bendix germanium transistor replacement? HP6823A "audio" amplifier *fixed, trace

[coppercone2]

I fried an audio amplifier doing something complicated, or maybe it broke by itself after light use, hard to tell

The transistor is B1654 in the manual and I found "Bendix 0 - 407" in the sister unit. Its from a HP6823A

I have a feeling I am fucked but I thought I would try the forum. I have a working sister unit so I found that its open on the diode test and the manual seems to point me to the same place.

https://www.keysight.com/us/en/assets/9018-03150/user-manuals/9018-03150.pdf?success=true

I just need something that works I don't care if its authentic or requires mods etc

Can't find any reference about these transistors at all

[floobydust]

schematic download didn't work, it's not at BAMA either. Can you post the sch or portion. Likely need to change the bias circuit for something silicon.

[coppercone2]

ok, i uploaded it

I don't really know about this type of circuit and it is pissing me off because I need the amplifier to continue experiments, unless I want to build something ridiculous by brute force with unknown characteristics, feel like i am getting constantly fucked by old gear 

I think it was on the way out because when I was lookin at the scope trace I had some nasty wiggling going on which I thought was a trigger/noise issue but I think the transistor was just fading out

I don't mind buying old germanium but I have no idea wtf to put in there

[bob91343]

An old popular small signal germanium transistor is the 2N404.  I might have one.  If you need some power dissipation or high voltage, you will have to use something else.

[coppercone2]

no, its a power device, its the to oval package

I don't mind buying an old transistor, this is not a modernization request, unless thats the only way, I just have no idea what parameters I want.


confused myself, not sure how that circuit works, if someone wants to teach me how to rebias that thing for a modern transistor..

[floobydust]

I would wonder about the crossover point, if Q1 and Q2 are ever both on it will cause current-limit I think.
But the circuit looks like it doesn't care if Q2 is Ge or Si. CR18, CR19 seem to both be Si leaving overlap actually    What does CR18 read for Vf?

I'd put in a vanilla PNP like 2N2955; 2N6609, MJ15023. What's in your junk box? Higher speed parts should not be a problem because this is an EF.
Q2 PNP 1850-0407 160V 25A 100W Ge Delco, Bendix, Gpd
Q1 NPN 1854-0225 60V 15A TO-3 like a 2N3055.

[coppercone2]

the datasheet says the CR18 is 1.0 and the cr19 is 1.25 (for the other side)

[coppercone2]

in circuit cr19 reads 0.550 and cr 19 reads 0.135, the working unit is very close at 0.554 and 0.139

cr18 looks like a silver can with a red stripe and cr19 looks like black bell (very weird).

i think the 0.139 = germanium?

[coppercone2]

lifted one leg on cr18, it measures 0.118 when hot so I think the in circuit measurement is accurate

lifted c19 and its 0.496

0.1 for a diode is low, I think thats germanium, I expect between 0.4-0.7 for a silicon? I think their right because the working supply measures the same

measured with 34401A

says its germanium
https://www.datasheets360.com/part/detail/1n91/-7020426734529444701/

[floobydust]

The schematic says:
CR18 Rect. Si 150mA 100PIV 1N91 G.E. but this is really germanium.
CR19 Rect. Si 500mA 200PRV 1N3253 RCA

They're both oddball diodes to give some bias current to Q1 and Q2, by CR18 and CR19 Vf adding up should be a bit less, and at most add up- to the VBE of Q1 and Q2.

If Q2 gets a Si replacement, CR18 as Ge would give a deadzone in the crossover region. The PSU might go unstable at that point around 0V out and squeal I'd guess.
So I would change CR18 to Si but with a bit less Vf than Q2 VBE. I have no idea what idle current goal is, likely very small there and the two diodes aren't thermal tracking Q1, Q2.
Measure your new Q2 VBE and try a 1N4004?

[coppercone2]

Ok I need to buy all the parts, I don't have a power PNP in my junk bin.

If I am buying new parts I might as well buy a germanium one, if you can suggest it, otherwise I should go for a 2N2955; 2N6609, MJ15023.

Not sure which one is the best bet?

What has a higher chance of success? I put alot of work otherwise restoring this amp so I am not trying to get the best dollar value


NTE121  ?

[floobydust]

The original Q2 PNP Ge is hand-selected by HP, I think for high hFE. This is why the part is exclusive and hard to find, as well as going back to 1966, and having a 25A 60V rating.
Maybe look at HEP644, 2N1166, NTE179 https://vetco.net/products/nte179-pnp-transistor-ge-af-amp lists the many parts of the day that are similar.  Vintage smaller Ge transistors suffer from shorts due to tin whiskers.

