Daniel, did you change the feedback resistor without changing the feedback capacitor?
Yes.
I did not change R16 or C3 or C4.
The treble errata exists no matter if R17 is at 10k or 22k.
I have to twist (and isolate for preventing short-circuit) C-E legs of Q4 (left one) for doing that. Otherwise, it is not possible bond them face-to-face as shown in photo.
In TO220: BDT85
SOT93: BDV95
In their new line of products, NXP doesn't seem to have power Bjt's anymore.
Gain adjustments should be made by altering R17 only. R20 follows R17, and R16, C3, C4 remain identical.
I was referring to the upper border of the audio band being disproportionate regardless of playback at modest levels.
Clipping? Must be kidding? Because of curiosity I managed to turn up the cute little Circlophone until clipping. Perhaps I should have either put the 94db efficient speakers far away from the tall bookshelves or used a helmet. Have you built a higher voltage edition of Circlophone? Gosh man, clipping isn't a secret because everything in the house will be shaking violently with audio like a gale force wind. My sample also has the weird errata of power output depending on operating temperature as it is not very strong when cold, eventually warms up to prodigious, and after that, it will change output by itself during normal/modest levels playback concurrently to when the orchestra decided to play a bit louder but instead you may get concert levels by surprise. It does this reliably, but I think it is unusual. A concert actually sounds like that if you're on the front row.
I hadn't tried it extremely loud before, but apparently the cute little amplifier is unexpectedly capable.
If it isn't clipping, there are two possibilities: either there is some malfunction, but I doubt it because it would impact the reproduction negatively and unpleasantly, or that's simply the sonic signature of the amplifier.
If you have some test equipment, put a 1KHz sinewave at the input, and check the output on a scope, just to make sure everything is healthy.
I didn't build high power versions, my neighbours are too close for that....
If you have some test equipment, put a 1KHz sinewave at the input, and check the output on a scope, just to make sure everything is healthy.
I didn't build high power versions, my neighbours are too close for that....
I would put the feedback resistor back to normal value. It, and the feedback capacitor, are calculated. Changing them effects stability margins.
Also i would go over your circuit again. Doesn't sound at all like what I hear from mine or from the other's descriptions. Maybe you have a capacitor the wrong value somewhere?
Also i would go over your circuit again. Doesn't sound at all like what I hear from mine or from the other's descriptions. Maybe you have a capacitor the wrong value somewhere?
Thank you.
I was worried about the possibility of a construction error. I will gently filter this one (a bit of extraneous RF RC's will get it) and attempt increasing the onboard compensations, and then build another somewhat differently. I've got to learn how to make PCB's. The little compact perfboard Circlophone is mighty entertaining, but it was laid out tight to avoid trace length, so I can't swap components without a partial rebuild. Right now it has an 3R+150nF (polyester dip cap) zobel added at the speaker jack and while the treble is a bit noticeable, it isn't overwhelming.
The ratio between transformer voltage and output power at RMS is similar to a standard bridge amp. I did manage to learn that my big radio project probably needs a 250w rated output device, the transmitter heatsink, and less voltage.
A violin sounds perfectly realistic and you could estimate the brand of violin and the brand or at least the gauge of strings on it, but a cello sounds like a voila and a piano like a keyboard, and so 100uF is too small. Terranigma latest photo shows 470uF, and probably for a "big enough" approach. Next, I will attempt "the middle ground" with 220uF per rail, although may actually do something more creative in addition to those. In any case, 100uF is too small.
I was worried about the possibility of a construction error. I will gently filter this one (a bit of extraneous RF RC's will get it) and attempt increasing the onboard compensations, and then build another somewhat differently. I've got to learn how to make PCB's. The little compact perfboard Circlophone is mighty entertaining, but it was laid out tight to avoid trace length, so I can't swap components without a partial rebuild. Right now it has an 3R+150nF (polyester dip cap) zobel added at the speaker jack and while the treble is a bit noticeable, it isn't overwhelming.
