Member
Joined 2009
Paid Member
Having finished (mostly...) a big speaker with 15" single full range driver I now want to build a new amplifier to drive it.
I believe every DIY nut should have a 2A3 SET at some point in their journey and this thread is where I get mine. I know it's a 'been there done that' kind of amplifier but the devil is in the details and I hope some of you folk with share your knowledge as I progress.
Things I have decided on:
Output tube: 2A3 [1]
Output transformer: Sowter SE05 2.5k primary into 8 ohm (a gift from my parents)
Input tube: 12AX7 [2]
Topology: Direct Coupled
[1] and maybe I will also want to try 6C4C, 6P31S...
[2] and maybe I will also want to try 6G7, 6SF5, 6SL7...
I'll work up a schematic or two.
I believe every DIY nut should have a 2A3 SET at some point in their journey and this thread is where I get mine. I know it's a 'been there done that' kind of amplifier but the devil is in the details and I hope some of you folk with share your knowledge as I progress.
Things I have decided on:
Output tube: 2A3 [1]
Output transformer: Sowter SE05 2.5k primary into 8 ohm (a gift from my parents)
Input tube: 12AX7 [2]
Topology: Direct Coupled
[1] and maybe I will also want to try 6C4C, 6P31S...
[2] and maybe I will also want to try 6G7, 6SF5, 6SL7...
I'll work up a schematic or two.
Last edited:
big? you could most peoples ENTIRE stereo setup inside it
Driving it with a teenyweeny wattage toob amp sounds like fun!
Subscribed
Member
Joined 2009
Paid Member
big? you could most peoples ENTIRE stereo setup inside it
Yes, when visitors see this thing it takes a bit of explaining
So, low wattage. How do we get big sound ? - I believe one key thing is the power supply - it has to be designed not to hold the amplifier back, when it needs juice then it should flow freely.
To this end I have been advised to design a power supply which has low-ish DCR components and good transient response. Yet simple.
Pictures, please! 
Member
Joined 2009
Paid Member
Well let's see what we got...
The speaker was designed to be free of cross-overs and other encumbrances and the amplifier should be approached the same, where it makes sense to do so.
The direct coupled approach means no transformers in the signal chain except at the output. Not saying trafo's are bad, but they are not a 'straight' wire and good one's cost real money.
Also means no capacitors. The component everyone loves to hate but we have them everywhere. In a SE Class A amp you can't fully eliminate them too easily and usually we have at least one in the power supply. This being the infamous 'last cap' which the signal current must flow through to complete it's circuit.
Electrolytics are the most unpopular because 'they' say the sound is bad, but I see a benefit in eliminating a technology that is known to have a limited life span and hence avoid having to replace them down the road. We can eliminate electrolyitics and design the amplifier to last a lifetime. To do that we need to avoid needing large caps without generating excessive ripple from the power supply. We do that using the Loftin-White topology (the secret sauce being the cap + resistor connecting the cathodes of the two types) and by using a supply with good regulation.
Key elements of the power supply owe their origins to several sources but it was a gentlemen called Jeff Medwin that put me on this particular path. This being the low-DCR and low energy storage approach. The low DCR approach is desirable because we would like to arrange for the impedance of the power supply to be much less than that of the amplifier. The 2A3 has a plate resistance of 800 ohm (nominal) so we'd like to achieve something around 1/10th of that, or at least close. We don't want to resort to large electrolyits remember and simplicity means no heat generating shunt regulators here. This favours SS rectifiers and a big power trafo. The filter design also optimizes for high regulation and good transient response. If we size the caps well, we may be able to do the whole thing without electrolytics.
The speaker was designed to be free of cross-overs and other encumbrances and the amplifier should be approached the same, where it makes sense to do so.
The direct coupled approach means no transformers in the signal chain except at the output. Not saying trafo's are bad, but they are not a 'straight' wire and good one's cost real money.
Also means no capacitors. The component everyone loves to hate but we have them everywhere. In a SE Class A amp you can't fully eliminate them too easily and usually we have at least one in the power supply. This being the infamous 'last cap' which the signal current must flow through to complete it's circuit.
Electrolytics are the most unpopular because 'they' say the sound is bad, but I see a benefit in eliminating a technology that is known to have a limited life span and hence avoid having to replace them down the road. We can eliminate electrolyitics and design the amplifier to last a lifetime. To do that we need to avoid needing large caps without generating excessive ripple from the power supply. We do that using the Loftin-White topology (the secret sauce being the cap + resistor connecting the cathodes of the two types) and by using a supply with good regulation.
