Peavey CS800 Doorstop to Decent

Hi Folks;
Now at the 'acceptance' stage, I'm faced with re-ordering parts. Newark in Canada seems to have the next best prices on these items at the moment. I decided on the MJ21196G because I have used these with success in my PL700B. I have studied two other possible substitutes and thought I would ask for opinions (since my options are unfortunately open again!). I will attached the pdf's of each, if anyone cares to cross-examine. I can see virtually no difference between these transistors and wonder why there are separate product lines.

Also, using the Peavey cross-reference guide and following a few 'degrees of separation', one could conclude that outputs could be used as drivers. I am hoping to get away with it. How bad is this? Economics will start to be an issue.

Before I jump into another shipment, perhaps I should assemble the amp with just drivers and one pair of outputs in each channel to verify function of the rest of the amp. Since I am ordering, I should add replacement small electrolytics for the driver and input boards, and quieter op-amps were mentioned? I will have to make a shopping list---I'm glad the wife doesn't read this!
Thanks again, everyone!
 

Attachments

  • MJ21196G .pdf
    125.4 KB · Views: 86
  • MJ21194-G.PDF
    122.4 KB · Views: 65
  • MJ15024-D.PDF
    111.6 KB · Views: 63
I don't know why ON sells both 194 & 196 either.
15024 has a bigger die and is reputed by Enzo to be a "tougher" part, I think he said. Costs more. How many hours are you going to play at 800 W? How tough do you need?
Note best Vbe match comes when you buy whole rails of 25. You saw the scattered Vbe of the 6 196 I bought with a asterisk by them in 2013. Out of different ends or rails I imagine. If buying small quantities, buy a couple of extras to not use in parallel.
As the CS800 matured, Peavey started using faster Ft drivers than the Ft of the output transistors. I guess this has something to do with high frequency response. If your speakers respond up to 14 khz and you have wood piano or bells or cymbals in your source, you might hear a little less IM distortion IMHO with fast MJ15020/21 as drivers as they used in the CS800s. I can't buy those anymore, so for drivers I bought some 2SA1294 Allegro Sanken and 2SC whatever the NPN complement is. They are faster but not as much soa as the 21194,21196, or 15024. I do play piano and have fantasies of playing a bigger crowd than the 100 in the fellowship hall Sunday before Christmas. I needed a louder piano that performance, a $200 Acrosonic wood piano would do it but an electric keyboard plus a PV-1.3k + an SP2 speaker would be easier to carry. I complain about piano sound a lot. I heard Sun Studio Sessions last night at 1 AM on KET, and they miked the piano with a SM58 voice mike. Blue something band. Piano sounded *****y, she the vocalist was playing lead but the piano was buried by the guitar. No high frequencies to cut through the mix.
I like to load all the parts I need in one box since newark charges for freight. Or used to. They were giving free UPS freight over $50 last month. I changed just the 2 e-caps that were bad in the CS800s last time I was into it, ( in switcher power supply) pous a blown fuse, and now one channel is 20 db softer than the other channel. So depends on whether you want an all new box with good reliability, or a cheap fix. Old pots get iffy too, also fans.
Finding an op amp that will plug in for that old moldy one will be a trick. I haven't downloaded any datasheets for single op amps, maybe they are still selling that pinout. RCA3080 in the DDT circuit sure didn't match any new op amp.
If the wife complains, point out how much new Inukes cost. And how 4000 watts in one of those is about equal to 800 in the Peavey's.
 
Last edited:
  • Like
Reactions: 1 user
The real reason for faster drivers is charge suckout at high frequencies. This will reduce heating due to cross conduction if the amp is run at tens or hundreds of KHz at war volume. But nobody ever runs war volume into their tweeters and most of the output power is a couple kHz or below, where even 2MHz output transistors as drivers don’t cause any problem reliability wise. Faster charge suckout also reduces crossover distortion, but amps like this tend to be under-biased per Self so it is of less benefit than if it were optimally biased. You’d need 50 mA or more in each output transistor to eke out the last drop of performance and no one ever runs em that high in an amp meant for war volume.
 
