Another Discrete Op-Amp (Pics)
- Thread starter [silent:arts]
- Start date
Help Support GroupDIY Audio Forum:
PRR
I have something almost like that drawn. Alot of opamps share the diode(s) for the current source parts. The part I need to look at is the diodes often the SMDs diodes are two diodes in series.
tab anode cathode tab anode cathode tab or
tab anode cathode tab cathode anode tab. or
tab cathode anode tab anode cathode tab
IIRC.
I have something almost like that drawn. Alot of opamps share the diode(s) for the current source parts. The part I need to look at is the diodes often the SMDs diodes are two diodes in series.
tab anode cathode tab anode cathode tab or
tab anode cathode tab cathode anode tab. or
tab cathode anode tab anode cathode tab
IIRC.
> Alot of opamps share the diode(s) for the current source parts. The part I need to look at is the diodes often the SMDs diodes are two diodes in series.
Yes, I just don't know which way that is. If these are single diodes, they are probably stacked as a single reference. (but then why two 60K resistors?) If these are dual-diodes, then there is some point in isolating the two references. I guess the way I drew it, I've assumed dual-diodes.
I'm just glad I don't have to squint these teentsy parts.....
Yes, I just don't know which way that is. If these are single diodes, they are probably stacked as a single reference. (but then why two 60K resistors?) If these are dual-diodes, then there is some point in isolating the two references. I guess the way I drew it, I've assumed dual-diodes.
I'm just glad I don't have to squint these teentsy parts.....
![[silent:arts]](https://cdn.imagearchive.com/groupdiy/data/avatars/m/0/172.jpg?1620371656){kind=avatar}
[silent:arts]
Well-known member
[quote author="PRR"]> don't know if this makes sense
It doesn't.
This is my best guess:
[/quote]
thanks PRR,
I will retrace and check this.
It doesn't.
This is my best guess:
thanks PRR,
I will retrace and check this.
They are dual diodes, wired like a BAV-99.
They're not so tiny under a stereo microscope.
PRR's circuit is correct.
Not 100% sure on the output emitter resistors, the two samples I looked at were toasted. I got 22 on one that looked undamaged, but no confimation.
My buddies board eats these things like jellybeans, like 10 or so in the last two years, so we're looking to build some improved replacements.
They're not so tiny under a stereo microscope.
PRR's circuit is correct.
Not 100% sure on the output emitter resistors, the two samples I looked at were toasted. I got 22 on one that looked undamaged, but no confimation.
My buddies board eats these things like jellybeans, like 10 or so in the last two years, so we're looking to build some improved replacements.
..in transistor sockets.. :shock:
Naw, probably still surface mount, it isn't that tough, but maybe two small boards with output stage improvements in dissipation, like parallel devices with current sharing resistors, and some heatsinking.
Just started thinking about it Wednesday, traced the circuit out Wednesday night after work, took a 12 pack between three of us, and then lo and behold this thread shows up!
We also sussed out the neccesary parts to buy to rebuild a Farfisa organ. The beer wasn't all used up on the opamp...
Just started thinking about it Wednesday, traced the circuit out Wednesday night after work, took a 12 pack between three of us, and then lo and behold this thread shows up!
We also sussed out the neccesary parts to buy to rebuild a Farfisa organ. The beer wasn't all used up on the opamp...
rafafredd
Well-known member
here is another version of one of our DIY forum friends... I´ve got it and redrawn, so here it is:
The 1pF caps are not 1pF. It´s just I don´t have the correct values, but we should be able to figure it out.
The 1pF caps are not 1pF. It´s just I don´t have the correct values, but we should be able to figure it out.
SonsOfThunder
Well-known member
Dan,
Any ideas on those unidentified transistors?
As far as heatsinking goes, it would be supercool if you could do this on T-Clad! Anyone familiar?
Peace!
Charlie
Any ideas on those unidentified transistors?
As far as heatsinking goes, it would be supercool if you could do this on T-Clad! Anyone familiar?
Peace!
Charlie
> My buddy's board eats these things like jellybeans, like 10 or so in the last two years,
I noticed it had no short-protection, but assumed it was deep inside the box where it would not get fatal abuse. 5/year? I'd cobble a 5534 or even a TL071 onto the SIP-pins: a plain-vanilla amp that works is better than an amp that dies when the musicians are hot.
Obvious first step is some current limiting on the output. Back-diodes on the output B-E junctions, with the Vbe bias-net, would limit the voltage across the emitter resistors to about 1.2/Re. If Re is 33Ω, 36mA. If it stalls in that condition, 0.54W in the transistor. But then you have to check the volt-amp stage... hmmmm, also 0.56W worst-case, kinda hot (though kinda unlikely). Another 100Ω in the collectors would cut that in half, without much effect on normal operation.
Is that resistor-ratio on the Vbe multiplier right? It looks like 1.6*Vbe, which underbiases the outputs. They may be idling at 0.050mA, awful poor. Output impedance will fall from 600Ω teeny-signal to 35Ω large-signal, too much variation for typical 1K-10K loadings. A richer bias might take some grit out.
