Hello,
Just finished sweeping through more than 200 BJTs for my MC/MM phono preamp, measuring Hfe and Is indirectly through Vbe with diode function of a DVM. A bit of a PITA if you ask me, but you gotta do what you gotta do right? Or do we? Througout this endeavor I couldn't help but wondering, does any audio manufacturer goes through this trouble ? Still have 200 more of the BC549/BC556 to go through... 😡
Thank god for spreadsheets, I can sort the set by Hfe then Vbe or inversely. I also produced statitics from the 2SC2547D/2SA1085D set and right off the bat I can tell you I can't get Hfe matching for the two polarities out of 50 NPN and 100 PNP for the input transistors. I read here that this is normal for the epitaxial transistors.
Descriptive statistics also put forward the following properties:
1. Hfe is generally higher for NPN than PNP;
2. Hfe varies much more than Vbe regardless of polarity;
3. Hfe variance is much higher for NPN than PNP (300.2 vs 47.8)
4. Vbe variance is also higher for NPN than PNP (1.75 vs 0.81), but less variable than Hfe like mentioned in point 2;
5. In my sample, PNPs have Hfe range lower than specified in the datasheet... Bad batch, bad luck, fakes ????
That being said, I'm wondering what to aim for in different parts of the circuit (see below), keeping in mind that Hfe match may be impossible between polarities :
- Joris
Just finished sweeping through more than 200 BJTs for my MC/MM phono preamp, measuring Hfe and Is indirectly through Vbe with diode function of a DVM. A bit of a PITA if you ask me, but you gotta do what you gotta do right? Or do we? Througout this endeavor I couldn't help but wondering, does any audio manufacturer goes through this trouble ? Still have 200 more of the BC549/BC556 to go through... 😡
Thank god for spreadsheets, I can sort the set by Hfe then Vbe or inversely. I also produced statitics from the 2SC2547D/2SA1085D set and right off the bat I can tell you I can't get Hfe matching for the two polarities out of 50 NPN and 100 PNP for the input transistors. I read here that this is normal for the epitaxial transistors.
Descriptive statistics also put forward the following properties:
1. Hfe is generally higher for NPN than PNP;
2. Hfe varies much more than Vbe regardless of polarity;
3. Hfe variance is much higher for NPN than PNP (300.2 vs 47.8)
4. Vbe variance is also higher for NPN than PNP (1.75 vs 0.81), but less variable than Hfe like mentioned in point 2;
5. In my sample, PNPs have Hfe range lower than specified in the datasheet... Bad batch, bad luck, fakes ????
That being said, I'm wondering what to aim for in different parts of the circuit (see below), keeping in mind that Hfe match may be impossible between polarities :
- The parallel transistors of the input cascode need to be matched for Is; that is have the same collector current for a given Vbe for equal current sharing between them;
- By which parameter to match the cascode's top transistors?
- Hfe influences the device's dynamic emitter resistance (re), but when emitter degeneration resistors (Re) are present, Hfe matching is more relaxed. So do I need to place the closest matches where the re/Re ratio is highest, e.g. where the degeneration resistor has the least effect on total output impedance?
- Do the CCS transistors need to be matched from one rail to the other? Matched for which parameter?
- Do the current buffer transistors (discrete Darlington configuration) need to be matched, and if yes by which parameter, total Darlington combined gain from rail to rail? I feel these positions are less critical since gain will be high anyways...
- By which parameter to match the legs of long-tailed pairs?
- Joris
Last edited:
That would make equipment unaffordable.
Manufacturers adapt the design to the market: the design should be stable when using unselected stock.
This exhaustive selection method is a hobbyist thing.
1: type dependend
2: they're related, but no strict law
3/4/5: type dependend
Have you considered thermal issues? What guarantee do you have that they will remain exactly the same over the operational temperature range?
If you avoid the emitter resistors (E-degen), other issues will rise (stability, noise). It's a balancing act.
I've used 5200/1943 in Sziklai's, the product of the beta's is more or less equal (5200: B=105, 1943: B=73 ; 1943: B=93, 5200: B=81 ; combo's: 7665 & 7533, a difference of <2%). Have the highest beta in front.
You are over-thinking the matching requirements. Really only the three parallel transistors without emitter resistors need to be closely matched.
