Hello!
I decided to build power amplifier that would run with my
Martin Logan electrostatic speakers. Due to the fact that
it's not easiest load I decided to go solid-state and push-pull.
So, some questions first:
1. From what I've heard 2SA5200/2SC1943 is one of the most
beta-linear pair of BJTs? Is it true? Are there any better BJTs
dedicated especially for audio that would have better parameters
AND better sound?
2. Assuming that I'd go 2SA5200/.. for power stage, there's other
important factor - the number of matched pairs needed to get
reasonable output swing/power/stablility. My ML's are rated
at 200W, 4 Ohms, however their impedance drops below 3 ohms
at 15 KHz. With my current amplifier the result of that is
the attenuation of highest frequencies.
So, how many pairs? I am sure that two pairs per channel are
minumum. The more the transistors the more difficult it it to
match them, so I need to achive balance between getting
a resonable Ooomp reserve/stability and relatively gimmick-free
matching of these transistors.
My bet is 3-4 pairs for channel. What would you say?
3. Power supply considerations. Here we have three elements,
trafo, caps and bridge.
a. as for transformer, what is the difference between toroidal
and double-C core (like ones used by YBA). I was told to go
double-C core because it is better owing to the fact that this
core acts like a filter between primary and secondary. Is is true?
If, so what VA rating of transformer would be suitable for
let's say one monoblock consisting of 4 pairs of BJTs stated
earlier?
Filtering caps. Does anyone knows the brand Kendeil?
Two BIG caps or, let's say, 4/6/8 smaller ones? I was told that
smaller capacitances in parallel sound better (faster, 4 being
precise). I am going to go class A/B with relatively high idle
dissipation (so that it runs class A up to 20W/channel), so
it will draw some current constatly. On the other hand I thing
about making external pwr supply 4 each monoblock and to
divide filtering capacitanse into two parts:
- one, smaller, that would act as a filtering only, located in the
PSU case
- second, bigger, say, one pair of caps per one transistor pair, as close
as possible to output transisors. Is this a good idea?
What about rectifiers? Slow or fast ones?
Well, that's all. I don't consider other things, because I want
to have output stage rethinked, altogether with it's psu (I plan
driver and input stage to have regulated rails, I plan to discuss
this ASAP).
Thanks,
esl
I decided to build power amplifier that would run with my
Martin Logan electrostatic speakers. Due to the fact that
it's not easiest load I decided to go solid-state and push-pull.
So, some questions first:
1. From what I've heard 2SA5200/2SC1943 is one of the most
beta-linear pair of BJTs? Is it true? Are there any better BJTs
dedicated especially for audio that would have better parameters
AND better sound?
2. Assuming that I'd go 2SA5200/.. for power stage, there's other
important factor - the number of matched pairs needed to get
reasonable output swing/power/stablility. My ML's are rated
at 200W, 4 Ohms, however their impedance drops below 3 ohms
at 15 KHz. With my current amplifier the result of that is
the attenuation of highest frequencies.
So, how many pairs? I am sure that two pairs per channel are
minumum. The more the transistors the more difficult it it to
match them, so I need to achive balance between getting
a resonable Ooomp reserve/stability and relatively gimmick-free
matching of these transistors.
My bet is 3-4 pairs for channel. What would you say?
3. Power supply considerations. Here we have three elements,
trafo, caps and bridge.
a. as for transformer, what is the difference between toroidal
and double-C core (like ones used by YBA). I was told to go
double-C core because it is better owing to the fact that this
core acts like a filter between primary and secondary. Is is true?
If, so what VA rating of transformer would be suitable for
let's say one monoblock consisting of 4 pairs of BJTs stated
earlier?
Filtering caps. Does anyone knows the brand Kendeil?
Two BIG caps or, let's say, 4/6/8 smaller ones? I was told that
smaller capacitances in parallel sound better (faster, 4 being
precise). I am going to go class A/B with relatively high idle
dissipation (so that it runs class A up to 20W/channel), so
it will draw some current constatly. On the other hand I thing
about making external pwr supply 4 each monoblock and to
divide filtering capacitanse into two parts:
- one, smaller, that would act as a filtering only, located in the
PSU case
- second, bigger, say, one pair of caps per one transistor pair, as close
as possible to output transisors. Is this a good idea?
