nah, I can visualize the fets and resistors pressed down against the board..
Perhaps the one sink is the way to go, and I'll case the boards upside down. I need square footage for the toroids. Gravity says they want to be on the bottom! My wallet is saying it doesn't want a sink large enough for everything on the bottom. My eyes are saying it needs to look pretty atop my Aleph.
Perhaps the one sink is the way to go, and I'll case the boards upside down. I need square footage for the toroids. Gravity says they want to be on the bottom! My wallet is saying it doesn't want a sink large enough for everything on the bottom. My eyes are saying it needs to look pretty atop my Aleph.
Hi Guys,
The basic board design allows quite a bit of flexibility, so there shouldn't be a need to run wires to side mounted fets. If you absolutely have to, then it will work, but I'm not sure how well. Keep in mind that although this dissipates as much power as an amplifier, it's still a low level circuit, and should be treated as such.
I've attached a sketch of exactly how I will be mounting my I/V in a chassis. You'll have to excuse the terrible perspective drawing, but it gets the point across.
I have two 4"x12"x1" heatsinks that are going to be the sides of the chassis. I'll be using a 1/4" thick aluminum base, and connecting it to the two side heatsinks with 1" aluminum bar stock. I have done this in the past, and it results in a chassis that is both solid, heavy, and transfers heat quite well to the side sinks. I'll be using three bolts into the heatsink, three holding the base to the bar stock, and then I slather both joints with heatsink compound and bolt it all together. If you have a non-ferrous blade and a table saw, you could build something like this in an afternoon for less than $50.
I'll probably be mounting my transformers either on the front plate, or on the base depending on how much room is left. Either way works.
As for mounting the board itself to the heatsink, it's pretty straightforward. Either use thermal pads (thin ones) or use mica pads and thermal grease. The Ohmite resistors are exactly the same height as the FQA fets, which works really nicely. They don't have mounting holes though, which means they need to be mounted such that a bolt can still be put in above them.
For the Caddock resistors, I used a scrap piece of FR4 PCB as a shim, which makes them exactly the same height as the fets. I drilled a hole in it, and glued it to the top of the resistor. Use a good amount of clamping force to hold all the parts down.
For mounting, I just took the PCB before mounting any PTH parts, and traced the holes with a pencil. I then used a centre punch to set each hole, and then drilled an appropriate size hole for the tap. You can use either M3 or 4-40 hardware. M3 is a perfect fit, which requires accuracy in drilling and tapping, and 4-40 is a loose fit which gives some wiggle room if your holes aren't exact.
Cheers,
Owen
The basic board design allows quite a bit of flexibility, so there shouldn't be a need to run wires to side mounted fets. If you absolutely have to, then it will work, but I'm not sure how well. Keep in mind that although this dissipates as much power as an amplifier, it's still a low level circuit, and should be treated as such.
I've attached a sketch of exactly how I will be mounting my I/V in a chassis. You'll have to excuse the terrible perspective drawing, but it gets the point across.
I have two 4"x12"x1" heatsinks that are going to be the sides of the chassis. I'll be using a 1/4" thick aluminum base, and connecting it to the two side heatsinks with 1" aluminum bar stock. I have done this in the past, and it results in a chassis that is both solid, heavy, and transfers heat quite well to the side sinks. I'll be using three bolts into the heatsink, three holding the base to the bar stock, and then I slather both joints with heatsink compound and bolt it all together. If you have a non-ferrous blade and a table saw, you could build something like this in an afternoon for less than $50.
I'll probably be mounting my transformers either on the front plate, or on the base depending on how much room is left. Either way works.
As for mounting the board itself to the heatsink, it's pretty straightforward. Either use thermal pads (thin ones) or use mica pads and thermal grease. The Ohmite resistors are exactly the same height as the FQA fets, which works really nicely. They don't have mounting holes though, which means they need to be mounted such that a bolt can still be put in above them.
For the Caddock resistors, I used a scrap piece of FR4 PCB as a shim, which makes them exactly the same height as the fets. I drilled a hole in it, and glued it to the top of the resistor. Use a good amount of clamping force to hold all the parts down.
For mounting, I just took the PCB before mounting any PTH parts, and traced the holes with a pencil. I then used a centre punch to set each hole, and then drilled an appropriate size hole for the tap. You can use either M3 or 4-40 hardware. M3 is a perfect fit, which requires accuracy in drilling and tapping, and 4-40 is a loose fit which gives some wiggle room if your holes aren't exact.