Ed
Ed
Sorry for that... I felt the forum wasn't appropriate the first time because it didn't generate answers and I can't find how to change a post's forum
Hi SomeJoe,
I designed a transistor matching jig over a decade ago and gave it to the members of this forum. Pairs must be at exactly the same temperature to match them, so beta or VBE measurements are not that useful (which is why I designed that jig).
Three or four PCB designs have been made for it by now. Matching does matter, a lot at times. Search around for it. I think my first post was in an Adcom GFA565 thread near the end.
I designed a transistor matching jig over a decade ago and gave it to the members of this forum. Pairs must be at exactly the same temperature to match them, so beta or VBE measurements are not that useful (which is why I designed that jig).
Three or four PCB designs have been made for it by now. Matching does matter, a lot at times. Search around for it. I think my first post was in an Adcom GFA565 thread near the end.
Thanks Chris for your reply. During my initial research I did stumble on your posts about the matching jig, nice of you to have shared this impressive design with the community. I just wasn't sure if I needed to go to that level since most of the transistors in my circuit have fairly large emitter degeneration resistors.
For what it's worth, all the transistors I measured where "planted" in a large piece of soft foam and stored in the same tray:
I let them rest an hour after placing them in the foam so they get to an uniform temperature and handled them with plastic tweezers when doing the measurememts. I keep my room temperature stable with an electronic thermostat, these things are precise to less than a degree Celcius (verified with an infrared thermometer).
Still I'm sure your matching jig can get much more precise matches. I might get one of them PCBs if the end result is not satisfactory.
For what it's worth, all the transistors I measured where "planted" in a large piece of soft foam and stored in the same tray:
I let them rest an hour after placing them in the foam so they get to an uniform temperature and handled them with plastic tweezers when doing the measurememts. I keep my room temperature stable with an electronic thermostat, these things are precise to less than a degree Celcius (verified with an infrared thermometer).
Still I'm sure your matching jig can get much more precise matches. I might get one of them PCBs if the end result is not satisfactory.
The easy way to get good temperature matching is forced airflow (a small fan). This greatly reduces the effect of radiant heat from your body that can affect a black epoxy package by a degree C or so easily - people always forget about radiant heat! A metal foil screen to place the devices behind will help. In a room at 21C, 35C body (like your face) emits about 90W per square metre (perhaps 6W for a face), yet a small transistor will heat up noticably with only a few mW extra heat input. It just has to absorb perhaps 0.1% of the heat radiation from you to matter.
The plastic tweezer's don't protect from the radiant heat from your hands, note.
Radiant heat equation at room temperature: power per square metre ~= 6 * (T - Tambient) W/m^2. (T, Tambient in degrees C)
Derived from Stefan's law.
Radiant heat is how you can sometimes detect someone looming behind you - your neck feels the slight temperature increase.
The plastic tweezer's don't protect from the radiant heat from your hands, note.
Radiant heat equation at room temperature: power per square metre ~= 6 * (T - Tambient) W/m^2. (T, Tambient in degrees C)
Derived from Stefan's law.
Radiant heat is how you can sometimes detect someone looming behind you - your neck feels the slight temperature increase.
Last edited:
No surprise here. Electron mobility is higher than hole mobility in silicon.
Hfe is dependent on the width of the base region in the transistor, which for a vertical transistor is determined by the way the doping is done. We're talking dimensions in nm or maybe even angstrom here so that's not a huge surprise either. Vbe depends on the doping concentrations (and starting material properties) which is easier to control.
Or a difference in the measurement setup. Hfe depends on Ic.
Generally PNPs perform worse than NPNs. I've worked in a few IC processes where the opposite was true, but those were the exception rather than the norm.
Discrete circuits are generally designed to rely on resistor tolerances as those can be pretty tight. If your circuit requires matching of semiconductors I'd redesign it so it doesn't.
The three parallel input devices could benefit from matching so they share the current evenly. Or... You could redesign the circuit such that each transistor had its own emitter resistor. Then they'd share the current pretty evenly and you don't need to match them.
That doesn't look like electrostatic dissipative foam. If it's just plain plastic foam (i.e., not the ESD dissipative kind) you run the risk of wrecking the devices with ESD strikes.
Semiconductors are sensitive to electrostatic discharge. Keep in mind that it's easy to build up an electrostatic charge in the several 10s of kV range just by moving around, especially if you're on carpet, wearing synthetic clothing, or sitting on a chair that contains synthetic materials.
