Kustom amp question.

[wavley]

Also - a light bulb limiter will help you ensure not frying things while messing with the output stages.

Everybody that works on amps (or anything that plugs into the wall really) should have one of these.

[R.G.]

Get your light bulbs now. Countries (including the USA) are making it illegal to manufacture them because they waste too much energy.

Soon we'll be contending for NOS incandescent bulbs.

Note that I did not put a smiley on this posting.

[digi2t]

@Larry;

Mouser has the Toshibas. Just ordered them. Thanks bro.

@R.G; No the drivers seem to be fine, but the drivers were fine in my same era Ampeg when I bought it as well. I would rather put some time and money into preventive maintenance/upgrade, than be faced with something that might fail at an inopportune time, like my Ampeg did.

Yes, I'm well aware of "If it ain't broke, why fix it" booha. That may work well in government, but it doesn't work for me.

One thing that the power section uses for the biasing is not one, but two diodes clipped to the power transistor heatsink. Their casing looks like Japanese germanium transistor's does. On the schematic, it states that only CR3 is clipped to the chassis, but there are two. I haven't determined which one is the other one yet, I'm just digging into it now. I'll update when I find out which one it is.

Now... off to study light bulb limiters. 

[R.G.]

Yes, I'm well aware of "If it ain't broke, why fix it" booha. That may work well in government, but it doesn't work for me.

One version of that saying goes - if it ain't broke, fix it til it is. 

One thing that the power section uses for the biasing is not one, but two diodes clipped to the power transistor heatsink. Their casing looks like Japanese germanium transistor's does. On the schematic, it states that only CR3 is clipped to the chassis, but there are two. I haven't determined which one is the other one yet, I'm just digging into it now. I'll update when I find out which one it is.

There are two main output stage types in bipolar-output amplifiers - the complementary feedback or Sziklai pair, and the Darlington. In the Darlington, you have the driver's emitter connected to the base of the output(s) directly, and the output comes off the emitter of the final transistor.

In the CF stage, you have the driver's collector connected to the base of an opposite-polarity output device, and the emitter of the driver connected to the collector of the output. Both these connections form a Darlington-gain-ish follower stage. You can use either stage for either the top-rail driver, which will invariably use an NPN, and for the bottom-rail driver, which invariably uses a PNP driver.

The trick played in this amp is to use a Darlington for the top-rail driver and a CF for the bottom rail driver, making all of the outputs be NPN, which were better and cheaper than PNP in silicon at one time. Only the drivers had to be complementary.

In the Darlington stage, the base-emitter voltage drops of both the driver and output have to be thermally compensated in the bias circuit. In the CF/Sziklai stage, only the driver's Vbe matters, as the Vbe of the output device is hidden, all wrapped up in a tight little 100% feedback stage. So for thermal tracking a quasicomplementary output stage with one of each kind of output stage, you need three total junctions to be thermally tracked: two drivers and only one output device, the one on the Darlington side.

Drivers typically stay a lot cooler than the output stage, unless they're mounted on the output-device heat sink. So ideally, one would have a thermal compensation device INSIDE THE CASE of all driver and output transistors. This was a good idea forty years ago that was rousingly ignored. Only in the last ten years have power devices with internal temp sensing diodes become available.

But I digress. In this amp, you would ideally have a compensation diode on the NPN of the Darlington top-side output device and also on the top-side driver. The third diode would be on the bottom-side PNP driver. In practice, I bet it's not set up that way.

There's a lot of finesse involved in getting the compensation devices where they detect the right temperatures to compensate.

As to the compensation diodes themselves, I bet they're the metal-cylinder kind.

A TO-220 driver transistor of any kind makes a GREAT single-diode for temp sensing if you solder the base to the collector. The package is even set up to make the chip be thermally coupled to the metal tab, which is not always true in diodes. Most diode temp is mediated by the lead temp.

Um... I've probably spent too much of my life on power amps. 

Now... off to study light bulb limiters.

From 2000: http://www.geofex.com/article_folders/spo_test.htm

[digi2t]

Well, I'll tell you R.G., I could sit and read you write ups until the cows come home. The way you write it up is a happy medium of techno jargon, and layman's talk. Enough so, that I can get my noggin around most of it. Whatever's left, allows me to do the research to get the rest.