Overall, the only thing I really like is the logo. The circuit seems vanilla. There's little excuse for underbiasing an AB output in discrete audio. The input stage current is large enough to be a noise issue in the balanced line-in circuits posted by silent:arts. In fact a 5532 would give lower noise, fine sound, and be failure resistant. This 0.25mA input emitter current is more for 1K-2K sources than those 13K sources. The 5532 minimizes at 9K and will still be low noise at 13K.
> output stage improvements in dissipation, like parallel devices with current sharing resistors, and some heatsinking.
Or that. Not knowing the load, I dunno if it needs more output, or just protection from "accidents".
> sussed out the necessary parts to buy to rebuild a Farfisa organ.
Now, that's going TOO FAR!!! You will give DIY a bad name. There is no excuse for that. Let dead dogs lie.
I noticed it had no short-protection, but assumed it was deep inside the box where it would not get fatal abuse. 5/year? I'd cobble a 5534 or even a TL071 onto the SIP-pins: a plain-vanilla amp that works is better than an amp that dies when the musicians are hot.
Obvious first step is some current limiting on the output. Back-diodes on the output B-E junctions, with the Vbe bias-net, would limit the voltage across the emitter resistors to about 1.2/Re. If Re is 33Ω, 36mA. If it stalls in that condition, 0.54W in the transistor. But then you have to check the volt-amp stage... hmmmm, also 0.56W worst-case, kinda hot (though kinda unlikely). Another 100Ω in the collectors would cut that in half, without much effect on normal operation.
Is that resistor-ratio on the Vbe multiplier right? It looks like 1.6*Vbe, which underbiases the outputs. They may be idling at 0.050mA, awful poor. Output impedance will fall from 600Ω teeny-signal to 35Ω large-signal, too much variation for typical 1K-10K loadings. A richer bias might take some grit out.
Overall, the only thing I really like is the logo. The circuit seems vanilla. There's little excuse for underbiasing an AB output in discrete audio. The input stage current is large enough to be a noise issue in the balanced line-in circuits posted by silent:arts. In fact a 5532 would give lower noise, fine sound, and be failure resistant. This 0.25mA input emitter current is more for 1K-2K sources than those 13K sources. The 5532 minimizes at 9K and will still be low noise at 13K.
> output stage improvements in dissipation, like parallel devices with current sharing resistors, and some heatsinking.
Or that. Not knowing the load, I dunno if it needs more output, or just protection from "accidents".
> sussed out the necessary parts to buy to rebuild a Farfisa organ.
Now, that's going TOO FAR!!! You will give DIY a bad name. There is no excuse for that. Let dead dogs lie.
rafafredd
Well-known member
PRR,
What do you think about the last schem I posted. The guy who sent me said it was reverse engeneered from a working unit. Any guesse on the cap values? I was thinkig 47pF for the output cap and 10pF for the others...
What do you think about the last schem I posted. The guy who sent me said it was reverse engeneered from a working unit. Any guesse on the cap values? I was thinkig 47pF for the output cap and 10pF for the others...
> Any guess on the cap values? I was thinkig 47pF for the output cap and 10pF for the others...
The cap across the bias transistor might not be needed at all. It suggests that bias transistor current might fall very low, which isn't good for the sound. But transients might do that. I think a minimum size is much-much larger than transistor base capacitance, over 1,000pFd. 0.1uFd is common in fat-finger work. I've seen 4.7uFd in loudspeaker output stages.
C3 across the top PNP B-E junction looks wrong. B-E caps tend to be a cheap fix for a misunderstood problem. I'd put no more than 50pFd there, probably nothing at all until I had a 50MHz scope and low-C probe to poke around.
Assuming the transistors are common "400MHz" types, and the opamp has to be unity-gain stable, the overall gain-bandwidth has to be less than 400MHz. Since there are only 3 current-gain stages, probably a lot less. If we need an effective β of 20 in each stage so gain does not collapse, and since some stages work at low current where Ft won't be 400MHz, I'd say 10MHz-20MHz (do we know what it claims?). Input 1/Gm is about 240Ω, so I get compensation cap C2 of about 40pFd. Slew-rate around 6V/uS. The 1K2 resistor will break the pole around 3MHz, which I guess is where the output devices' input impedance turns capacitive from Ft β-falloff. But I'd need a lot more thought to understand where this is going.
And on the face of it, I'm not sure how it is better than some readily available ICs (except the logo and the fact it fits an API socket). I don't think it is bad, but if there is greatness in there I'm missing it.
The cap across the bias transistor might not be needed at all. It suggests that bias transistor current might fall very low, which isn't good for the sound. But transients might do that. I think a minimum size is much-much larger than transistor base capacitance, over 1,000pFd. 0.1uFd is common in fat-finger work. I've seen 4.7uFd in loudspeaker output stages.
C3 across the top PNP B-E junction looks wrong. B-E caps tend to be a cheap fix for a misunderstood problem. I'd put no more than 50pFd there, probably nothing at all until I had a 50MHz scope and low-C probe to poke around.