What about rectifiers? Slow or fast ones?
Well, that's all. I don't consider other things, because I want
to have output stage rethinked, altogether with it's psu (I plan
driver and input stage to have regulated rails, I plan to discuss
this ASAP).
Thanks,
esl
Four pairs per channel will be sufficient. I suggest you find a copy of Douglas Selfs "Audio Power Amplofier Design Handbook". There is an explanation in the Output Stage section that tells how to use the SOA chart from the spec sheet to determine if your proposed configuration can handle the possible combinations of voltage and current. Just remeber that the example in the book is for a single pair so you need to multiply the current on the chart by 4 for a 4 pair output stage.
Note also, the discussion on beta droop distortion when the load drops well below the nominal 8 ohms. If you are a fanatic about squeezing out the lowest THD, you may want to consider doubling the number of pairs to 8 !!
I've run through this thought process in regard to a project I'm thinking of undertaking in about a year. At this point, given a desire for 200W/channel into a 4 ohm nominal load, lateral MOSFETs are looking better and better. They don't experience beta droop and double die versions are available which could make physical construction much simpler.
Regarding external powe supplies, I think they are a PITA from a human factors consideration, and of little value if have good PSRR and layout the amp reasonably well. Getting all clever and exotic with power supplies makes some sense if we are dealing with line level devices, signal processing or phono amps, but with largish power amps it just diverts attention from more critical issues.
No doubt you will get opinions different than mine. Listen to all and make you own call.
However, if you are committed to BJT, 2sa1943/2sc5200 looks that the top choice. (2sa1302/2sc3281 are an earlier, out of prodction version of these which are best avoided due to a lot of reports of counterfeits on the market.) There is also a new pair of devices MJL4281A/MJL4302A from Onsemi which look good on paper. It's not clear to me how available these are as yet, but if you do the SOA analysis you may find you can reduce the numger of output pairs you need.
If this were my project, for a single 200W channel, I would be looking for a toroidal 45V-0V-45V with VA rating one step up from 500VA, i.e., 625VA. I'm conservative about this, in real life you could get by with less, but I like knowing I could (assuming big enough heatsinks) put a sine wave into it full power, walk away and have it still perking a long fine a couple of weeks later.
Similarly for filter caps, I would go with 20,000uF per rail. One could get away with less but this is so conservative that it is doubtful anything more would be wasted.
I wouldn,t bother with a fast bridge. Rather a standard 35A, 400V(or greater) conventional bridge rectifier in a metal case plus snubbers should be sufficient. In a power amp of this size you will hopefully find yourself using a design with very good PSSR (such as mirror image IS and VAS stages and adequate bypass caps) which means you can keep the power supply pretty much non-exotic.
Please do search on this site regarding imrush limiting, and speaker turn-on delay. You, your amp and your speakers will be happier if make provision for these.
Also if you check out Self's book, look at the crtique of class AB. I'm not qualified or experienced enough to say he is right, but you ought to at least be aware of the concerns and make your own call.
Note also, the discussion on beta droop distortion when the load drops well below the nominal 8 ohms. If you are a fanatic about squeezing out the lowest THD, you may want to consider doubling the number of pairs to 8 !!
I've run through this thought process in regard to a project I'm thinking of undertaking in about a year. At this point, given a desire for 200W/channel into a 4 ohm nominal load, lateral MOSFETs are looking better and better. They don't experience beta droop and double die versions are available which could make physical construction much simpler.
Regarding external powe supplies, I think they are a PITA from a human factors consideration, and of little value if have good PSRR and layout the amp reasonably well. Getting all clever and exotic with power supplies makes some sense if we are dealing with line level devices, signal processing or phono amps, but with largish power amps it just diverts attention from more critical issues.
No doubt you will get opinions different than mine. Listen to all and make you own call.