Cheers,
Owen
yes mate, your sorted, still cant know for certain how many precision matched fets I can get, but there will be devices either way, if not ideal match, probably still better than buying precision matched from elsewhere, because they will be matched at the correct operating point. pretty sure its all good though, you would hope so matching out of 170 fets
qusp:
You won't have a problem finding matched pairs with a sample size that large. Keep in mind that you're not actually matching to a set value, you're just matching between devices, so I wouldn't be surprised if you were able to use at least 150-160 of the 170 devices. If you want to go to the extreme, then you could match into groups of four, but channel to channel variation will likely be dominated by the resistors, and not the fets themselves. Just the same, if you're using 0.1% resistors everywhere, then you could probably make an argument for matching all 4 devices carefully.
If you're running dual mono, then you should definitely match all four devices.
You'll get a feel for it when you start matching, but I expect you'll see pretty good results. Just be meticulous with labeling each fet, and noting the values you measure in a spreadsheet. It's best to hold off on matching until you've measured everything you have. Then use the spreadsheet to separate the parts into groups.
Cheers,
Owen
You won't have a problem finding matched pairs with a sample size that large. Keep in mind that you're not actually matching to a set value, you're just matching between devices, so I wouldn't be surprised if you were able to use at least 150-160 of the 170 devices. If you want to go to the extreme, then you could match into groups of four, but channel to channel variation will likely be dominated by the resistors, and not the fets themselves. Just the same, if you're using 0.1% resistors everywhere, then you could probably make an argument for matching all 4 devices carefully.
If you're running dual mono, then you should definitely match all four devices.
You'll get a feel for it when you start matching, but I expect you'll see pretty good results. Just be meticulous with labeling each fet, and noting the values you measure in a spreadsheet. It's best to hold off on matching until you've measured everything you have. Then use the spreadsheet to separate the parts into groups.
Cheers,
Owen
thanks owen, yeah hard to say how many i'll get, but indeed many of the guys with me have just sprung for 0.01% zfoils so best get them as close as I can, but thanks thats encouraging.
i've done up the excel file for gm match from VDS so just need to work out how i'm going to label them, maybe little sticker spots or something. thats pretty much what I said to the guys too, that pairs was more important than quads, i'm sure there'll be a few quads though.
bought some larger pin sockets too for connecting up the DMMs to, so all set for when everything gets here
i've done up the excel file for gm match from VDS so just need to work out how i'm going to label them, maybe little sticker spots or something. thats pretty much what I said to the guys too, that pairs was more important than quads, i'm sure there'll be a few quads though.
bought some larger pin sockets too for connecting up the DMMs to, so all set for when everything gets here
Attachments
That spreadsheet looks great! Just add a column for the FET number, and one for the Vgs DC measurement.
You'll want to measure the DC voltage on the centre leg of the pot to ground once you have exactly 24VDC across the 200R resistor. You might get quite a bit of variance in this value, but it's nice to try and match it along with transconductance. Just add a column and measure it before measuring the other two values. You have a large enough number that you can probably get at least a few groups of four that match exactly in both TC and Vgs. A group of four like that would be worth it's weight in gold, since you can only get matching that close with a very large supply of fets. If it were me, I'd be keeping those for myself
It's probably also worth mentioning that there are a few threads on here regarding some PC connected matching boards. They actually measure each device and give you a curve for each at your selected operating conditions. This would be the ultimate since it would save a lot of time and allow near perfect matching. I don't know how far along they are though:
http://www.diyaudio.com/forums/pass-labs/25730-diy-mosfet-tester.html
For identification of fets, I used little sticky paper circles about 3/8" in diameter. They come in different colours, so you could just do something like blue 1-99 and red 1-71 and stick one to the front of each fet.
Cheers,
Owen
You'll want to measure the DC voltage on the centre leg of the pot to ground once you have exactly 24VDC across the 200R resistor. You might get quite a bit of variance in this value, but it's nice to try and match it along with transconductance. Just add a column and measure it before measuring the other two values. You have a large enough number that you can probably get at least a few groups of four that match exactly in both TC and Vgs. A group of four like that would be worth it's weight in gold, since you can only get matching that close with a very large supply of fets. If it were me, I'd be keeping those for myself
It's probably also worth mentioning that there are a few threads on here regarding some PC connected matching boards. They actually measure each device and give you a curve for each at your selected operating conditions. This would be the ultimate since it would save a lot of time and allow near perfect matching. I don't know how far along they are though:
http://www.diyaudio.com/forums/pass-labs/25730-diy-mosfet-tester.html
For identification of fets, I used little sticky paper circles about 3/8" in diameter. They come in different colours, so you could just do something like blue 1-99 and red 1-71 and stick one to the front of each fet.
