To get the bias up with two Si transistors you just need two diodes. Actually, a regular 1N4001 and a schottky in series ought to be about right. Two 1N4001’s is too much drop - you’d have to go to 6A diodes (or even 10) to get the right drop. A small schottky is easier to add. That resistor (22 ohm) that goes in parallel is used to trim the current. The whole works will end up being more reliable, and easier to source.
No reason to go special ordering 2N types. I just happen to have a big stash that cost me very little. I use them in low voltage circuits like these so I don’t use up ones suitable for bigger amps that cost more. A couple years before COVID, all sort of 60-80V epi-base TO-3 flooded the surplus market. Dirt cheap. I took advantage. I passed on all the horizontal output transistors that showed up around the same time. No use for them. When buying new transistors at full price, there is no reason to go lower than MJ15003 for any reason. If you have good pulls it’s a judgement call. The customer has to understand that’s what they are getting, and approve.
The TIP41 driver probably isn’t hurting anything with that .001 uF miller cap. If I was revamping the circuit I’d try to make it (the cap) smaller and go with a better transistor. Even a BD139. I have a bunch of old D40-D44 series.
BTW “sustained beta driver” = C3296/A1306, or similar, current production part….
No reason to go special ordering 2N types. I just happen to have a big stash that cost me very little. I use them in low voltage circuits like these so I don’t use up ones suitable for bigger amps that cost more. A couple years before COVID, all sort of 60-80V epi-base TO-3 flooded the surplus market. Dirt cheap. I took advantage. I passed on all the horizontal output transistors that showed up around the same time. No use for them. When buying new transistors at full price, there is no reason to go lower than MJ15003 for any reason. If you have good pulls it’s a judgement call. The customer has to understand that’s what they are getting, and approve.
The TIP41 driver probably isn’t hurting anything with that .001 uF miller cap. If I was revamping the circuit I’d try to make it (the cap) smaller and go with a better transistor. Even a BD139. I have a bunch of old D40-D44 series.
BTW “sustained beta driver” = C3296/A1306, or similar, current production part….
That s what i thought as well, reducing the IPS current by 4x and hence increasing R96 to 4.7k and the attached 470R resistance in the Pcd6 to 2.2k as well as R95 to 680R all while reducing the huge 1nf miller cap to 100pF, this should keep the DC operating points unchanged, and of course increase the output cap to something like 4700uF.
Also increase by 10x R99/R100, use 4 1N4004 diodes for bias and a pair of cheap complementary darlingtons as OS,
if TO220 can be adapted then a pair of BDW93/94 wich are enough for a 41.5V PSU, or else any pair of cheap
enough T03s that can be found at low price, dunno if MJ901/1001, or better MJ2500-2501/3000-3001, are still available.
I’m definitely good to make any changes needed to the circuit. For the bias diodes, one being schottky, I can place an order if necessary once I have a full list, but I have a bunch of 1N5819 (40v 1A schottky). Keep the 22 ohm in parallel alone? I’m totally good with leaving the MJ15003 in there if you think they’ll do the job well, they’re of course brand new. I’ll swap out the TIP41 for the BD139, have plenty of them. I have a fairly large collection of transistors, so if you think something might be better…
Should I drop the miller cap to 100 pF?
I really appreciate all of the tips. If I’m merging your suggestions with wg_ski I guess I’d just want to make sure they mesh together well. I really appreciate you giving me the value changes along with the device number.
For Darlington I have a crap load, 110 pair of MJ11015 and MJ11016, would those work well? Never use them really for anything else so would be cool, but if they’re not a good fit I get it.
Dan
The shunt 22 ohm resistor would go up if the bias needs to go up, down if it needs to go down. The lower the current that goes through the bias diodes, the lower the Vf. Any time you swap out transistor types it must be addressed. That schottky will be fine, in series with a 1N4001,2,3 it will put you in the right range to bias two Si devices (about 1 to 1.1 volts for these big boys).
Make sure that BD139 goes on the little provided heat sink - it will be dissipating 2W. Fine, but not naked. You can probably drop the miller cap value with BD139 driving MJ15003/4. Just make sure the small signal square wave doesn’t have a bunch ringing.