However, if you are committed to BJT, 2sa1943/2sc5200 looks that the top choice. (2sa1302/2sc3281 are an earlier, out of prodction version of these which are best avoided due to a lot of reports of counterfeits on the market.) There is also a new pair of devices MJL4281A/MJL4302A from Onsemi which look good on paper. It's not clear to me how available these are as yet, but if you do the SOA analysis you may find you can reduce the numger of output pairs you need.
If this were my project, for a single 200W channel, I would be looking for a toroidal 45V-0V-45V with VA rating one step up from 500VA, i.e., 625VA. I'm conservative about this, in real life you could get by with less, but I like knowing I could (assuming big enough heatsinks) put a sine wave into it full power, walk away and have it still perking a long fine a couple of weeks later.
Similarly for filter caps, I would go with 20,000uF per rail. One could get away with less but this is so conservative that it is doubtful anything more would be wasted.
I wouldn,t bother with a fast bridge. Rather a standard 35A, 400V(or greater) conventional bridge rectifier in a metal case plus snubbers should be sufficient. In a power amp of this size you will hopefully find yourself using a design with very good PSSR (such as mirror image IS and VAS stages and adequate bypass caps) which means you can keep the power supply pretty much non-exotic.
Please do search on this site regarding imrush limiting, and speaker turn-on delay. You, your amp and your speakers will be happier if make provision for these.
Also if you check out Self's book, look at the crtique of class AB. I'm not qualified or experienced enough to say he is right, but you ought to at least be aware of the concerns and make your own call.
I know that this will not answer your question, but......
About filter caps:
You can not generalise that small vs. big vs. lots vs. few is better. All caps sound different, and will most likely sound different in different topologies. Finding the right combination that goes with your amp, and power transformer can be maddening. Trust me......I've more of this than I care to, and every time that I think that I know the answer, I find that I don't.
It is possible to have too much. Low ESR types may not sound good in one way, but good in others. Some combinations that you think will sound good won't, and at other times, the complete opposite is true.
Wish that I could give you a direct answer, but there really is not one. You will just have to try all the solutions that you can.
BTW......the transistors are good.
Jocko
About filter caps:
You can not generalise that small vs. big vs. lots vs. few is better. All caps sound different, and will most likely sound different in different topologies. Finding the right combination that goes with your amp, and power transformer can be maddening. Trust me......I've more of this than I care to, and every time that I think that I know the answer, I find that I don't.
It is possible to have too much. Low ESR types may not sound good in one way, but good in others. Some combinations that you think will sound good won't, and at other times, the complete opposite is true.
Wish that I could give you a direct answer, but there really is not one. You will just have to try all the solutions that you can.
BTW......the transistors are good.
Jocko
Hi,
I've developped an instinct with this...it takes years of experience as I'm sure Jocko has plenty of.
At the end you browse a catalogue and pick this for that and that for that...
It's a moving target really, some of it gets better some worse.
All in all I feel component quality has gone up if I'd were to ask to make a balance.
You'd still need the same experience to pick the right one for the job though....
Non e facile,
I've developped an instinct with this...it takes years of experience as I'm sure Jocko has plenty of.
At the end you browse a catalogue and pick this for that and that for that...
It's a moving target really, some of it gets better some worse.
All in all I feel component quality has gone up if I'd were to ask to make a balance.
You'd still need the same experience to pick the right one for the job though....
Non e facile,
Hello,
I am going to order 2SA5200/2SC1943 next week. I found them
pretty cheap and I was assured that the company I am going
to buy from has the originals.
The question is - assuming that I am going to use 4 pairs per
channel - how many pairs shoud I buy, bearing in mind that
I would like to match them pretty close ?
GreetZ
esl
I am going to order 2SA5200/2SC1943 next week. I found them
pretty cheap and I was assured that the company I am going
to buy from has the originals.
The question is - assuming that I am going to use 4 pairs per
channel - how many pairs shoud I buy, bearing in mind that
I would like to match them pretty close ?
GreetZ
esl
Hmmm. That depends on whether you are buying from the same batch.
I buy 1000 at a time, and find each box has IIRC about 200.
From within each box, I can assume all are of the same batch.
