I know about thermal distortion, but I am so satisfied with the sound that I am not going to change anything. Maybe I'll put it in a solid block of epoxy
I'm on the way, but I have to finish my speakers in couple days.
While preparing pcb's, did you realize that situation issued at post#274 ?
Yes, I did, and, looking at the schematic told me the quick easy fix. Otherwise, PCBs are just fine.
I cannot complain, other than the holes for the power supply input and the output transistor connections are somewhat small. Some careful enlarging of the holes and careful soldering and you can use larger diameter wire though.
What output transistors/drivers are you using terra? Just curious. I used 2SB649A's and MJW21196's. And what compensation do you plan to start with? I started right out with 270pF for C12 and build the circuit with C11 in place. I had no issues and still haven't, and it sounds better than I thought possible. I could lower C12 and/or try without C11 but I do not wish to change it, at least not for a while.
I have read of dominant 2nd order distortion sounding great and musical but lacking fine detail. I have heard of dominant 3rd having "detail" (or the feeling of?) but not sounding very musical. Then I hear the Circlophone, and I have more detail in the music than I have ever heard, and I am talking about songs I have been listening to since I was a kid. On top of that it is so musical and "sweet" than I always want to switch it on and play music, even in the most inappropriate situations.
But, good that you're not just taking my word for it and you are going to build it and find out for yourself.
What speakers are you building?
Last edited:
I'm planning to use oldschool pcb pins on those holes. Thus, hole diameter will not an issue for me anyway. I like PCB's with old looking fashion and well applied plate grounding feature. I think plate grounding helps for amp's silence.
I want to use japan bjt's on small signal side as much as possible. I just edited pdf file for fitting 2sa devices on front side (Q1, Q3, Q4 in original schematic). That may not necessary if you know what type of transistors you are using but I could'nt help myself on this. Four matched pairs (I'm planning to build 4 channels in total if everything goes right) of 2sa970's are ready. Q12 and Q13 are will be 2sc2240.
I found gold plated Motorola 2N3020's at a local supplier and I want to use them in place of 2N3019's. I think 1.8pF Ccob 2sc2705's are perfect in place of Q5,Q6. But I like metal can devices, then those 2sc2705's may wait for chance.
As drivers, I prefer fabulous devices.. and 2sa1930's are ready.
I got plenty of NPN TO3 metal case power transistors. Those are Moto, Toshiba 2N3773 , MJ15003, 2SD110 and maybe some others. I like tweak with parts. I think I will choose output trannies after tweaking. Circlophone is a great project for these king of power transistors. These devices on my hand are generally 1-4mhz range slow but robust devices. Then my C12 will be 180pF, maybe 270pF. I think another important issue with C12 is its material. Silver mica or C0G/NP0 ceramic types must be used in my opinion.
Thank you for encouraging comments. When you look at the schematic, you capture that Designer's (Thanks for Elvee) free minded sensation and point of view easily.. No emitter resistors, no complementary devices, no bias trimmers, no fabulous boutique parts..
I'm trying build a vented box with well regarded B&C 12PE32 mid-bass' and other B&C gem, DE-250 as compression drivers. 2nd order Linkwitz-Riley crossovers just finished. Box is ready except vents and horn cutouts. This driver combination is great for acoustic and live music..And they are waiting for their Circlophone!
Last edited:
not to offend, but i did some minor redrawing to improve readability, also based on physical location of components. keep in mind that Q7 should be close to Q2, but that would mess up the schematic.
i'm also almost finished with a single-layer PCB
An externally hosted image should be here but it was not working when we last tested it.
i'm also almost finished with a single-layer PCB
Last edited:
V0.1
i left out the supply caps.
happy about:
- almost fully closed ground plane
- small (7.7x5.7 cm)
- only one connection has to be made manually (red)
- components are physically close to each other when preferred, see schematic
- output-stage can be easily modified
unhappy about:
- location of ground connector
An externally hosted image should be here but it was not working when we last tested it.
i left out the supply caps.
happy about:
- almost fully closed ground plane
- small (7.7x5.7 cm)
- only one connection has to be made manually (red)
- components are physically close to each other when preferred, see schematic
- output-stage can be easily modified
unhappy about:
- location of ground connector
Last edited:
corrected schematic, updated PCB (with improved ground)
edit: silly me. need more coffee! it was allready correct. still, the 0.2 PCB is better
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.
edit: silly me. need more coffee! it was allready correct. still, the 0.2 PCB is better
Last edited:
No offense taken: my abilities to produce clear and legible artwork are more than limited, and any help is most welcome.
Apart from that, nice design.
Since you make modifications, it would be nice to take the feedback directly from the output pad: this will avoid asymetries and distortions caused by the track resistance.
I still think more could be done to clarify this schematic.
I prefer your (cleanup of Elvee's original) way drives are
routed, over my re-drawn. Much less confusing to keep
the drives together as the pair that they are.
But I still prefer my way of drawing diodes straight down
the middle, as they are all in a series. Also current mirror
style of drawing transistors paired near the top, better
shows how the emitter voltages are forced to be equal.
http://www.diyaudio.com/forums/atta...tle-cheap-circlophone-circlophonem-redraw.gif
Obviously old -M ref designators won't match current PCB.
If we merge these best of both ways of drawing this circuit,
I think it would become much easier to introduce to those
yet unfamiliar with its rather unusual theory of operation.