I'd put in a Si part and go through the pain of setting up the bias current with various Si CR18's, On-Semi MJ2955 or MJ15023, MJ15016G is quite tough. There's no protection diodes across Q1, Q2 so inductive loads could be a problem for this PSU. What are you using this for it seems a bit unusual.

[coppercone2]

will try nte179, looks like high gain, i was fooling around with a motor and dummy loads tryin to do something (I noticed the sine wave looked fuzzy on dummy load, compared to the other one, I think it was a sign it was going to blow)

i had germanium confused with selenium technology, i was thinking I was up shit creek but it looks like those parts are not as terrible

not connected to motor, like a break on a electromagnet dragging on a motor flywheel

[Kleinstein]

I would go for a silicon transistor. They are more tolerant to high heat sink temperature. A germanium transistor may have a slight advantage at saturation, but thus should not be a problem here.

Instead of hunting for diodes with a suitable VBE, one could replace one diode with transistor as a VBE multiplier, so one can adjust the voltage, a bit like in class AB audio amplifiers.

The missing revere diodes to the power transisors could be added for some additional protection.

[coppercone2]

I will try a direct substitution of the high HFE germanium part and if it does not work or blows again following minor abuse I will replace it with a Silicon transistor

I also hope changing that diode does not effect anything on the current limiting circuit?

[floobydust]

The current-limiting is not affected by changing Q2 from Ge to Si.
How did Q2 fail? Sometimes nearby parts get taken out as well so I would check.

If you are using this with inductive or motor loads, add two protection diodes.
Typically across the the C-E of Q1 and Q2 but the current-limiter resistors are in the way, best from output to collector of Q1, Q2 each. Even a 1N4004 will protect the beast.
I notice this PSU is special in that it does not have an output cap and faster than expected 20kHz even with the slow fT 500kHz of the Ge part.

[bob91343]

I have a number of very old transistors, including some germanium.  Some are power transistors.  I suspect the detailed parameters are not critical, so nearly any TO-3 germanium should work.

[coppercone2]

it does say selected for reasons unknown, and its the only part I could find easily that is different when comparing the two PCBs with a diode tester, but its obvious because it says when you measure the voltage in step 2 of the PCB low that does not involve desoldering the transistor I don't have on my revision (which is step 1) (there are 2 revs of the PCB, each of my units are slightly different, I could just not find one of the Q's on my PCB), it says check for Q2 open. and there is a  200mV drop on the working unit but a open circuit / infinity on the non working unit, right at Q2 which is what the manual attributes to that voltage failure. I assume if I fix that the other problems will go away.. I am pretty sure I have my man

[coppercone2]

Hmm, I remeasured the same circuit points yesterday as today. The voltage drop changed. I don't want to take the chassis apart until I get my supposed replacement transistor. It was reading nothing now its reading 0.9V and drifty, other circuit is quick to jump to 0.6V and gives the reading I expect from a proper semiconductor junction

[Kleinstein]

There is some frequency compensation in the stage before the ouput transistors. So I would no consider the power transistor so critical - though some of the Ge power transistors can be awfully slow and this could be a point to check in the selected section.  The old Ge transistors can also vary quite a bit in the initial gain and over time the gain tends to go up.

[coppercone2]

I have a dual HP supply that has which one side which seems to be adjusting to setpoint faster then the other supply, would higher gain in the transistor in a similar configuration from gain drift cause this ? I thought it was capacitance shrinkage but maybe its both.

[coppercone2]

is it possible the matching is related to the silicon NPN transistor driving the positive rail? not related to total gain?

[Kleinstein]

I don't think the transistor gain should be that important in this circuit. It can effect the maximum current, as the base current is limited by the current source around Q3. So a lower gain transistor may not reach the full current.

The speed is mainly set by the capacitors in the compensation (C5,C6, C7, maybe the cap at Q4) - these should be film / keramik types and thus not much aging expected there.

I see only a need to have matching for the sum if the VBE of the power transistors to the 2 diodes. This is to ensure the power transistors are no on at the same time, fighting each other and also keeping the dead zone around zero current small (or keeping a small standing current). There is quite some emitter resistance. So the matching should not be that critical.

