Nelson Pass said:
--I remember that you've mentioned having Harris and Fairchild on hand. Setting aside the use of IRF in followers, do you allocate specific MOSFET manufacturers' parts for particular products in your lineup?
--Yup, that's me. Guilty as charged.
--Well...yes, but no...you buy by the thousand, possibly even ten thousand, depending on the part. That gives you a larger pool to match parts from than someone who buys ten or twenty and has to take best available match rather than reach into a bin of .01V parts and pull out a fist full of closely related devices.
Part of the problem I've run into is that, even buying moderate quantities, the MOSFETs have at least two and often more series of stamped numbers. That kinda knocks the same-wafer, lower-distortion strategy into the weeds. I got lucky this past fall and received 100 devices in "only" two series .25V apart from one another. Hmmm. We takes what we can get.
Why things have gotten this way, I cannot say. I used to see a single series if I ordered twenty parts. No longer. From where I sit, it seems that the problem began about a year, maybe eighteen months ago. (Certain people...ahem, ahem...certain people have had their parts since the Jurassic and might not have this problem. [Okay, okay...so your Fairchild parts date from the Pleistocene...picky, picky, picky...])
So what to do? The obvious thing is to start matching at more than one current. That way, I no longer have to worry about wafers. If the boogers match, they match and I'm good to go. Yes, that will take more effort and more time and I confess that I haven't started yet--I'm just barely keeping my head above water time-wise at the moment. For the time being, I've still got some decent looking Vgs matches from the same batches and I'm getting by on those. The next time around, I'm going to at least two point matching and possibly more, because I'm a little frustrated with this broken series problem and (given my current approach of using as little feedback as possible) it'll give me lower distortion, anyway.
Using fewer devices gives me the chance to use devices that match closely, thus lowering distortion. To use more devices would necessitate loosening the matching criterion, raising the distortion.
Maybe others are having better luck with their parts buying than I am. I sure hope so.
Grey
I took a moment and looked at Digikey and Mouser. Both report having the 2SC4793. Digikey has the 2SA1837; Mouser has it on order, but with no delivery date specified. A few other places had them as well but were more expensive. I'm in no position to buy anything at the moment and won't be for quite a while--just thought it would be interesting to see of they were available without too much hassle. They are.
Grey
Grey
Ian Macmillan said:
Perhaps you are.
There is no bootstrapping in the PL products. If I want a higher
load impedance for the UGS I suppose I would either bootstrap
the resistors which source current to the bias network or use
CCS's, but I don't want the higher load impedance - my resistors
are chosen for double duty.
GRollins said:
I use either Harris or Fairchild for the P channel Source applications,
I have no doubt that there are others equally qualified with
comparable parts. I have tried Samsung in the past with good
results and expect that Toshiba parts would also be fine.
GRollins said:
It's a fairly ordinary transistor - you can sub comparable bipolars.
gl said:
There is a slight tendency for hand wringing and sleepless nights
over matching Mosfets. The primary concern is for equal current
sharing, and once the Vgs is reasonably matched at operating
temp, you depend on the Source resistors to enforce the
sharing at higher currents. I use same lot codes and tight Vgs
tolerances because I can do so easily. This should not prevent
you guys from going ahead with different lot codes if that's what
you've got.
Mr.Pass, is there a prefered voltage type for the front end MOSFETs?
I'm searching for an alternative to the hard to find fairchild 200V parts in Germany. :-( Because the gain devices are common source the easy to get IR devices will not be my first choice.
Also I would like to ask, if you can give me a hint again what is the purpose of the 3W resistor between relay and meter-led-switch connector and of the 2,2kOhm / 3W type? Perhaps both for LED or coil circuitry?
Dirk
I'm searching for an alternative to the hard to find fairchild 200V parts in Germany. :-( Because the gain devices are common source the easy to get IR devices will not be my first choice.
Also I would like to ask, if you can give me a hint again what is the purpose of the 3W resistor between relay and meter-led-switch connector and of the 2,2kOhm / 3W type? Perhaps both for LED or coil circuitry?
Dirk
gl said:
Yes.
Let me pound on this one again in case anyone's confused.
1) MOFETs are poorly controlled when it comes to Vgs. It's the nature of the beast. Bipolars vary, JFETs vary, MOSFETs vary...if you want something tightly matched, you'll need to look at tubes--they're much more uniform in their characteristics than any of the solid state competition.
2) In the real world, this means that if you supply two MOSFETs with the same voltage at their Gates, one will turn on harder than the other if they're not perfectly matched.
3) This leads to one device doing more work than the other (aka "current hogging").
4) Nelson has long suggested that--when possible--DIY folks use a simple Vgs match on MOSFETs manufactured at the same time, on the same wafer of silicon. There is usually (not always!) a tiny embossed number on the body of the device. This tells the tale.
5) If you can get devices manufactured at the same time, they will generally have similar characteristics.
6) What characteristics? Well, this is where things begin to get annoying. Suppose you were to supply two MOSFETs (matched for Vsg, but from different lots) with a voltage at their Gates. If you use the voltage that gives you the current you matched for, then the devices will both give you that same current. Marvelous! Just what we wanted! Er...not so fast, bub...what happens if you change that voltage at the Gate to something higher or lower? Yikes! The bloody things don't act the same--one will give higher current than the other. But I thought they were matched, you say. Well, yes they were...at the one current you chose to match them at. This does not necessarily mean that they're matched at any other current.
(We will now pause until the rioting in the streets calms down.)
7) This is why Nelson suggests using devices from the same batch. There is a high probability that the devices from the same batch will react similarly to an increase or decrease in voltage at the Gate.
8) But what do you do if you only have devices from different batches? First off...DON'T PANIC...relax, have a home brew...it's going to be okay. You simply match the devices in question at more than one current. The more data points you have, the more accurate the match. If you want to get really into it, you can drop some money and buy or build a curve tracer.
9) If you really want to go nuts, you can match for noise or most any other parameter you might be able to think of. I used to know a couple of guys who decided to go into business building head amps. They bought an entire batch of Siliconix parts and matched them for six different parameters. Don't worry, nobody's going to ask you to do that. For the vast majority of DIY uses, just nailing the Vgs for a couple of currents will do the trick.
Grey
I didn't mean you, specifically, but there are any number of other people reading this thread. As Nelson noted above, there's a lot of angst out there about part matching. Note the number of threads started because someone built amplifier X, Y, or Z, measured the voltage across each Source resistor...then had a mild case of hysteria because the voltage wasn't equal across all the resistors. "But I matched all the MOSFETs to within .00001V!" they cry. "What have I done wrong?"
Actually...nothing. The amp is fine. Sit down and listen and quit worrying. If you really want to get serious about evening up that discrepancy, then match you devices at the idle current you intend to use instead of at 10mA. If you really, really want to go whole hawg, match at more than one current.
Since I've gone and opened this particular can of worms, I want to make sure that no one comes back three months from now, having taken something out of context, and wants to know why I said that it's okay to run MOSFETs from different batches when their amp smoked the first night because one device was doing all the work. (I get e-mail from people who seem to have some pretty free-form interpretations of things I've said.) I hope to forestall some of that by flogging the idea that you've got to do a little more work during the matching stage of the game if you want to make it come together properly.
I'll still get the e-mails, just hopefully not as many.
Grey
Actually...nothing. The amp is fine. Sit down and listen and quit worrying. If you really want to get serious about evening up that discrepancy, then match you devices at the idle current you intend to use instead of at 10mA. If you really, really want to go whole hawg, match at more than one current.
Since I've gone and opened this particular can of worms, I want to make sure that no one comes back three months from now, having taken something out of context, and wants to know why I said that it's okay to run MOSFETs from different batches when their amp smoked the first night because one device was doing all the work. (I get e-mail from people who seem to have some pretty free-form interpretations of things I've said.) I hope to forestall some of that by flogging the idea that you've got to do a little more work during the matching stage of the game if you want to make it come together properly.
I'll still get the e-mails, just hopefully not as many.
Grey
Nelson Pass said:
Many thanks for the answer, I will stop pursuing bootstrapping as a way to achieve "... approximately 20% greater power output for a given supply voltage" (taken from the XA.5 literature). I couldn't really see how bootstrapping was going to do it for me anyway but then I dont' (yet) see another way either. Time to get the thinking cap on
Ian.
(i know this thread isn't about matching...) but,
when i built my Aleph 2, i matched via VGS to get similar fets, then i
observed the fets chosen in the working amp. Taking VGS readings
again while all warmed up and running in the circuit, i swapped out
until i got good reading all around... within .01 or something like that.
(the Aleph 2 still sounds good today)
when i built my Aleph 2, i matched via VGS to get similar fets, then i
observed the fets chosen in the working amp. Taking VGS readings
again while all warmed up and running in the circuit, i swapped out
until i got good reading all around... within .01 or something like that.
(the Aleph 2 still sounds good today)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.