I cringe when I see ads in the Swap Meet where the supposedly ULTRA-RARE, NOS, blah, blah, blah devices have been stored in plastic Tupperware bins. GET YOUR SUPER-RARE ESD DAMAGED DEVICES HERE!! L@@K!!
Tom
Thank you Tom for this very informative reply. I learn a lot here thanks to good folks like you who take the time to write thorough answers.
I though that was mainly the FET/MOS/CMOS technology devices who were sensitive to ESD due to their insulated gate ?
MOS devices are more sensitive, yes. The gate oxide is pretty thin, especially in modern devices. Power MOS devices can be a bit more rugged because they tend to be large, but I wouldn't count on them being tolerant of ESD by any stretch of imagination.
Diodes, BJTs, and J-FETs are bit more rugged and some are large enough that they can survive an ESD strike. I still wouldn't count on them being self-protective.
If a device comes in ESD packaging from the manufacturer I'd keep it in ESD packaging until I'm ready to use it. At Texas Instruments we were taught that two layers of ESD packaging was needed for shipment. That's why a reel of parts (that are on ESD dissipative tape) arrive in an ESD bag.
The worst part about ESD damage is that it can be latent damage. I.e., the device might work fine now but it'll suffer a premature failure. Or, even worse, its performance has shifted so the overall circuit doesn't perform as well.
Also note that if you're in a humid environment you can afford to worry slightly less about ESD as there's enough leakage resistance through the air to keep the electrostatic buildup somewhat in check. Visit Calgary on a winter day after a week or two below -20C when the humidity is down around 15-20 % RH and you'll see what static electricity is. 🙂
I was playing around with that a few years ago where we didn't get much above -25C for the month of February. It turns out that just the static that's generated from me getting up from my desk chair results in enough voltage to jump from one end of a 1 MΩ, 1/4 W resistor to the other. I was hoping the current would flow through the resistor but the spark was visible along the entire resistor body. That hurt, by the way.
I use a grounded electrostatic dissipative mat on my desks and lab benches to make sure I don't damage anything before it goes to a customer. Digikey/Mouser has the best prices on those, by the way. You can also get ESD foam there.
Tom
Diodes, BJTs, and J-FETs are bit more rugged and some are large enough that they can survive an ESD strike. I still wouldn't count on them being self-protective.
If a device comes in ESD packaging from the manufacturer I'd keep it in ESD packaging until I'm ready to use it. At Texas Instruments we were taught that two layers of ESD packaging was needed for shipment. That's why a reel of parts (that are on ESD dissipative tape) arrive in an ESD bag.
The worst part about ESD damage is that it can be latent damage. I.e., the device might work fine now but it'll suffer a premature failure. Or, even worse, its performance has shifted so the overall circuit doesn't perform as well.
Also note that if you're in a humid environment you can afford to worry slightly less about ESD as there's enough leakage resistance through the air to keep the electrostatic buildup somewhat in check. Visit Calgary on a winter day after a week or two below -20C when the humidity is down around 15-20 % RH and you'll see what static electricity is. 🙂
I was playing around with that a few years ago where we didn't get much above -25C for the month of February. It turns out that just the static that's generated from me getting up from my desk chair results in enough voltage to jump from one end of a 1 MΩ, 1/4 W resistor to the other. I was hoping the current would flow through the resistor but the spark was visible along the entire resistor body. That hurt, by the way.
I use a grounded electrostatic dissipative mat on my desks and lab benches to make sure I don't damage anything before it goes to a customer. Digikey/Mouser has the best prices on those, by the way. You can also get ESD foam there.
Tom
Oh, and one more thing: Modern ICs have ESD protection built in. This protection is intended for protecting the devices during assembly in an ESD-controlled environment. You'll need additional ESD protection on anything that the end user can touch.
Topping learned that the hard way when their L30 started exploding in customers' hands: https://www.head-fi.org/threads/topping-l30-exploding.958806/
Tom
Topping learned that the hard way when their L30 started exploding in customers' hands: https://www.head-fi.org/threads/topping-l30-exploding.958806/
Tom
Now that's just depressing... I might have wrecked 50$ worth of transistors. I live in Quebec, baseboard heating and my humidifier running 24/7 can't even get to 20% humidity so yeah, I know what static is 😎
I keep my parts in regular plastic pill bottles (heeeek...!), including MCUs and I have projects here that used these parts running 24/7 for a couple years but that doesn't mean they will not fail eventually. Looks like I'll need to order some of those ESD protection mats and foam.
