For output stage transistors Hfe mismatch causes a discontinuity in the crossover region as the effective Hfe of the pair jumps from one transistor to the other. This is closer to the 1st watt, and high order harmonics are more audible, so I would match Hfe at the same Vce and bias current as they will be used in the amplifier. Maybe a little hotter if the amp will be played loud. At this point what happens at higher powers may suffer a little more, but that lower order harmonic distortion will probably not be as audible as a discontinuity at crossover. I would look for transistors that naturally have flat Hfe at high currents and try to match those.
When you do the matching, you give both transistors the same Ic and you select a pair that has the same Vbe. Ensuring Ic and Vbe both match is EXACTLY what you are doing in the matching jig. I don't need statistics to tell me that if I cut two wires to be 1" long, they will still both be 1" long the next time I measure them.
Vbe is the single most important factor in determining Ic. Early effect, Ib and temperature were all taken care of in the test jig to eliminate them as confounders. Beyond those, maybe cosmic rays or radiation could have an effect? Do you keep a stash of uranium under your bench?
I can't help but think all this FUD must make building an amp a harrowing task for you. Don't curse yourself with a fear of the bench. You need to get your hands dirty and see for yourself what matching can and cannot do.
Reasons the measured Hfe might be different from the matched Hfe used in the original design:
1: Not measured at the same Ic that was used for matching (many transistors have low Hfe at low Ic)
2: Both transistors are at a lower temperature than was used for matching (tempco is different for NPN and PNP).
3: The temperature of the transistors was different than the temperature used for matching.
4: The Vce of the transistors was lower than was used for matching (Early effect causes Hfe to change with Vce). This is usually not a problem because power outputs tend to have very high Early voltage, which makes them less sensitive to Vce. Vce will increase temperature however, leading to #2.
When you do the matching, you give both transistors the same Ic and you select a pair that has the same Vbe. Ensuring Ic and Vbe both match is EXACTLY what you are doing in the matching jig. I don't need statistics to tell me that if I cut two wires to be 1" long, they will still both be 1" long the next time I measure them.
Vbe is the single most important factor in determining Ic. Early effect, Ib and temperature were all taken care of in the test jig to eliminate them as confounders. Beyond those, maybe cosmic rays or radiation could have an effect? Do you keep a stash of uranium under your bench?
I can't help but think all this FUD must make building an amp a harrowing task for you. Don't curse yourself with a fear of the bench. You need to get your hands dirty and see for yourself what matching can and cannot do.
Reasons the measured Hfe might be different from the matched Hfe used in the original design:
1: Not measured at the same Ic that was used for matching (many transistors have low Hfe at low Ic)
2: Both transistors are at a lower temperature than was used for matching (tempco is different for NPN and PNP).
3: The temperature of the transistors was different than the temperature used for matching.
4: The Vce of the transistors was lower than was used for matching (Early effect causes Hfe to change with Vce). This is usually not a problem because power outputs tend to have very high Early voltage, which makes them less sensitive to Vce. Vce will increase temperature however, leading to #2.
Last edited:
I asked Bob Cordell his opinion on the subject
https://www.diyaudio.com/forums/sol...lls-power-amplifier-book-921.html#post5752035
https://www.diyaudio.com/forums/sol...lls-power-amplifier-book-921.html#post5752035
Opamps cant take the high voltages, dosnt this make them as much work as discrete? Not to mention discrete is more flexible, like if you want to change the LTP emmiter resistors, etc.
And they do make matched transistor pairs in one package that are as good as the ltp pair in most opamps.
Last edited:
Dont think theres a dicontinuity unless the amp is class B. In A/B amps the bias current fills in the gap, so not as big a deal.
How many great amps have been designed in the past when there was no such thing as complementary transistors, even the ones that were called that had NPN hfe's twice that of PNPs. So Sure matching helps, but, how much is the real question.
Hi cbdb,
Hi Rick,
Now, for our learned friends talking about matching input transistors, every single matched pair sold has Hfe match listed and as a defining parameter on price. This is done by people way above my pay grade, so I'll trust them on this.
