Today I checked again, and the 5U 400 chassis has plenty room to just mount an extra OS on top section of sinks, with top row MOSFETs at the top row of pre drilled holes. All that is needed is a few precision drilled holes for mounting the OS.
So, I now run the sinks at more than max limit, meaning dissipation around 120 watts per channel. I figure 108 watts is pretty safe. At 22v5 under load with new PSU, that means I can bias each device at approx 400mA running 6 deep, or approx 800mA 3 deep.
The question then, is whether it is better to run 12 devices at a lowish 0,4 amps, or 6 devices at 0,8? 0,4 is the F4 spec anyways, so can’t be bad I figure. I guess what I wanna know is how low a bias is too low.
Mighty: I know, I should post IQ figures per vertical pair and all. Please bear with me: approx max bias will be 2A4. Best distributed across 12 devices or 6?
Think I know the answer, but here we go.
I figure more devices can actually take more temp with lower dissipation per device, before reaching the never to cross junction limit, so win win I guess. At the same, what about distortion figures? 6L6s F4 guide showed reduced distortion above 400mA, with diminishing returns above 600mA. PL often biases around 14 watts. My case will be approx 9 and in any case sub 10. not taking into account Papas typos, the BA-2 stock bias is 500mA per device 6 deep. So i will not reach that figure, by approx 100mA.
Hugz and kissez,
Andy
So, I now run the sinks at more than max limit, meaning dissipation around 120 watts per channel. I figure 108 watts is pretty safe. At 22v5 under load with new PSU, that means I can bias each device at approx 400mA running 6 deep, or approx 800mA 3 deep.
The question then, is whether it is better to run 12 devices at a lowish 0,4 amps, or 6 devices at 0,8? 0,4 is the F4 spec anyways, so can’t be bad I figure. I guess what I wanna know is how low a bias is too low.
Mighty: I know, I should post IQ figures per vertical pair and all. Please bear with me: approx max bias will be 2A4. Best distributed across 12 devices or 6?
Think I know the answer, but here we go.
I figure more devices can actually take more temp with lower dissipation per device, before reaching the never to cross junction limit, so win win I guess. At the same, what about distortion figures? 6L6s F4 guide showed reduced distortion above 400mA, with diminishing returns above 600mA. PL often biases around 14 watts. My case will be approx 9 and in any case sub 10. not taking into account Papas typos, the BA-2 stock bias is 500mA per device 6 deep. So i will not reach that figure, by approx 100mA.
Hugz and kissez,
Andy
Last edited:
simple as this: once when you're going paralleling output devices, more of them, means greater balls
not necessary needed, but in general and roughly speaking , greater balls giving greater sense of authority of sound or, if you wish, greater sense of ease
illustration - parallel between things I made : Babelfish XJ I made are not able to give immediacy and Green-evoking little tiny details, which smaller (FW format) amps can give
but, B. XJ are having some grandiose easiness in performance which is eluding, so I admit that I started thinking about smallest possible iteration for my needs
worth mentioning that B. XJ were thoroughly developed/voiced to have same 2nd dominant nature, as FW format amps are having; well, most of them, so this is apple-apple comparison
so, conclusion - number of pairs - go wild
regarding definitions - pairs, Iq, etc. - as long it's clear what you think and wrote, define it as you wish
not necessary needed, but in general and roughly speaking , greater balls giving greater sense of authority of sound or, if you wish, greater sense of ease
illustration - parallel between things I made : Babelfish XJ I made are not able to give immediacy and Green-evoking little tiny details, which smaller (FW format) amps can give
but, B. XJ are having some grandiose easiness in performance which is eluding, so I admit that I started thinking about smallest possible iteration for my needs
worth mentioning that B. XJ were thoroughly developed/voiced to have same 2nd dominant nature, as FW format amps are having; well, most of them, so this is apple-apple comparison
so, conclusion - number of pairs - go wild
regarding definitions - pairs, Iq, etc. - as long it's clear what you think and wrote, define it as you wish
Think I have to try that, then.
1: You use some really cool self-hinging sockets when you drill sinks. What are they called/where can they be found? I’ll do some searching
2: Need for more degen with more parallel devices? Spec is 1R. 3 deep I use 0R5, but some of the Store matched fets have up to 19% pockets. Wanna reduce that.
