An Attempt to Quantify Semiconductor Differences

[rodevil]

the last 2 bars of the first sample appear to be "louder" , and the last 2 of the second sample appear more heavy on the lowend

but maybe i really should not judge hearing on the mbp speakers

[bcbox]

...let's get a couple more opinions here and I'll let you know what this particular test is about.

[Altitude]

Maybe it's my blown out hearing but I gave this a good listen on cans and I cant hear anything dramatic. Maybe a slightly different whistle to the resonance but I wont swear to that in court.

Are you being sneaky and doing a whats the difference when there is no difference to make a point thing by any chance?

[antto]

bcbox: simply no audiable difference.

[bcbox]

ok fair enough.
I have many tests like this that I've done while qualifying certain parts as substitutes. Most of it is for my own use for my own designs but I plan to share much of my results. This was simply a test of BA662 versus BA6110, with the open loop gain of the BA6110 adjusted to match the BA662. This is done by adjusting R127/R128/R129. I'll update with the exact value tonight.

I can not hear a difference myself.
I have also done numerous measurements on my AP system, gain linearity, THD, etc.. and everything is the same except the BA6110 is actually a little more noisy than the BA662. Also, depending on the matching of the transistors used for the current mirrors at the control input the gain linearity can change slightly.. nothing that would really be audible though.

As for the samples.
The first and third bars are the BA662.
The second and fourth bars are the BA6110.

[Altitude]

Very cool indeed. How are you matching the transistors for the mirror? By Hfe or Vbe?

[bcbox]

How are you matching the transistors for the mirror?

just by Hfe. Technically the closer they are and higher the beta the better, but it's really not that important. Just using a hand held meter with Hfe function to match is fine. Even worst case wouldn't be bad. Some 303 clones just use the gate to the VCA (no decay envelope) and people can't even tell much difference with patterns at typical tempo.

[Brassteacher]

Hi guys, Sorry for not getting many updates posted, I've been working on other things, as well as salvaging dead gear trying to get more space in my work area. This has been kind of helpful to this project though, I found enough 1SS133 diodes inside a dead Korg DW-8000 to refit the analog section of the x0x board, so that test can be done later as well. Also, I've been transistor hunting on quite a few surplus sites, and came into a batch of 2SA733AP transistors, among others. I've also found a few more flavors of K30A to play with, both GR and R suffix. I've also been playing around with different NPN transistors, so there will be a few tests with what I'll call "mystery transistors" in key positions.

I'm getting cabin fever from being trapped inside by heavily iced-over roads, so tonight might be a good night to play with the scope and Audacity.

antto, if you see this, I'm going to find a clean LFO and record a low-frequency square wave tonight so that you can do the adjustments you were talking about.

EDIT: antto, I guess I lied. The cleanest square wave generator (not a synth, but a function generator) I have only goes down to 146Hz . All my analog stuff that's currently operational can go subsonic, but less than 100Hz on any of them looks definitely un-square on the 'scope. I'll breadboard a little flip-flop based circuit soon and record/take screen shots of that.

[antto]

uhm, wait
i'm not sure what you're talking about, i needed pictures of the square wave on oscilloscope
..at different notes
no one helped, so i recorded the VCO directly into my soundcard, and later - i "restored" the DC signals in the recording, not perfectly, but it works

i have a few new ideas but electronics are involved and i can't do it on my own, so i guess i can post it in the forum if someone likes to make em happen

one of them is about a PC oscilloscope
yes, there are already lots of apps, but so far - all of them don't handle the AC-coupling of the soundcard (and it looks like 90% of the soundcards are ac-coupled)
this means - with a soundcard scope - you can't see very low frequencies correctly (like 1Hz) because of the effect of the DC-Filter in the soundcard which is usually set to 1Hz to 20Hz (mine was 16Hz)

but there is a way to "restore" the DC content of the recorded signal afterwards
i wonder why no one made such an app that does this..
well, i guess this is not the place


here's what i did
i regret that i didn't take even lower freqs in the recording
i no longer have the same soundcard (the motherboard burned) and my current soundcard sux, it's pointless to record at 192K with it..

