Okay, good to know. I though as much, but it's good to see that confirmed.
I'm preparing to build the "Honey Badger" power amp. I'm willing to take my time and do my research and I'm learning new things every day.
Going with what you said in the post above, would I be perfectly safe buying the transistors needed and not paying attention to matching? Does the above also apply to the output devices?
The Honey Badger has a trim-pot allowing to set the DC-offset. Is this such a measure to compensate for small differences in individual semiconductors?
[EDIT]: Please correct me when I'm wrong, but I assume you will need to get "ballpark" matching done. Meaning that two transistors in a pair have beta's more or less close to eachother. So, for example, one beta of 800 and one of 900 will be better than 100 and 1500.
Also I've read somewhere on this forum that manual selection and matching can make an amp progress from "good" to "excellent". What would be needed to achieve that, with easy to obtain tools and methods? Is it at all necessary, or another example of the law of diminishing returns?
Last edited:
Where devices are used in parallel (to increase the current capability), some parameters must be matched. Otherwise current hogging and overheating can occur.
hFE is just one parameter.
For an input pair using an Long Pair Tail (LTP) arrangement, the two inputs are compared and the output is a function of the difference.
that comparison of inputs relies on the two devices correcting reading the two input signals. That requires at least TWO parameters to be closely matched, hFE and Vbe.
It also requires the two nearly matched devices to be operated in nearly matched power/voltage/temperature/current. You the Builder has to ensure you do your best to meet ALL these operating requirements to allow the "difference" output signal to portray the difference between the TWO Inputs.
hFE is just one parameter.
For an input pair using an Long Pair Tail (LTP) arrangement, the two inputs are compared and the output is a function of the difference.
that comparison of inputs relies on the two devices correcting reading the two input signals. That requires at least TWO parameters to be closely matched, hFE and Vbe.
It also requires the two nearly matched devices to be operated in nearly matched power/voltage/temperature/current. You the Builder has to ensure you do your best to meet ALL these operating requirements to allow the "difference" output signal to portray the difference between the TWO Inputs.
Have a look at this: http://www.diyaudio.com/forums/equipment-tools/231840-bipolar-transistor-matcher.html
Mine is shown in post #10, and it works fine.
Mine is shown in post #10, and it works fine.
Yaiks! You make building circuits sound like rocket science!
When I closely follow the supplied build guide and make sure my set up and measurements are okay, I should be fine. Right? This is, ofcourse, assuming my components are good and on-spec. In this case, will matching only make for smaller tolerances? Sort of like the icing on the cake?
It is science.
If you have two supposedly identical devices measuring two inputs and generating a "difference signal", then the quality of the difference signal will depend on the operating conditions of the TWO MEASURING INSTRUMENTS.
That is fundamental to getting good performance from the amplifier.
It's not just about selecting good matching components and it's not just about selecting a good schematic.
It's about how you build it.
A good "Build Guide" will get you 98% there.
But it needs some skill in reading and understanding and assembling to fill in the last 2%.
The Bonzai guide and others are very good. But the Builder does need to think.
If you have two supposedly identical devices measuring two inputs and generating a "difference signal", then the quality of the difference signal will depend on the operating conditions of the TWO MEASURING INSTRUMENTS.
That is fundamental to getting good performance from the amplifier.
It's not just about selecting good matching components and it's not just about selecting a good schematic.
It's about how you build it.
A good "Build Guide" will get you 98% there.
But it needs some skill in reading and understanding and assembling to fill in the last 2%.
The Bonzai guide and others are very good. But the Builder does need to think.
Last edited:
This makes perfect sense to me.
Applying this to the Honey Badger: hypothetically, having some transistors (the parameters of which I have no influence on), I can fine tune their performance (absolute and relative to each other), by tweaking temperature, resistors and, with that, power supply voltage. Something like that?
Of course, this is hair splitting taken to the extreme. I'm only trying to grasp what is needed to make a good amp better. Or, if possible, excellent.
I have the time and the patience. I just really like to do this right. If I don't need any of this extra work to make a very capable amp, please let me know. Remember I'm not very knowledgeable and experienced. So all this extra work may be beyond my skills. This could be something you guys can judge better.
I want to stay with the LTP, particularly since this is inside most Power Amplifiers and since it is regularly abused by the Designer and by the PCB layout designer and by the Builder.
The two halves of the LTP are the difference amplifier.
The biggest influence of all the variations is Tj.
If Tj for the two halves is identical, there is a very good chance that the output signal will be good.
What does one need for the Tj to be identical?
The starting temperature, Ta, needs to be the same. One transistor facing towards a hot component will ruin this.
The power dissipated in the two needs to be the same so that deltaT is the same from that common start point (Tj=Ta+deltaT). This requires the product sum of {Ic*Vce plus Ib*Vbe} to be the same. Ib*Vbe is usually very much smaller and can safely be ignored in first stage matching.
If Ie and Vce of both devices are the same, then both will dissipate the same power and thus deltaT will be the same.
What does one need to make Ie equal and to make Vce equal.
a.) one needs to be able to measure Vce
b.) one needs to be able to measure Ie.
c.) it is preferable to have built in matched resistors in both sides of the LTP to enable a. & b.
One could have the same Pj with different Vce and different Ie.
Now consider that implication:
One could adjust Ic a little bit lower to take account of a slightly higher Vce, so that the Pj & deltaT now become equal. What happens to the other transistor parameters, when Ic is deliberately made different?
hFE may change, Cob may change, and many other parameters rely on the identical devices to operate at the same conditions.
An example of a schematic error is the omission of one collector resistor (Rc) in the feedback side of the LTP. Technically it is not required for correct operation of the diff amp. But that directly changes Vce. That changed Vce brings with it changes in all the other parameters. The diff amp no longer gives good and accurate output difference signal.
Have a read of D.Self and what a 1% change in Ic does to diff amp performance, before feedback.
If the VAS pulls a current, or extracts a current from ONE side of the LTP, what does that do to Tj?
This is not found in a build guide. This is background information that the Builder will use in assessing the PCB layout and the schematic.
It is relatively easy to Thermally Couple a pair of LTP transistor, if the PCB layout allows such.
It is often possible to insert a pair of SMD low value resistors into the Re position where they are omitted in the layout.
It is possible to insert the missing Rc.
Balancing unequal draw due to VAS current is not easy. SymaSym is one sch that allows it. But at a cost, two LTPs to match and an extra transistor, or four.
Oh
and read post49 link.
Leach's papers on the Lo Tim and Double Barreled should be compulsory reading for us SS Members.
The two halves of the LTP are the difference amplifier.
The biggest influence of all the variations is Tj.
If Tj for the two halves is identical, there is a very good chance that the output signal will be good.
What does one need for the Tj to be identical?
The starting temperature, Ta, needs to be the same. One transistor facing towards a hot component will ruin this.
The power dissipated in the two needs to be the same so that deltaT is the same from that common start point (Tj=Ta+deltaT). This requires the product sum of {Ic*Vce plus Ib*Vbe} to be the same. Ib*Vbe is usually very much smaller and can safely be ignored in first stage matching.
If Ie and Vce of both devices are the same, then both will dissipate the same power and thus deltaT will be the same.
What does one need to make Ie equal and to make Vce equal.
a.) one needs to be able to measure Vce
b.) one needs to be able to measure Ie.
c.) it is preferable to have built in matched resistors in both sides of the LTP to enable a. & b.
One could have the same Pj with different Vce and different Ie.
Now consider that implication:
One could adjust Ic a little bit lower to take account of a slightly higher Vce, so that the Pj & deltaT now become equal. What happens to the other transistor parameters, when Ic is deliberately made different?
hFE may change, Cob may change, and many other parameters rely on the identical devices to operate at the same conditions.
An example of a schematic error is the omission of one collector resistor (Rc) in the feedback side of the LTP. Technically it is not required for correct operation of the diff amp. But that directly changes Vce. That changed Vce brings with it changes in all the other parameters. The diff amp no longer gives good and accurate output difference signal.
Have a read of D.Self and what a 1% change in Ic does to diff amp performance, before feedback.
If the VAS pulls a current, or extracts a current from ONE side of the LTP, what does that do to Tj?
This is not found in a build guide. This is background information that the Builder will use in assessing the PCB layout and the schematic.
It is relatively easy to Thermally Couple a pair of LTP transistor, if the PCB layout allows such.
It is often possible to insert a pair of SMD low value resistors into the Re position where they are omitted in the layout.
It is possible to insert the missing Rc.
Balancing unequal draw due to VAS current is not easy. SymaSym is one sch that allows it. But at a cost, two LTPs to match and an extra transistor, or four.
Oh
and read post49 link.
Leach's papers on the Lo Tim and Double Barreled should be compulsory reading for us SS Members.
Last edited:
Matching Hfe for predrivers/ drivers/ output devices is not necessary.
According to D. Self a VAS buffer is a solution as is using an emitter follower enhanced VAS to reduce the impedance at the VAS output which increases the local feedback around the VAS.