The real reason for faster drivers is charge suckout at high frequencies. This will reduce heating due to cross conduction if the amp is run at tens or hundreds of KHz at war volume. But nobody ever runs war volume into their tweeters and most of the output power is a couple kHz or below, where even 2MHz output transistors as drivers don’t cause any problem reliability wise. Faster charge suckout also reduces crossover distortion, but amps like this tend to be under-biased per Self so it is of less benefit than if it were optimally biased. You’d need 50 mA or more in each output transistor to eke out the last drop of performance and no one ever runs em that high in an amp meant for war volume.
Cross conduction is bad so I'm glad I found 2SA1294 2SC3263 for drivers since MJ15020/21 is gone.
We're converting Peavey equipment for home use, because they are a lot cheaper on e-bay/craigslist than a blown MacIntosh or Parasound. Title of thread is "doorstop to decent". Which means in my case, I'll be listening to 5 output transistor pairs at 2 W a lot. Since decreasing the crossover distortion cost 2 each $.04 1/4 watt resistors in the predriver diode stack, my question is - do you set the resistor value to make ALL five output transistors have ~40 ma idle current, or do you set the resistor value to have ONE output transistor pair have 40 ma idle current? The ones with lowest Vbe I guess? So one transistor pair makes the lowest part of the waveform then the others cut in later as the waveform reaches their turn on point? Enzo said set resistor value lower until vertical edges disappear on the scope, but I have 3 broken scopes and am not in the mood to buy anything new from *****.
As far has having the power reserve of 85 V rails & 5 pair MJ15024/25, I do play piano out in public, and might need 20 W/ch of good high frequencies + 300 W/ch woofers for a crowd of 600 (in my dreams). I played for 100 before Christmas, a bigger crowd is not total fantasy.
 
Last edited:
Optimal bias is actually a voltage, not a current. Optimal bias is 25 millivolts across the emitter resistor. Lower Re, more current. More transistors, more current. The lower the value of Rbe on the output transistor, the faster the charge suckout, but the more bias you end up with in the driver. Both of these factors mean more idle dissipation if optimized. And then you’re going to want a quiet fan. This adds up to running hot with 85 volt rails. And with the original diode biasing scheme, the bias may not be all that temperature stable. If you crank up the bias you may want a real Vbe multiplier circuit.
 
I'll run the fan as necessary. Power amp lives behind an organ. PS, I've been calculating I=V/R since 1967. ST70 idle current was set via a voltage.
Peter, going out to look for a substitute for the hissy SC5278 op amp. Newark doesn't have a clue on the original part. Is it TO-99, a round TO-5 metal case with 10 pins? Runs on +-16v, which is pretty standard, but power pins 2 & 4 is a bit old fashioned. I'm thinking you can fly in a 8 pin DIP socket, twist in 30 ga wire to fit in the old PCB holes, I looked at what was in stock in single op amps with pins that will take +-16 v. CA3140 looks really noisy, everybody hates 741 in audio. LF411 and LF061 don't have enough drive current, 10kohm loads minimum. NE5534a from TI has enough slew rate, drives 600 ohms, costs $1, but you have to do something with the comp & bal pins. 5532 has an extra op amp; atl east you could terminate the extra ones input to not oscillate without a lot of thought.
that 180 pf cap between 47k feedback resistor and negative op amp input is weird. I think you would short across that then put 22 pf parallel the feedback resistor to prevent oscillation.
 
Last edited:
Hi Folks;
Just a little update; with some 'smooth talking' (and sending pictures), I have actually received a refund for the "incorrectly labeled" MJ21196 transistors!
Thanks Indianajo, wg_ski, and this forum, for yet another lesson. I will re-order the parts from a supplier with outlets based in Canada. I will order 26 pieces again. I will subject them to the same tests and regardless of test results, one of the least matching devices will be cut open for visual inspection.

Work, unfortunately, is interfering with my life again and I have not had time for Peavey-play. As mentioned, I will hold off on ordering to combine the shipment with any front-end or fan control parts that may be required.

At my first opportunity, I would like to revive the fan control shown above, and address it's design flaws.
Regards, Peter
 
Congratulations on the refund. E-bay star ratings are actually good for something I suppose.
Sorry about your having to work. At my age I've got my volunteer job down to 3 hours a day, and I'm taking the day off today because it might snow. Hooray for retirement. No way I could ** on the internet when I was working.
 
Merry Christmas! I'm still trying to put together the old CS800 on the cheap. I am matching Vbe as close as possible from some leftover transistors from other projects. I am trying to have each bank fitted with matching batch numbers. I also had a concern about matching the emitter resistors. Would it be an idea to place the transistors with the lowest Vbe with the highest measured emitter resistors? .... or have I got that backwards? Or does this not make any sense either way..
Hi To all and yes, this is the same amp from 3 years ago! There have been a lot of changes in this time, I'm sure for everybody!
I wish you all the best for 2023 and hope we all have more time for our favourite hobbies.
 

Attachments

  • DSCN2510[1].JPG
    DSCN2510[1].JPG
    394 KB · Views: 69
I suppose a lower vbe would benefit from more degeneration (higher R). But you are likely getting a better device to device match than anything they would have back in the 1980’s (Or 70’s).

