MOS-FET matching

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Is it possible to match the source resistors instead of the output MOS-FETs when building an Aleph 3?

The resistors are typically +/-5%. If you in you test circuit test for at certain current (2A) and adjust the resistor until this is reach....? Starting with 1 ohms and paralleling those with small 10 ohms until current is equal....

Drawbacks?
 
Hi,
I asked this question many months ago, but it fell on deaf ears.

I matched and managed to get 8 quads out of my 16 pairs.
But one quad had one errant FET. I built it up with unmatched source resistors and sure enough the source currents appear to be different. But I suspect most of the voltage error is due to the resistor variation and not Vgs variation. I cannot hear a problem.

I have learned my lesson. I am currently matching the emitter resistors for the BJT amps that fill in the remaining gaps in the arsenel of bi-amped speakers for the 5.1 surround sound that is still under construction.

My recommendation;- match the source resistors and match the Vgs at the operating quiescent current.
Then measure the variation. Trimming, if you consider it necessary, can follow later but the maths is more complicated if you then take account of the different source resistance.

Good luck and let's hope someone enlightens us.
 
My idea is to match the FET and resistor as a pair at the actual application current.

Starting with ex. 0.68 ohm in a test circuit, and then adjust the resistor by parallel resistance until the needed current is there. The mount the complete set as it is....
 
Hi,
I think your proposal will work.
But how do you know what the source currents are after you have completed the project? All the source resistors will be different and all the different voltages may mislead you into thinking the source currents, as installed, are also different.
 
Hi

From my experience, a normal matching procedure (within 10mV and matched at the operating current and at another one to get a transcondutance) will raise no particular problems, even if the source resistors are not that matched.

The best and cheapest solution I've found to "match" the source resistors is to use paralleled small 0.6W metal film 1% resistors, to achieve the desired resistance and wattage. The resulting power resistors will be statistically matched. More soldering, but worth the effort ;) Not any problem in my AX that's been continuously on (or in standby @ Ibias/4) for two years now...

Regards
 
CheffDeGaar said:
The best and cheapest solution I've found to "match" the source resistors is to use paralleled small 0.6W metal film 1% resistors, to achieve the desired resistance and wattage. The resulting power resistors will be statistically matched.

And you will also get a lower inductance with this solution, which is good. I think it was Upupa Epops who said he always do so in his power amps to get a lower inductance.
 
Christer said:
I think it was Upupa Epops who said he always do so in his power amps to get a lower inductance.

Rendons donc à César ce qui est à César ;)

And, yes this is an appreciable collateral effect ;)

Skorpio :
As a side note, I may add that the matching in an Aleph is less critical than in an X one. Match the upper fets together, the lower ones together, use a lot of small resistors, and you're done ;)
 
AndrewT said:
Hi,
I think your proposal will work.
But how do you know what the source currents are after you have completed the project? All the source resistors will be different and all the different voltages may mislead you into thinking the source currents, as installed, are also different.

I will monitor the current in each leg with a 0.1% resistor temporaly mounted in series with the drain...
 
Hi Skorpio,

it is always good to keep things simple........your method doesn´t seem to be simple, will cost a lot of time and can be a source of mistakes.

The sollution with lots of 0,6watt 1% resistors is nice and simple and cheap (used it in my X). The matching in an Aleph is not that critical. Current per Fet can be of by 10-20% without any problem (unless you run them at 50watts each).

William
 
If I'm not mistaken, you'll have to build as many aleph current sources as the number of mosfets you have, i.e. add a bipolar transistor to monitor the current through each resistor. As you don't want to match the Id=f(Vgs), each mosfet will have a different Vgs for the wanted current, thus you can not drive them with a single BJT. And this will work for a constant current, but you can't be sure that the dynamic behaviour will be the same, and that the AC current gain will vary from one source to another.

And the same goes for the amplifying mosfets (negative rail). The input diff pair drives the gates in parallel. At idle, you can manage to have the same current through all the mosfets, but you can't be sure that variying the gate voltage will induce the same current variation in all the mosfets.

Putting it alltogether, the current drawn by the load is the difference between the current sourced by the upper current sources and the one sinked by the lower mosfets. If the mosfets are not matched, the variations of the sourced and sinked current won't be as linear and cummulative as in the matched case...

I usually match the mosfets at two different currents, thus having an idea of their transconductance. It's a bit anal, but it allowed me to verify that two mosfets having the same Vgs at a given current will quite likely have the same Vgs at another current (roughly the same transconductance). But for random mosfets, I've noticed that the discrepancies can be quite high. It's up to you.

Sure you can have a nice idle point adjusting the resistors. And a bias point doesn't make music...

But I use to be wrong ;)
 
15 odd years ago the story was that pretty boy Levinson had his girlfriend measure thousands of transistors to get perfect matched couples for his Cello undertaking.
To get really well matched BJT sets it is not sufficient to measure Hfe at one output current.
I gathered from Lumunauw's story that Manuel Huber from Swiss FM acoustics still does the same which makes them so costly despite inferior looking interiors.

The same applies to Mosfets, you need to measure them at least at 2 different current levels to get a decent match.
According to Mr Pass a static measurement is sufficient for the Aleph zone, at the nominal bias level.
At the price level of vertical mosfets the real cost is the extra time invested.
In the lateral mosfet years it was money, those devices then were 10 times more expensive than verticals are now.
Matching has never been so cheap for diy-ers.
(but maybe i am anal too)
 
Skorpio said:
Interesting viewpoint.....but in fact you are saying that a fet and a source resistor i combination does not work together dynamically?
One Mosfet/resistor combination will work, two combinations will also work, but two unmatched mosfets with their respective resistors won't behave the same way dynamically.

Skorpio said:
And a prober matching of currents will need more that the recomended Ugs static matching?
If the fets are Vgs matched, it's quite sufficient for an aleph. The differences will still exist, but in a lesser way than with unmatched ones...

O.T. : Jacco, did you receive my scans ?
 
Very O.T. :

Oui, Cheff.
And read the article more than once, very nice to see the Chef of NRDS do the review of the Pass Labs himself.
I had a little accident and did not come round to answering emails:
http://www.diyaudio.com/forums/showthread.php?postid=767673#post767673

Thank you very much for taking the trouble to scan and email the article.
Its nice to see NRDS do reviews of Pass gear again as they did on Thresholds before the magazine started focussing so much on tubes and surround systems.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.