Working at +/-30V. No signs of instability. Idle current is tracking perfectly.
Amoung the new parts from DK I have new multiturn 1K pots. The ones I have been using are from Sure Electronics on Ebay. They look like Bourns 3296W, but they are "bourts"3296W. They haven't given me trouble, but trouble I don't need so I have new Cermet ones. Also the new 47pF silver mica for the Miller cap.
Stability into the dummy load looks good, clipping looks good. Still have the slight sticking on the bottom wave, but I'm not sure that can be cured anyway.
Squares at 20k look sharp with no overshoot and no ringing. Still testing without a Zobel.
Amoung the new parts from DK I have new multiturn 1K pots. The ones I have been using are from Sure Electronics on Ebay. They look like Bourns 3296W, but they are "bourts"3296W. They haven't given me trouble, but trouble I don't need so I have new Cermet ones. Also the new 47pF silver mica for the Miller cap.
Stability into the dummy load looks good, clipping looks good. Still have the slight sticking on the bottom wave, but I'm not sure that can be cured anyway.
Squares at 20k look sharp with no overshoot and no ringing. Still testing without a Zobel.
Attachments
I like digikey too, it takes 2 or 3 days for me to get my stuff - to Sweden! Free shipping on orders over $100 or so too. I can get 4 FJL4x15 for the same price as one TIP36C locally. (Yes TIP36C is the best output transistor in plastic package that is easily obtainable here)
megajocke said:
I don't get that deal! But it's only $8.00 so no big expense.
I went ahead and prepped the big heatsink for mounting of this board so that I could do some full power tests.
I set the idle (which was only 17mA difference from the 30V supply) to 55mA per device and left it with a signal to warm up.
Good and warm the idle current is reset and I let it sit again for a while. After about 20 minutes of the outputs drawing 250mA, I check the idle again. It's rock solid. No more issues there.
It's warmed up, I'll do some tests. Stability is still fantastic - no sign of any ringing in spite of the open arrangement and the heatsink not being grounded.
Clipping.
Sine wave clipping at 20k is sticking again on the top. Below that, at 10k and lower it's fine.
Square waves look good and sharp at 20k, but have a small overshoot on the bottom.
So what has changed? For one, the diodes are different - I had been using the UF1003. Also the Miller cap had been reduced to 47pF, with little influence from the clamping diodes capacitance.
Could I have damaged the clamping diodes with my soldering job? I pretty much just dabbed it with my iron, very quickly. I don't know how much heat these can take though.
Attachments
Greetings...
John, i guess it's almost "redundant" for me to congratulate you on this... genesis 🙂 It's been quite a ride just popping in to your thread(s) every now and then...
One quick question though: would any major mods be necesary to make this jewel of yours work with more "earthly" active devices?
(Read: BC / BD / MJ(L) type stuff)
Yes, i'm aware there might be some performance limitations, but keep in mind not everyone has access to japanese stuff. Not to mention that some of us are still student, so... 🙄
Keep up the wonderful work... The DIY community will thank you for it, given time
John, i guess it's almost "redundant" for me to congratulate you on this... genesis 🙂 It's been quite a ride just popping in to your thread(s) every now and then...
One quick question though: would any major mods be necesary to make this jewel of yours work with more "earthly" active devices?
(Read: BC / BD / MJ(L) type stuff)
Yes, i'm aware there might be some performance limitations, but keep in mind not everyone has access to japanese stuff. Not to mention that some of us are still student, so... 🙄
Keep up the wonderful work... The DIY community will thank you for it, given time
Khron said:
Hi Khron,
No, I can never get enough or too much congratulations! Lay it on me!
The amp likes fast devices, that's for sure. You can use the BC550 for the diff pair, BC546 and 556 for the cascode and mirror.
the EF needs to be high voltage, so pick one with reasonable speed and gain.
The VAS: I would not recommend anything else for this. The 2SA1381 made such a vast difference it would be hard to find something that's comparable and more available. This is in current production from Fairchild.