Oy. Fun...
I use these for storage: https://www.flambeaucases.com/special-order-conductive-and-static-dissipative#conductiveboxes
Not cheap, but safe for ESD-sensitive parts.
You can also get folding cardboard boxes that are ESD safe: https://www.uline.ca/Grp_499/ESD-Storage
I love the ESD safe trash can, by the way. 😀 That's something you'd see in a production environment. In such environment the furniture would be ESD safe and the floor would be treated periodically with ESD safe wax. Also note how the ESD Restocking Cart has a chain that drags along the floor to dissipate any charge collected.
Of course, it's up to you how far you want to go with this. It's a hobby after all. But if you build for others I would definitely follow ESD-safe procedures.
Tom
I use these for storage: https://www.flambeaucases.com/special-order-conductive-and-static-dissipative#conductiveboxes
Not cheap, but safe for ESD-sensitive parts.
You can also get folding cardboard boxes that are ESD safe: https://www.uline.ca/Grp_499/ESD-Storage
I love the ESD safe trash can, by the way. 😀 That's something you'd see in a production environment. In such environment the furniture would be ESD safe and the floor would be treated periodically with ESD safe wax. Also note how the ESD Restocking Cart has a chain that drags along the floor to dissipate any charge collected.
Of course, it's up to you how far you want to go with this. It's a hobby after all. But if you build for others I would definitely follow ESD-safe procedures.
Tom
Of course, but we put a good deal of effort in those builds, would hate to have them malfunction because of improper procedures... I will pay more attention to ESD from now on, thanks for the good info.
Hi Mark,
The parts should be in thermal contact. We want them to be at the same temperature. Place a foam hood over them, and a box over the entire jig to stop air currents. That works well and you can get sub 1% matches. When you install the parts, diff pairs should have thermal compound to aid temperture match, and be ni heat shrink to keep them in good contact, while also blocking some air current effects. You could also place foam over the pair if you wanted. In test equipment these also go inside an enclosure when it is important. That can cut subsonic noise. If you consider the physics, everything makes sense.
Hi SomeJoe,
You're doing things the way I did it in 1980.
No, that isn't correct. I tried and one board was designed with a fan, turns out it was a bad idea. It doesn't work.
The parts should be in thermal contact. We want them to be at the same temperature. Place a foam hood over them, and a box over the entire jig to stop air currents. That works well and you can get sub 1% matches. When you install the parts, diff pairs should have thermal compound to aid temperture match, and be ni heat shrink to keep them in good contact, while also blocking some air current effects. You could also place foam over the pair if you wanted. In test equipment these also go inside an enclosure when it is important. That can cut subsonic noise. If you consider the physics, everything makes sense.
Hi SomeJoe,
You're doing things the way I did it in 1980.
You can't tell unless you then check with the jig, or measure DC offsets and compare that to the calculated value from the circuit design. Distortion is reduced by many things, but this also has a direct impact on distortion. Whatever is your largest issue for distortion will dominate.
You'll either be surprised (or depressed) as to how much variability there is with how you're doing things. Sometimes it works out, and this is better than how others try to match parts. The entire reason I went to the trouble to design the matcher was by seeing how variable my, and other methods were. I thought about it and realised that temperature was the variable. The jig solved this and allows good matches. The best part is your meter becomes a null detector as long as your zero is correct. I should try an old tuner centre tune meter.
Hi Tom,
You are bang-on. Semiconductor damage does not have to be catastrophic at all. It sometimes doesn't take much to cause excessive noise, or complete failure at a later date. Oxide devices like Mosfets, or Jfets have very high impedances, so staic can build up to extremely high levels easily. Even BJTs can be "zapped".
That's why proper distributors use anti-stat bags and foam to transport devices. It isn't overkill.
As for circuit design, your error amps must be matched closely. Other device matching always matters, we can talk about degree. If your error amp is an op amp with over 100 dB of gain and a high turnover frequency you don't have to match the other parts as closely for a given distortion level, but matching always improves performance. The less nonlinearity your front end has to correct for, the better the overall performance will be. In manufacturing you can't easily match parts economically, so you design to make it less important. But if someone comes in behind you and does match the parts that need it, performance should be improved in a linear circuit.