[Brassteacher]

uhm, wait
i'm not sure what you're talking about, i needed pictures of the square wave on oscilloscope
..at different notes
no one helped, so i recorded the VCO directly into my soundcard, and later - i "restored" the DC signals in the recording, not perfectly, but it works

i have a few new ideas but electronics are involved and i can't do it on my own, so i guess i can post it in the forum if someone likes to make em happen

one of them is about a PC oscilloscope
yes, there are already lots of apps, but so far - all of them don't handle the AC-coupling of the soundcard (and it looks like 90% of the soundcards are ac-coupled)
this means - with a soundcard scope - you can't see very low frequencies correctly (like 1Hz) because of the effect of the DC-Filter in the soundcard which is usually set to 1Hz to 20Hz (mine was 16Hz)

but there is a way to "restore" the DC content of the recorded signal afterwards
i wonder why no one made such an app that does this..
well, i guess this is not the place


here's what i did
i regret that i didn't take even lower freqs in the recording
i no longer have the same soundcard (the motherboard burned) and my current soundcard sux, it's pointless to record at 192K with it..

Yeah, that was it, you wanted to restore the DC content and figure out where my computer's input rolled off. Ok, screenshots of the 'scope on a square wave recording, that I can easily do. I mean, if we're going to this much trouble, we may as well be as accurate as possible. Besides, this IS a bass machine we are talking about. I should pop this old computer open also, and trace the filter from the input (if possible, damned multi-layer boards) to see if we're looking at a 6dB, 12dB, 18dB, or 24dB input filter, assuming it's discrete and not digital. There would have to be at least one blocking cap regardless.

I think I'll run all the tests with different JFETs at Q28 at the same time, recordings and screen shots off the 'scope. That'll give me something to do tonight to keep from going stir-crazy, a 300-yard long sheet of ice on the street in front of my house that's been there all week has quite discouraged going out...

[Brassteacher]

Sorry about the apparent lack of progress, unfortunately many other things have been taking up my time lately. The delay has not been without some progress though, since one of the things that has kept me busy was dismantling old equipment in the garage, which has added quite a few interesting transistors to play with.

I do need a bit of info from someone, or more than one if possible, that has a real TB-303. I need to know if there is any variation in the Hfe ratings on the C2291 and C1583 transistors. The service manual lists the parts as having the "F" suffix (the last letter of the "date code" is the Hfe rating). I've managed to collect C2291 transistors with F,G, and H suffixes, and C1583 transistors with F and G suffixes.

If any F, G or H rated transistors made it into any of the production TB-303s, it could possibly make a HUGE difference in sound, depending upon where the transistor wound up. For instance, my lowest measuring C1583 F has an Hfe of 349/344, while the highest measuring C1583 G has an Hfe of 675/681! That just may make a wee bit of difference in sound, possibly, perhaps something on the order of the difference in performance between a Yugo and a Ferrari. The differences in the F,G, and H rated C2291 transistors in my stash range from 281/283 to 660/668.

If no reports of anything other than F suffix transistors come in, that actually opens up the number of possible substitutions for those hard-to-get parts.

Anyhow, the soldering iron is currently out of commission, which will leave time to play with the x0x a bit over the next day or two.

[bcbox]

If any F, G or H rated transistors made it into any of the production TB-303s

I'm 100% confident that all 303's used 1583F & 2291F. I've worked on dozens, never seen anything else. For the exponential converter, Q26, F/G/H doesn't make a difference, other than offsetting Ic slightly which will change the pitch of the VCO a few cents. Q21 of the differential amp is the most sensitive; there's not much difference with 1583G but 1583H makes a small audible difference. Q12 in the filter ladder doesn't sound any different regardless of which part is used. I do have samples of all these tests etc in my files, I just don't have the time to share it all at this point. It would be good to do these same type of tests to compare results.

[Brassteacher]

If any F, G or H rated transistors made it into any of the production TB-303s

I'm 100% confident that all 303's used 1583F & 2291F. I've worked on dozens, never seen anything else. For the exponential converter, Q26, F/G/H doesn't make a difference, other than offsetting Ic slightly which will change the pitch of the VCO a few cents. Q21 of the differential amp is the most sensitive; there's not much difference with 1583G but 1583H makes a small audible difference. Q12 in the filter ladder doesn't sound any different regardless of which part is used. I do have samples of all these tests etc in my files, I just don't have the time to share it all at this point. It would be good to do these same type of tests to compare results.