The drivers can be MJE1503x and the outputs can be MJL21193/94 or others.
It's here...snow. First for the season:
Attachments
One more thing... Is there any chance you'll be posting the pcb in a more "universal" format? (say, a pdf, for instance) 😀
Khron said:
I want to iron out all of the wrinkles first. There are a couple of spacing issues with the current layout that I'd like to fix also. When it's complete to my satisfaction I'll provide a pdf of the board.
I have bypassed D1 with a UF1003 and clipping is back to (nearly) normal. I think that maybe reducing the miller cap to 47pF is compromising clipping performance. I saw the same behavior before when I reduced it.
I will try it with the Miller cap doubled up to 100pF again, leaving the BAV103 in place.
Yes, that's it. Cdom needs to be 100pF to get that good clipping performance. The overshoot on the lower half of the 20k square wave is gone also.
This means that my soldering job on the BAV103's did not damage them.
I'm wondering what the effect would be if I were to increase the emitter resistors in the differential pair, leaving the Miller cap at 100pf? Even better clipping recovery? What does that do to my slew rate? It must be high enough anyway, there's almost no discernible slope in the vertical of the 20k square wave.
I'm not going to bother with pics of the clipping or squarewave, as they are no different from before.
Measured DC offset with the input plugged into my FG, turned on but the amplitude set to "0" is ~10mV. Maybe a lower figure could be had if the diff pair transistors were matched (I didn't do).
All in all very good. Some real life distortion tests will come soon, be those only with the RMAA. This gives a very accurate indication of what's going on anyway.
This means that my soldering job on the BAV103's did not damage them.
I'm wondering what the effect would be if I were to increase the emitter resistors in the differential pair, leaving the Miller cap at 100pf? Even better clipping recovery? What does that do to my slew rate? It must be high enough anyway, there's almost no discernible slope in the vertical of the 20k square wave.
I'm not going to bother with pics of the clipping or squarewave, as they are no different from before.
Measured DC offset with the input plugged into my FG, turned on but the amplitude set to "0" is ~10mV. Maybe a lower figure could be had if the diff pair transistors were matched (I didn't do).
All in all very good. Some real life distortion tests will come soon, be those only with the RMAA. This gives a very accurate indication of what's going on anyway.
hi MJL21193,
Why the concern about the sticking clipping behaviour? Surely the clipping part of the curve is a bigger problem than the sticking. Is it normal to run a Hifi amp into clipping?
Increasing Cdom will probably have a negative impact on the sound.
regards
Why the concern about the sticking clipping behaviour? Surely the clipping part of the curve is a bigger problem than the sticking. Is it normal to run a Hifi amp into clipping?
Increasing Cdom will probably have a negative impact on the sound.
regards
Greg Erskine said:
Hi Greg,
I don't want to use this amp into clipping, and at 153 watts into 8 ohms this shouldn't happen. It can happen occasionally.
It's not just sticking that I see when the Cdom is lower, it's signs of instability in the upper wave that worry me. I would like to get a picture but I don't want to leave it in this state for any more than a split second or so. Suffice to say that it turns ugly, with the upper wave shifting sideways slightly and "quivering".
I could leave cdom at 47pF and not have a problem. There are no signs of instability at all and the chances of clipping, especially at high frequency, are slim. These amps are to drive my active 3-ways, so there will be some headroom.
Try using a small cap say 10pf to 15pf from vas collector back to ltp feedback, should cure some of those problems.
when you do this you may find you can reduce Miller comp cap slightly. This usually benefits the sound. So listen to a "before and after".homemodder said:
MJL21193 said:
Sanyo 2SA1352/2SC3416, rulers of the low Cob/Cre universe, for pennies (as Mr Popa/Syn08 would say) at MCM.
(i see you too are a member of the endless pool cleaner and intoxicater club)
homemodder said:
Hi homemodder,
Yes, you suggested that before, and I tried it in the sim but didn't on the actual amp. I will give it a try.
