Replacing TO220 with TO126, am I asking for trouble? Markthefixer and anyone else willing.

saabracer23

AK Subscriber
Subscriber
Found another issue with this driver board I’m working on. The Marantz 2330b drivers (Q709,Q710 and Q739,Q740) are faulty. In one channel the 2SA913 is measuring with a gain of 2 and the 2SC1913 is measuring with a gain of 0. One option is to go with MJE15032 and MJE15033 and roll with it. Though some are not big fans of this option due to the fact that the MJE devices are much slower at 30 MHz vs the originals which were quite fast at 120 MHz.

There is a thread here https://audiokarma.org/forums/index.php?threads/is-there-a-sa-7800-recap-list.795519/#post-11010875
where markthefixer discusses where he wish he had stocked up on some faster parts when they were available, since they are no longer available he was toying with the idea of swapping in some KSA1220/KSC2690. They’re much closer at 175 MHz.
9B48150C-3A6C-4AD8-85D0-C43105D3724F.png

I was curious if you got anywhere with that Mark and what your conclusion was. Anyone else that wants to give their thoughts please do. Specs look to be fairly close with some being better for the KSA/KSC devices and some being better for the 2SA/2SC.
271A0947-B5D9-4372-A128-902BC0D19767.jpeg
95D59F0A-5EE9-45A7-BE62-D8CD465DC527.jpeg

For the PNP it seems the specs for the TO126 seem to be better in all cases except Vcb and Vce which won’t matter in this case. Well and the fact that it’s in a smaller case and can’t dissipate heat as quickly.

6D953284-7DB9-46A8-9A46-8CB0EA11383B.jpeg
9F215F35-814F-4352-87B3-56C559A03666.jpeg

For the NPN it looks like the only issue is Ic, the original being rated for an additional 300 mA.

As Mark mentioned there must be adequate heat sinking and in this case there is.
70D1D2AD-D253-4844-B2BA-0C3FF1194395.jpeg
8E70D472-65C2-4658-B52A-4739B834EE2F.jpeg

I’m guessing that in the case of the 2230b that I’d be asking way too much of the smaller devices, but if someone thinks I’m wrong please say so. Basically I’d like to see if maybe Mark had any luck with these swaps in anything or if he gave up on the idea as it didn’t work well and also to get the thoughts of others on this idea. If I should just use the MJE devices then I guess a 2330 with slower drivers is better than a 2330 that doesn’t work at all.

Dan
 
Last edited:
Every time a get a Pioneer Spec 2 in for repair those blue NECs are always bad, shows up as leakage on my ancient transistor analyzer. Always weary now when I see those and their counterpart.
 
In some SAE I’ve used 2SA1859A / 2SC4883A for those.

Toshiba 2SA1837 and 2SC4793 make good subs also but are hard to find. BDent might have some.

Craig

Thanks guys. Yeah they’re definitely bad. Bdent does have the 2SA1837 and 2SC4793, but for two pair (just enough for the drivers and no extras, I’m looking at over $30 with shipping. That seems crazy to me for a TO220.

I’m going to post a new thread if I can’t figure this out, but I have MJE15032/33 in there now, is it possible that they’re too slow and causing this?

1EFACC1A-82BE-4D7C-B11D-B17DF8526C39.jpeg

this is just my function generator (1khz) going directly into the amp driver board and into an 8 ohm load. Both channels. Turn the input down and it cleans up.
432D9657-BD69-420B-A580-B5F1ADF2CAC9.jpeg

I was hoping I could maybe run the MJE drivers, but not if they’re causing this. Outputs are now MJ21195G/21196G.

I was thinking about swapping in KSA1220/KSC2690 just to see what happens, but I’d hate to have them fail catastrophically and take out the new outputs. I do have an ad up in bt looking to purchase some of the TO220s we’re discussing, so hopefully something pops up. First time I’ve been really halted by a transistor replacement.

Thank you again,
Dan
 
Oh and I thought I should note, I did talk to Mark and basically he hasn’t used up his stash of these transistors so he hasn’t gotten to the testing of the TO126 devices as subs yet. I wonder if I should be a Guinea pig, maybe not with this amp lol. Thinking maybe add a heat sink to the front side of the transistor as well, sandwich it between two large pieces of metal to draw as much heat off of it as possible.

Dan
 
the fuzz looks more like oscillation to me. I'd expect that from a too-fast part, not a too-slow part.
 
the fuzz looks more like oscillation to me. I'd expect that from a too-fast part, not a too-slow part.

That was my exact thought, it’s oscillating. I knew it would happen with a device too fast, but wasn’t sure if a device too slow would cause issues as well.

