I know a lot has been written on the topic of parallel output devices.
But, after repairing over 100 power amps of varying makes in the past two years, i have started once again questioning the matching issue again.
I have been matching power BJT 's on a home made matcher and the process has been really cumbersome, not to mention expensive.
I was reading a post on FB where a guy fixed some hifi amp without matching, because he reasoned that current sharing between parallel devices was insured by the emitter resistors.
This makes sense to me and its something I want to try out. Thoughts?
I however still think mosfets used as output devices need to be matched because of the really large Vgs spread on these devices.
Can someone please concur my thoughts on this topic.
But, after repairing over 100 power amps of varying makes in the past two years, i have started once again questioning the matching issue again.
I have been matching power BJT 's on a home made matcher and the process has been really cumbersome, not to mention expensive.
I was reading a post on FB where a guy fixed some hifi amp without matching, because he reasoned that current sharing between parallel devices was insured by the emitter resistors.
This makes sense to me and its something I want to try out. Thoughts?
I however still think mosfets used as output devices need to be matched because of the really large Vgs spread on these devices.
Can someone please concur my thoughts on this topic.
Contrary to BJTs a mos fet's gain goes down as it warms up, so if you have say four in parallel then the one with the most gain warms up (in the unlikely event they are not from the same die), the gain goes down allowing the others to 'self balance', avoiding the need to match.
That is why there is an emitter resistor in most quasi and complimentary output stages.
Negative feedback takes care of any abnormalities in gain, as long as there is enough gain to drive the required load, that is all that matters.
That is why there is an emitter resistor in most quasi and complimentary output stages.
Negative feedback takes care of any abnormalities in gain, as long as there is enough gain to drive the required load, that is all that matters.
I agree with this.
( but that is just MHO)
In 40 years of building amplifiers I have never matched transistors whether they are bi-polar or mosfets. I haven't ever had an amp blow up.
I tripped over a speaker lead once and it yanked out shorting the amp but that just blew the output fuse and not the amp.
An over-built pro. amplifier can survive a lot of abuse, including poor current sharing. Still, it won't be able to deliver its best performance without some form of matching of the parallel devices, so you are talking about a crippled amp. and the OP question becomes; Is it OK to return a crippled amplifier to a customer and still call it repaired?
As far as I am aware, BJT amplifiers should nowadays be acceptable without matching, provided you buy your parts in the original manufacturer's tube or box packaging and fit them as such, so that the chips will be closely similar, likely being from the same zone of the wafer. In my experience though, neither vertical or lateral mosfets are near as well matched as they come. Anyway, You still need to check distortion is up to spec. under load before signing off on the job and this should clear up any doubts you may have.
I think that a repairer has a reputation to keep and it if it is a good one, you don't risk it and other people's gear by lowering standards. Sure, do a few cheap repairs at agreed cut rates and see how it works out but I suggest you don't suddenly drop a careful test routine before discussing this with important customers.
As far as I am aware, BJT amplifiers should nowadays be acceptable without matching, provided you buy your parts in the original manufacturer's tube or box packaging and fit them as such, so that the chips will be closely similar, likely being from the same zone of the wafer. In my experience though, neither vertical or lateral mosfets are near as well matched as they come. Anyway, You still need to check distortion is up to spec. under load before signing off on the job and this should clear up any doubts you may have.
I think that a repairer has a reputation to keep and it if it is a good one, you don't risk it and other people's gear by lowering standards. Sure, do a few cheap repairs at agreed cut rates and see how it works out but I suggest you don't suddenly drop a careful test routine before discussing this with important customers.
I have always used transistors matched as closely as possible. This gives a lower distortion than using unmatched pairs, even for amplifiers with high OLG. Generally the tighter Vbe of BJT's seems to allow parallel pairs where unmatched devices are used if the emitter balance resistors are high enough. Buying several at a time I find that most are matched, as others have said. But there could always be a device at the edge of a wafer that is different from the others... For parallel pairs or triples I still aim to match however, but for low distortion it is important, I think, to still try to match the two halves (NPN-PNP).
MOSFETs may have a reasonably tight spread if also purchased several at a time, but the wider Vgs spec suggests matching is necessary. Early literature suggested that parallelling MOSFETs was simple. In reality, relying on one to heat up and self-limit is asking for trouble. A source resistor value should be chosen on good engineering principles of evaluating typical Vgs and gm spreads, and choosing an optimum value to control the currents between devices.
MOSFETs may have a reasonably tight spread if also purchased several at a time, but the wider Vgs spec suggests matching is necessary. Early literature suggested that parallelling MOSFETs was simple. In reality, relying on one to heat up and self-limit is asking for trouble. A source resistor value should be chosen on good engineering principles of evaluating typical Vgs and gm spreads, and choosing an optimum value to control the currents between devices.