If you do go to darlingtons the bias stack will need more fiddling. The typical thing that’s done is to use three diodes and have it deliberately underbiased. It’s a no muss no fuss solution that sounds “ok-ish”. To those who actually build amplifiers it’s TERRIBLE, usually worse than the EF1. Get it up to 10 or 20 mA, and it sounds like it SHOULD. Getting the proper starting point for bias is tricky without good spice models. Three diodes is not enough. Four is WAY too much, and might burn right up before you can even get the current measured. The right value is therefore something in between. Thats right, swap out one or two for schottkys. I’d build two stacks. Try the lower drop one first, and if you are underbiased with no parallel resistor then try the higher drop one (With the resistor, then trim). This is the reason they started doing vbe multipliers - adjustment is a simple.
If darlingtons are used, it opens up the option of reducing the VAS (previously driver) current down to what it usually is. I like to run about 6 mA, maybe 10. 100 mA is silly, but necessary for EF1. This current does affect how much bias drop is needed for the output stage, so if you do decide on darlingtons, and are considering VAS current change, do this FIRST, not after the bias is dialed in.
MJ11015,11016 would be my go-to. I don’t have 110 pairs, but probably more than I need in a lifetime. There are a lot of vintage receivers that these will work in. My bigger stash is actually the 11032/11033. MJ2500/3000 just didn’t exist when this thing was made, or they just would have designed around them.
The reason you have a big stash of transistors is so you don’t have to go ordering parts every time you turn around.
Make sure that BD139 goes on the little provided heat sink - it will be dissipating 2W. Fine, but not naked. You can probably drop the miller cap value with BD139 driving MJ15003/4. Just make sure the small signal square wave doesn’t have a bunch ringing.
If you do go to darlingtons the bias stack will need more fiddling. The typical thing that’s done is to use three diodes and have it deliberately underbiased. It’s a no muss no fuss solution that sounds “ok-ish”. To those who actually build amplifiers it’s TERRIBLE, usually worse than the EF1. Get it up to 10 or 20 mA, and it sounds like it SHOULD. Getting the proper starting point for bias is tricky without good spice models. Three diodes is not enough. Four is WAY too much, and might burn right up before you can even get the current measured. The right value is therefore something in between. Thats right, swap out one or two for schottkys. I’d build two stacks. Try the lower drop one first, and if you are underbiased with no parallel resistor then try the higher drop one (With the resistor, then trim). This is the reason they started doing vbe multipliers - adjustment is a simple.
If darlingtons are used, it opens up the option of reducing the VAS (previously driver) current down to what it usually is. I like to run about 6 mA, maybe 10. 100 mA is silly, but necessary for EF1. This current does affect how much bias drop is needed for the output stage, so if you do decide on darlingtons, and are considering VAS current change, do this FIRST, not after the bias is dialed in.
MJ11015,11016 would be my go-to. I don’t have 110 pairs, but probably more than I need in a lifetime. There are a lot of vintage receivers that these will work in. My bigger stash is actually the 11032/11033. MJ2500/3000 just didn’t exist when this thing was made, or they just would have designed around them.
The reason you have a big stash of transistors is so you don’t have to go ordering parts every time you turn around.
That is indeed why I have so many parts on had, I really do hate waiting on parts. Nothing worse than ordering the exact number of parts you need and then something happens to kill a part. Speaking of parts, does anyone know the brand of these transistors? Are they Signetics?
I certainly want to try with the Darlingtons, but as I mentioned I’d need my hand held lol, so I’ll have questions. Everything so far seems pretty straight forward. For the biasing of outputs, instead of a single 1N4001 and a schottky (three is not enough) maybe start with two 1N4001 and two schottky and go from there?
Reducing VAS current (previously driver), are you referring to Q15 and Q19? Which will become the BD139 mounted to the heatsinks? That’s what I see when you say driver. How would that current be dropped, by replacing the 150 ohm resistor? If so a recommendation where to start in value?
Thank you very much!
Dan
I certainly want to try with the Darlingtons, but as I mentioned I’d need my hand held lol, so I’ll have questions. Everything so far seems pretty straight forward. For the biasing of outputs, instead of a single 1N4001 and a schottky (three is not enough) maybe start with two 1N4001 and two schottky and go from there?
Reducing VAS current (previously driver), are you referring to Q15 and Q19? Which will become the BD139 mounted to the heatsinks? That’s what I see when you say driver. How would that current be dropped, by replacing the 150 ohm resistor? If so a recommendation where to start in value?
Thank you very much!
