ostripper said:
...but with modern parts i might get something better..
OS
I think that's pretty much a guarantee.
I have done a layout ugly and abominable enough for testing. I won't have time to make it until next week - Monday maybe. By then you will have yours up and running no doubt.
Here it is. I keep it small so as not to hurt some members eyes with it's hideousness.
Attachments
Dang, for the first time I've seen it here ,(I may be wrong)
the reader builds the amp before the designer.
Anyways, good work with board, I see you have the differential
"face to face" but you should try to line up the VAS "single file" to use 1 big heatsink for thermal coupling. try the 2 - 1 - 2
approach (2 active - 1 cascode - 2 CM) for VAS.
[all 5 VAS trannies above W/ to-92 widlar Buffer in middle]
This amp's VAS should dissipate 30mA average so 5sq. inch flashing (the "L" above") ,should do the job. FA1 with 2.5 sq. in. ,stays just warm with "party levels"
@ 8.5mA per device.
Layout is a B****!!, so I'm glad I just had to incorporate the
FA2 VAS into FA1's board design.
OS
the reader builds the amp before the designer.
Anyways, good work with board, I see you have the differential
"face to face"
but you should try to line up the VAS "single file" to use 1 big heatsink for thermal coupling. try the 2 - 1 - 2approach (2 active - 1 cascode - 2 CM) for VAS.
An externally hosted image should be here but it was not working when we last tested it.
[all 5 VAS trannies above W/ to-92 widlar Buffer in middle]
This amp's VAS should dissipate 30mA average so 5sq. inch flashing (the "L" above") ,should do the job. FA1 with 2.5 sq. in. ,stays just warm with "party levels"
@ 8.5mA per device.Layout is a B****!!, so I'm glad I just had to incorporate the
FA2 VAS into FA1's board design.
OS
Generally, I think many DIY try to make too small PCB.
Small parts gets cluttered up and makes it difficult to exchange e.g. one resistor.
It is possible to make very good boards with some space around parts of PCB and easy to modify
and still keep it all electrically good.
Destroyer X make some very good boards.
Perfect for DIY modifications and experimenting.
Same bad with modern cars.
Today they are so guttered up you can't hardly find a way to use one normal screwdriver/wrench.
Not so with them old Volvo cars I have had.
Lot's of space to easily be able to do your repair work
Lineup
Small parts gets cluttered up and makes it difficult to exchange e.g. one resistor.
It is possible to make very good boards with some space around parts of PCB and easy to modify
and still keep it all electrically good.
Destroyer X make some very good boards.
Perfect for DIY modifications and experimenting.
Same bad with modern cars.
Today they are so guttered up you can't hardly find a way to use one normal screwdriver/wrench.
Not so with them old Volvo cars I have had.
Lot's of space to easily be able to do your repair work
Lineup
lineup said:
In an environment with so much radiation (telco, radio and TV stations, WiFi and so on) every millimeter of coper trace in high impedance nodes is a small antenna. Combine this with modern pre/amplifiers with high topology bandwidth ...
That is the main reason of crowded PCB layout
Cheers,
Mihai
Actually, diff pair are not face to face but the VAS and the CM are sharing common heatsinks.
Since you has trouble seeing the details I gave a bigger view.
As for making boards too small, well board material costs and generally you have the bugs worked out in a prototype before you do up a tight layout. To change an R or two can still be done easily.
Since you has trouble seeing the details I gave a bigger view.
As for making boards too small, well board material costs and generally you have the bugs worked out in a prototype before you do up a tight layout. To change an R or two can still be done easily.
depends a bit on the type of amplifier
for RF frequencies I can see the benefit of reducing 'hit area' for the circuit
for example regarding MC phono preamplifiers with very high gain processing very low level signals (uV)
for Audio Band (20-20.000 Hz) we do not amplifiy high very high frequencies
the gain is very low, 20-30,
and the level we use for input is higher
and so we can many times use layouts that are DIY-friendly
now, there are many sort of hobby audio designers
where some are more at NASA space level
while other are more down to earth
.. and so there are as many preferences when doing Printed (plastic) Circuit Boards
Making Excellent Printed Circuit Boards
for RF frequencies I can see the benefit of reducing 'hit area' for the circuit
for example regarding MC phono preamplifiers with very high gain processing very low level signals (uV)
for Audio Band (20-20.000 Hz) we do not amplifiy high very high frequencies
the gain is very low, 20-30,
and the level we use for input is higher
and so we can many times use layouts that are DIY-friendly
now, there are many sort of hobby audio designers
where some are more at NASA space level
while other are more down to earth
.. and so there are as many preferences when doing Printed (plastic) Circuit Boards
Making Excellent Printed Circuit Boards
I agree with you lineup,in most cases I like to be able to easily
replace a resistor or two so I tried to hit the middleground in
size (3x6) also I bought 20 6X6 boards real cheap (2 amps
per board) so I can make many..