Hi mchambin,
I took physics and semiconductor physics in university, but I didn't follow that path through. It has helped enormously for what I do, but so does my empirical knowledge built up over a long period of time.
I have jigs for mosfet power transistors, and bipolar power transistors. They are designed to match parts closely, and it works. Believe me, if there was a less expensive route I could take, I would as long as it didn't impact the quality of what I do.
Get some thermal compound under your nails for a change, and really experiment with real commercial products. Sometimes a little practical knowledge really helps to understand what is going on out there.
-Chris
I'll have to read things more carefully instead of scanning quickly. Thank you for pointing that out. I have a mountain of service to get through and want to participate here as well. But, I do have to focus on the service work as much of it concerns failed service attempts in other shops. These are more difficult as you might imagine. The one I'm on now suffered from panicking technician syndrome. He twisted all kinds of pots unrelated to the original fault (which is fixed now).
Hi Rick,
Well, that's where my confusion actually begins. Yes, I was taught about transistors being voltage controlled devices, but in service work it is often easier to think in terms of current flow and approximations to come to an understanding of what the circuit is actually doing. I'm fixing, not normally designing these repairs. Although, lately it seems that things need some redesign to make them work correctly and reliably. Some of our engineers should be forced to repair their designs for a couple years. That will either make them better, or make them quit.Maybe Chris should focus on repairs vs semiconductor physics
It can be useful as it makes the load on the Vas more equal on positive vs negative swings. This will also depend on the exact circuit you are working on - as always. I found that it often reduces distortion a bit for whatever reason.
Obviously we disagree. And DIY has nothing to do with manufacturing high volumes of equipment. We can afford the time. Often, you are looking at things from a QA department by asking why it failed for starters. If you are trying to improve the performance of well designed amplifiers, then it pays to match components within reason. The degree of match depends on circuit (as Bob mentioned) and component values (Carver used 0R05 emitter resistors). You can't close your eyes and dictate some degree of match to fit every situation either, and I didn't suggest that. Where it matters, match the parts. For higher performance, match them tighter. Using various jigs is normally the only way to keep temperatures the same between members of a group. Yes, it can be a PITA. Example, Counterpoint output mosfets.
Now, for our learned friends talking about matching input transistors, every single matched pair sold has Hfe match listed and as a defining parameter on price. This is done by people way above my pay grade, so I'll trust them on this.
Hi mchambin,
Actually, running the pair at higher tail currents makes them converge more quickly, then you can drop the current to what your circuit uses. That is why my jig has adjustable tail currents. The jig makes an impossible job far easily accomplished and at greater degrees of match
I have. We are not dealing with transistors on the same substrate often, or with thermal gradients (that are designed to be a minimal problem). What you can do with devices on the same patch of silicon doesn't apply too well to the real nasty world of discrete parts.
Okay, so what would you do. I have found that on some test jigs, the low current match can hold to higher currents, and it sure beats doing nothing. Besides, I think it matters most for low currents or for parts that are working into very low emitter resistor values.
I took physics and semiconductor physics in university, but I didn't follow that path through. It has helped enormously for what I do, but so does my empirical knowledge built up over a long period of time.
I have jigs for mosfet power transistors, and bipolar power transistors. They are designed to match parts closely, and it works. Believe me, if there was a less expensive route I could take, I would as long as it didn't impact the quality of what I do.
Get some thermal compound under your nails for a change, and really experiment with real commercial products. Sometimes a little practical knowledge really helps to understand what is going on out there.
-Chris
Is it because you are using different devices than were originally used? usual case these days, unless you have reliable stock or supply.
I meant incoming inspection QA or sorting for hfe etc.
Show me one.
I see in
DMMT5551S (Matched Pairs)
| Diodes Incorporated
The DC Current Gain, hFE, (matched at IC = 10mA and VCE = 5V) Collector Emitter Saturation Voltage, VCE(SAT), and Base Emitter Saturation Voltage,
VBE(SAT) are matched with typical matched tolerances of 1% and maximum of 2%.
To me that implies if you measure/compare Vbe you are going to get a matching hfe as well.
Last edited:
Just looked at this matched pair http://www.sycelectronica.com.ar/semiconductores/LM394.pdf
The data shows one parameter for hfe matching and 3 for Vbe matching. So there both important depends on the aplication.