[floobydust]

The PSU has a closed-loop gain of 20dB with 20kHz bandwidth, so it doesn't have a lot of gain to spare and a lazy Q2 might be something they wanted to avoid. Is an hFE of 10 going to work there? It's lucky to have fT of 200-500kHz.
But it's strange output current is 0.5A but there are TO-3 transitors rated 10A or more, at about 40V for worst-case SOA.
It might be from a problem with shoot-through current, or the designers mid-1960's just found transistors unreliable or HP being extra conservative. Well, it does weigh in at 16lbs.

[Kleinstein]

The power rating with germanium transistors is a bit tricky, as the maximum temperature for germanium is at some 70-90 C. So if the case temperature is not 25 C as for the nominal power, but more like 50 C for a large heat sink, the permissible power is down 50%. Going higher temperaure it gets worse quite fast.
Germanium power transistors have no real problem with a high current, but are more power dissipation limited. So the > 20 A current rating comes for free when you need a 150 W nominal (maybe 50 W real world) power rating.

For transisents the worst case voltage may reach some 60 V, if the load is reactive, like a capacitor / inductor. I can imagine the SOA could be a problem at higher temperature and voltages of more than some 30 V, as there seems to be very little internal series resistance to ensure even currrent distribution.
The Si transistor may be a bit over-kill, following the rules for the more sensitive germanium transistors.

Selecting the Gemanium transistor for high gain is relative - worst case gain can be really low. There is not that much series resistance at the base - so curent gain of the transistor would not effect the loop gain (mainly voltage based) very much. A low current gain would need a higher current from the constant current source and thus more power loss in the stages in front.

[coppercone2]

Do you guys think its possible to increase the maximum current if I
1) replace the germanium with a silicon
2) change component values

You make it sound like the old transistors are a weak point. If I go with 3055's and some kind of PNP (replace the 4 power transistors), and possibly other transistors.

I did notice thought even running doing nothing for a while that thing gets pretty toasty. The transformer is smallish though

[floobydust]

Why does it weigh in at 16lbs but +/-40V 0.5A output, it's not the power transformer?

I think you could increase the output current with modern transistors and mods to the current-limit controls.
The rectifier diodes are not likely great, 1966 was a bad year for PNP as well. Have you re-capped this PSU?
Old HP designs command respect, they performed very well. Even today, who makes a push/pull PSU?

The MJ15016G and 2N3055A supposedly have better SOA than the 2N3055 but look the same 40V 3A on the SOA curve. I would say even TO-220 parts are enough here.
Ge transistors do not suffer beta droop, I have no idea why.

[coppercone2]

it has a large heatsink, but I can't exactly tell the weight distribution. The transformer is ~3x3 inch and 0.75 inches thick EI core

[1audio]

On vintage stuff like this I would built an LTspice model to understand how it works. Its pretty easy to swap parts then and see what happens with only virtual smoke. The Germanium power devices were never that sturdy. A silicon TO3 upgrade  would be well worth the effort.

Still since Jameco says they have 86 of these: https://www.jameco.com/webapp/wcs/stores/servlet/ProductDisplay?storeId=10001&productId=2286039&catalogId=10001&langId=-1&CID=GOOG&gclid=CjwKCAjw55-HBhAHEiwARMCszraLmrtYI-QbqJtbPsCXpFFQl3FoDDb9XjR9P-pbKnm36nANI7frfBoCFa4QAvD_BwE  I would just get 2 (one for a spare) and get it running again. Adding protection diodes is a really good idea. I have blown circuit with inductive loads as well.

[coppercone2]

I don't know if I would go that far, I have a nice transformer, I think I would rather make my own amp then try to reverse engineer a germanium silicon hybrid with a circuit I can bearly read, this would need to be easy.

I always turn it down without abruptly disconnecting it so I don't think inductance killed it acutally, the trace on the scope was just nasty from this one. One unit looked nice the other unit looked like superimposed with tons of fuzz, I think it was on the way out, like jitter and noise

[1audio]

Easy option 1, $12 germanium transistor and you are back where you started. Easy option 2, one of the chipamp based PCB's from eBay. Some have on board rectifiers and filter caps. Add a reference supply for the power supply function and you are there with a lot less hassle than redesigning something.