If I match on Vbe (to within a millivolt) and beta (within 2%), I can normally pull about half into matched pairs. If I match on triples, about a third of a box. And if I match on quads, about 20% will match up.
For makers of large amplifiers, these are not good figures, and they really require a big initial purchase. Nuthin's easy.......
Cheers,
Hugh
I buy 1000 at a time, and find each box has IIRC about 200.
From within each box, I can assume all are of the same batch.
If I match on Vbe (to within a millivolt) and beta (within 2%), I can normally pull about half into matched pairs. If I match on triples, about a third of a box. And if I match on quads, about 20% will match up.
For makers of large amplifiers, these are not good figures, and they really require a big initial purchase. Nuthin's easy.......
Cheers,
Hugh
ESL,
I'd try with 10 pairs of each to begin with, and focus first on Vbe at the same collector current/temperature, then test within each sample on beta. Measure Vbe at the same collector current as the bias setting, and at a constant, low Vce voltage (say 6V); this is important. And do it quickly; it changes 1.5mV/degree C as the devices warm up. Use a rapid sampling DMM like a Fluke.
Of course matching also depends on the tolerance you choose; looser tolerance will mean less devices need be purchased. I try for only 1mV on Vbe (like devices) and 2% on beta, but if you relax the beta tolerance, possible with larger 0R47 emitter resistors, you might find matching easier....
These are very good transistors, extremely linear, very fast, acceptable SOAR, and inexpensive. I like 'em a lot, and feel that they easily best the MJL21193 series. You will find most contemporary high end amps use these devices.
JCX,
I don't even try to match NPN versus PNP. The argument is spurious in my view; I use the Self Type II DEF output stage with charge suckout and believe it's fatuous to match unlike devices. However, matching like devices is VERY important, as the slightest mishap as the musical baton is passed along at crossover is very readily audible. Any differences in beta or Vbe with Ic are absorbed relatively well in the NFB loop; but crossover disjunctions are poorly corrected, and thus all four NPNs should turn off at the same time as the four PNPs are turning on (and vice versa!). This is one reason why large power amplifiers with multiple output pairs never sound as good as little amps with only a single pair.
Cheers,
Hugh
I'd try with 10 pairs of each to begin with, and focus first on Vbe at the same collector current/temperature, then test within each sample on beta. Measure Vbe at the same collector current as the bias setting, and at a constant, low Vce voltage (say 6V); this is important. And do it quickly; it changes 1.5mV/degree C as the devices warm up. Use a rapid sampling DMM like a Fluke.
Of course matching also depends on the tolerance you choose; looser tolerance will mean less devices need be purchased. I try for only 1mV on Vbe (like devices) and 2% on beta, but if you relax the beta tolerance, possible with larger 0R47 emitter resistors, you might find matching easier....
These are very good transistors, extremely linear, very fast, acceptable SOAR, and inexpensive. I like 'em a lot, and feel that they easily best the MJL21193 series. You will find most contemporary high end amps use these devices.
JCX,
I don't even try to match NPN versus PNP. The argument is spurious in my view; I use the Self Type II DEF output stage with charge suckout and believe it's fatuous to match unlike devices. However, matching like devices is VERY important, as the slightest mishap as the musical baton is passed along at crossover is very readily audible. Any differences in beta or Vbe with Ic are absorbed relatively well in the NFB loop; but crossover disjunctions are poorly corrected, and thus all four NPNs should turn off at the same time as the four PNPs are turning on (and vice versa!). This is one reason why large power amplifiers with multiple output pairs never sound as good as little amps with only a single pair.
Cheers,
Hugh
Hello,
thanks Hugh 4 your input. Yes, I am going to order 10 pairs
of Toshibas for starters and then match them as you said.
There's however one thing that I don't get. The fact, that when
the signal crosses the point where "NPNs should turn off at the
same time as the four PNPs are turning on". Does it means that
the more the amp is biased into Class-A the less is it susceptible
to these distortions. I mean - can I use more less precisely
matched devices biased at higher quiescent current ot not?
Could also anyone provide me with a shematic of power supply
consisting of two transforers per channel? (ie. one transformer
per rail as far as I remember, there was some trich used by
Electrompaniet in their higher-end amps with double transformers,
could anyone have any input on this?)