Dan
No, not Signetics. It’s Solid State Inc. Its sort of an industrial-military version of NTE. Same parent company, same address. Are they any good, that is the $64000 question. They are not outright fakes, but may not be up to what would expect from an original ON, or old Motorola like I have on hand. Good enough to use in an amplifier with only a single 44V rail, for sure. If they were in one of the old Marantz with +/-50 volt supplies that had the old aluminum house numbers originally, I’d be cautious for sure.
Two IN4001 and two schottky is where I would start. With a couple hundred ohms in parallel. Increase R91 and 94 by a factor of ten. That drops the current in Q15 to something more traditional. Current in Q14 is set by the 150 ohm resistor and could be left alone. If the center voltage moves, I’d correct for that - need to see by how much. It should stay close to 20.8V. Then get the bias dialed in.
Two IN4001 and two schottky is where I would start. With a couple hundred ohms in parallel. Increase R91 and 94 by a factor of ten. That drops the current in Q15 to something more traditional. Current in Q14 is set by the 150 ohm resistor and could be left alone. If the center voltage moves, I’d correct for that - need to see by how much. It should stay close to 20.8V. Then get the bias dialed in.
That’s one thing that kind of sucks, they’re not matched. All of the MJ11016 I have are Motorola and then these are the MJ11015. Would it be wire to measure them on a curve tracer? Obviously the Peak Atlas wouldn’t stress them at all, but I do have a Heathkit IT-3121 and a Hickok 440 which will test at much higher voltages and currents. Hopefully being an industrial-military version they’re built to higher standards.
Of course with their company name it makes it difficult to find opinions on them.
Dan
Last edited:
The only thing that could “go wrong” is if the emitter-base resistors that are built in end up being different values between brands. On those parts ST’s and ON (Motorola) are the same. On TIP142/147 they are not. Really depends where SS Inc. got them/had them made. The NTE “ECG number” line often comes from questionable sources, as they are intended for use in consumer equipment. SS offers only a handful of types, legacy devices that are relatively low volume but just refuse to go away. Probably because they are designed into a lot of old infrastructure and need to still be available. If they were the junk that the regular NTE line is, they would have gone away a long time ago. Measure the base to emitter resistance, with B-E reverse biased so it doesn’t interfere. See how close they are. The two resistors that make up the chain would be ratioed the same, so not getting to that middle node isn’t the end of the world. If Rbe on the drivers match, the ones on the outputs (a small but constant fraction of the total) will. If the Rbe’s don’t match, it will be harder to judge when you’ve got enough output stage bias, since all you can measure is the sum of output plus driver.
I’m trying to learn here, don’t get this depth into component testing usually. So I’m going off of what you said, did some reading and hopefully got it correct.
So measure the resistance with reverse bias. I measured the junction of B-E to determine the cathode and the anode of that given “diode”. I applied voltage from a power supply, didn’t know how much to use so I used 3v.
I applied + to the cathode or the “N” and - to the anode or “P” and measured. First was the MJ11015 Solid State devices and I measured 13.06 Mohm across the base and emitter. I measured 5 of the PNPs and all were dead on 13.06. Adjusting the 3v changed this resistance.
I then grabbed one of the Motorola MJ11016s and did the the same way, + to the cathode and - to the anode. At the NPNs supposed to be measured in the same was as the PNPs? Anyways the resistance on the Motorolas interestingly enough is the same, exactly 13.06 M.
Did I do this correctly? Should have I used a different voltage? If I did then it would seem okay to move forward with this as a pair? I did notice that the polarity on the two transistors were opposite, there the + of the voltage went to the emitter and on the other it went to the base and vise-versa.
Thanks again!
Dan
Hmmm. Base to emitter should be measuring about 8k. Reverse biased so the BE diodes don’t go into forward conduction and the only path is thru the resistors. Megohm readings indicate lack of these resistors, which is wrong for these transistor types.
Attachments
Should I be applying voltage from the power supply? if I remove the power supply voltage and just apply the meter across the pins, so that the + lead of the meter is on the cathode still I get closer readings.
The Motorolas I’m seeing anywhere from 10.6k to 11.6k in the group of five I measured.
For the MJ11015s I was seeing from about 6k to 6.6k in the group of 5.
Is it the “Rshunt”?
The Peak Atlas meter pics it up and is pretty close to what I’m seeing on the fluke. The Motorolas are measuring close to double of the Solid States.
Dan
I do have a number of these PNPs, Mospec which I’m sure aren’t great, but they match closer to the Motorolas.