Will do, as even among the F rated parts the Hfe can vary by over 200 from one to another, at least with the ones I have.

Thanks for the response, it was helpful. I've been working on audio/synth gear long enough that I no longer assume that the part called for in the service manual will be in whatever unit I'm working on, or even that the service manual itself is correct.

[Brassteacher]

Finally, some useful data. The screen shots are taken with my 'scope set to 0.2V per vertical division. For all the different JFETs, this setting was not changed in order to show amplitude differences clearly, even though in one case the trace ran off the top of the screen a bit.

The time base for each photo was adjusted slightly to fit one wave on the screen, and to have that one wave fill the screen. In all the photos, the trace on the left is at 100Hz (between 100.0 - 100.5Hz), and the trace on the right is at 200Hz (between 200.0 - 200.5Hz).

After testing, I measured the Idss of each JFET actually used in the test. This is Idss measured with a drain-source voltage of 9V, actual measurements, not datasheet specs. All JFETs were measured at room temperature, and powered up only long enough to get a good reading. In other words, "cold".:
J201: 0.56mA
K30A-O: 1.00mA
K30A-R: 0.54mA
K30A-Y: 1.81mA
K30A-GR: 4.52mA

EDIT: In all these tests, Q28 was the only JFET swapped out. A J201 was used for Q39 in a cases. Why? Q39 is a constant current source for the envelope generator. It makes no difference here, as the output of the VCO was taken right off the waveform selector switch.

First up, the J201:

J201

SQ-J201.jpg (65.67 KiB) Viewed 2383 times

K30A-O:

K30A-O

SQ-K30A-O.jpg (67.58 KiB) Viewed 2383 times

K30A-R:

K30A-R

SQ-K30A-R.jpg (67.08 KiB) Viewed 2383 times

Continued in next post...

[Brassteacher]

...Continued from above.

K30A-Y:

K30A-Y

SQ-K30A-Y.jpg (71.43 KiB) Viewed 2381 times

K30A-GR:

K30A-GR

SQ-K30A-GR.jpg (61.65 KiB) Viewed 2381 times

Results based on oscilloscope measurements:
Going simply from the visual data, it appears that a JFET with an Idss between 0.5 and 1.0mA is in the "sweet spot". The J201, K30A-O, and the K30A-R all fall within this range. All three of these give nearly a 50% duty-cycle, all three show a slight decrease in pulse width as the frequency increases, with the J201 showing the least change in pulse width vs. frequency. The K30A-O shows a slight increase in amplitude overall. This appears to be related to Idss as we'll see in a moment. By the way, I forgot to mention I had the 'scope's input AC coupled to save time by not tweaking the V/division setting. In all cases, even with the scope AC coupled, the bottom of the square does not descend as far as the frequency increases.

In the case of the K30A-Y and the K30A-GR, these both have a significantly higher Idss than any of the previous JFETs. The most apparent difference is the increase in amplitude. The higher the Idss, the higher the amplitude.

NOTE: Do not take the 0.2V per division setting as "gospel", the 'scope is a Techtronix 922R, which has a pot in the center of the selector switches that allows you to vary the V/div. anywhere between the two adjacent settings, so that you can adjust the trace to fit the screen. However, this was not touched after the first test, so the relative amplitude from one JFET to another is valid.

Also with the K30A-Y and the K30A-GR the pulse width starts narrow relative to the others, and gets even narrower as the frequency increases.

My bottom-line opinion, based on visual data: If you can't get a 2SK30A-O for Q28, find either a 2SK30A-R or a J201. Forget about using a 2SK30A-Y, and certainly stay away from the 2SK30A-GR.

Why did I not test any other JFETs? Based on the parametric search function on the Mouser website, the ONLY currently available JFET that's even close to the 2SK30A-O is the J201. I had intended to test 2SK381, since Korg used those in place of K30A in certain applications, but K381 is even MORE difficult to locate now than K30A!

How much difference does this make in sound? We'll know as soon as I get the recordings posted!