Assuming the transistors are common "400MHz" types, and the opamp has to be unity-gain stable, the overall gain-bandwidth has to be less than 400MHz. Since there are only 3 current-gain stages, probably a lot less. If we need an effective β of 20 in each stage so gain does not collapse, and since some stages work at low current where Ft won't be 400MHz, I'd say 10MHz-20MHz (do we know what it claims?). Input 1/Gm is about 240Ω, so I get compensation cap C2 of about 40pFd. Slew-rate around 6V/uS. The 1K2 resistor will break the pole around 3MHz, which I guess is where the output devices' input impedance turns capacitive from Ft β-falloff. But I'd need a lot more thought to understand where this is going.
And on the face of it, I'm not sure how it is better than some readily available ICs (except the logo and the fact it fits an API socket). I don't think it is bad, but if there is greatness in there I'm missing it.
SonsOfThunder
Well-known member
I've seen 100uF there before. Not that unusual when you have unregulated switching supplies.
If PRR's assessment is correct (and it very probably is), this DOA is quite a disappointment. And we get back to what my mentor says..."what's the point of a DOA if there are IC amps that are good enough?"
I understand it is about flava'
Peace!
Charlie
StephenGiles
Well-known member
How much better would one of these be than a top notch OPA opamp like the 2228? Would you hear the difference..........is it that much better than a 4558...........really at line level..........be honest.
Stephen
Stephen
> How much better would one of these be than a top notch OPA opamp like the 2228?
I don't know if it is better. Charlie says flava', you ask "Would you hear the difference..."; the ear is the only true judge and I have mot heard one.
> ..is it that much better than a 4558
Ought to be a whole lot better than a 1MHz 0.5V/uS '558. Unless you hold levels down to 1V, the '558 will slew-limit most music, and can't work at high gain without running out of NFB at 20KHz. The '558 may be equal-or-better for noise from 50K sources, and a lot better for DC error, but the API ought to be the better audio device in most real uses. (However, it is darn hard to kill ten '558s in two years.....)
Compared to 5534... depends on how bad the 5534 is. The older ones seem to be good, the current batch may be a bit careless about AB current or something.
But if API used 5534, LM833, TL071, NJM558, etc, they'd be "same as" B-ringer et al. Sometimes you just have to do it different.
What does this thing cost? I mean, any of us can flip through DigiKey and see it is about $2 of parts on a $2 board (and that's generous). In quantity 10,000, the assembly is cheap. Dunno if they do any adjust-on-test... seems difficult?
I don't know if it is better. Charlie says flava', you ask "Would you hear the difference..."; the ear is the only true judge and I have mot heard one.
> ..is it that much better than a 4558
Ought to be a whole lot better than a 1MHz 0.5V/uS '558. Unless you hold levels down to 1V, the '558 will slew-limit most music, and can't work at high gain without running out of NFB at 20KHz. The '558 may be equal-or-better for noise from 50K sources, and a lot better for DC error, but the API ought to be the better audio device in most real uses. (However, it is darn hard to kill ten '558s in two years.....)
Compared to 5534... depends on how bad the 5534 is. The older ones seem to be good, the current batch may be a bit careless about AB current or something.
But if API used 5534, LM833, TL071, NJM558, etc, they'd be "same as" B-ringer et al. Sometimes you just have to do it different.
What does this thing cost? I mean, any of us can flip through DigiKey and see it is about $2 of parts on a $2 board (and that's generous). In quantity 10,000, the assembly is cheap. Dunno if they do any adjust-on-test... seems difficult?
[silent:arts]
Well-known member
Thank you very much PRR
I will exchange the SMD thingy to NE5534 at two channels.
I will measure original and NE5534 units.
I will wait a few weeks if the golden ears of the engineers will notice a difference.
If not I will ask them to compare the units (original to NE5534).
We will see (or hear) if there is any magic. If not ? why should I repair the stupid SMD stuff, or pay 70 $?
A*P*I didn?t even respond to any email I send because of repair / replacement etc (and it is still under warranty...)
I will let you know.
I will exchange the SMD thingy to NE5534 at two channels.
I will measure original and NE5534 units.
I will wait a few weeks if the golden ears of the engineers will notice a difference.
If not I will ask them to compare the units (original to NE5534).
We will see (or hear) if there is any magic. If not ? why should I repair the stupid SMD stuff, or pay 70 $?
A*P*I didn?t even respond to any email I send because of repair / replacement etc (and it is still under warranty...)
I will let you know.
[silent:arts]
Well-known member
too busy to do the test yet, sorry.
have to do it the next few weeks ...
have to do it the next few weeks ...
[silent:arts]
Well-known member
this thread just came back to my mind. :roll:
ok, I did it another way ("one wire repair"): I connected the output of the TL084 sidechain input to the audio path. measured no difference. nobody heard a diffference yet. and it's in daily production ... :green:
thanks, PRR :thumb:
ok, I did it another way ("one wire repair"): I connected the output of the TL084 sidechain input to the audio path. measured no difference. nobody heard a diffference yet. and it's in daily production ... :green:
thanks, PRR :thumb:
