Cause Musher wants something like +/-40V for the ouputs and +/-60V for the front end supply from the same transformer.... the transformer only gives +/-40V or so when rectified and a voltage doubler is a reasonably easy way for him to get the extra voltage for the front end .... sure its noisy but it will probably give about +/-75 which could then be regulated down to a nice quiet +/-60V
AudioFreak
Yes, I know what the voltage doubler would be used for. I asked the question because Musher indicated in his original post that he intended to use a separate transformer for the driver stages which would be a much better solution. Grey followed up with some further suggestions, including the voltage doubler, and then Joe Berry gave a fairly comprehensive coverage of the use this arrangement. I was wondering if you had any additional information to contribute or whether you were just reiterating that which had been said four posts earlier.
Geoff
Yes, I know what the voltage doubler would be used for. I asked the question because Musher indicated in his original post that he intended to use a separate transformer for the driver stages which would be a much better solution. Grey followed up with some further suggestions, including the voltage doubler, and then Joe Berry gave a fairly comprehensive coverage of the use this arrangement. I was wondering if you had any additional information to contribute or whether you were just reiterating that which had been said four posts earlier.
Geoff
Re: Then to the problems with front-end
Musher,
using LED for the current sources has some advantages and some disadvantages as well. LED's
produce very less noise compared to zener diodes. On the other hand if you like to use mosfets for
the current sources you will need at least 3 LED's in series to create enough voltage drop for the
current source (4V for the mosfet, 2 volt for the resistor). In that case you have to deal with the
temperature drift of the 3 LED's.
From the manual of the smaller X-series amps like the X250 you can guess which type of input
device Mr. Pass is using. The manual describes it with as a double JFET with 0.02ms transconductance
figure. I guess it is a 2SK389 which is a really nice transistor for a differential pair. It has a
maximum drain/gate voltage of 50V. If you like to use it with +/-60V power supply you need to put
an additional cascode around it. This technique is very well described in the article about
cascoded amplifier design by Mr. Pass.
I do not really see a need to use parallel JFET for the front end of the X-amp design. It depends
on how much current variation you will need to drive your output section. a 2SK389 can be used
up to 10mA. But you should keep the voltage low on them because of the power dissipation.
Regards,
Maik
Musher,
using LED for the current sources has some advantages and some disadvantages as well. LED's
produce very less noise compared to zener diodes. On the other hand if you like to use mosfets for
the current sources you will need at least 3 LED's in series to create enough voltage drop for the
current source (4V for the mosfet, 2 volt for the resistor). In that case you have to deal with the
temperature drift of the 3 LED's.
From the manual of the smaller X-series amps like the X250 you can guess which type of input
device Mr. Pass is using. The manual describes it with as a double JFET with 0.02ms transconductance
figure. I guess it is a 2SK389 which is a really nice transistor for a differential pair. It has a
maximum drain/gate voltage of 50V. If you like to use it with +/-60V power supply you need to put
an additional cascode around it. This technique is very well described in the article about
cascoded amplifier design by Mr. Pass.
I do not really see a need to use parallel JFET for the front end of the X-amp design. It depends
on how much current variation you will need to drive your output section. a 2SK389 can be used
up to 10mA. But you should keep the voltage low on them because of the power dissipation.
Regards,
Maik
What would you think of Semelab MOSFETs in the power stage? BUZ900 and 905? Those are quite expensive, but I think I'd be ready to get some if they are really good, which they much seem to be.
IRF9610 are obtainable here. 2SK389 also.
2SJ109 need to be hunted for a bit. Same with IRF610.
-Kimmo S.
IRF9610 are obtainable here. 2SK389 also.
2SJ109 need to be hunted for a bit. Same with IRF610.
-Kimmo S.
What is the deal with Finland?
Musher,
I don't understand the deal about those components not being available in Finland. Have you tried ANY of the multinationals such as RS Components, Future Electronics etc?
Now, if these components truly are unavailable in Finland, I suggest you get them from overseas. All these components are available in Norway -- perhaps we are just a much more advanced society than Finland
What is the cost of shipping to Finland when ordering from the US or UK? Surely it cannot be as bad as Norway!
Petter
Musher,
I don't understand the deal about those components not being available in Finland. Have you tried ANY of the multinationals such as RS Components, Future Electronics etc?
Now, if these components truly are unavailable in Finland, I suggest you get them from overseas. All these components are available in Norway -- perhaps we are just a much more advanced society than Finland
What is the cost of shipping to Finland when ordering from the US or UK? Surely it cannot be as bad as Norway!
Petter
Obtaining transistors and so on...
It costs extra to get transistors by mail from Norway, Sweden or any other place than nearby Helsinki. It also more often than not requires getting a large batch at once. That really matters since I am a student. I am determined to get enough resources to finish the project, but I have no extra cents or percents to lose.
Then to the matching business...
My distant dream is to match within 0.5% as Mr. Pass does. I think that means the Vgs' of the transistors are within 15mV, am I right or not? I think the other possibility is that the Vgs' are within 30mV. I'm very afraid I will have to get 100 for every single matched.
Looking at Petter's schematic, you see one channel has one output section at the left and one at the right. And both sections have an upper and a lower row. Now, should the four rows be matched internally, or upper left also with lower left, or upper left with lower right, lower left with lower right, or every single transistor in the channel with every other?
And how many IRF240/9240 do you think I will have to get if I want to match 6+6 (or 3+3+3+3) within 15mV? How much easier will it get if I make four channels with 6+6? Or one channel with 24+24 (or 12+12+12+12)? International rectifier hasn't answered this yet, and it's been a while since I asked.
Maybe I will end up doing serious business selling matched transistors after all this. Unless Mr. Pass would like to start selling matched sets again. I know how this could help his business and maybe did earlier, but I also guess there are good reasons why he doesn't do it anymore.
-Kimmo Sundqvist
It costs extra to get transistors by mail from Norway, Sweden or any other place than nearby Helsinki. It also more often than not requires getting a large batch at once. That really matters since I am a student. I am determined to get enough resources to finish the project, but I have no extra cents or percents to lose.
Then to the matching business...
My distant dream is to match within 0.5% as Mr. Pass does. I think that means the Vgs' of the transistors are within 15mV, am I right or not? I think the other possibility is that the Vgs' are within 30mV. I'm very afraid I will have to get 100 for every single matched.
Looking at Petter's schematic, you see one channel has one output section at the left and one at the right. And both sections have an upper and a lower row. Now, should the four rows be matched internally, or upper left also with lower left, or upper left with lower right, lower left with lower right, or every single transistor in the channel with every other?
And how many IRF240/9240 do you think I will have to get if I want to match 6+6 (or 3+3+3+3) within 15mV? How much easier will it get if I make four channels with 6+6? Or one channel with 24+24 (or 12+12+12+12)? International rectifier hasn't answered this yet, and it's been a while since I asked.
Maybe I will end up doing serious business selling matched transistors after all this. Unless Mr. Pass would like to start selling matched sets again. I know how this could help his business and maybe did earlier, but I also guess there are good reasons why he doesn't do it anymore.
-Kimmo Sundqvist
You would be surprised how many matches you
will get. First off, it is only necessary to match
devices in parallel, so for a typical X amp in stereo,
you have 8 batches of devices to match. Out of 50
devices, you will have a good chance of getting 8 X 6
sets within .1 volts, an probably quite a bit better.
We match to .01 V, but only because we can easily
do so. There is no big performance advantage to
getting this close, so do the best you can.
will get. First off, it is only necessary to match
devices in parallel, so for a typical X amp in stereo,
you have 8 batches of devices to match. Out of 50
devices, you will have a good chance of getting 8 X 6
sets within .1 volts, an probably quite a bit better.
We match to .01 V, but only because we can easily
do so. There is no big performance advantage to
getting this close, so do the best you can.
Out of 100 IRF644s, I got eight sets of six matched devices--enough for two pairs of Aleph 2.60s. The others will find their way into Mini-As and such things as time goes on, and there are smaller matched sets of two and three that can be used for smaller projects.