[coppercone2]

defiantly not going with chipamp, tomorrow my transistor comes in so we will see what happens

[coppercone2]

the desoldered bendix transistor leads 0.9V/OC while the other transistor reads open circuit/open circuit.
-
when read with a multimeter the new one reads 290 ohm and 6k, the other 1.3k and 1.3k from emitter to collector (both ways). the old bendix transistor I noticed also has a clipped leg for some reason, when I desoldered it, most of the legs were sticking out past the pad, but the bendix leg was like flush with the pad, but only one leg. not sure if its manufacturing feature or if someone did something to it.. the nte has equal legs.. would be kinda weird if TI used a part that does not solder in well

just a comparison with a 34401A but it might say something, even if the values depend on what the meter is doing, they seem substantially different, I hope thats enough to locate the problem

I can't use the manual because the one I am working on has a substantially different circuit, there are missing transistors, different layout.. the manual is just a hunch

need to paste it in later and test it out

[coppercone2]

damn, thats not it
same problem

it seems the reference circuit does not have voltage on one of the rails.  oh boy

[coppercone2]

it measured 0.35V, now I pulled R17 and it measures 7V. It should be 60V unregulated. It must be a problem with something the 60V unregulated is connected to

[coppercone2]

son of a bitch there was a perfectly fine looking trace broken on a rivet. added a bit of solder and that shit skyrocketed.

went from 50 across the transformer to 10 meg across the emi caps to 50 across the emi caps. i need to resolder the feeder resistors to see what happens next

[coppercone2]

works, some fucked up intermittent must have been driving me crazy, and the transistor might be bad too

lesson: don't take too much from confidence or experiences to try to rule out factors in old gear, its utterly chaotic as to what you will run into.

also get a long cord, I was fucking turning the unit so I can't look at the panel meter when im testing it because the cord on my work bench is too short, it was probobly working and not working as I pushed broken contact point on reference circuit, so you won't get a good glue on that low draw thing.

I used to paint the entire trace of a PCB with solder till somewhere on the forum I started reading too much about peoples experiences and I stopped doing it, turns out it would have prevented this problem.. try not to listen to the lucky gamblers and cover all points in old equipment if you want to take it easy.. I read about reliable solder joints so many times I was utterly ruling it out

I am a little dyslexic so I think I was getting different readings on positive and negative rails when I was measuring them.. probobly was measuring the correct points but the voltage was randomly changing because it was a high resistance put in a low current draw trace that was making shit jump around (normally on a normal rail this kind of problem would probobly open itself).. got confused C12 and C13 many times.

the trace measured at times 50 ohms, 10k ohms, 1 meg, 10 meg (seemed to change each time, jumping around somewhat randomly). now reads correctly 0 ohms with remelt and added solder

[coppercone2]

had this thing running for a few months, it started making static. I used a needle to put flux (rosin activated) on every solder joint and hit it with an iron from the bottom. The popcorn static noise stopped.

So for the harrison owners, its a solution.

I was not able to detect any thing from tapping on the PCB or whatever.

So requirements
1) 1 episode of star(gate)?, preferably about being stranded
2) soldering iron, flux and hope.

If you have the stomach to flux both sides of the PCB, it will work better, but its alot of cleaning work. The joints wont look nice on top if you just flux the bottom. You want nice burned flux with lots of bubbles to ensure that the surface got nice and clean.

One joint made a 'ping' noise when I hit it, clearly there was some strain building up some where. I was sure I was gonna be dealing with Mr. Bendix again.

[Dave Wise]

Update us if it comes back.  My guess is your heating transistor leads has temporarily "fixed" a noisy transistor.  I've had this happen.

[coppercone2]

well its a screw up on my part, I bridged a trace which muted it but it did not stop the noise. Maybe it changed a bit after I fixed it, because I just had a speaker plugged in as a sensor without a input, but when I tried a input..did not work until I fixed the bridge

I changed the potentiometer and it did nothing, so now I am waiting on freeze spray to try to find the offending transistor  I was so happy thinking it was that easy

But Yes, I think it got quieter when I reflowed the joints, or less crackly.

[coppercone2]

Well Ok, I finally got my freeze spray.

The amp was on the side powered off without cover since the last post 3 weeks ago.

I turned it on and plugged in a speaker and it was quiet. I methodically shot every transistor with a healthy dose of freeze spray and the noise did not come back.

I dunno why it fixed itself. Before the amplitude of the output noise was increasing as I decrease the gain, and when I increase the gain it got quieter (no input).. also tested without chassis.

The freeze spray is pretty strong. (mg chemicals super cold).



Why do I have a feeling this problem is not solved. 


I wonder if it was some kinda weird RFI shit going on maybe. Ima put the chassis back on before it returns plasma fire later tonight

[coppercone2]

been another week, still no noise.

I wonder if there was a whisker inside that got fused by a surge or something.