The ratio between transformer voltage and output power at RMS is similar to a standard bridge amp. I did manage to learn that my big radio project probably needs a 250w rated output device, the transmitter heatsink, and less voltage.
A violin sounds perfectly realistic and you could estimate the brand of violin and the brand or at least the gauge of strings on it, but a cello sounds like a voila and a piano like a keyboard, and so 100uF is too small. Terranigma latest photo shows 470uF, and probably for a "big enough" approach. Next, I will attempt "the middle ground" with 220uF per rail, although may actually do something more creative in addition to those. In any case, 100uF is too small.
Last edited:
Daniel,
Regarding my experiences, I may add some trial list for you as follows:
* Try many ac-coupling caps including non-polar polyester ones. Mine is 5uF 350V polarised electrolytic.
* Trial with other brand input pairs with minimum 300-350 hfe.
* Use bigger sized and valued decoupling caps. Bigger caps comparing small ones with same rating/value generally means more ripple current. This approach prevents some bottlenecks on supply side. My previous builds were using other brand, also bigger 470uF's. In photo, the small black cap below is also 470uF/50V..then compare their size
* Take attention for if there is any compensation cap heated by driver and servo transistors.
I'm just another witness for how this amplifier capable of things.
Regarding my experiences, I may add some trial list for you as follows:
* Try many ac-coupling caps including non-polar polyester ones. Mine is 5uF 350V polarised electrolytic.
* Trial with other brand input pairs with minimum 300-350 hfe.
* Use bigger sized and valued decoupling caps. Bigger caps comparing small ones with same rating/value generally means more ripple current. This approach prevents some bottlenecks on supply side. My previous builds were using other brand, also bigger 470uF's. In photo, the small black cap below is also 470uF/50V..then compare their size
* Take attention for if there is any compensation cap heated by driver and servo transistors.
I'm just another witness for how this amplifier capable of things.
Last edited:
Does the much greater capacitance of the modern Chinese BD140, by Fairchild of China specs, cause any potential problems? This one transistor on my amplifier does not correspond reasonably to post 1.
Also, my capacitor//feedback resistor may be a concern. Is it okay to try 33pF?
And my C12 is not all the way up to 820p and visually, it does look like it could help cut my lisping treble problem. Is that true?
I'm not sure about what, if anything, C11 would do; however, there is no guarantee that ST TIP35C is actually running at 2.5mhz~3mhz even if perchance that TIP35C is authentic. It may be a common something else conveniently re-labled--convenient for the manufacturer, not convenient for me due to invalid datasheet. I'm guess that at least one variant runs 30mhz and should run lower voltage than the datasheet suggests. The construction of these removes some possibilities. Not exploding despite high power armstrong boost removes some more possibilities (NTE source would have exploded). But my voltage and 8R load is within range for the NEC fab variant, 30mhz. That factory didn't stop on the day the NEC brand name was removed but rather uses other brand names, which is a shame because good parts need accurate datasheets. So, does C11 possibly help stability and/or uppermost treble?
So after this and after yet another RF filter added at input, may I try a capacitor set at + to - inputs of the input pair? That probably should make the treble take a step back, but if it is okay to do this, I sure would like a "ballpark" capacitor value so that I can begin the search in a reasonable area.
The explosion that I had previously, emitted a blast at Q11, R12, and R15. This happened because I was up too late and mistakenly assumed the pinout was the same as the smaller transistors. V- went into base of Q11 by mistake. R12 was between emitter and collector of Q11 by mistake. Q6 was hooked to collector of Q11 by mistake. Do I need to replace Q6? So far I have only replace Q11, Q10, R12, R13.
Also my SR504 is really SR506, but I knew that would happen since a modern fab doesn't actually make every variety and you can tell which one is "host" by seeing which "steps" in a lineup are either vastly available or greatly less expensive (ST says "Preferred" right on the datasheet to indicate a "host" core and if an ST datasheet doesn't say "Preferred" then that datasheet is possibly not matching the part). For example 2n5550's are all 2n5551's and all of the modern larger size bat diodes are Bat86 although only a few manufacturers would discontinue the inactive model numbers--the majority just paint near match labels onto the one model they make, because a greater amount of model numbers equals a greater amount of customers. And, if you special order transistors, they'll paint on whatever label you require onto the least expensive near match and increase the price dramatically. The only good news there, is that I got authentic 2n5551 and authentic Bat86.