Cheers,
Owen
Last edited:
thanks, yeah OK will do that.
I came across that project some time ago and they were still stuttering along, seems they are getting there, but now Patrick has got a hold of it, its likely to be a very excellent, but also very expensive device made with somewhat difficult to procure sense resistors and sand, in a fully CNC enclosure not really a direction I want to get side tracked with at the moment.
I dont plan to make a business out of making amps, or matching devices, so just knocking up a test circuit with adjustable power supply, socketed devices and source resistors so I can test and match jfets as well, will do for my purposes for the moment. if it turns out to be cheaper than I think, or there is a PCB on the offing i'll rethink that. at the moment there is only a couple of options that will take the current we want for testing power devices for signal use.
yep sticker dots, a large benchtop and a marker
I came across that project some time ago and they were still stuttering along, seems they are getting there, but now Patrick has got a hold of it, its likely to be a very excellent, but also very expensive device made with somewhat difficult to procure sense resistors and sand, in a fully CNC enclosure
not really a direction I want to get side tracked with at the moment. I dont plan to make a business out of making amps, or matching devices, so just knocking up a test circuit with adjustable power supply, socketed devices and source resistors so I can test and match jfets as well, will do for my purposes for the moment. if it turns out to be cheaper than I think, or there is a PCB on the offing i'll rethink that. at the moment there is only a couple of options that will take the current we want for testing power devices for signal use.
yep sticker dots, a large benchtop and a marker
oh and yeah this
**** I hadnt thought of that...... best match highest gm quads are for myself and Flocchini, you guys really oughta thank this man too for making this all possible. it was just a fleeting idea I had and was hesitating about, until he offered to buy some and send them over. so I will either be thanking him for the angels coming down to my listening room from above and visiting upon me as I weep tears of joy, or cursing his name as I wake sweating in my bed, with nightmares of 5 decimal place LCDs 10 years from now
opc said:
**** I hadnt thought of that...... best match highest gm quads are for myself and Flocchini, you guys really oughta thank this man too for making this all possible. it was just a fleeting idea I had and was hesitating about, until he offered to buy some and send them over. so I will either be thanking him for the angels coming down to my listening room from above and visiting upon me as I weep tears of joy, or cursing his name as I wake sweating in my bed, with nightmares of 5 decimal place LCDs 10 years from now
Last edited:
psu electrolytics
opc,
which make/rating of smoothing caps are you using/recommending for the PSU section of the bipolar rails? I cannot see them in the BOM.
I guess a 65V rating is on the limit and I'm therefore inclined to use Pana FC's (100V, 1000uF). I'm not really on a budget so every suggestion is welcome.
Thanks,
Nic
opc,
which make/rating of smoothing caps are you using/recommending for the PSU section of the bipolar rails? I cannot see them in the BOM.
I guess a 65V rating is on the limit and I'm therefore inclined to use Pana FC's (100V, 1000uF). I'm not really on a budget so every suggestion is welcome.
Thanks,
Nic
Could you elaborate a bit on this? What are you referring to with respect to the "other parts"?
well not really, I should clarify that, there is 1800uf before and after the regulator, but remember, these are floating regulators, they are not specified to cope with the voltages we are feeding them, so shorting them to ground is a BIG nono. I havent looked into the onboard supply much, so forgive me if i'm missing something here. aha, you are right, owen hasnt specified a cap for this position in the BOM. C2, C24, C1, C23, C4, C26, C5, C27 are all 1800uf as you have noticed, I really think you can probably get away with 1000-2000 depending on quality, but remember, none of these caps connect to ground, so they do not have to cope with the voltage to ground and pana FC is going to be fine IMO
wow thats something that really should be GB'd, sorry i've made my order at DK, i'm making a PCX order right now though or at least in the next 24hrs, I suppose everyone is going to have a favorite here, so maybe not the best thing to GB, what you guys need tpo know though is the pattern for the placement, who has the board and can tell us what the lead spacing is?
sorry Owen, I really should have picked that up
sorry Owen, I really should have picked that up
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.