The data shows one parameter for hfe matching and 3 for Vbe matching. So there both important depends on the aplication.
If the hfe vs Ic curves are the same shape ( dosnt matter if one is higher than the other which it will be, this is what were matching) from one 2nxxx to the next 2nxxx ( they may not even need to be the same shape but just not cross each other ) which is very likely, than Vbe matching has to equal hfe matching. Because Ic is proportional to Vbe and Ib is also proportional to Vbe hfe must also relate to Vbe.
We dont need to know actual parameter values just which transistor has more. I may be wrong, please let me know, but I cant see how Vbe rating ( you rate 10 transistors before you match 2 ) as in highest to lowest, would not show you hfe rating if other factors like Ic are the same.
Another way to look at it. To measure hfe you meaure Ic/Ib. But Ib is Vbe/re ( the base emitter impedance ) so hfe is Ic*re/Vbe. If the re curves ( exponential ) of different samples dont cross each other ( the part im not sure of but intuitively makes sense ) than using Vbe or hfe for comparing samples is the same thing. The is that you have to have enoungh Ic flowing to make the measurement, you cant just measure the Vbe with the diode setting of your multimeter. So you still need a jig which makes this almost the same amount of work as measuring Ic/Ib. So why not measure both, one more meter.
We dont need to know actual parameter values just which transistor has more. I may be wrong, please let me know, but I cant see how Vbe rating ( you rate 10 transistors before you match 2 ) as in highest to lowest, would not show you hfe rating if other factors like Ic are the same.
Another way to look at it. To measure hfe you meaure Ic/Ib. But Ib is Vbe/re ( the base emitter impedance ) so hfe is Ic*re/Vbe. If the re curves ( exponential ) of different samples dont cross each other ( the part im not sure of but intuitively makes sense ) than using Vbe or hfe for comparing samples is the same thing. The is that you have to have enoungh Ic flowing to make the measurement, you cant just measure the Vbe with the diode setting of your multimeter. So you still need a jig which makes this almost the same amount of work as measuring Ic/Ib. So why not measure both, one more meter.
Hi Rick,
Look here : https://www.onsemi.com/pub/Collateral/NST65011MW6-D.PDF on page 2.
Your link shows a data sheet that is missing a lot of what you would need. I would recommend that you use On Semi for your parts in that case.
Hi cbdb,
Yes, in certain designs, the vbe parameter would supersede an Hfe match, although I suspect the Hfe would be close under those conditions (a match).
As far as price determined by how close a match is, that is found on datasheets sometimes as different grades of device. Certainly when you look at pricing for that general part. It's been that way for years now, but maybe they are dropping the grading aspect - to our detriment as we can't choose a very close pair when it matters. You would be back to buying a bunch of devices and sorting through them. Ever sort smt devices? It is a normal PITA X 100! I have smt adapters for my jig. Ever sneeze when you have parts sitting there graded? I have. Many can't even be found, never mind that the grading is now completely gone. It's funny until it happens to you. What unit do I do this for? Cyrus or other high quality equipment that is surface mount. It is a supreme pain, words can't do it justice. The biggest issue is to keep the leads in contact with the contacts reliably while not heating up the parts.
I'm not going to run around the internet looking at further examples. The pages shown with minimal information are second source suppliers, just an observation.
-Chris
Look here : https://www.onsemi.com/pub/Collateral/NST65011MW6-D.PDF on page 2.
Your link shows a data sheet that is missing a lot of what you would need. I would recommend that you use On Semi for your parts in that case.
Hi cbdb,
Yes, in certain designs, the vbe parameter would supersede an Hfe match, although I suspect the Hfe would be close under those conditions (a match).
As far as price determined by how close a match is, that is found on datasheets sometimes as different grades of device. Certainly when you look at pricing for that general part. It's been that way for years now, but maybe they are dropping the grading aspect - to our detriment as we can't choose a very close pair when it matters. You would be back to buying a bunch of devices and sorting through them. Ever sort smt devices? It is a normal PITA X 100! I have smt adapters for my jig. Ever sneeze when you have parts sitting there graded? I have. Many can't even be found, never mind that the grading is now completely gone. It's funny until it happens to you. What unit do I do this for? Cyrus or other high quality equipment that is surface mount. It is a supreme pain, words can't do it justice. The biggest issue is to keep the leads in contact with the contacts reliably while not heating up the parts.