Cautious good news. Operation "Whistle Kill" seems to be successful. I pulled both preamp boards, and the voicing board tonight. Swapped out most ceramic caps for metal film, replaced the electros, including a nice 10uF bipolar for the two22uF back to back (Paul would approve). I installed sockets for the IC's, and exchanged the RC4558's for NE5532's. The transistors are now MPSA18's, and PN2907's. I also found one of the fat caps had broken it's solder joints, which explained the intermittent tremolo operation. I ran over all the solder joints, reflowing, and beefing them up. As several other Kustom owners have pointed out on other forums, the Molex pins on the boards are susceptible to solder cracks, and corrosion on the pins, so I took care of those as well.

At low volume (the wifey-poo was watching TV), everything sound good. Very clean, hiss free, and actually has a bit of a warmer, less clinical feel, than my Ampeg. It responds very nicely to my Bastard-caster's neck pup. This baby might have a soul after all. The voicing adjustment, on either channel, works perfectly throughout the range, with no oscillation present whatsoever.

The true test of course will be tomorrow, when the castle is mine.  

Thanks for the light bulb tester link. My partner at work is the licenced electrician, time to call in a favor.

[PRR]

> Get your light bulbs now

At present, "Halogen" incandescents squeak under the law, having XX% better efficiency than bad/old incandescents. For lamp-limiter purpose, Halogens are dandy, also tougher. They are a bunch more costly, but a lamp-limiter ought to last many-many try/fry sessions, so a 3-pack is a lifetime buy.

LED spotlamps are still expensive, looong payback, and I suspect halogen spotlight bulbs will be around a long time for uses like yard and garage lights that don't burn enough hours to pay-off LED efficiency in a few years.

Back when I brought-up sick amps for a living, I liked big resistors in the power rails. You can _compute_ the fault-current, while following through the circuit to see where it goes. 100 Ohms 10W (each rail) is a starting value for very-sick amps. Have spares; this amplifier will smoke 100r10W if it is dead-short. When the amp seems working no-load, 10r 10W will slow the damage when testing into light load.

[tubegeek]

There's quantity, and there's quality.

Says the man wearing legendary overalls!

At present, "Halogen" incandescents squeak under the law, having XX% better efficiency than bad/old incandescents. For lamp-limiter purpose, Halogens are dandy, also tougher.

I've had terrible luck with halogens from IKEA that screw into the standard Edison socket (as opposed to the more-common-in-my-experience tubular shaped halogens that use the spring-loaded sockets.) I have had numerous infant-mortality problems with them - this is in regular lamp service, not limiter service.

What's up with that?

[R.G.]

Back when I brought-up sick amps for a living, I liked big resistors in the power rails. You can _compute_ the fault-current, while following through the circuit to see where it goes. 100 Ohms 10W (each rail) is a starting value for very-sick amps. Have spares; this amplifier will smoke 100r10W if it is dead-short. When the amp seems working no-load, 10r 10W will slow the damage when testing into light load.

There is something to be said for that. The old saw is that the man whose only tool is a hammer will approach every problem as though it were a nail. There is a reason machinists have large, well stocked tool chests.

The resistor-in-the rail approach is well suited to class AB power amps, because they have such small idling currents. The light bulb limiter's forte is that it is a variable resistance, from very small when the filament is cold, to much larger when the filament is hot, and the full fault current is modestly well defined. This lets you get almost normal operation with low powers but full limiting with high powers. A light bulb limiter is GREAT for testing issues with AC power wiring and/or faulty power transformers. It's good for tube amps and also good for most AC powered things under about 250W. All tools have limits.

One reason I quit using per-rail fusing is that it's hard to tell what the amp will do to the output if one side of the power supply is opened. This is not such an issue with resistor dummy loads, but one can kill speakers this way. Err - ask me how I know that last one. 

[digi2t]

There is something to be said for that. The old saw is that the man whose only tool is a hammer will approach every problem as though it were a nail. There is a reason machinists have large, well stocked tool chests.

I'll back that one up my friend. My old man was a machinist/mechanic, and one of his favorite expressions was, "If you're working too hard, it's because you have the wrong tool". That one got drilled into my brain at an early age.

OK, full power test confirms that Operation Whistle Kill is a bonafide success. The amp rocks out nicely. Preamps and voicing are done.

Moving over to the power section, I did notice (and mention before) that not one, but two diodes are clipped to the heatsink rail. The schematic we have in hand shows a plan for board number PC5065, but the board itself is etched "PC5065 rev.2". The difference being that CR1 is replaced by a 33R resistor, and CR2 and CR3 are RCA D1300A (1N3754) can diodes. Including some other differences, marked in red, the schematic should look like this;



Insofar as the diodes are concerned, am I correct in assuming that this was a cheap and nasty way of improving the bias compensation (two heads are better than one  )?

