headphone amp with 440uF oscon bypassed with mica which almost drastically improved the sound quality...bass tightness, low level resolution and very delicate highs!
I'd like to try led bias in my amp since it would come way less expensive than those oscons...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.