Key elements of the power supply owe their origins to several sources but it was a gentlemen called Jeff Medwin that put me on this particular path. This being the low-DCR and low energy storage approach. The low DCR approach is desirable because we would like to arrange for the impedance of the power supply to be much less than that of the amplifier. The 2A3 has a plate resistance of 800 ohm (nominal) so we'd like to achieve something around 1/10th of that, or at least close. We don't want to resort to large electrolyits remember and simplicity means no heat generating shunt regulators here. This favours SS rectifiers and a big power trafo. The filter design also optimizes for high regulation and good transient response. If we size the caps well, we may be able to do the whole thing without electrolytics.
Attachments
Last edited:
I'd move the fuse to the primary. Add a soft start circuit if the fuse keeps blowing. Right now the fuse is in the signal path. The impedance of a fuse is highly non-linear and dependent on the current through the fuse. The Loftin-White topology saves the day as it makes most of the signal current flow through the cathode-anode cap, but still.... The fuse will also not protect anything in case one of the reservoir caps short out and blows (this is actually a rather common failure mode).
~Tom
~Tom
Member
Joined 2009
Paid Member
Thanks Tom, that's a good point - my thought was to have something to protect the rectifiers and secondary in case of something nasty. I will include a fuse in the primary for sure.
To explore a little further the potential benefits of this approach to the B+ supply I did a quick comparison between two options using the PSUD simulation tool - Jeff uses this approach to fine tune the component values in his designs. I'm not that familiar with PSUD but it sure is simple to use and fast compared with using LTspice. Jeff's approach has been controversial amongst some, but it does appear to offer less ripple without large inductors or large capacitors and if that's the case I'm all for it.
On the right (attached image), I set up the proposed design. I used Hammond's data sheet as guidance for the rms voltage on the secondary of the 363CX. I don't have a model for the SiC rectifiers but I think we're close enough with the 1N4007 for now. I introduced a step in the current draw at 1.5 s after turn-0n so that I could see something of the transient response.
On the left, I set up something closer to what I used on my current CELLINI amplifier. This uses a large choke, the so-called Critical Inductance needed for good regulation at the design current level and it ensures the diode conduction angle is relatively large. This puts less stress on the diodes and the transformer. Of course I had to use a higher secondary voltage in order to achieve the desired B+ voltage after the filter and I increased the DCR of the secondary to reflect the likelihood that the higher voltage would mean more wire and more resistance. The total capacitance in both cases is the same, but in CELLINI I used a lot more; I don't think the L-W needs so much.
What do I see ?
The more conventional approach (left) has around +/-1V of ripple. It takes 50ms to respond to the step change in current and doesn't settle that well.
The proposed scheme (right) has around +/-0.15V of ripple. And it's response to the current step looks a little cleaner, maybe slightly faster too.
Food for thought.
To explore a little further the potential benefits of this approach to the B+ supply I did a quick comparison between two options using the PSUD simulation tool - Jeff uses this approach to fine tune the component values in his designs. I'm not that familiar with PSUD but it sure is simple to use and fast compared with using LTspice. Jeff's approach has been controversial amongst some, but it does appear to offer less ripple without large inductors or large capacitors and if that's the case I'm all for it.
On the right (attached image), I set up the proposed design. I used Hammond's data sheet as guidance for the rms voltage on the secondary of the 363CX. I don't have a model for the SiC rectifiers but I think we're close enough with the 1N4007 for now. I introduced a step in the current draw at 1.5 s after turn-0n so that I could see something of the transient response.
On the left, I set up something closer to what I used on my current CELLINI amplifier. This uses a large choke, the so-called Critical Inductance needed for good regulation at the design current level and it ensures the diode conduction angle is relatively large. This puts less stress on the diodes and the transformer. Of course I had to use a higher secondary voltage in order to achieve the desired B+ voltage after the filter and I increased the DCR of the secondary to reflect the likelihood that the higher voltage would mean more wire and more resistance. The total capacitance in both cases is the same, but in CELLINI I used a lot more; I don't think the L-W needs so much.
What do I see ?
The more conventional approach (left) has around +/-1V of ripple. It takes 50ms to respond to the step change in current and doesn't settle that well.
The proposed scheme (right) has around +/-0.15V of ripple. And it's response to the current step looks a little cleaner, maybe slightly faster too.
Food for thought.
Attachments
Last edited:
Member
Joined 2009
Paid Member
You should tune the size of the 47u capacitor to and the alst psu capacitor to inject the corect amount of ripple in the cathode. the 2 capacitors are a volage divider, but it is not clear the way you drew it. The ratio should be arioound 1:10 (the cathode capactior should be the smaller one) too get the hum ti null.
Have a look here
Lowering SE Amplifier Noise (page 4)
Have a look here
Lowering SE Amplifier Noise (page 4)
I don't think the 12AX7 with a resistor plate load will be able to very cleanly drive the 2A3 ie. deliver current to 2A3 grid.