As I showed above, I’d the output is low it isn’t there, then the oscillations come, but once driven to a higher output it seems to clean up again.
E1AF479D-C085-4CBF-8393-87D2D9D6F528.jpeg

Here at 34v output into 8 ohms it doesn’t look horrible.

Dan
 
and it doesn't do that with the stock parts? Or do you not have a channel-worth of stock parts to compare to?

can't look up a datasheet but the other thing that maybe would tell something is the capacitance values. B-E or B-C probably. If the new part is lower, you should be able to tack a cap in externally to make up the difference, possibly that would change the behavior?

i'd sort of expect a slower part to have more capacitance but *shrug*
 
the fuzz looks more like oscillation to me. I'd expect that from a too-fast part, not a too-slow part.
I knew it would happen with a device too fast, but wasn’t sure if a device too slow would cause issues as well
The 'too fast' case (when replacing), is by far the commonest reason for instability, but I understand it is also possible with a part that is too slow.
 
Last edited:
Bdent does have the 2SA1837 and 2SC4793, but for two pair (just enough for the drivers and no extras, I’m looking at over $30 with shipping. That seems crazy to me for a TO220.

I was thinking about swapping in KSA1220/KSC2690 just to see what happens, but I’d hate to have them fail catastrophically and take out the new outputs. I do have an ad up in bt looking to purchase some of the TO220s we’re discussing, so hopefully something pops up. First time I’ve been really halted by a transistor replacement.
If Pd is not a big problem, I would also consider Toshiba's TTA004B (PNP) and TTC004B (NPN). Lower Cob, fT closer to the original, and cost less than 1 dollar each.

Main specs, compared with the original 2SC1913:

TTC004B (TO-126) - 160V / 1.5A / 10W (on heatsink) / 100MHz / 80~280 hFE / 12pF
2SC1913 (TO-220) - 150V / 1A / 15W (on heatsink) / 120MHz / 65~330 hFE / 14pF
KSC2690A (TO-126) - 160V / 1.2A / 20W (on heatsink) / 155MHz / 130~320 hFE / 19pF
 
and it doesn't do that with the stock parts? Or do you not have a channel-worth of stock parts to compare to?

can't look up a datasheet but the other thing that maybe would tell something is the capacitance values. B-E or B-C probably. If the new part is lower, you should be able to tack a cap in externally to make up the difference, possibly that would change the behavior?

i'd sort of expect a slower part to have more capacitance but *shrug*

Unfortunately no, when I got it had already had maybe half of the original silicon replaced with a mismatch group of parts and I’d say the other half were not functioning correctly. Not to mention half of the electrolytics were replaced with some random valued Chinese cheapos. So I never had a chance at an all original channel even the outputs were a random selection of NTE and ECG parts. So got all of the caps on the board replaced and replaced most all of the transistors.

Q701, 702: 2SC1775 to KSC1854 matched
Q703: 2SC1775 did not replace
Q704, 705: 2SA912 to KSA1013
Q706: 2SA914 did not replace
Q707: 2SC1953 did not replace
Q708: 2SC1568 to MJE243
Q709: 2SA913 to MJE15033
Q710: 2SA1913 to MJE15032
Q723: 2SC945 to KSC1815
Q725: 2SA733 to KSA1015
The same as above was done in the other channel.

And then outputs to MJ21195 and MJ21196. I’ll look up the data sheet and see if I can’t find where the extra capacitance would go.


I think gadget 73 was right I had the same problem with MJE15032/33 drivers
I solved the problem by putting between the bases of the driver transistors and the ground a capacity between 200 and 500pF for me it worked.

Because even taking transistors from a reliable source it was exactly the same!

https://drive.google.com/file/d/1uFYGTxabj3kblXGVVzUp3GOMxfkaTws1/view?usp=sharing

Thank you for posting that, that’s the exact same issue. After you added the capacitance did you just keep the MJE parts in there? Off the top of your head you don’t happen to remember the two legs you put the extra cap across do you?


The 'too fast' case (when replacing), is by far the commonest reason for instability, but I understand it is also possible with a part that is too slow.

Great, that’s what I like to hear. I’m going to add the capacitance to see what happens, but luckily Ray is being a super hero and selling me a couple pair of 2SA1837/2SC4793 so that I can at least get drivers in there with close to equal speed. Thanks again Ray!


If Pd is not a big problem, I would also consider Toshiba's TTA004B (PNP) and TTC004B (NPN). Lower Cob, fT closer to the original, and cost less than 1 dollar each.

Main specs, compared with the original 2SC1913:

TTC004B (TO-126) - 160V / 1.5A / 10W (on heatsink) / 100MHz / 80~280 hFE / 12pF
2SC1913 (TO-220) - 150V / 1A / 15W (on heatsink) / 120MHz / 65~330 hFE / 14pF
KSC2690A (TO-126) - 160V / 1.2A / 20W (on heatsink) / 155MHz / 130~320 hFE / 19pF

Awesome suggestion, I looked at those as well and I actually happen to have a couple hundred of each on hand already. Now the question is how would a TO126 handle the job?