So aside from emulating OpAmp circuitry, complimentary device Hfe matching seems mostly as a waste of resources in most cases. I have not run across many Amplifier designs where matching is used or discussed ?
Most VAS run into a constant current load.
That constant current is fed from the CCS (or bootstrapped resistor pair) through the Vbe multiplier to the VAS device.
From either side of the Vbe multiplier we connect the output stage.
A BJT output stage draws current from the VAS constant current line and re-injects back into the VAS constant current line.
With no signal we can measure the quiescent currents through the CCS, through the VAS, through the Vbe multiplier transistor and through the multiplier resistor string.
Now vary the signal.
What happens next?
Now change the hFE of ONE driver.
What effect does that have on the constant current loaded VAS?
That constant current is fed from the CCS (or bootstrapped resistor pair) through the Vbe multiplier to the VAS device.
From either side of the Vbe multiplier we connect the output stage.
A BJT output stage draws current from the VAS constant current line and re-injects back into the VAS constant current line.
With no signal we can measure the quiescent currents through the CCS, through the VAS, through the Vbe multiplier transistor and through the multiplier resistor string.
Now vary the signal.
What happens next?
Now change the hFE of ONE driver.
What effect does that have on the constant current loaded VAS?
Ok, so I broke down and built the first circuit to test NPNs:
http://www.musicfromouterspace.com/analogsynth_new/TRANSISTORMATCHER/TRANSISTORMATCHER.html
but the last two digits of the 5 1/2 digit HP meter drift slowly up and down. I don't know what is the frequency but the period is in minutes.
I caught what seems to be a low of .61722 mV and a high might be .61788.
Is that normal?
http://www.musicfromouterspace.com/analogsynth_new/TRANSISTORMATCHER/TRANSISTORMATCHER.html
but the last two digits of the 5 1/2 digit HP meter drift slowly up and down. I don't know what is the frequency but the period is in minutes.
I caught what seems to be a low of .61722 mV and a high might be .61788.
Is that normal?
You're right about all those things, but what I was getting at was this "rough batching" is still an improvement over the original circuit in the OP, which only tells when one transistor has more gain than the other, or whether they "match" within some unknown percentage.
Measuring a number of transistors, N PNP and N NPN, requires 2N measurements with the DMM, and these numbers can then be sorted and used to pair up transistors to whatever approximation the DMM and careful (temperature-awate) handling of the parts gives. The circuit that only compares one transistor against another only gives binary info (okay, trinary, if they match within some window), and one must compare every NPN against every PNP, requiring N squared comparisons.
I wasn't sure "how far" the OP wanted to go in matching transistors, but the OP has obviously done some good research and found the Holy Grail of transistor measurements, and is now ready to make a ladder filter* with a handful of inexpensive transistors:
That's really neat - I've seen the schematic of Moog's famous ladder filter, which is the apparent use of these transistors, but hadn't thought of how well the transistors need to be matched.
I'm guessing this is slight temperature variation - perhaps from a heater or A/C in the room turning on and off?
Put your finger on the transistor for about a second, and see if that makes the voltage go down near-instantly with the heat of your finger, and then over several seconds go back to what it was.
You might make two of these jigs, and put the two sockets where two (preferrably matched) transistors glued together can go into both sockets. I have little doubt the voltages will go up and down in sync.
There's always the possibility that the base is picking up some local radio signals and rectifying them, changing the voltage slightly. Putting everything in an electrical shield would prevent that, but making everything even more complicated.
* Patent US3475623 - Electronic high-pass and low-pass filters employing the base to emitter ... - Google Patents
Yes.
you have 0.617Vbe±tolerance, the fourth digit (4 significant figures) is rarely needed.
Now you can measure all your transistors, then ask yourself "do the Tj for each test actually match?"
Room temperature variations, handling variations, time to take and record the measurement varies, all these make a nonsense of the fifth digit and have a very significant effect on the 3rd and 4th digits.
Benb has given a few good pointers in his reply.
He rubbishes the idea of REFtoDUT, because it adds to the work required. In that respect he is right. I measured Idss for 600 jFETS, that took a couple of hours. I then matched the jFETs to find good pairs that took a couple of weeks. And I did not charge for my time unlike some others who offer services on this Forum.
But if you want two devices that need to be matched then they also need to operate in the same conditions for the matching to be of any value.
Thermally coupling the devices and matching (Ref : Dut) in an LTP does keep the devices in the same operating environment and with care the jig then ensures that matched pairs do have the same Ie & Vce and thus the same Tj while the measurements are being obtained.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.