[Mark Hammer]

Peter would be happy.  (the amp was from the collection of former forum member, and all-round nice guy, Peter Snow)

[R.G.]

Moving over to the power section, I did notice (and mention before) that not one, but two diodes are clipped to the heatsink rail. The schematic we have in hand shows a plan for board number PC5065, but the board itself is etched "PC5065 rev.2". The difference being that CR1 is replaced by a 33R resistor, and CR2 and CR3 are RCA D1300A (1N3754) can diodes. Including some other differences, marked in red, the schematic should look like this;
...
Insofar as the diodes are concerned, am I correct in assuming that this was a cheap and nasty way of improving the bias compensation (two heads are better than one  )?

Ah. They put in a Zobel, but no damped inductor.

The crux of the issue on the compensation diodes is - where are they? The point of those diodes is to reduce the bias voltage by an equal amount to what a transistor is going down as it gets hot. The ideal position is right on the transistor being sensed. As I mentioned, the ideal thing for such amps would be to have a diode inside the transistor case on every transistor that gets hot. Ideally, the drivers and output devices would all have five leads, two of which were an internal diode running at the same temp as the transistor chip. Sanken and perhaps On Semi made some in the last ten years.

But you want one of the compensation diodes on every device that gets hot. The one touchy thing about Vbe multipliers for thermal compensation is that it's easy to get too much compensation. In this case, the designers may simply have wanted to reduce the amount of change in bias voltage as the outputs got hot. The resistor pretty much is not thermally sensitive, so the change in bias voltage per degree is reduced from three diodes' worth to two diodes' worth. In this setup, I would put probably one diode on the top-side output transistors and on on the bottom side *driver*.

In speculating on this amp, there are some things I would do if I were trying to make it ultra reliable. I would be severely tempted to replace the outputs (Q10,11,12,13) each with two transistors in parallel. The MJL3281A is a very, very tough animal, and it has good hfe at lower currents as well as great hfe hold-up at higher currents. What's in there is almost certainly a variant of the 2N3055 and successors, for which the beta falls off at high currents. With sustained-beta devices like the MJL3281A, you can drive more paralleled devices with the same drivers, so you can easily (I think...) double up power devices, as well as current capability, power capability, and heat sink efficiency by spreading the heating out over more devices.

Note that this kind of thing is very much in the category of unnecessarily polishing - um, something that doesn't need polish and won't look much better for it, but it's an interesting intellectual exercise. 

The NJL3281D has those nifty onboard compensation diodes I keep yammering about if you REALLY want to keep polishing this t... er, object.   

[digi2t]

The crux of the issue on the compensation diodes is - where are they?

They are both slathered in heatsink compound, and clipped to the heatsink rail of the power transistors.

Personally, swapping in the Toshiba drivers, and the MJ15003's is definitely on the agenda, as is the new power caps.

Insofar as the TO-220 style bias circuit is concerned, I'm just wondering what value resistor and trimmer would be needed within that scheme.

I'll add in the output and input diodes as well, and call it a day. That should be enough polishing for one day.

[digi2t]

I did a copy paste of the bias scheme that my Ampeg uses;



Insofar as the resistor values are concerned, that's where my math skills fail me.

In the Ampeg, a 2N3403 is called for, but at the time of it's rebuild, I found that they were hard to get. 2N3405's, on the other hand, were quite plentiful. Same gain range as the 2N3403, but 50v rating, instead of 25v.



It works perfectly in the circuit. It comes already glued into a clamp bracket, and is screwed into the power transistor heatsink.

The only adjustment I had to make on the Ampeg was the 18K resistor above the trimmer. I found that I was right at the end of the trimmer range, and was still a bit off the required bias voltage. Lowering it to 15K centered the trimmer, and I could dial in the bias voltage perfectly.

[R.G.]

I did a copy paste of the bias scheme that my Ampeg uses;

That is the early, disastrous form of Vbe multiplier. What happens if the pot wiper gets intermittent? When it opens, the bias transistor goes open, and both output stages conduct fully at the same time; shortly (oops, sorry  ) after, the output transistors change into Darkness Emitting Diodes (DEDs).

The adjustment pot has to be set up so if the wiper opens, the bias transistor does not go fully open circuit. That's why I attached the wiper to one end of the pot. It's also why I said that stuff about paralleling the chain with four diodes to slow down the smoke generation.