I'm wondering if it is appropriate to parallel the big shiny schottky with a small schottky or an MR to "clip" straighten out the voltage curve.
Or, if you think it wouldn't cause added noise, an 1n5407//SR could hold the indicated voltage curve rather than the weak schottky leaning over. Some system like this seems that it could regulate the cost versus variances issues at D4, D5. Anyway, do you think that I may have some audio artifacts from a really saggy schottky at D4, D5? Or do the 1R remove/reduce that problem? Possibly my 1R need to be increased to 1w for handling the extra current?
My power supply has an LED on both rails. At unplug the V+ goes out immediately. The V-stays on for 20 seconds while the amplifier continues to play at low output. The light bulb tester, if connected, never flashes bright, even if high output is requested, up until the Circlophone becomes annoyed and decides to spontaneously engage its interesting high gear for audio like gale force wind, and THAT will light up the bulb, although, with the house nearly wrecked, current usage wasn't a secret anyway. A visual indicator LED is just hilariously useless. I love listening to classical music with this amp. Its like a collegiate orchestra on crack! Wow, it really rocks, but I would at least learn the cause of the errata even though I probably won't repair that problem, because it is just too much fun. I was blasting some Enya yesterday to see if Circlophone could do anything boring. It certainly cannot. It was playing along sort of normal output levels and then suddenly I ended up gripping the chair just like that Memorex commercial.
Also, my capacitor//feedback resistor may be a concern. Is it okay to try 33pF?
And my C12 is not all the way up to 820p and visually, it does look like it could help cut my lisping treble problem. Is that true?
I'm not sure about what, if anything, C11 would do; however, there is no guarantee that ST TIP35C is actually running at 2.5mhz~3mhz even if perchance that TIP35C is authentic. It may be a common something else conveniently re-labled--convenient for the manufacturer, not convenient for me due to invalid datasheet. I'm guess that at least one variant runs 30mhz and should run lower voltage than the datasheet suggests. The construction of these removes some possibilities. Not exploding despite high power armstrong boost removes some more possibilities (NTE source would have exploded). But my voltage and 8R load is within range for the NEC fab variant, 30mhz. That factory didn't stop on the day the NEC brand name was removed but rather uses other brand names, which is a shame because good parts need accurate datasheets. So, does C11 possibly help stability and/or uppermost treble?
So after this and after yet another RF filter added at input, may I try a capacitor set at + to - inputs of the input pair? That probably should make the treble take a step back, but if it is okay to do this, I sure would like a "ballpark" capacitor value so that I can begin the search in a reasonable area.
The explosion that I had previously, emitted a blast at Q11, R12, and R15. This happened because I was up too late and mistakenly assumed the pinout was the same as the smaller transistors. V- went into base of Q11 by mistake. R12 was between emitter and collector of Q11 by mistake. Q6 was hooked to collector of Q11 by mistake. Do I need to replace Q6? So far I have only replace Q11, Q10, R12, R13.
Also my SR504 is really SR506, but I knew that would happen since a modern fab doesn't actually make every variety and you can tell which one is "host" by seeing which "steps" in a lineup are either vastly available or greatly less expensive (ST says "Preferred" right on the datasheet to indicate a "host" core and if an ST datasheet doesn't say "Preferred" then that datasheet is possibly not matching the part). For example 2n5550's are all 2n5551's and all of the modern larger size bat diodes are Bat86 although only a few manufacturers would discontinue the inactive model numbers--the majority just paint near match labels onto the one model they make, because a greater amount of model numbers equals a greater amount of customers. And, if you special order transistors, they'll paint on whatever label you require onto the least expensive near match and increase the price dramatically. The only good news there, is that I got authentic 2n5551 and authentic Bat86.
I'm wondering if it is appropriate to parallel the big shiny schottky with a small schottky or an MR to "clip" straighten out the voltage curve.
