Trying for the same lot codes in matching the parts
gives the effect that when the two Vgs values are
the same, the rest of the characteristics tend to also
be matched. Lot codes reflect the variations across
silicon wafers, and when pulling these out of the tubes
we see repetitive patterns generated by the row/column
positioning of the die on the wafer.
There is, however, considerable variation of Vgs across
a wafer, and the closer Vgs figures appear to come from
adjacent parts.
gives the effect that when the two Vgs values are
the same, the rest of the characteristics tend to also
be matched. Lot codes reflect the variations across
silicon wafers, and when pulling these out of the tubes
we see repetitive patterns generated by the row/column
positioning of the die on the wafer.
There is, however, considerable variation of Vgs across
a wafer, and the closer Vgs figures appear to come from
adjacent parts.
Grey,
Your lust for precision is admirable, but goes well beyond what is required for +/- .5% matching. Measuring to two decimal places is a lot easier, and the numbers don't fly by nearly so fast.
I think +/- .5% is a good target, not just because it's pretty easy to hit, but also in view of the overall "error budget" I'm typcally working with. I don't know about yours, but my source resistors are only +/- 5% tolerance (Panasonic metal oxide), so at .5% the MOSFETs are already ten times better matched than the resistors.
As for university work, I did a three-year stint earlier on at another fine Southern institution, and can relate all too well. After that I looked forward to working in the business/industrial sector as I was sure it would brook no such nonsense. BWAHAHA! At least the pay is better.
Your lust for precision is admirable, but goes well beyond what is required for +/- .5% matching. Measuring to two decimal places is a lot easier, and the numbers don't fly by nearly so fast.
I think +/- .5% is a good target, not just because it's pretty easy to hit, but also in view of the overall "error budget" I'm typcally working with. I don't know about yours, but my source resistors are only +/- 5% tolerance (Panasonic metal oxide), so at .5% the MOSFETs are already ten times better matched than the resistors.
As for university work, I did a three-year stint earlier on at another fine Southern institution, and can relate all too well. After that I looked forward to working in the business/industrial sector as I was sure it would brook no such nonsense. BWAHAHA! At least the pay is better.
A girl I knew used to get highly variable results when she cooked. I suggested that, instead of putting in 'some' of each ingredient, she consider the use of measuring spoons, etc. in the manner suggested by the recipes she was using.
She looked at me as though I'd lost my mind. "But look at this. These two eggs aren't the same size. You can't control the the size of eggs. So what does it matter?"
My reply was that, yes, eggs vary, but that's no reason to introduce any more variables than are necessary. She never got my point...and the last I knew of her, she still wasn't a very good cook.
I used the 5% Panasonic resistors for the source resistors. Just for fun I measured a few. They were clustered more tightly than the 5% would suggest. Furthermore, having the MOSFETs as tightly grouped as possible allows me the option of putting in 1% resistors (or whatever) in the future without feeling,"Aw, what's the use? The MOSFETs are all over the map, anyway..."
Grey
She looked at me as though I'd lost my mind. "But look at this. These two eggs aren't the same size. You can't control the the size of eggs. So what does it matter?"
My reply was that, yes, eggs vary, but that's no reason to introduce any more variables than are necessary. She never got my point...and the last I knew of her, she still wasn't a very good cook.
I used the 5% Panasonic resistors for the source resistors. Just for fun I measured a few. They were clustered more tightly than the 5% would suggest. Furthermore, having the MOSFETs as tightly grouped as possible allows me the option of putting in 1% resistors (or whatever) in the future without feeling,"Aw, what's the use? The MOSFETs are all over the map, anyway..."
Grey
I've sent some queries now about possible deals on IRF240/9240. No answers yet. Do they laugh at me when I want only 50+50? Partco has IRF240 at 38FIM apiece, and SHP (a Finnish supplier for electronics professionals) has IRF9240 at 34FIM apiece. Divide by 6 to get euros.