Dan
Dan
No, you shouldn’t use a power supply at all. I just meant to be sure the ohm meter applied voltage backwards, as to not bias the diode. Opposite the polarity one would use for a “quick diode test” of a transistor. If your tester gives you Rshunt directly then that simplifies things.
Nothing any more wrong with Mospec than SS, I’ve been using them since the 90’s. Mouser used to sell them then. Central semi, Multicomp, NTE, SS, and others use them now. I’d use a pair that match more closely regardless of whose they are. The problem is the smaller resistor on the “output” will be off by 40% as well, resulting in 40% difference in bias in the drivers when the output with the lower Rbe finally kicks in.
Mospec will NOT have the overkill capacity than many Motorola/ON parts have, so use them within ratings. I’ve run real Motorolas way outside before - as have many early manufacturers did.
Nothing any more wrong with Mospec than SS, I’ve been using them since the 90’s. Mouser used to sell them then. Central semi, Multicomp, NTE, SS, and others use them now. I’d use a pair that match more closely regardless of whose they are. The problem is the smaller resistor on the “output” will be off by 40% as well, resulting in 40% difference in bias in the drivers when the output with the lower Rbe finally kicks in.
Mospec will NOT have the overkill capacity than many Motorola/ON parts have, so use them within ratings. I’ve run real Motorolas way outside before - as have many early manufacturers did.
Last edited:
Okay great, I’ll use the Mospec and Motorola. I figured they weren’t as robust, I wish they all were Motorola, but I didn’t originally buy them. I got them from a local tech that retired. He was super paranoid about fakes and purchased from authorized vendors only, so I know they’re good, but it’s too bad the PNPs are a bit cheaper. By using the Mospec there won’t be a 40% difference right? Now to work.
Dan
10.7 vs. 11k. Which would but Rbe on the outputs at 53 and 55 ohms. A difference you can probably live with. At 11 mA (0.5V vbe) you would start to see output stage bias. I’d shoot for about 15mA total by measuring drop in the 0.47 ohmers. That’s 2/3 of a watt in the output heat sink, per channel.
Make other adjustments (VAS current, center voltage, etc) before dialing it up. You can run underbiased while doing other adjustments, and should, for safety.
MJ11015/6 are overkill here, as the MJ2500/3000 or 900/1000 would have been used in early years. The old SAE parts from their little 60 watter (MK31B) were house numbered versions of those. I ran the hell out of that little amp and it never gave moment’s trouble. No concern about them blowing up at all, unless you somehow got China fakes.
Make other adjustments (VAS current, center voltage, etc) before dialing it up. You can run underbiased while doing other adjustments, and should, for safety.
MJ11015/6 are overkill here, as the MJ2500/3000 or 900/1000 would have been used in early years. The old SAE parts from their little 60 watter (MK31B) were house numbered versions of those. I ran the hell out of that little amp and it never gave moment’s trouble. No concern about them blowing up at all, unless you somehow got China fakes.
Last edited:
First hiccup. I’m guessing with the metal heat spreader to the heatsink would be optimal. The BD139 is reverse pinout of the originals. So the legs were wrapped in silicone and crossed.
Just to confirm on the miller caps, those would be C75 and C80, correct? The originals are ceramic which I do have, any benefit in a piece like this to use films there? I have some Wima that are 100 pF.
Dan
Last edited:
R91 and R94, increase by a factor of 10, going from 120 ohm to 1.2k. With the increase should I go with larger resistors? Originals being 2 watt. Or sticking with 2 watt okay?
Thank you,
Dan
Thank you,
Dan
Last edited:
So many questions, D4 and D5 which are going to be replaced with the 1N4001 and schottkys, the originals are mounted to the heatsink between the outputs. Is that to keep them cooler or are they heat sensing diode like many in bias circuits? Not sure how I would get these four attached to the heatsink well enough.
Do I need to not worry about that?
Dan
Do I need to not worry about that?
Dan
If you do drop the VAS current to 10 mA, the heat spreader becomes unnecessary. Not bad to have though. The resistors go to 1.2K or thereabouts. 1k, 1.5k, or one of each is fine. 2 watt is unnecessary as the dissipation will go down. Just use regular half watt. The miller cap is that .001. It doesn’t need to be that big with more normal VAS current and EF2 output. I wouldn’t waste Wima caps - 100 or 220 pf NPO and call it good.