You are bang-on. Semiconductor damage does not have to be catastrophic at all. It sometimes doesn't take much to cause excessive noise, or complete failure at a later date. Oxide devices like Mosfets, or Jfets have very high impedances, so staic can build up to extremely high levels easily. Even BJTs can be "zapped".
That's why proper distributors use anti-stat bags and foam to transport devices. It isn't overkill.
As for circuit design, your error amps must be matched closely. Other device matching always matters, we can talk about degree. If your error amp is an op amp with over 100 dB of gain and a high turnover frequency you don't have to match the other parts as closely for a given distortion level, but matching always improves performance. The less nonlinearity your front end has to correct for, the better the overall performance will be. In manufacturing you can't easily match parts economically, so you design to make it less important. But if someone comes in behind you and does match the parts that need it, performance should be improved in a linear circuit.
Hello anatech,
I am a bit too tired right now to search and browse the relevant threads so I will just ask, does your matching jig do power transistors matching as well or just small signal level? I have read on Elliot Sound Product's site that proper measurement of power devices require a different setup than small signal - mainly to recreate the very different operating conditions, but I think one can make a kit that does both.
I am a bit too tired right now to search and browse the relevant threads so I will just ask, does your matching jig do power transistors matching as well or just small signal level? I have read on Elliot Sound Product's site that proper measurement of power devices require a different setup than small signal - mainly to recreate the very different operating conditions, but I think one can make a kit that does both.
In concept, the jig I designed will work for power devices. Just increase the test current to 50 ~ 100 mA and run leads to sockets. You could increase the current higher to deal with more devices. Restrict the size of heat sink so they stabilize sooner, but don't run hotter than they normally would. A 10 volt or so test voltage is still valid. Your time to thermal equilibrium will be longer with power devices, but it is anyway with other methods. You must allow the temperatures to settle when matching signal transistors as well.
Right now I have heat sinks drilled for TO-204 devices and can also use it for plastic packages. I run the base currents to be equal and measure collector currents. I don't measure emitter currents because I don't want to desensitize the jig for differences, they would act as emitter resistors. Each base has a resistor so I can measure base currents individually. I let them cook and then measure them four at a time. Thankfully they don't need to be as close as signal transistors.
The Counterpoint mosfet amps needed extremely close matching to remain together, no source resistors in those (stupid), but running the amp "open loop" they couldn't use source resistance. There is feedback they didn't know about, but it is distorted. Eventually I designed a bipolar output stage, making them repairable again. More reliable by far with much less distortion. Some Carver amps used 0.05 ohm emitter resistors. The lower the emitter resistor, the more critical the transistor matching is when you have multiple outputs per rail.
With output stages, matching is far more important at low currents. Emitter / source resistors force current sharing at higher currents, and if they match at low current, the resistors will have the final word at higher currents. Lateral Mosfets do tend to self balance, but they do work better matched. So while you don't have to match them to keep them from going kaboom (within reason of course), they do operate with lower distortion if matched.
Right now I have heat sinks drilled for TO-204 devices and can also use it for plastic packages. I run the base currents to be equal and measure collector currents. I don't measure emitter currents because I don't want to desensitize the jig for differences, they would act as emitter resistors. Each base has a resistor so I can measure base currents individually. I let them cook and then measure them four at a time. Thankfully they don't need to be as close as signal transistors.
The Counterpoint mosfet amps needed extremely close matching to remain together, no source resistors in those (stupid), but running the amp "open loop" they couldn't use source resistance. There is feedback they didn't know about, but it is distorted. Eventually I designed a bipolar output stage, making them repairable again. More reliable by far with much less distortion. Some Carver amps used 0.05 ohm emitter resistors. The lower the emitter resistor, the more critical the transistor matching is when you have multiple outputs per rail.
With output stages, matching is far more important at low currents. Emitter / source resistors force current sharing at higher currents, and if they match at low current, the resistors will have the final word at higher currents. Lateral Mosfets do tend to self balance, but they do work better matched. So while you don't have to match them to keep them from going kaboom (within reason of course), they do operate with lower distortion if matched.