Which rectifier diodes are better: slow ones or fast ones? Is it
dependable on PWR supply configuration etc. etc. ?
Thanks,
esl
thanks Hugh 4 your input. Yes, I am going to order 10 pairs
of Toshibas for starters and then match them as you said.
There's however one thing that I don't get. The fact, that when
the signal crosses the point where "NPNs should turn off at the
same time as the four PNPs are turning on". Does it means that
the more the amp is biased into Class-A the less is it susceptible
to these distortions. I mean - can I use more less precisely
matched devices biased at higher quiescent current ot not?
Could also anyone provide me with a shematic of power supply
consisting of two transforers per channel? (ie. one transformer
per rail as far as I remember, there was some trich used by
Electrompaniet in their higher-end amps with double transformers,
could anyone have any input on this?)
Which rectifier diodes are better: slow ones or fast ones? Is it
dependable on PWR supply configuration etc. etc. ?
Thanks,
esl
Matching
In my 400 W amp I am using six pair of these transistors without any matching ( only take this from one plastic transport tube ). Balast resistors are 0,47 R. Maximal current difference between each currents is about 2 % (measured as voltage on balast resistors at full load ) = all transistors ( from one series ) are practicaly the same.
In my 400 W amp I am using six pair of these transistors without any matching ( only take this from one plastic transport tube ). Balast resistors are 0,47 R. Maximal current difference between each currents is about 2 % (measured as voltage on balast resistors at full load ) = all transistors ( from one series ) are practicaly the same.
Finally...
... I got some output transistors, form the same batch, gonna
match them ASAP... some new ideas...
Since my Martin Logans are rated at 200W/channel with
impedance ranging form 4 ohm (1KHz) to 2.5 ohm (15 KHz),
I wonder how many pairs to go for a single channel so that:
- it goes into Class - A as far as it's reasonable (30 W??? 50W???)
- I get 80 W@8ohm/160@4ohm and about 200@2 ohm and
still transistors work in their SOA
What about using a tube in the input stage?
What about drivers? FETs or BJTs? And what topology?
As for the power supply - I can but some 68000uf/63V Kendeil
caps (four of them)... would it be good or maybe such a huge
capacitance would be an overkill? Shouldn't I stick to smaller
values conected in parallel?
Is there anything more than a trafo, caps and bridge required
in a good power supply? Any filtration?
Thanks
esl
... I got some output transistors, form the same batch, gonna
match them ASAP... some new ideas...
Since my Martin Logans are rated at 200W/channel with
impedance ranging form 4 ohm (1KHz) to 2.5 ohm (15 KHz),
I wonder how many pairs to go for a single channel so that:
- it goes into Class - A as far as it's reasonable (30 W??? 50W???)
- I get 80 W@8ohm/160@4ohm and about 200@2 ohm and
still transistors work in their SOA
What about using a tube in the input stage?
What about drivers? FETs or BJTs? And what topology?
As for the power supply - I can but some 68000uf/63V Kendeil
caps (four of them)... would it be good or maybe such a huge
capacitance would be an overkill? Shouldn't I stick to smaller
values conected in parallel?
Is there anything more than a trafo, caps and bridge required
in a good power supply? Any filtration?
Thanks
esl
Yes, but bear in mind that I want to use tube as an input stage, followed by fet and then these 2sa5200/2sc1943...
... is there any design that I could use and make some changes so
that it can have a tube as an input stage?
Sorry for so lame question, but I just want to know whether does
it take a lot of reworking some orginal schematics to get rid of
(usual) diff input pair and use tube instead...
Or maybe sth like that... say, 6992 input tube, then jfet diff pair and
then Toshibas as a follower?
What do you think?
esl
... is there any design that I could use and make some changes so
that it can have a tube as an input stage?
Sorry for so lame question, but I just want to know whether does
it take a lot of reworking some orginal schematics to get rid of
(usual) diff input pair and use tube instead...
Or maybe sth like that... say, 6992 input tube, then jfet diff pair and
then Toshibas as a follower?
What do you think?
esl
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.