My other option is to use Magnatec BUZ900/905 parts. How are lateral better or worse than hexfet?
For obtaining BUZ900/905, Partco is far too expensive. The Magnatec price from UK equals 38FIM if I take 100 pieces, Partco has it 88FIM _and_ VAT of 22% apiece. I don't know if Magnatec has VAT included or not.
I don't know if I will have the same lot code in my parts, since nobody has answered yet. (Oops, I have also forgotten to ask Partco about IRF, but I'll do that soon)
Then, if anyone has any information on Magnatec double die versions BUZ900D and BUZ905D. What I'm most afraid of is that one double-die fet would equal two randomly picked single-die equivalents. If the dies have practically identical parameters, then I just need to find out if I will get more power for the buck with double or single dies.
I just learned that there are two types of mos-fets, lateral and vertical, and that the former are better for linear and latter for switching applications. Now IGBT is not really a MOS-FET, but IRF240/9240 are IGBT? Yet called hexfet too. Why? My guess is that hexfet is a trademark for some IGBTs, but someone please set this straight.
Gotta go now, more later
-Kimmo Sundqvist
My other option is to use Magnatec BUZ900/905 parts. How are lateral better or worse than hexfet?
For obtaining BUZ900/905, Partco is far too expensive. The Magnatec price from UK equals 38FIM if I take 100 pieces, Partco has it 88FIM _and_ VAT of 22% apiece. I don't know if Magnatec has VAT included or not.
I don't know if I will have the same lot code in my parts, since nobody has answered yet. (Oops, I have also forgotten to ask Partco about IRF, but I'll do that soon)
Then, if anyone has any information on Magnatec double die versions BUZ900D and BUZ905D. What I'm most afraid of is that one double-die fet would equal two randomly picked single-die equivalents. If the dies have practically identical parameters, then I just need to find out if I will get more power for the buck with double or single dies.
I just learned that there are two types of mos-fets, lateral and vertical, and that the former are better for linear and latter for switching applications. Now IGBT is not really a MOS-FET, but IRF240/9240 are IGBT? Yet called hexfet too. Why? My guess is that hexfet is a trademark for some IGBTs, but someone please set this straight.
Gotta go now, more later
-Kimmo Sundqvist
In Hungary I sourced IRF240/9240 50+50 about 1.5 Euro/1.3 USD a piece from Future Electronics and from Spoerle. Deliveries were reasonably quick via curier.
Lot code was also ok.
I would go for devices from International Rectifier. I have another favorite, IRF044N if it is a single ended amp (but up to 55Vds only...).
Lot code was also ok.
I would go for devices from International Rectifier. I have another favorite, IRF044N if it is a single ended amp (but up to 55Vds only...).
Pwer Mosfet
The lateral Mosfets have lower transconductance (change in Drain to Souce current for a given change in the Gate to Source voltage). The input capacitance is usually much lower than the vertical type such as a Hexfet. An IGBT is a biplor with mosfet input, sort of like a darlington only with a mosfet input. They have been several amps using them at the output. I think the most popular laterals for audio are from Hitachi, Exicon, Tosiba, and Magnatec. For more info:
http://perso.wanadoo.fr/jm.plantefeve/comp.html
http://www.borbelyaudio.com/Classa.pdf
H.H.
The lateral Mosfets have lower transconductance (change in Drain to Souce current for a given change in the Gate to Source voltage). The input capacitance is usually much lower than the vertical type such as a Hexfet. An IGBT is a biplor with mosfet input, sort of like a darlington only with a mosfet input. They have been several amps using them at the output. I think the most popular laterals for audio are from Hitachi, Exicon, Tosiba, and Magnatec. For more info:
http://perso.wanadoo.fr/jm.plantefeve/comp.html
http://www.borbelyaudio.com/Classa.pdf
H.H.