AndrewT said:
Hi Andrew,
I don't know if it's my imagination, but I think this latest version has a difference in the midrange/treble. Seems more detail is coming through. Could this be attributed to the lower Miller cap value? This is the first listen with it at 47pF.
most of my best commercial amps have no Miller comp cap across the VAS.MJL21193 said:
They use the LTP/VAS collector cap instead, but I don't have the skills to make any of my versions stable with this arrangement.
The Miller comp capless sounds better in my opinion.
jacco vermeulen said:
Hi jacco,
That's a good replacement, for sure!
I need to remember to ask you next time - you are the defacto king of devices.
I tried at B&D and they have plenty of the 2SA1352 but there are only 10 of the 2SC3416 in stock. Do you know of another source? I'd like to have a "few" of these on hand...
Yes the pool is more trouble then it's worth, but on a hot summer day you soon forget that.
I also have a tendency to forget how much trouble it is when my delinquent 18 year old son brings a couple of his friends of the female persuasion over for a quick dip.
😀
I tried the cap from the VAS collector back to the LTP feedback and the result was wild instability.
I find myself in a quandary: If I leave the Miller cap at 47pF (or lower, I think it will run stable as low as 30pF), I sacrifice good clipping behavior. If I do this, what is the purpose of the clamping diodes? If I am willing to accept bad behavior while clipping to have a lower Cdom, why not get rid of the clamps?
In essence, what should I be shooting for? Ultimate sound quality (my hearing is way past the point where it can resolve the difference) or ultimate stability. It's still remarkable to me that this will run this smoothly without a Zobel or output inductor.
I find myself in a quandary: If I leave the Miller cap at 47pF (or lower, I think it will run stable as low as 30pF), I sacrifice good clipping behavior. If I do this, what is the purpose of the clamping diodes? If I am willing to accept bad behavior while clipping to have a lower Cdom, why not get rid of the clamps?
In essence, what should I be shooting for? Ultimate sound quality (my hearing is way past the point where it can resolve the difference) or ultimate stability. It's still remarkable to me that this will run this smoothly without a Zobel or output inductor.
Thats very odd about the wild instability, how large a cap did you try ??? I use the method mainly to get better phase response and do away with output coil and zobel. Like Andrew I find lower cdom amps to sound better.
From bdent you could get the 1381 brother the 1380, higher Ft same Cob and cheaper than 1352, look into 1406 too. Another important parameter is the early effect and the higher ft parts often perform better here. For drivers look into 2sa1249 and complement. MCM is a excellent source of Japanese transistors, often cheaper than bdent.
Decreasing miller ups the slewrate and as you mentioned better high frequency performance and sound quality. Trade offs, I hate them too. 150w is a lot of power for home use and 6 of them, will have to ask yourself whether you ever gonna use the amps into clipping.
Your wife might hate it, but your son will love the power, maybe hell bring more girlfriends around 😀 😀 😀
For your personal music enjoyment, think you should go for ultimate quality.
From bdent you could get the 1381 brother the 1380, higher Ft same Cob and cheaper than 1352, look into 1406 too. Another important parameter is the early effect and the higher ft parts often perform better here. For drivers look into 2sa1249 and complement. MCM is a excellent source of Japanese transistors, often cheaper than bdent.
Decreasing miller ups the slewrate and as you mentioned better high frequency performance and sound quality. Trade offs, I hate them too. 150w is a lot of power for home use and 6 of them, will have to ask yourself whether you ever gonna use the amps into clipping.
Your wife might hate it, but your son will love the power, maybe hell bring more girlfriends around 😀 😀 😀
For your personal music enjoyment, think you should go for ultimate quality.
homemodder said:
I used a 15pF first then a 10.
Yes, I didn't recognize MCM in jacco's post. They look like a better source.
homemodder said:
I'm leaning that way. I did some very loud listening last night with the scope connected. I couldn't see any clipping, even on the most powerful transients.
- Status
- Not open for further replies.