I find it interesting when they put a die into a device and rate it for more voltage and more current, yet put it into a smaller case. Won’t it produce more heat?


Which brings me to this point. What do you guys think, should I try a TO126 in there? I’m really curious if it would open up better options for us in the future as there a a few different TO126 options that have a higher Ft, while I won’t hold my breath on any TO220 being made in the future.

I know that the die is attached to the back plate metal, but heat will come off of the front so I was thinking I could sandwich the device in between the factory heat sink and an additional like this.
36EB42D5-7E9C-4E8A-8061-AF7A2DC58079.jpeg
A1904F5B-8AAD-40A7-AF76-4A0EF4B1673A.jpeg

I let the amp run with a 1khz sine wave for about 17 hours, only 1/4 watt output, but enough that the output heat sink was quite warm while the driver heat sinks were just slightly warm, a bit above ambient. Here you can see that they’re just below 90 degrees. Maybe that’s because the MJE devices are just cruising along.

EF893C8A-CCE1-4BCC-8C7F-CEB4345F3285.jpeg

Now Q706, Q707 and Q736 and Q737 which are TO126 devices are getting quite warm, close to 180 degrees and the resistors (R725, R775) below them are well over 200! I may actually swap in some larger resistors. Maybe even up to 5 watt. They even look a bit darkened. Since the 2SA914 and 2SC1953 have had 40 years of being this hot would you guys just replace them with a fresh set of KSA1381/KSC3503? Or leave them since they’re working?
C1BF3128-8F41-4EF4-B93D-8AA6E3B7A2EE.jpeg
EE656D6F-DE86-4488-8393-2005E1C2EEF7.jpeg


Anyways, the drivers look like they’re living a fairly easy life looking at the heat being given off, maybe I should try the 1220/2690 or maybe even the Toshiba devices and see how they do. In the end I’ll use the parts that Ray is sending me, but it may be good to know.

Dan
 
Last edited:
Awesome suggestion, I looked at those as well and I actually happen to have a couple hundred of each on hand already. Now the question is how would a TO126 handle the job?

I find it interesting when they put a die into a device and rate it for more voltage and more current, yet put it into a smaller case. Won’t it produce more heat?
Yes, it produces more heat, so the idea of sandwiching them is good. When it's possible, I prefer to use TO-126s to replace TO-220s. And even TO-3P's to replace the all-metal TO-3 cases, because of scarce availability.
Actually, we perhaps may adapt to this challenging scenario of smaller and smaller component parts.

Now Q706, Q707 and Q736 and Q737 which are TO126 devices are getting quite warm, close to 180 degrees and the resistors (R725, R775) below them are well over 200! I may actually swap in some larger resistors. Maybe even up to 5 watt. They even look a bit darkened. Since the 2SA914 and 2SC1953 have had 40 years of being this hot would you guys just replace them with a fresh set of KSA1381/KSC3503? Or leave them since they’re working?
View attachment 2765664
View attachment 2765662
If the 2SA914/2SC1953 combo is doing fine, I'd leave them mainly because of their fT (70MHz). The KSA1381/KSC3503 is more than twice faster (fT 150MHz) and may oscillate. Or maybe not. But Pd is the same.
Upgrading the resistors (R725, R775) to a higher wattage is a good practice here.
 
Looking at the original 2SA913 and 2SC1913 data sheet it looks like they had an output capacitance if 15 pF. I cannot find any capacitance value data for either the MJE15032 or MJE15033, does that mean they have no output capacitance? Nothing on input capacitance of any of them?

Being output capacitance does that mean I’d tack the capacitor across B-E? I’ve never had to add this before. I don’t have all that many values that low, I have 10 pF and 20 pF ceramics and 22 pF silver mica.

Dan
 
Looking at the original 2SA913 and 2SC1913 data sheet it looks like they had an output capacitance if 15 pF. I cannot find any capacitance value data for either the MJE15032 or MJE15033, does that mean they have no output capacitance? Nothing on input capacitance of any of them?

Being output capacitance does that mean I’d tack the capacitor across B-E? I’ve never had to add this before. I don’t have all that many values that low, I have 10 pF and 20 pF ceramics and 22 pF silver mica.