Well then... should I be worried about the length of wiring from the front end to the final stage? In my original plan I'd have to use leads as long as 8 to 10 inches. Would that cause a problem?
I have also settled on a solution to accept the offer on Fairchild IRFP240/9240, unless something extremely surprising turns up.
Can I just assume that the lot size of Fairchild IRFP240/9240 is 30, or is there something to gain in requesting that multiple tubes be from the same lot?
Then, where did I saw someone, I am not sure was it NP or not, comparing lateral mosfets and hexfets and exploring the transconductance issue in detail?
-Kimmo Sundqvist
I have also settled on a solution to accept the offer on Fairchild IRFP240/9240, unless something extremely surprising turns up.
Can I just assume that the lot size of Fairchild IRFP240/9240 is 30, or is there something to gain in requesting that multiple tubes be from the same lot?
Then, where did I saw someone, I am not sure was it NP or not, comparing lateral mosfets and hexfets and exploring the transconductance issue in detail?
-Kimmo Sundqvist
About transformers....
I will idle both channels at 100 or 120W. Do you think one 545VA transformer could take that and live long?
Also, if I put a 2x18V smaller transformer to the end of it, will that make enough voltage for the front end? Main transformer will have 2x24V secondaries and I will end up with +/-32.4V supplies. That 18V would give +/-49V supplies for the front end. Enough or too much, or what would you recommend?
-Kimmo Sundqvist
I will idle both channels at 100 or 120W. Do you think one 545VA transformer could take that and live long?
Also, if I put a 2x18V smaller transformer to the end of it, will that make enough voltage for the front end? Main transformer will have 2x24V secondaries and I will end up with +/-32.4V supplies. That 18V would give +/-49V supplies for the front end. Enough or too much, or what would you recommend?
-Kimmo Sundqvist
Musher said:
Dont do that.
If your going to use separate transformers for the front end then make sure they have the correct voltage rating to work on their own ... you shouldnt have the secondaries of seperate transformers connected in any way as it will almost certainly cause a number of problems. Generally speaking, the front end is usually given rails that are approximately 15V higher than the output stage. Also, you should probably expect that output stage rail voltage to sag somewhat under load.
Actually, tagging secondaries from one transformer onto another works quite well. Just make sure that the phase is correct and that you don't exceed the current draw of whichever transformer is smaller.
Two channels idling at 100-120W would be 200-240W total--quite safe for a 545VA transformer.
The math on the 24V secondaries yielding approximately 32.5V rails looks okay, but I'm not quite sure how the 49V rail was calculated for the front end. If I understood the original question correctly, I'd expect the rail to be more along the lines of 57V, thus:
24VAC+18VAC=42VAC
42VAC*1.414=~59.4VDC
59.4VDC-(diode losses, etc.)=~57.5VDC
Grey
Two channels idling at 100-120W would be 200-240W total--quite safe for a 545VA transformer.
The math on the 24V secondaries yielding approximately 32.5V rails looks okay, but I'm not quite sure how the 49V rail was calculated for the front end. If I understood the original question correctly, I'd expect the rail to be more along the lines of 57V, thus:
24VAC+18VAC=42VAC
42VAC*1.414=~59.4VDC
59.4VDC-(diode losses, etc.)=~57.5VDC
Grey
If you're pushing the circuit hard enough that the rail collapses--or at least bends a bit--then I can see your point. I don't load my transformers that heavily. I lean towards really stiff power supplies. Those who build otherwise will need to take into account the possibility of saggy rails. I suppose you could send 'em to a plastic surgeon for a bit of a nip 'n tuck, but I'd rather have the young smooth ones.
Even so, with roughly 57 volts to play with and a target somewhere in the 40s, you've got plenty of headroom for a regulator. Just watch the heat dissipation in the pass device.
Grey
Even so, with roughly 57 volts to play with and a target somewhere in the 40s, you've got plenty of headroom for a regulator. Just watch the heat dissipation in the pass device.
Grey
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.