Ive had amplifiers sitting a lot more than 3 years. Consider yourself lucky to have gotten back to it in 3.
 
Thanks wg_ski, I was hoping it was something like building a performance engine, put the biggest pistons in the biggest holes! I have heard that tolerances have improved over the years but my collection of outputs are not exactly spring chickens!
I am measuring Vbe with a Peak Atlas DCA55. I gather that this is a very low level test ( 2.5 mA ) but is it not better than nothing? Having measured the E resistors, could I check the current sharing by measuring V drop across them with the amp under load?
Merry Christmas! Peter in Canada
 
Any MJ2119x you might have are spring chickens compared to anything that was in use when the CS800 was developed. I still have a handful left from 1990, when I got half a tray of each from the sample program before they showed up in the Newark catalog. Being affiliated with a university had its perks.

I match output transistor vbe with the diode function on my DMM. Low current is more indicative of actual “diode” behavior. At high base currents, the parasitic resistance dominates (IV characteristic goes linear, and far exceeds the diode drop). You tend to get a better and better match at very high current because of the parasitic resistances (and the intentional sand box Re). Altogether different types of transistor have different values of those internal resistances which throw things off if you parallel them. That’s why you can’t mix old and new types, or one of the old RCA’s with Motorolas. Two random 2119x will generally be good enough, but out of the same batch the idle bias will match as well. If you put one new 21194 in with a batch of the old Moto selected 2N3773 (SJ6343) or RCA1B05(even worse), the higher gain larger die 21194 will hog current. It won’t take ALL the current, but enough that it will be the early failure.
 
Thanks again, wg_ski. That is interesting and helpful. So if I understand you correctly; the transistor's internal resistance helps with current sharing at higher levels.
I also gather that the Vbe measurements from the Peak Atlas tester at low levels are a reasonable method of matching (?).
In this particular low budget rebuild, I am planning to use all TP9054's in one channel, MJ15024's in the other. I do not have enough of each batch number to fill a bank of five, but each bank will have closest matching Vbe.
BTW, is there a percentage range of Vbe that would be considered a good match? In the amp channel with TP9054's; the neg bank measured from 0.570 V to 0.576 V. The pos bank varies from 0.561 V to 0.570 V Is this good/bad/ugly?
Thanks again, Peter in Canada
 
So there is one channel. 45.9 Volts into 8 ohms is about 263 Watts. Just to be 'OCD', would measuring the voltage across each Emitter resistor with the amp under load provide an indication of proper current sharing?
 

Attachments

  • DSCN2514.JPG
    DSCN2514.JPG
    530.5 KB · Views: 60
  • DSCN2515.JPG
    DSCN2515.JPG
    374.9 KB · Views: 53
I like to see how well they drive up, too.

Using MJ15003 for drivers in that board? Did you check for Vceo first, or were they already known good at 160+ volts? Modern ones usually well over 200, but I always check anyway.

They did that with the original house numbered 2N3773’s (which you have one version of). I’m used to seeing the aluminum case SJ6343 but there were others. The top line in the part # (pretty blurry in the pics) ends in “0073”. Those might not have been tested and labeled exclusively for Peavey. Everybody and their dog used them - the MJ15024 was just coming out, and expensive as hell.
 
I used the MJ15003's as drivers because I had 4 of them handy. The Vceo is only 140 V. Is this not sufficient? Should this value be higher than the sum of the rail voltages?
Peavey suggested the MJ13330 which was a 20 Amp Ic and 200 V Vceo. I used the 15003's because they also had a 20 Amp Ic.
It was suggested to avoid used aluminum T03's because of a risk of failure due to mechanical fatigue caused by higher thermal expansion rates of the case. Could I use MJ15024's as drivers? or SJ6343's? (This is a budget rebuild and I would like to use what's at hand)
Thanks again for the help.
 
MJ13330 is a now-unobtainium switching transistor. The idea was to have the drivers with a higher fT. This is really unnecessary, as it will run fine with the same driver as output.

The 15003‘s 140 V rating is not really enough, but real ones will almost always go much higher. Lower SOA at 80 volts than the 3773 or 15024, but about 4X that of the 13330. The driver doesn’t need to be as rugged. But handling full voltage is necessary. Just put a 100k resistor in series with the collector-emitter, and check leakage current at full supply voltage. If you measure basically none (ie, nano amps), they‘re good to go. That’s how the original types were selected back in 1978. You could use 15024’s as drivers. If you have spare unmatched ones (say a couple with different batch #’s) use them for drivers.

The problem with aluminum TO-3’s is that they will eventually fail from thermal cycling even if never overloaded. If you’ve got an easy/cheap way to avoid them, do so. Especially used ones, which may have a lot of hours on them.