Dan
Parasitic capacitances are always present in semiconductors' P-N junctions.
Usually, the higher the collector current (Ic) with a slower fT, the higher Cob. So the MJE15032/33 (8A, 30MHz) must have a much higher Cob than the 2SA913/2SC1913 (1A, 120MHz). They are not always present in datasheets (and that's annoying), but we can guess an approximate value by knowing Ic and fT.
Looking at the 2SC3955 datasheet: 100mA (Ic) and 300MHz (fT), and a Cob of just 1.9pF... The internal die size also plays a role here: the larger the die, the larger the Cob. That's why the TTC004B (no metal tab) has a rather low Cob for a 1.5A device.
There were through-hole transistors with wonderful specs being made until around 1998. Now they became pretty rare (and $$$) to find.
Gallium Nitride semiconductor technology looks promising though.

paracapa.jpg
 
Last edited:
Parasitic capacitances are always present in semiconductors' P-N junctions.
Usually, the higher the collector current (Ic) with a slower fT, the higher Cob. So the MJE15032/33 (8A, 30MHz) must have a much higher Cob than the 2SA913/2SC1913 (1A, 120MHz). They are not always present in datasheets (and that's annoying), but we can guess an approximate value by knowing Ic and fT.
Looking at the 2SC3955 datasheet: 100mA (Ic) and 300MHz (fT), and a Cob of just 1.9pF... The internal die size also plays a role here: the larger the die, the larger the Cob. That's why the TTC004B (no metal tab) has a rather low Cob for a 1.5A device.
There were through-hole transistors with wonderful specs being made until around 1998. Now they became pretty rare (and $$$) to find.
Gallium Nitride semiconductor technology looks promising though.

View attachment 2765885

That makes sense regarding the capacitance being higher on the MJE components. Can’t really subtract the capacitance without series wiring, could potentially do as John did (missed it the first time reading through) and add 200 to 500 pF between the bases of the driver sets. Is dielectric too big of a concern here? I have ceramic, silver mica and polypropylene, 220 and 330 pF.

I thought it odd that the Toshiba devices done have metal tabs, figure it wouldn’t be as efficient at heat transfer.

Of the TO126 devices which would you use? The Toshiba had a bit higher Ic, but the 2690 has a higher Ic than the original as well and higher ratings than the other specs as well. If the generally recommended transistor has an Ft of 200 MHz then the Ft of the 2690 shouldn’t be a problem.

I guess the only things that concern me about the Toshibas is the no metal tab (I’m guessing they won’t get too hot anyways) and the lower Pd. Though I don’t know that would be an issue, I figure better to be 5 watts more than 5 watts less than the original. As for capacitance I’m not really sure how much a difference of 12, 14, and 19 pF would make. I’m guessing that the capacitance of the MJE devices is higher than 19 pF.

Unfortunately in 1998 I was a sophomore in high school and not buying up transistors. Really wish I could have had the foresight to do so though.
Dan
 
The KSC2690 is fine, indeed better for the application. Pd is 20 watts on H.S., higher than the originals. Cob difference here is negligible.
The MJE's are essentially power devices in small cases, designed for other purposes. They can perfectly "work" as audio drivers, or not. Their Cob should be around 150pF or higher, not really good since the larger power outputs have already 500 or 600pF...
The ZTX753 is a 100V 2A small case power transistor that can dissipate 1 watt with no heatsink.
 
The KSC2690 is fine, indeed better for the application. Pd is 20 watts on H.S., higher than the originals. Cob difference here is negligible.
The MJE's are essentially power devices in small cases, designed for other purposes. They can perfectly "work" as audio drivers, or not. Their Cob should be around 150pF or higher, not really good since the larger power outputs have already 500 or 600pF...
The ZTX753 is a 100V 2A small case power transistor that can dissipate 1 watt with no heatsink.

Odd that I didn’t remember it off the top of my head, but the 2690/1220 pair don’t have the exposed metal tabs either lol. Oh well, I did get them sandwiched as I intended. I had some thicker aluminum TO220 heatsinks I wanted to use, but other components and test points wouldn’t let me. Oh and anyone reading this wanting to do something similar, another reminder to check pinout, these new devices had to be installed backwards.

F7FB1CCE-336F-4E5F-BE49-943A5FDD60F9.jpeg
842E0584-6FB8-4033-805E-43446ECF1CE4.jpeg

All installed.

E7B915BE-A16D-4C3E-8EBF-3B1285BC5860.jpeg

Powered it up and got bias and offset reset and fed a signal. All of that silliness with the oscillation on the output is completely gone with the 2690/1220s installed, because of the capacitance I’m sure.

E78E33A4-6F00-4152-809D-45F1C8EA31C6.jpeg

I’ve had it outputting about 2 watts into an 8 ohm dummy load for about 2 hours now and like before the driver heatsinks are just barely above ambient temp, so far. Once I find that it’s completely stable I’ll run some music through it pretty loudly into actual speakers. I’m guessing the extra heatsinks probably weren’t all that necessary.

Dan
 
Last edited:
Back
Top Bottom