I'm not going to run around the internet looking at further examples. The pages shown with minimal information are second source suppliers, just an observation.
-Chris
DMM Diode test constant current source is usually 1mA which is close to what you use for a LTP in a audio amp IPS.
I need to make a jig and do some more testing to be able to add anything more.
NST65011MW6
Current Gain Matching to 10%
• Base−Emitter Voltage Matched to 2 mV
That is what I am doing by hand, matching to 1mV and assuming that hfe follows suit.
I understand smt, I have soldered 0.5mm DFN's by hand. you are correct, you need good fixtures to test smt.
FJL4315/FJL4215 are hfe graded, MJL3281/MJL1302 are not.
I need to make a jig and do some more testing to be able to add anything more.
NST65011MW6
Current Gain Matching to 10%
• Base−Emitter Voltage Matched to 2 mV
That is what I am doing by hand, matching to 1mV and assuming that hfe follows suit.
I understand smt, I have soldered 0.5mm DFN's by hand. you are correct, you need good fixtures to test smt.
FJL4315/FJL4215 are hfe graded, MJL3281/MJL1302 are not.
Last edited:
If one transistor has Hfe=100 and the other has Hfe=200, then the effective Hfe transitions from 100 to 200 over the course of the crossover region. At low output it's not a discontinuity per se, but as output increases it looks more and more like a discontinuity.
An amp can be made so that THD doesn't respond to Hfe mismatch. But this just means some other distortion is dominant, the mismatch is just buried in some other distortion. Not that this is a bad approach, it's just another way to skin the cat. Although I would argue that an amp that takes full advantage of each transistor would both respond well to Hfe matching and also have lower distortion than an amp where the mismatch is masked by the distortion of another stage.
It depends on your objectives. What kinds of distortions do you think are worse, how many transistors are you willing to tolerate, do you prefer matching or do you prefer a more elaborate design. Is the design that results something that you would actually want to build or is the pin count so high that you run out of wits and solder?
I assume when someone asks for advice on Hfe mismatch that they have determined for themselves that matching will improve the amp. I think that's a reasonable platform to have this discussion on. Otherwise the scope of the discussion inflates precariously...
There are nasty non linearities in a class AB output stage.
In case of a 200 Hfe NPN with a 100 Hfe PNP, these figures are far from constant over the full swing in the B regions.
Then near cross over, where it works in the A region, you have gm doubling where the Hfe is some additive mix of the two transistors.
At cross over, you have base charge issues when turning off transitors.
Feedback will force linearity over these defects wathever they are, including all those no one is aware of.
You end up with residual distortion that depends of frequency, signal level and bias that is shaken by thermal effects..
Then it's more a matter of taste.
This opens a can of worm, where there is no point IMO to focus on one linearity independantly of the others.
In case of a 200 Hfe NPN with a 100 Hfe PNP, these figures are far from constant over the full swing in the B regions.
Then near cross over, where it works in the A region, you have gm doubling where the Hfe is some additive mix of the two transistors.
At cross over, you have base charge issues when turning off transitors.
Feedback will force linearity over these defects wathever they are, including all those no one is aware of.
You end up with residual distortion that depends of frequency, signal level and bias that is shaken by thermal effects..
Then it's more a matter of taste.
This opens a can of worm, where there is no point IMO to focus on one linearity independantly of the others.
Hfe is independent from Gm. The transistors are in parallel, so effective Hfe does not double (in series the Hfe multiplies, minus losses from Early effect). If both transistors have Hfe=100, then the effective current gain at the crossover point will be 100. You should have observed this in your own experiments at some point.
With modern transistors, the distortion contribution from base charge is actually not very large. Usually the rising THD with frequency is due to the preceeding stages and stability compensation.
With modern transistors, the distortion contribution from base charge is actually not very large. Usually the rising THD with frequency is due to the preceeding stages and stability compensation.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.