Regarding matching, Hfe alone is pretty useless: a more useful metric is the Vbe when the DUT is diode-connected (base and collector shorted).
That said, for BJT's, if they are of the same origin and have sufficient emitter resistors, matching is just icing on the cake.
Due to the fundamental exponential I vs. V characteristic of a semiconductor junction, junctions of he same area will always behave very similarly.
For Mosfets, things are very different: lateral types have a relatively low transconductance and a positive tempco and can benefit from matching but can tolerate the absence thereof, but switching types have a high transconductance, a wide spread of Vth and a negative tempco = a recipe for disaster if you don't match them.
This means that for popular, low-cost types like IRF240/9240, matching is absolutely essential
That said, for BJT's, if they are of the same origin and have sufficient emitter resistors, matching is just icing on the cake.
Due to the fundamental exponential I vs. V characteristic of a semiconductor junction, junctions of he same area will always behave very similarly.
For Mosfets, things are very different: lateral types have a relatively low transconductance and a positive tempco and can benefit from matching but can tolerate the absence thereof, but switching types have a high transconductance, a wide spread of Vth and a negative tempco = a recipe for disaster if you don't match them.
This means that for popular, low-cost types like IRF240/9240, matching is absolutely essential
With irfp240/9240 I have only just used 0r22 source resistors and not had any problems.
Just don't buy them off ebay.
The last one I bought lasted 10 minutes until I turned off my soldering and it blew.
So I bought some from RS Components and deliberately tried to blow it up plugging in and unplugging my soldering iron and couldnt blow it.
If given the choice I would go for IR.
It was "allegedly" a Vishay one that blew up.
Hi
In solid-state output stages if there is a blown - usually shorted - device, you have to do a brute force repair. This means replacing al the outputs, drivers and predrivers. The reason for this is that where you might have only one absolutely dead device by resistance checks or isolation, other devices may only show their stress dynamically.
BJTs are like super-heroes. An amp designed for 8R with a 2R load attached will say "I can drive that!": and proceed to blow up its outputs. The drivers will then say "I can drive that" and blow up as well, since the whole circuit is direct-coupled.
An amp with defective outputs can display a clean sine wave with no load (open load), but then show half the wave clamped with just a 1k load. No shorted devices but dynamically falilng ones present.
Doing the brute force repair puts you into the position all the other posters advise, where you use a set of devices all from the same rail, which are in their production sequence and likely close in performance to each other.
Getting an amp functional is what most repair guys are happy with. For PA or other industrial use this is likely good enough. However, for a hifi amp that has any pedigree, matching is preferred, although the same-rail-grouping will be pretty good. Open-loop for a symmetric EF output stage goes down by an order of magnitude when the devices are matched to within just 10%.
In solid-state output stages if there is a blown - usually shorted - device, you have to do a brute force repair. This means replacing al the outputs, drivers and predrivers. The reason for this is that where you might have only one absolutely dead device by resistance checks or isolation, other devices may only show their stress dynamically.
BJTs are like super-heroes. An amp designed for 8R with a 2R load attached will say "I can drive that!": and proceed to blow up its outputs. The drivers will then say "I can drive that" and blow up as well, since the whole circuit is direct-coupled.
An amp with defective outputs can display a clean sine wave with no load (open load), but then show half the wave clamped with just a 1k load. No shorted devices but dynamically falilng ones present.
Doing the brute force repair puts you into the position all the other posters advise, where you use a set of devices all from the same rail, which are in their production sequence and likely close in performance to each other.
Getting an amp functional is what most repair guys are happy with. For PA or other industrial use this is likely good enough. However, for a hifi amp that has any pedigree, matching is preferred, although the same-rail-grouping will be pretty good. Open-loop for a symmetric EF output stage goes down by an order of magnitude when the devices are matched to within just 10%.
When you use a single pair, unmatched P and N devices will not pose problems.
When you use multiple pairs, there will be sharing issues between same sex transistors, and that is problematic
Depends what you mean. As I mentioned a matched pair will give lower distortion. So unmatched pair will increase distortion. But a single pair in an amp does not have a current sharing problem.
Regarding matching, you can take this to extremes. Vbe matching is, I take it universally accepted as the most appropriate measurement to make. And simple hfe testing on a cheap DMM is uncontrolled in the sense that a base current is applied and the hfe read as a collector current, whereas hfe needs to be measured at the same collector current.
A very simple and effective test of matching is to connect the two bases and two emitters together, and measure the collector currents in each device.
And for input stages (OK this thread was about OP stages) hfe matching at a given collector-base voltage (probably not zero) is probably adequate for diff pairs.
So the matching method depends on the required application, or you could characterise each transistor with a gummel plot at several collector voltages ... which is probably unnecessary.
That's what I meant.
Intersex matching should in principle reduce even order harmonics, but P and N MOS are so dissimilar for technological reasons that it is a lost cause anyway
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.