Or, if you think it wouldn't cause added noise, an 1n5407//SR could hold the indicated voltage curve rather than the weak schottky leaning over. Some system like this seems that it could regulate the cost versus variances issues at D4, D5. Anyway, do you think that I may have some audio artifacts from a really saggy schottky at D4, D5? Or do the 1R remove/reduce that problem? Possibly my 1R need to be increased to 1w for handling the extra current?
My power supply has an LED on both rails. At unplug the V+ goes out immediately. The V-stays on for 20 seconds while the amplifier continues to play at low output. The light bulb tester, if connected, never flashes bright, even if high output is requested, up until the Circlophone becomes annoyed and decides to spontaneously engage its interesting high gear for audio like gale force wind, and THAT will light up the bulb, although, with the house nearly wrecked, current usage wasn't a secret anyway. A visual indicator LED is just hilariously useless.
I love listening to classical music with this amp. Its like a collegiate orchestra on crack! Wow, it really rocks, but I would at least learn the cause of the errata even though I probably won't repair that problem, because it is just too much fun. I was blasting some Enya yesterday to see if Circlophone could do anything boring. It certainly cannot. It was playing along sort of normal output levels and then suddenly I ended up gripping the chair just like that Memorex commercial.
Last edited:
http://www.diyaudio.com/forums/atta...little-cheap-circlophone-circlophonekpmod.gif
Q14?
We wouldn't have to source various zeners? One transistor instead of a mess? Now, that is nice! Has this been tested? Can Q14 be the 2n5551 or a generic BD139 (est~30pF) or the 2sd669a (14pF)?
Be careful with that: the compensations are very nearly optimum, and tampering significantly with more than one of them could have disastrous effects.
If you go into that direction, you should first make a test on a dummy load, with a scope and a test signal, otherwise you might smoke your tweeters.
The Zobel for this amplifier is just R25 and C7.
The Circlophone will tolerate about any impedance at its output, and your additional Zobel will do no harm, but it will bring no benefit either: it will simply dissipate some more HF in your 3R, but other than that, there will be no effect since the output impedance is in the 2 to 3 milliohm range.
No, not a bridge amp, a standard PP amplifier.
To simplify: the output voltage (rms) will be ~15% lower than the transformer's voltage: for example, if you use a 24-0-24 transformer, the rms output would be 20.4V ~=100W/4R
If you do not cumulate too many "bad apples" in your amplifier, it will work: I have tested generic unbranded BD140's, and they work, provided you have decent VAS/phase-splitter transistors, like 2N3020.
If you cumulate bad VAS and bad drivers, you will run into problems.
33pF is still OK.
C12 is the least critical cap in this amplifier, you can do pretty much what you like.
Only where it is required. If you use it with low frequency transistors, it will on the contrary cause some ringing in the step response (~= more treble, sort of).
You can increase the RF filter as much as you like.
If it works like that, your components are most probably OK: it is normally a case of all or nothing (or it becomes one very quickly).
That is completely unimportant.Also my SR504 is really SR506, but I knew that would happen since a modern fab doesn't actually make every variety and you can tell which one is "host" by seeing which "steps" in a lineup are either vastly available or greatly less expensive (ST says "Preferred" right on the datasheet to indicate a "host" core and if an ST datasheet doesn't say "Preferred" then that datasheet is possibly not matching the part). For example 2n5550's are all 2n5551's and all of the modern larger size bat diodes are Bat86 although only a few manufacturers would discontinue the inactive model numbers--the majority just paint near match labels onto the one model they make, because a greater amount of model numbers equals a greater amount of customers. And, if you special order transistors, they'll paint on whatever label you require onto the least expensive near match and increase the price dramatically. The only good news there, is that I got authentic 2n5551 and authentic Bat86.
No see above. 0.25W is plenty for these resistors, they will normally dissipate a fraction of that.
That is a bit strange, I do not see where the V+ current could be dumped.
It has not been physically tested , but there is little that could go wrong there. And since the transistor is cascode-operated, its characteristics are almost irrelevant, you just need transistor-action with the right voltage and power ratings.