Insofar as the resistor values are concerned, that's where my math skills fail me.

In the Ampeg, a 2N3403 is called for, but at the time of it's rebuild, I found that they were hard to get. 2N3405's, on the other hand, were quite plentiful. Same gain range as the 2N3403, but 50v rating, instead of 25v.

Most transistors work as long as they have enough voltage rating, current rating, and enough gain to make the resistors small. It's nice to have the bias transistor have a gain more than 100, so you can make the bias resistors conduct less than 1/10 of the collector current and have that still be 10x the base current. The choice of resistors is then straightforward.

It works perfectly in the circuit. It comes already glued into a clamp bracket, and is screwed into the power transistor heatsink.

The only adjustment I had to make on the Ampeg was the 18K resistor above the trimmer. I found that I was right at the end of the trimmer range, and was still a bit off the required bias voltage. Lowering it to 15K centered the trimmer, and I could dial in the bias voltage perfectly.

That certainly will work.

I did a quick thumbnail of a bias circuit. I think I'd use a TIP29 in a TO-220 or if I could find them, a 2SC3902 in a TO126. I'd use a 2.4K for the collector-base resistor, and a 1K plus a 1K trimpot for the base-emitter resistor, with the pot connected as a variable resistor with the wiper connected to one end.  The 2N3405 has a quite high gain, 180 to over 400, so they can use higher resistors for the Vbe multiplier. The TIP and 3902 have lower gains, but over 100 at the low currents this would work at. The heat flow into the metal tab of the to-220 and to-126 is faster. Faster is better. Of course, you have to insulate the TO-220 and TO-126 tabs so it's more of a PITA.

[digi2t]

or if I could find them, a 2SC3902 in a TO126

I have BD139's. Lower voltage/amp rating than the 2SC3902. OK, just checked, and Mouser carries them. Both the S and T gain ranges. I have to order some stuff from them anyway, might as well tack a couple on. No insulation required with the TO-126 sounds good to me.

That is the early, disastrous form of Vbe multiplier. What happens if the pot wiper gets intermittent? When it opens, the bias transistor goes open, and both output stages conduct fully at the same time; shortly (oops, sorry  Smiley ) after, the output transistors change into Darkness Emitting Diodes (DEDs).

The adjustment pot has to be set up so if the wiper opens, the bias transistor does not go fully open circuit. That's why I attached the wiper to one end of the pot. It's also why I said that stuff about paralleling the chain with four diodes to slow down the smoke generation.

In that case, maybe the Ampeg should be revisited once I get done with this one. I would hate for it to go snap crackle fizz again. Or simpler yet, remove the trimmer, measure the two sides, and insert two resistors in it's place? Much safer than a trimmer, no?

[R.G.]

I have BD139's. Lower voltage/amp rating than the 2SC3902. OK, just checked, and Mouser carries them. Both the S and T gain ranges. I have to order some stuff from them anyway, might as well tack a couple on. No insulation required with the TO-126 sounds good to me.

I think you misunderstood me. The TO-126 does need insulation. I looked for a full-pack (completely epoxy-covered) TO-220, but didn't find one with enough gain.

The BD139 is fairly low gain. The lower the hfe, the smaller you have to make the values of the setting resistors to provide enough base current to make the thing work. To anticipate the next question, I've never seen one of these made with an integrated darlington. It may be tricky to make it work right with two junctions inside the package being used for sensing and compensation.

I personally always preferred the TO-220 package because I could epoxy the trimmer pot to the top of the body, run resistors to the legs, and have the whole thing in a single lump.

In that case, maybe the Ampeg should be revisited once I get done with this one. I would hate for it to go snap crackle fizz again. Or simpler yet, remove the trimmer, measure the two sides, and insert two resistors in it's place? Much safer than a trimmer, no?

Yes, safer. Once you get it right, you probably will not go back and rebias it until other things happen. Output transistors get replaced much less frequently than output tubes.

[digi2t]

I think you misunderstood me. The TO-126 does need insulation. I looked for a full-pack (completely epoxy-covered) TO-220, but didn't find one with enough gain.

Oh sorry. What threw me was what is written on the first page of this datasheet;

http://www.onsemi.com/pub_link/Collateral/EN2101-D.PDF

I figured where it says «The plastic covered heaksink eliminates the need for an insulator», that Bob was my uncle.