1: You use some really cool self-hinging sockets when you drill sinks. What are they called/where can they be found? I’ll do some searching
2: Need for more degen with more parallel devices? Spec is 1R. 3 deep I use 0R5, but some of the Store matched fets have up to 19% pockets. Wanna reduce that.
speaking of more pairs, I presumed source resistors same in all cases
biasing not being critical in BA3, you can stay with lowest possible value for Rs
(in some other amps, to be able to decrease Iq per device - increasing their number but staying at same sum Iq, you need to increase value of Rs; not case her, luckily)
self hinging sockets?
not sure I understand question
maybe you're asking about sheet metal stepped drills?
or centering drills ?
linked content is just for visual identification, didn't look where nor price nor quality
biasing not being critical in BA3, you can stay with lowest possible value for Rs
(in some other amps, to be able to decrease Iq per device - increasing their number but staying at same sum Iq, you need to increase value of Rs; not case her, luckily)
self hinging sockets?
not sure I understand question
maybe you're asking about sheet metal stepped drills?
or centering drills ?
linked content is just for visual identification, didn't look where nor price nor quality
Last edited:
I see. That’s good news. I just thought, reading 6L6s guide (using 1Rs) and the manual, that 6 deep «requires» 1R. But guess that’s due to the fets being handed out at BAF weren’t that closely matched. I know Itsallinmyhead uses 0R33 in 6 deep, if I remember correctly. So that’s smooth.
The build I was referring to is this one: Babelfish XJ , or JX …….. or whatever – Zen Mod Blog
Though I see that you used countersunk hex screws. Smooth technique! And letting the PCB float on the MOSFET legs was a smart move. Somehow I think I read your blog, a build were you drilled larger holes than M3, and then put some kind of socket into that hole, and that socket had an M3 predrilled hole in it. Maybe I remember incorrectly.
The build I was referring to is this one: Babelfish XJ , or JX …….. or whatever – Zen Mod Blog
Though I see that you used countersunk hex screws. Smooth technique! And letting the PCB float on the MOSFET legs was a smart move. Somehow I think I read your blog, a build were you drilled larger holes than M3, and then put some kind of socket into that hole, and that socket had an M3 predrilled hole in it. Maybe I remember incorrectly.
Last edited:
I'm never using self-tapping screws (as for sheet-metal), except in case they came as part of enclosure - example - Modushop heatsinked cases are having these screws for top and bottom panels
everything else is matter of moment's decision - choice between ideal solution, what I have on hand or have to go to nearest hardware store or order and wait for it; all pretty much usual and routine things
regarding threaded inserts - I'm rarely doing that; sometimes when I get pre-fabricated front plates, they're having that; in fact - that's perfect solution for wooden elements - when you want to put machine screws in wood, though, there is difference in outer thread for these elements made for wood application and for metal application
just google for threaded inserts ........ though, as I said - I see their usage when combining wood with metal, for all metal you don't need them except in case of salvaging results of your own sloppiness
everything else is matter of moment's decision - choice between ideal solution, what I have on hand or have to go to nearest hardware store or order and wait for it; all pretty much usual and routine things
regarding threaded inserts - I'm rarely doing that; sometimes when I get pre-fabricated front plates, they're having that; in fact - that's perfect solution for wooden elements - when you want to put machine screws in wood, though, there is difference in outer thread for these elements made for wood application and for metal application
just google for threaded inserts ........ though, as I said - I see their usage when combining wood with metal, for all metal you don't need them except in case of salvaging results of your own sloppiness
Last edited:
I used 0R22. I chose the tightest matches I had from two tubes of Ns and Ps each. Now... whether MY matching actually means that they are good matches is another thing altogether.
Two batches of my store N-channel FETs are respectively in 12% and 19% bracket, even with 1% Vishay source resistors. I tried to complain, but they answered Papa matched them so the fault most likely is mine. Never found a fault on my side. But that doesn’t mean it is not there.
From the same store batch the N channels are within 3%. Go figure...
How tight are the voltage drop readings across those with only 0R22? That’s some low degen you got going there. Kudos!
From the same store batch the N channels are within 3%. Go figure...
How tight are the voltage drop readings across those with only 0R22? That’s some low degen you got going there. Kudos!
Last edited:
MOSFETs, the ones we discussed last autumn on PM (not expecting you to remember that, even with a mind the size of a basketball).
Batch 1: all P-chans within 3% at 600mA. 3 N-chans within 3%. 3 N-chans within 19% (1 x MOSFET is misbehaving, the two others are very close).
Batch 2: Bought to replace the «bad» ones from batch 1. from This batch, I have put to use 3 Ps and 3 Ns, one OS. Ps are fine, within 5%-ish. N’s are a bit off. Drops across 0R5 are now 305/325/342.
They are matched for F4 working conditions, so at 22v5 and not 30 like I have right now (new PSU not assembled).
Too speedy for you? Wonder if I shouldn’t take that as a compliment
Regards,
Andy
Batch 1: all P-chans within 3% at 600mA. 3 N-chans within 3%. 3 N-chans within 19% (1 x MOSFET is misbehaving, the two others are very close).