R14 (new ref) probably deserves special comment added
to the drawing, PSRR function is not at all self-evident.
I prefer your (cleanup of Elvee's original) way drives are
routed, over my re-drawn. Much less confusing to keep
the drives together as the pair that they are.
But I still prefer my way of drawing diodes straight down
the middle, as they are all in a series. Also current mirror
style of drawing transistors paired near the top, better
shows how the emitter voltages are forced to be equal.
http://www.diyaudio.com/forums/atta...tle-cheap-circlophone-circlophonem-redraw.gif
Obviously old -M ref designators won't match current PCB.
If we merge these best of both ways of drawing this circuit,
I think it would become much easier to introduce to those
yet unfamiliar with its rather unusual theory of operation.
R14 (new ref) probably deserves special comment added
to the drawing, PSRR function is not at all self-evident.
Last edited:
i do realy like your drawing, i must have overlooked it when scrolling this thread. not much to improve there!
if we would merge our versions the whole readability of the original schematic would be much improved.
i now see how to group the "mid-section" with the series diodes. my PCB could be improved by that
Great work!
Last edited:
I missed that.
I am not sure it is a good idea: without a local decoupling, it won't be possible to achieve the transient cleanliness displayed in the oscillograms at the beginning of thread.
There is still plenty of room in the middle of your PCB, and it would certainly be possible to include at least a minimal on-board decoupling: nowadays, 10µ/63V caps are tiny, and they would make a big difference if there some length of wiring to the supply caps.
If you choose to include them, place the supply pads in close proximity, to avoid creating low impedance loops in the tracks.
I reiterate also about the respective placement of the output pad and the feedback take-up point: as they currently are, they will easily cause a doubling of the THD.
will place the decoupling capacitors back in, and will try to re-route the feedback.
however, i don't think that the trackresistance makes the difference.
my main problem with placing the caps is that the central ground plane at the output doesn't have a large trace to the rest of the ground. but i will fix that.
however, i don't think that the trackresistance makes the difference.
my main problem with placing the caps is that the central ground plane at the output doesn't have a large trace to the rest of the ground. but i will fix that.
I can assure you it does, and it even has a more dramatic effect than what I had first guesstimated.
I ran two simulations to compare the situations: first, a "normal" one, where the track routing is state-of-the-art, and the other one based on the actual layout, including the tracks resistances and configuration estimated from the drawing, assuming standard PCB process.
You can see the effect of several misplaced milliohms: the even order distortions shoot up by a factor of more than 10!!!!
The overall THD figure increases from ~39ppm to 289ppm.
On the first version of my prototype, the initially measured THD was in the 40ppm range.
After a careful clean-up and some component rearrangements, I arrived at 15ppm, better than in the sim.
This means that the effect of the layout is not benign at all, it can very easily ruin an otherwise excellent amplifier.
That is the reason why I am so finnicky about those seemingly secondary "details".
PS
If you tin the tracks, or if you use heavier than standard PCB copper, the observed effects will be attenuated, but they will remain significant.
Even with superconducting tracks, an inductive effect would remain. I didn't model the inductance in the sim, this means that in reality, the situation will be even worse than predicted
Attachments
Last edited:
I hope to make a discrete amplifier some day soon. This design has peaked my interest. Aprox 15 CAD of silicon per channel and a forgiveness of build quality and part selection make this very interesting.
I have a few questions, some of them smarter then others:
1- Would a pair of MJE3055T work for power transistors? I ask because Heat sinks for TO-220 are easier to build. (I suspect the answer is "yes, but only for less then 20v rails")
2- Does the PSU need anything more fancy then transformer, rectifier bridge and caps?
3- (The newbie question) how does one turn the PNG that powerflux posted into a PCB. More to the point, how does one extract the bits that are copper and print that at the right scale? I gave a stab in the GIMP selecting various bits by colour, filling with white or black, but this seems ackward. What's more how do I get the scale exactly right? Powerflux hasn't posted dimensions.
I plan to do the laser-toner-flimsy-paper-iron-on transfer method for creating the PCB.
I have a few questions, some of them smarter then others:
1- Would a pair of MJE3055T work for power transistors? I ask because Heat sinks for TO-220 are easier to build. (I suspect the answer is "yes, but only for less then 20v rails")
2- Does the PSU need anything more fancy then transformer, rectifier bridge and caps?
3- (The newbie question) how does one turn the PNG that powerflux posted into a PCB. More to the point, how does one extract the bits that are copper and print that at the right scale? I gave a stab in the GIMP selecting various bits by colour, filling with white or black, but this seems ackward. What's more how do I get the scale exactly right? Powerflux hasn't posted dimensions.
I plan to do the laser-toner-flimsy-paper-iron-on transfer method for creating the PCB.
Yes, they would work, and you don't need to limit yourself to 20V: the dices are that of 3055's, and they share the same SOAR.
With 4Ω loads, you can go up to +/-30V safely, provided you don't use your amp all day long at full throttle with heavily compressed music: that would require impossible heatsinks, but for a normal use, that's OK.
There are many cheap, higher power plastic alternatives too: like BD249 or TIP35
No. The PSRR is not huge, but sufficient for practical purpose, and the amp works perfectly, even with lightly filtered unregulated PSU's.
You will have to wait Powerflux is satisfied enough with his work to post the source file and/or a printable file.
I just hope he wasn't put off by my comments, that was only constructive criticism, and I appreciate any contribution, including his....