I personally always preferred the TO-220 package because I could epoxy the trimmer pot to the top of the body, run resistors to the legs, and have the whole thing in a single lump.

I was envisioning a small piece of vero, with a short stand-off between it and the transistor. Run a screw through the works, and into the heatsink, transistor side down, of course. Solder the transistor legs up to the vero strips, place my resistors and trimmer on the mini vero, and then jumper over to the main board.

Yes, safer. Once you get it right, you probably will not go back and rebias it until other things happen. Output transistors get replaced much less frequently than output tubes.

That`s what I thought, but I feel more comfortable if it has your...um, benediction. It`s set up perfectly now, so it`ll be a snap to pull the trimmer and swap in the equivalent resistors.

[R.G.]

You could also do the sring-of-diodes safety net instead of rewiring the bias pot.

The function of the bias string is to keep the bases of the two output drivers a specific number of volts apart, all the time. If the voltage between them is zero, or too small, the amp works fine, stays cool (unless something else is making it overheat) but has excess crossover distortion. If the voltage between the two bases is increased judiciously, there is a place where crossover distortion is at a minimum.

What that voltage is doing is slightly turning on both output devices at the same time. If it's let get too big, they conduct too much, and try to drag the entire power supply down. The ugliest state of affairs is if the bias network goes completely open. This is disaster time for the outputs.

What you can do is note how many volts exist across the bias network where you have it set now. This is probably between 2 and 4 diode-drops, depending on whether the output stage is complementary darlingtons (4 diodes), quasi complementary (3 diodes) or dual Sziklai pairs (2 diodes). The bias network is setting things to, say 3.027... diode drops for a quasi complementary output stage to turn them just slightly on. If you parallel that bias network with 4 diodes independent of the bias network, the 4 diodes do nothing at all.

... until the bias network opens. Then they conduct at 4 diodes, not 3.027... diodes, and while the output stage is dramatically overbiased and will overheat quickly, it's not nearly as bad as a full open with outputs conducting massively. It's a crude safety net. You may never need it, as the wiper of a pot doesn't always open. It just might.

[digi2t]

To answer Mark's request (I didn't forget about you);

I take it the chip labelled 80848 in the schematic is a 3080, given the pinout, and especially given the 150k/4k7 attenuator on the input.  If you can scope the LFO output waveform, I'd be curious about its basic shape.

People over at the Kustom forum say that it's an LM3080 equivalent.

Here you go my friend. Screen shots of the LFO waveform.

Slow speed, minimum, medium, and maximum depth;





Medium speed, minimum, medium, and maximum depth;





Max speed, minimum, medium, and maximum depth;





It's got a nice looking wave, if I do say so myself.

I see that here it's running on a +/-12vdc supply, but I'm wondering if it would work as well on a +/-9vdc supply. Just package this baby into a pedal, and call it the "Kustomolo".

[pinkjimiphoton]

dino my brother, if ya need some trannys for this, let me know, i may have some that will work for ya kicking around.
i have a couple kustom boards i'm de-populating, and you know, anything i got bro... mi casa, su casa

Yep, I'd add a Zobel and a damped inductor in series. And tinker a bit with the  protection. If you want, I'll hack on the schemo a bit.

Yes please, if you would sir. Like I said, this amp, despite the aforementioned technical shortcomings, has potential. It also provides a bit of a platform for me to work with, and gain a bit of amp experience, which is almost nil.

Would the MJ15003's, with the Toshiba 2SA1837 / 2SC4793 be OK in this amp? I really can't find any data on the original RCA transistors in this unit. These were recommended by you R.G., a while back, and they worked smashingly well in my Ampeg. To my untrained eye, looking at voltages, wattage, and whatnot, it seems to me like they would do. I figure if I'm going to recap, I might as well beef up the power end as well, as I did the Ampeg. Small price to pay for many more years of service.

Any opinions on the 4558 opamps? Leave them, change them? Let's hear it. This amp has struck a heart nerve with me, so I'm going to upgrade whatever may improve it's performance.

Teemuk; thanks for chirping in here. The way you described the voicing circuit operation is spot on. That's the way I hear it, you accurately wrote what I heard (if that makes any sense  ).

You're correct again on the oscillation. It ranges from just below dog whistle frequency, to up around something that drives my dog insane. I realize it's oscillation, but I'm using the term "whistle" as a layman term.

EDIT: Many of the polarized caps in here are not cans, but rather, they look like black jellybeans. I'm assuming that they are tantalums, some are up to 100uF/20V. Change them, or leave them?