Batch 2: Bought to replace the «bad» ones from batch 1. from This batch, I have put to use 3 Ps and 3 Ns, one OS. Ps are fine, within 5%-ish. N’s are a bit off. Drops across 0R5 are now 305/325/342.
They are matched for F4 working conditions, so at 22v5 and not 30 like I have right now (new PSU not assembled).
Too speedy for you? Wonder if I shouldn’t take that as a compliment
Regards,
Andy
Been trying to avoid that, though always - since our chat last autumn - I have known this must be the next step of my DIY career.
I do have a quite good el half cheapo adjustable bench PSU that even has current limiting. So that’s one step of the way. Getting fets at the moment from Mouser/Digikey is more difficult, so currently I am actually doing some searches IOT find suitable devices. Harrises are available from Rochester, but minimum buy of 84 pcs. So considering 9240/240 PBFs. The 2nd harmonic P channel «problem» provides some lovely warmth, so it is the easiest way to land this thing =)
Gonna look up some matching jigs, I think I’ll make one on a breadboard or something. Or extend the pads on one of the unused OS’es I have to give them even more real life working conditions during matching.
Gotten good advice from Dennis on which devices to choose, so this has been in the makings for some time.
Cheers, and thanks for putting up with me! =)
I do have a quite good el half cheapo adjustable bench PSU that even has current limiting. So that’s one step of the way. Getting fets at the moment from Mouser/Digikey is more difficult, so currently I am actually doing some searches IOT find suitable devices. Harrises are available from Rochester, but minimum buy of 84 pcs. So considering 9240/240 PBFs. The 2nd harmonic P channel «problem» provides some lovely warmth, so it is the easiest way to land this thing =)
Gonna look up some matching jigs, I think I’ll make one on a breadboard or something. Or extend the pads on one of the unused OS’es I have to give them even more real life working conditions during matching.
Gotten good advice from Dennis on which devices to choose, so this has been in the makings for some time.
Cheers, and thanks for putting up with me! =)
Last edited:
2 reels of 240s on the way, as well as parts for making jig. Thanks for the kick in the beeehiiiiind. 9240s to come later, different source. Plus I want to see if I can do this
Hi, all!
Long time no posts in this thread I made. I have put this project on a temporary halt due to building the Aleph J Zen (prototype boards hopefully arriving in a little while), as well as other things in life needing priority. Project though is still i process, in my mind.
None of your posts are forgotten, all very appreciated, and all stored here and in memory, and this thread will be revisited once I am ready.
Long story short, I got hold of 100 MOSFETs to be matched, choice of depth to be determined later, chassis too,
But what I am most pleased about, is getting original Toshiba J74/K170’s matched within 0.01mA @ 9.99mA. I will hold on to these and treat them like treasure.
For those unaware, I have a running BA-3 allready. To this I want to share that increasing FE bias steadily, almost until melting point of fets, gave very, very audible results. A shame this amp will now be dismantled, on to other projects, but never to be forgotten, and to be revisited - again.
What bias current was that? 60mA / 1,8 watts per device made me not think that anymore. Grit gone, more depth, more palpative. Yanno, all of that class A stuff just there.
Sinks a bit small. So: Try at your own risk, but be sure: it might be well worth it. At least it was in my circuit with Fairchild FE output MOSFETs, 30v rails under load, and JFETs with lowish approx 6mA IDSS.
Regards,
Andy
Long time no posts in this thread I made. I have put this project on a temporary halt due to building the Aleph J Zen (prototype boards hopefully arriving in a little while), as well as other things in life needing priority. Project though is still i process, in my mind.
None of your posts are forgotten, all very appreciated, and all stored here and in memory, and this thread will be revisited once I am ready.
Long story short, I got hold of 100 MOSFETs to be matched, choice of depth to be determined later, chassis too,
But what I am most pleased about, is getting original Toshiba J74/K170’s matched within 0.01mA @ 9.99mA. I will hold on to these and treat them like treasure.
For those unaware, I have a running BA-3 allready. To this I want to share that increasing FE bias steadily, almost until melting point of fets, gave very, very audible results. A shame this amp will now be dismantled, on to other projects, but never to be forgotten, and to be revisited - again.
What bias current was that? 60mA / 1,8 watts per device made me not think that anymore. Grit gone, more depth, more palpative. Yanno, all of that class A stuff just there.
Sinks a bit small. So: Try at your own risk, but be sure: it might be well worth it. At least it was in my circuit with Fairchild FE output MOSFETs, 30v rails under load, and JFETs with lowish approx 6mA IDSS.
Regards,
Andy