The diodes just sense the temperature. 1N400x clamped to the sink tend to do well enough in these low power situations even though thermal contact is not the greatest. I would make a string of them, with heat shrink where the leads are soldered to one another. Clamp the assembly to the heat sink with a piece of plastic - there’s a hole you can use for the screw, assuming it doesn’t interfere with anything on the back side of the board.
A fancier solution would be to build a little vbe multiplier and put it in there. A BD139 could be mounted to that hole, and the leads run up to a little piece of perf board that also contains two resistors and a pot. Good thermal contact and fully adjustable. There’s a Dynaco ST120 mod that’s something like this.
The diodes just sense the temperature. 1N400x clamped to the sink tend to do well enough in these low power situations even though thermal contact is not the greatest. I would make a string of them, with heat shrink where the leads are soldered to one another. Clamp the assembly to the heat sink with a piece of plastic - there’s a hole you can use for the screw, assuming it doesn’t interfere with anything on the back side of the board.
A fancier solution would be to build a little vbe multiplier and put it in there. A BD139 could be mounted to that hole, and the leads run up to a little piece of perf board that also contains two resistors and a pot. Good thermal contact and fully adjustable. There’s a Dynaco ST120 mod that’s something like this.
Alright, my first power up and it went well enough, nothing blew up at least. One problem is that my bias is very uneven. Sorry that I’m being a tad slow, my wife is in Colorado for work and I’m watching the kids, they’re turning 3 this month so they’re keeping me busy, hopefully I can get more done.
So the work I’ve done so far since discussing changing the topology. Installed the MJ11015 and MJ11016 Darlington outputs. I matched them as closely as I could. The two PNPs had shunts measured at 11200 and 11000. Gain was closely matched as well. For the NPN the shunts measured 10800 and 10700. The Bias diodes D4 abs D5 were replaced with 2 1N4001 and two 1N5819 Schottky diodes in series. The diodes are clamped in the original metal clamps. The resistor in parallel with the diodes R111 and R112 were increased to 220 ohms, 10x original value. They were put in the reverse side of the board.
I increased the output coupling caps to 3300uF 35v, I couldn’t get the 4700 I have on hand to physically fit. Q15 and Q19 were changed from the TIP41c I had in there to BD139. The miller caps C75 and C80 were changed to 100 pF from 1 nF. Four resistors, R91, R94, R99, and R100 were changed from 120 ohm 2W to 1.2 kohm 1/2W.
My bias is all over the place, I’m about to go through and do some voltage checks, but wanted to post this to get some ideas. Here are the NPNs
And here are the PNPs
This is across 0.5 ohm emitter resistors.
Here is the bias current on each output if you were looking at it from the top of the board.
The right channel PNP is barely there at 0.08 mA while the NPN is a tad high of the 15 mA target at 19.4 mA. Vastly different.
The left channel is a bit closer to each other with the PNP sitting at 4 mA and the NPN at 7 mA.
Any recommendations as to where to go from here are greatly appreciated!
Thank you,
Dan
So the work I’ve done so far since discussing changing the topology. Installed the MJ11015 and MJ11016 Darlington outputs. I matched them as closely as I could. The two PNPs had shunts measured at 11200 and 11000. Gain was closely matched as well. For the NPN the shunts measured 10800 and 10700. The Bias diodes D4 abs D5 were replaced with 2 1N4001 and two 1N5819 Schottky diodes in series. The diodes are clamped in the original metal clamps. The resistor in parallel with the diodes R111 and R112 were increased to 220 ohms, 10x original value. They were put in the reverse side of the board.
I increased the output coupling caps to 3300uF 35v, I couldn’t get the 4700 I have on hand to physically fit. Q15 and Q19 were changed from the TIP41c I had in there to BD139. The miller caps C75 and C80 were changed to 100 pF from 1 nF. Four resistors, R91, R94, R99, and R100 were changed from 120 ohm 2W to 1.2 kohm 1/2W.
My bias is all over the place, I’m about to go through and do some voltage checks, but wanted to post this to get some ideas. Here are the NPNs
And here are the PNPs
This is across 0.5 ohm emitter resistors.
Here is the bias current on each output if you were looking at it from the top of the board.
The right channel PNP is barely there at 0.08 mA while the NPN is a tad high of the 15 mA target at 19.4 mA. Vastly different.
The left channel is a bit closer to each other with the PNP sitting at 4 mA and the NPN at 7 mA.
Any recommendations as to where to go from here are greatly appreciated!
Thank you,
Dan