I would use a gyrator plate load, say a IRF820, and take the output at the source. This can drive in my experience any ordinary (and many extraordinary) tube extremely cleanly.
Also I would definitely recommend regulated B+, simply for the sonics at the very least. This takes the pressure off from the PSU component choises before the regulator - doesn't much matter whether you use cheapest electrolytics (except for longevity of course) or fist sized film caps.
I would use a gyrator plate load, say a IRF820, and take the output at the source. This can drive in my experience any ordinary (and many extraordinary) tube extremely cleanly.
Also I would definitely recommend regulated B+, simply for the sonics at the very least. This takes the pressure off from the PSU component choises before the regulator - doesn't much matter whether you use cheapest electrolytics (except for longevity of course) or fist sized film caps.
Member
Joined 2009
Paid Member
Costis is correct, in this topology the supply ripple can actually be higher than a 'standard' good SE amplifier and still achieve a no-hum output.
This is where the 'Loftin-White' part of the design comes into it's own.
To have supply rail noise contaminate the speaker signal it has must appear in the current through the output transformer. This design takes the output stage, the 2A3 and OPT, and floats the whole shebang a.c.-wise on the supply rail so that the OPT never experiences any supply rail noise driven current flow. How does it do this ? Well there are three points of contact with the output stage - the top of the OPT, the cathode of the 2A3 and the grid of the 2A3. If we connect all three to the B+ rail a.c.-wise, then the output will see no nasties from the supply rail. And this is what we've done with this design.
The top of the OPT is connected to B+ - so that point of contact will ride on the supply rail noise. The cathode of the 2A3 is connected via a capacitor to the B+ rail too ("ultra-path" connection) so it too is now riding on the supply rail noise.
We can't do the same simple connection for the grid of the output tube or we'll be 'shorting' the signal out. To put the right amount of B+ noise onto the grid so that the output stage is floating on it, we have to inject it via the input tube. We do this by adding another ultra-path connection to the cathode of the input tube. This is accomplished with the capacitor + adjustable resistor that connects the two cathodes together. Rather than take the input cathode ultra-path connection to the B+ this approach has the benefit that it will also relay any other noise or distortion at the output tube cathode and null that out too
I do have to tweak the parts values so this works, but once set up it will be very effective.
Yeah, don't go sticking your fingers in the chasis !
Last edited:
Member
Joined 2009
Paid Member
I've been thinking about this for sometime. I started a thread regarding this very question and given that others have built very nice amplifiers with such a drive arrangement I'm now quite comfortable with this approach. Commercial success stories include Don Garber Fi 2A3 and Serious Stereo 2A3.
http://www.diyaudio.com/forums/tubes-valves/212673-why-do-wimpy-drivers-2a3-work-well-they-do.html
I agree that a regulator can be a very good choice. If I were to use a regulator, my personal preference would be to use a shunt regulator. I feel that a shunt regulator puts it's impact on the sonics at a minimum. And if I were to shunt the B+ I would want to use another 2A3. I've seen this approach more than once, the last time I think I read it in Sound Practices. I like it. But I'm not going to use it here because it adds a lot of power dissipation (whether tube or SS it's still a Class A regulator), expense and some complexity. I don't believe we need a regulator if we design the power supply and the amplifier right.
However, we will consider some regulation for the power feed to the input tube
DC level shifts should not be an issue. The Loftin-White noise cancellation is a.c. balanced and doesn't get affected by moderate changes in d.c.
Given that, for d.c., this is a cathode biassed amplifier (the output tube has a huge cathode resistor!), nominal changes in the rail voltage will have negligible influence on the operating points of the input or output tube.
Last edited:
Referring to post #10.
Yeah, with the 4th order system on the right, you're getting a -80 dB/dec slope after about 190 Hz (the pole of the first LC). I'd try to move that pole down a bit - at least below 120 Hz if you can.
I bet the reason for the poor ripple rejection in the circuit on the left is that the circuit is operating in the frequency range where the filtering is essentially accomplished by the resistive division of the ESR of the inductor and cap respectively. An AC sim in spice would show this easily.
I find that PSUD-II is a great tool for getting started. But for the intricate details and learnings, I set the circuit up in LTspice. Both approaches are valid.
~Tom
Yeah, with the 4th order system on the right, you're getting a -80 dB/dec slope after about 190 Hz (the pole of the first LC). I'd try to move that pole down a bit - at least below 120 Hz if you can.
I bet the reason for the poor ripple rejection in the circuit on the left is that the circuit is operating in the frequency range where the filtering is essentially accomplished by the resistive division of the ESR of the inductor and cap respectively. An AC sim in spice would show this easily.
I find that PSUD-II is a great tool for getting started. But for the intricate details and learnings, I set the circuit up in LTspice. Both approaches are valid.
~Tom