For Q14, I was trying to ask if I need the current handling of a heatsinked driver transistor, an un-heatsinked driver transistor or if a small signal transistor will do? I guess 2sd669a without heatsink. Correct?
Dissipation and current is identical to Q5/Q6. A 2SD669 without heatsink is suitable for supply voltages of up to +/-75V.
Layout:
Is the RC of C12, R18 most useful located with the drivers or with the phase inverter? Currently, I have it with the zeners at the drivers.
Thanks again!
Edit:
And an additional question. Do you have a Circlophonic headphone amp?
Last edited:
Good question. Logically, it has to be on the driver's side since that's where the effect is maximal.
No, if I made a Circlophonic headphone amp, it would probably be in pure class A. Next project perhaps....
I have made another type of low power amp, it is the Tringlinator, based on the Tringlotron topology:
http://www.google.be/url?sa=t&rct=j...lFb6uHQRg&sig2=oTF1jGTUaM1SUuf9RixcLg&cad=rja
This is much more stable in clipping with all BJT,
BJT distortion figure becomes nothing exceptional...
How to fix it to have best of both?
I wanted to use IRF9610, for the 200V standoff.
But that .model makes it crazy unstable! Maybe
high standoff not ncessary? As this one crosses
GND, easy to cascode the three affected devices.
M1,M2,Q5. Cascode not making any more or less
stable, nor changing the distortion that I can tell.
Ring of Two (BJT BJT) or (BJT MOSFET) also not
stable for threshold detectors on the right side.
Diodes references for unknown reason, not giving
me the same grief...
8 Transistors (including the MOSFETs). But the
supporting part count is still higher than I was
shooting for... Some of the stoppers clearly are
not stopping much, and could go if the main
problem with stabilty was solved. It goes full
triangle by 40KHz, but even thats not got it...
Those Taylors are made not to ever shut off.
I don't know if thats part of the problem?
BJT distortion figure becomes nothing exceptional...
How to fix it to have best of both?
I wanted to use IRF9610, for the 200V standoff.
But that .model makes it crazy unstable! Maybe
high standoff not ncessary? As this one crosses
GND, easy to cascode the three affected devices.
M1,M2,Q5. Cascode not making any more or less
stable, nor changing the distortion that I can tell.
Ring of Two (BJT BJT) or (BJT MOSFET) also not
stable for threshold detectors on the right side.
Diodes references for unknown reason, not giving
me the same grief...
8 Transistors (including the MOSFETs). But the
supporting part count is still higher than I was
shooting for... Some of the stoppers clearly are
not stopping much, and could go if the main
problem with stabilty was solved. It goes full
triangle by 40KHz, but even thats not got it...
Those Taylors are made not to ever shut off.
I don't know if thats part of the problem?
Attachments
Last edited:
Actually, a diode bridge between gates seems to fix the clipping instability.
Limits gate voltage difference to 1.4V, now hard clips and recovers clean.
Guess this also limits the maximum error stored in gate charges, thus the
quick recovery?
Still can't get this thing past 40K. If I ease up on comp, goes banannas.
Lessening the open loop gain helps stable bandwith, but hurts distortion.
Like for example: source resistors, or using a BJT limited by beta. Going
other way, to a MOSFET with even higher gain, stability is unobtanium.
Maybe I am pushing too few transistors for way too much open loop gain?
Feels like xquad again, though I doubt any negative resistance is at fault.
Limits gate voltage difference to 1.4V, now hard clips and recovers clean.
Guess this also limits the maximum error stored in gate charges, thus the
quick recovery?
Still can't get this thing past 40K. If I ease up on comp, goes banannas.
Lessening the open loop gain helps stable bandwith, but hurts distortion.
Like for example: source resistors, or using a BJT limited by beta. Going
other way, to a MOSFET with even higher gain, stability is unobtanium.
Maybe I am pushing too few transistors for way too much open loop gain?
Feels like xquad again, though I doubt any negative resistance is at fault.
Last edited:
