How critical is Mosfet matching in power amp output stages?

TheBrownNote

New Member
I have one and a half B&K ST-140 amps, and I'd like to convert them to a dual mono amp that would be similar in design (but not identical) to the older versions of the B&K EX-442. One of the amps I have now is the regular stereo model which I've already done some parts replacements and modifications to and sounds good. The other was converted to a monoblock by B&K before I bought it, the seller said about 8 years ago. To do this they just removed one of the two amp boards, keeping the two Mosfet output devices for that one in place, and jumpered over to the pair on the remaining stereo board. Here's my question. I'm hoping that I can just keep the ones in the monoblock together as they are and keep the four drivers in the stereo one as they are too, do the same thing to the stereo amp in terms of removing one board and jumpering the orphaned output devices with the proper partner in the pair that stays on the other board, then put both sides in the monoblock configuration back into the stereo case with the two improved power supplies. I want to make sure that the output devices would not be mismatched, though, and I'm not sure if I can be safe in this assumption. I've put some nice parts in the stereo one I don't want to have wasted this effort and to have sound and reliability problems that result from mismatched output devices. I've heard that most manufacturers just use devices from the same batch (presumably there's some way to tell this now?) and are satisfied with that. I've also read recommendaions and descriptions of equipment where devices were matched to the millivolt level using a big batch of potential candidates, and that this makes all the difference. This was in the context of input stagee on preamps, though. How much precision is needed in matching output devices, specifically Mosfets, in power amp output stages? Thanks is advance for any advice.
 
Yep, short answers:

1) If using matched MOSFETs, no drain resistors in paralleled sets.

2) If not using matched/batched MOSFETs, use drain resistors to keep 'current hogging' at bay.

Cheers,

David
 
Greetings and welcome.
We've had a couple of very recent threads on the ex-442 the 202 etc. on DIY forum.

pisser is the hifiengine schema's are pretty poor from I recall reading. might do a 'audiokarma included query in gloogle / model number etc.

Thanks for pointing me to the older posts, Binkman! I looked at those threads and found them very informative, but they didn't have anything specifically about matching of output mosfets. I have found schematic images that are fairly legible to work off of and have verified that the monoblock version of the amp I have on hand has no differences in the circuit besides having another pair of output Mosfets strapped onto one side of the amp and the board for the other side removed. So I know that I can just mimic that for the other side and stuff it all into one case, but first I need to know how critical it is to check the match of Mosfet pairs.
 
Yep, short answers:

1) If using matched MOSFETs, no drain resistors in paralleled sets.

2) If not using matched/batched MOSFETs, use drain resistors to keep 'current hogging' at bay.

Cheers,

David

David, that makes sense, but is there any way to know a set is matched already (batch numbers?) or is it safe to assume that they're already matched when they're put into a push pull Class A/B amp of this vintage and mid-grade quality? Do VGS values drift over time in older power amps? I think you're probably going to tell me that the only way to know is to check and not to take anything for granted but had to follow up and get your opinion. If you think I should check, is the procedure detailed by Pass in this 93 article the best way to go about it? http://www.firstwatt.com/pdf/art_matching.pdf Thanks for your reply.
 
My experience is with mid-80s Hafler amps (the DH and XL series). All of those had output devices grouped into batches. You'll see a stamp on the devices indicating that grouping. I don't have any data on how the groups were derived, however.

If in doubt, a procedure much like the Pass Labs approach is certainly something to consider. Current production devices are _far_ more pre-matched than 30 years ago so it still remains that you need to understand something about the circuit topology and the devices in use before making decisions about the parameters that need matching.

Sorry, but the question of whether to match or not to match isn't that binary. There's more to it.

Cheers,

David
 
Hope it helps but search for matched in batched mfg. Look for 'paired' as usually sold as in pairs but some buy lots and sort 100's.
Kuddo's to all above for good info-tech.
 
My experience is with mid-80s Hafler amps (the DH and XL series). All of those had output devices grouped into batches. You'll see a stamp on the devices indicating that grouping. I don't have any data on how the groups were derived, however.

If in doubt, a procedure much like the Pass Labs approach is certainly something to consider. Current production devices are _far_ more pre-matched than 30 years ago so it still remains that you need to understand something about the circuit topology and the devices in use before making decisions about the parameters that need matching.

Sorry, but the question of whether to match or not to match isn't that binary. There's more to it.

Cheers,

David

You're right, I do need to understand the circuit a bit better, and I also need to mark the parts and detach them so I can eyeball them for markings. Thanks for your input, David.

Jon
 
Hope it helps but search for matched in batched mfg. Look for 'paired' as usually sold as in pairs but some buy lots and sort 100's.
Kuddo's to all above for good info-tech.

Thanks Binkman, if you have any specifics or resources to consult about the markings please let me know. These are Hitachi J50 and K135 pairs.
 
You're right, I do need to understand the circuit a bit better, and I also need to mark the parts and detach them so I can eyeball them for markings. Thanks for your input, David.

Jon
Jon,

I re-read the original post - I skimmed it earlier. If those were my amps in the condition they are now, I'd be seriously considering the purchase of new driver boards from Musical Concepts or Audio by Van Alstine. There are a couple of others, too. I'd be checking with them to confirm that the output devices you have would work and go with one that could do the job.

It sounds like you have the chassis and possibly power supply already in place. That plus the output devices get you well down the road.

Cheers,

David
 
I did look at both of those websites when I was starting to think about this project and decided to go my own way and maybe learn a bit. My initial project, that is; I already have an ST-140 that I've put some effort and parts into, and that's going to be the case that houses the dual mono EX-442 wannabe that I want to create. The boards in this existing amp have the electrolytic input caps on each board removed and replaced with humongous 47uf inline Mundorf EVO oil film caps. I also removed the two 100 uf rail caps from each side and replaced them with nothing (on the advice of a now retired pro who does amazing work--he modified/completely recreated both of my Acurus amps). The two large can 15000uf 75v commercial filter caps in the amps were replaced by skinnier but heavier and more powerul Mundorf MLytic 15000uf 100v caps. I have Jensen film caps to replace the electrolytic output caps that are currently there (after that there will be no lytic caps), but I haven't installed those yet, and I also bought some Kimber cable to clean up the wiring, but haven't done that yet. At this point I'm too far down the road to buy new boards, I think. Anyway, I've been listened to it with the existing modifications for a few months now and did A/B comparisons with my existing main amp, the modified Acurus A-250, using this Niles A-B switch I bought. The modified B&K compares favorably and definitely has some warmth and mid-high definition that is alluring, but it lacks generally lacks low end definition and the ability to deliver full detail at multiple frequencies when being challenged by high volume music that's precise and percussive. My next step was to buy a stock and reconditioned B&K ST-140 and a monoblock ST-140 converted by the factory to add to my existing project. I stumbled across these on EBay and I initially was interested just so I could compare the stock ST-140 to the one I had worked on; I was unsure what to do with the monoblock. The comparison of the stock ST-140 with the one I modified was gratifying. The stock version has in your face, front of the stage power but no finesse, along with the tizzy highs and flabby bottom denigrators of the stock version of this amp focused on. The modified amp had a much wider and deeper soundstage and much better imaging overall, sweet bass, and didn't get as fatigued when being worked hard. However, my modified ST-140 still couldn't keep up with the Acurus A-250 at times and this made me wonder: what about making the ST-140 monoblock and the one I had on hand already into something similar to an EX-442 and then doing the comparison with both my main amp and also the stock ST-140, which is still intact? This is where I am in the process at this point, and although I've heard great things about both of those companies and have been very impressed by Van Alstine's amps when I heard them at an audio show it's too late to go down that path now. Thanks again, David, I appreciate your interest.

Jon
 
Thanks for the contact, Binkman. Today I bought a cheapo transistor tester based on a thread on this site that was started by Echowars. So now I can test gate to source threshold voltage with a cheapo level of accuracy, and I think that will be it for mosfet matching (other than doing a/b listening comparisons). I might try contacting Alan, though, to see what he thinks about other changes I intend to make.
 
You have to test threshold voltages at operating temperature and current for an optimal match. The parameters vary greatly with temperature.

If you need matched laterals, Profusion PLC sells the Exicon Hitachi lateral clones pre-matched for a slight premium.
 
Anyone has the schematic, I can't find it in hifi engine.

Ha ha, I never designed a MOSFET amp, so I don't know how much I can help. But thank you Binkman of thinking about me.
 
I looked at the schematics of ST140 and EX442, they are pretty much the same with exception of a few value of the components are different. If you have picture of EX442, you can copy how they put in the extra transistors.

Noted on the EX442 schematic, R3, R4 are different, C2 and C8 are different. The power transistors are different. You need to change those. You can try doing that to convert the ST140 to EX442.

As for matching. I never match MOSFET before, but I match BJT which is even more critical. This is the circuit I use and I redraw in MOSFET. I don't want to say this is the best way as I never try before.

MOSFET matching.jpg

This shows circuit to match N and P channel transistors. The circuit draws about 80mA or so through the transistor. I use protoboard to do this as shown. You measure the source of the transistor RIGHT ON THE PIN and record it down. You need to get everything ready before turning on the power. Have one side of the meter clip onto the ground, have the probe in your hand touching the source pin of the transistor before turning on the power. The moment you turn on the power, read the voltage and record it down for that transistor. You measure all and then pick the ones that are close.

I use this type of protoboard. Notice the 0.1uF cap right close to the legs of the transistor? This is critical to prevent oscillation. The picture shows BJT, but the idea is exactly the same for MOSFET.

Matching circuit 1.JPG


Make sure you buy at least 6 to 7 transistor to find 4 matched ones. Double check the voltage to make sure it's consistent. It is important to have consistent time delay from you turn on the switch to you read the measurement. This is because when the transistor heat up, the voltage change. Don't hold the transistor in your hand too long to prevent from heating up the transistor by your hand. The voltage is temperature dependent.
 
Last edited:
This might not be a good idea for MOSFET as you have to handle the transistors, you have to strap yourself with a ground strap if you want to try this.

I do 60 at a time using this protoboard fixture. I need to match sets of 10. I am getting quite good at this, using only 0.6V on the collector to reduce the power dissipation and keep the transistor from getting warm, I get let than 3mV matching in my amp.

Vbe test1.JPG

What I did is measure and record down the Vbe ( Vgs in this case), then I sort sort and group them next to each other and remeasure to confirm. Then pick out the set. To increase the speed of measuring, I actually have one meter probe on the emitter ( source) of one of the transistor, then I use the other probe to measure the emitter of the others. So instead of getting the Vbe, I measure the DIFFERENTIAL voltage between the reference transistor against the others. This way, I have less to write and faster. Also more importantly, they all heat up at the same time so the temperature are matched, it is not as sensitive with timing as measuring individual transistors.

the biggest variable is the temperature. the Vbe/Vgs varies a few mV/deg C while you are trying to match to within a few mV.
 
Wow.. Do you measure the whole bank or each one? of sets etc.. one by one? Very interesting your approach.. probably mimics a factory.
 
I measure the whole bank and write them down in the note book. I put one probe on the emitter of one transistor and use that as reference so I can use the most sensitive scale on the meter to read. So the rest of the transistors are label + or - voltage from the reference transistor. Also when the transistor heat up, they roughly heat up the same time( they draw the same current). Then roughly sort them and group them in the fixture to verify again. I sort in groups of 10. After I verify, they I put them in bags. I have no idea how factory do theirs. It is not easy, if you do this, you'll find they are change quite a bit.

I found the most important improvement came when I reduce the collect voltage ( drain voltage) down to only 0.6V to minimize the heat produced by the transistor.
 
Back
Top Bottom