...you never know what other circuits might pop up that need a MOSFET or two...
Grey
...you never know what other circuits might pop up that need a MOSFET or two...
Grey
Kimmo,
Here are some additional stats from a recent project that may give some idea of what to expect with regard to tolerance. I measured the Vgs for 100 IRFP240s and 100 IRFP9240s at an idle current of about 250mA. All devices of like type had the same lot code, and Vgs readings were taken after 5 seconds -- there is some downward drift over time with the bare device idling at 1W in free air. I've listed the counts below in groups of +/- .5%.
IRFP240 (Qty)
3.96 - 4.00 (02)
4.01 - 4.04 (06)
4.05 - 4.08 (07)
4.09 - 4.12 (18)
4.13 - 4.16 (20)
4.17 - 4.20 (13)
4.21 - 4.24 (16)
4.25 - 4.28 (04)
4.29 - 4.32 (04)
4.33 - 4.36 (09)
4.45 (01)
IRFP9240 (Qty)
3.92 - 3.95 (15)
3.96 - 3.99 (29)
4.00 - 4.03 (29)
4.04 - 4.07 (17)
4.08 - 4.11 (06)
4.12 - 4.15 (04)
I think these results are pretty representative. FWIW, the matching on a population of 100 IRFP140Ns (for a different project, and all with same lot code) was much better than this, with a spread of 3.83-3.93V (+/- 1.3%) for all 100 of them. Hope this helps.
Here are some additional stats from a recent project that may give some idea of what to expect with regard to tolerance. I measured the Vgs for 100 IRFP240s and 100 IRFP9240s at an idle current of about 250mA. All devices of like type had the same lot code, and Vgs readings were taken after 5 seconds -- there is some downward drift over time with the bare device idling at 1W in free air. I've listed the counts below in groups of +/- .5%.
IRFP240 (Qty)
3.96 - 4.00 (02)
4.01 - 4.04 (06)
4.05 - 4.08 (07)
4.09 - 4.12 (18)
4.13 - 4.16 (20)
4.17 - 4.20 (13)
4.21 - 4.24 (16)
4.25 - 4.28 (04)
4.29 - 4.32 (04)
4.33 - 4.36 (09)
4.45 (01)
IRFP9240 (Qty)
3.92 - 3.95 (15)
3.96 - 3.99 (29)
4.00 - 4.03 (29)
4.04 - 4.07 (17)
4.08 - 4.11 (06)
4.12 - 4.15 (04)
I think these results are pretty representative. FWIW, the matching on a population of 100 IRFP140Ns (for a different project, and all with same lot code) was much better than this, with a spread of 3.83-3.93V (+/- 1.3%) for all 100 of them. Hope this helps.
I went at the selection process from the back end forward. I bought 100 of the IRF644 with the intention that I would select the best groups and hopefully come out with extremely close matching from such a large starting population.
It didn't work out that way.
I ended up with six groups of six that were matched to within .01V (.005, .005, .006, .006, .007, .008). In pulling two more sets to complete the second pair of Alephs I ended up with .014 and .017. Note that these six sets were generally far removed from each other. From there it diverges quite rapidly--assuming that I want them matched in groups of six. If I want smaller groups, then I can pull several sets of three or two.
Some people seem to have much better matching results. There are numerous possibilities involved.
--I ordered my IRF parts from Digikey, specifying same lot code if possible. Perhaps they sent a bunch of tailings instead. I regard this as the most likely possibility.
--It could be that the IRF644 naturally has a wider spread than other devices, although I can't see why this should be. I chose the 644 because it had the same specifications as the 244 (albeit TO-220 casing rather than TO-3, hence somewhat lower power dissipation) and was cheaper, which enabled me to order 100, thus getting the price break and (hopefully) very tight matching.
--Matching conditions. I ran the devices at .5A (nominal bias for an Aleph 2) for ten minutes before taking measurements. I've seen where people measure them for anywhere from seconds to 5 minutes. It's been well over a year since I matched the parts, but I seem to recall that the voltage was still changing rapidly up until about seven or eight minutes; more slowly after that. Measuring them after seconds wouldn't have been practical--the numbers on the meter were flying by so quickly I wouldn't have been able to read them.
The matched groups I got weren't shabby, by any means, but it would have been nice to have more sets--and more closely related sets--out of that large a population. It may all devolve to a question of luck, perhaps if I ordered another set tomorrow things would work out better.
Unfortunately, Lady Luck has never left me her phone number; I usually have to bull my way through problems by pure persistance.
Grey
It didn't work out that way.
I ended up with six groups of six that were matched to within .01V (.005, .005, .006, .006, .007, .008). In pulling two more sets to complete the second pair of Alephs I ended up with .014 and .017. Note that these six sets were generally far removed from each other. From there it diverges quite rapidly--assuming that I want them matched in groups of six. If I want smaller groups, then I can pull several sets of three or two.
Some people seem to have much better matching results. There are numerous possibilities involved.
--I ordered my IRF parts from Digikey, specifying same lot code if possible. Perhaps they sent a bunch of tailings instead. I regard this as the most likely possibility.
--It could be that the IRF644 naturally has a wider spread than other devices, although I can't see why this should be. I chose the 644 because it had the same specifications as the 244 (albeit TO-220 casing rather than TO-3, hence somewhat lower power dissipation) and was cheaper, which enabled me to order 100, thus getting the price break and (hopefully) very tight matching.
--Matching conditions. I ran the devices at .5A (nominal bias for an Aleph 2) for ten minutes before taking measurements. I've seen where people measure them for anywhere from seconds to 5 minutes. It's been well over a year since I matched the parts, but I seem to recall that the voltage was still changing rapidly up until about seven or eight minutes; more slowly after that. Measuring them after seconds wouldn't have been practical--the numbers on the meter were flying by so quickly I wouldn't have been able to read them.
The matched groups I got weren't shabby, by any means, but it would have been nice to have more sets--and more closely related sets--out of that large a population. It may all devolve to a question of luck, perhaps if I ordered another set tomorrow things would work out better.
Unfortunately, Lady Luck has never left me her phone number; I usually have to bull my way through problems by pure persistance.
Grey
Ordering tips
When ordering, it is usually impossible to get same lot codes unless by this happens by chance.
Therefore, consider asking what the package size is and order exactly that number or multiples of it. At least the packages are likely to be from same lots.
NP and others who order substantial amounts have much greater chance of getting same lot code + they have a much larger population meaning that they are likely to get close to 100% utilization.
Matching of the most critical parts (input stage) was easy as I needed pairs (or quads at worst).
Petter
When ordering, it is usually impossible to get same lot codes unless by this happens by chance.
Therefore, consider asking what the package size is and order exactly that number or multiples of it. At least the packages are likely to be from same lots.
NP and others who order substantial amounts have much greater chance of getting same lot code + they have a much larger population meaning that they are likely to get close to 100% utilization.
Matching of the most critical parts (input stage) was easy as I needed pairs (or quads at worst).
Petter
Digikey lists the standard package size for the IRF644 as fifty. I had hoped that ordering one hundred would get me two unbroken fifties, but that doesn't mean that's the case. Once they fill a random quantity order, there's nothing to stop them from adding two or three tubes together at the end of the day to backfill to fifty. One would hope that if the order asked for the same lot code, they would pull two fresh tubes, but perhaps they hire workers the same way that the University of South Carolina does: For any give job opening, take ten applications, discard the five most obvious choices, then throw a dart at the ones that remain. I don't know how much Digikey pays, but it's unlikely that they ask for college degrees in their back-line folks...
Grey
Grey
Digi-key probably goes for speed. They certainly succeed, I've been very happy with their service. In fact, it's way too convenient, I usually order more than I really end up wanting.
But special requests are probably a problem. Hmm, does this sound like any other highly successful purveyor of goods for public consumption?
How about "McTronics"?
But special requests are probably a problem. Hmm, does this sound like any other highly successful purveyor of goods for public consumption?
How about "McTronics"?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.