More important than RF (we do use a metal chassis, don't we?)
is parallel small signal / high current traces (dirty ground,rails, vas)
that is why I placed LTP/Input/NFB return away from OPS/VAS/CCS.I even will run my NFB through a 'shielded' 2 inch
under the board jumper(the only one)right to the NFB return of the differential.FA1 is way more quiet than my sound blaster live with a"sloppier" layout than this one..
Dead silent. (with SBlive turned down)
Yes , 35db OLG at 40khz then unity gain at 600Khz,-20db after
1mhz, so It should not amplify RF at all.
Also, one might want to put 5 in the same case for HT.
I see .... 1. are r34/35 for additional ripple rejection??you could easily straighten them out as you have plenty of real estate.
(maybe add another decoupling cap to further reduce ripple)
2.Are you going to make a U shaped HS for 1381's/3503's ??
it shouldn't be hard to line them up in a nice straight line
(don't be afraid of a jumper or 2) as you have ample real estate
and PC based layout tools.
3 . rails,NFB, grounds seem to abide by accepted rules of engagement. C5 can be 100uf-63v, I left the 220 there in last simulation, found that 100/220 only gave uA's of improvement
in PSRR.
4. I see you use a mpsa42 (higher voltage) ,better than the bc556
as 50+ volts will bump the derating spec. I am using 2sc2682
(have a lot of them ,180v devices)for CCS.
5 . and since you are putting Vbias on the main HS you might have to play with the value of R14 to set the right tempco as
2K7 reflects one tailored for direct device feedback. (starts at
50mA cold per device builds to 70mA, then levels out)
BTW , feel free to use other OPS devices, MJL21193/4 or
2sc5200/1943, etc.. even T- tracks . I simmed diode biasing
replacing R14 (leach Vbias circuit) works good..
Good job, otherwise...
OS
roender said:
Please allow me a small improvement
No, NO, you've ruined it!! My vision!! You have desecrated my art!!
I can see what these changes are meant to do, but they will do nothing. I prefer to maintain as large as possible current path directly to it's drain. This other trace that picks up grounds from the front end carries very little current and it's length discourage the flow back of the larger current from the decoupling cap.
Slicing the output copper area to put the feedback closer to the output terminal is counter productive - it will do more harm than good as it restricts flow around the cuts. My pick off point is close enough to ideal.
ostripper said:
No, they are there more as jumpers to allow an easier layout. They would isolate the front end from large current flow in the output stage. Your amp has better PSRR than mine, as I used bootstrapped VAS, therefore I included the extra decoupling / bypass caps in mine to help with this.
ostripper said:
I think that it is important to thermally couple the VAS T's to each other and the CM T's, but it's probably not necessary to couple all four together.
I try to avoid jumpers as much as possible because they look krappy and take longer to put in than a trace that is already there.
ostripper said:
I'm just following your lead here...
ostripper said:
I have also changed the diff pair to the ZTX696B.
I have decided to run all of my future "serious" amps at higher rails to avoid clipping. The output will still be limited to a sane level be raising the input sensitivity.
ostripper said:
This is how it is on my Patchwork and really how it should be anyway. To maintain a steady idle current in an EF output amp, the Vbe should track the outputs. In a CFP, the Vbe needs to track the drivers. If these are all on the same heatsink, that's fine.
I'll probably build this with 2SC5200/A1943 as I have quite a few of these still. I will not be using the MJE drivers, but I'll use the 2SC4793/A1837 - these have a nice insulated case that avoids mucking with mica and insulating shoulder washers. These are great drivers anyway.
ostripper said:
Thanks.
MJL21193 said:
I have also changed the diff pair to the ZTX696B.
----
I'll probably build this with 2SC5200/A1943 as I have quite a few of these still.
I will not be using the MJE drivers, but I'll use the 2SC4793/A1837 - these have a nice insulated case that avoids mucking with mica and insulating shoulder washers. These are great drivers anyway.
Thanks.
ZETEX (ZTX696B) makes good transistors.
2SC4793/SA1837. Yes, these are topclass drivers.
Would not debate a thing with an old pro (sarcasm) except just
2...
1.
do a jumper than reroute 3 or 4 traces to go around something. I only
have 4 in my LPT/ VAS, but It enabled me to allow for real
short paths between stages.
All the big jumpers in front are the rails (poor mans double sided board) which could be eliminated with a larger board (4X6).
output device,not a driver like CFP.. this is as close as you can get to a T-trak without one (AKSA and many precision amps
do this, even D. self recommends this) A much tighter "spread'
of comp can be obtained as on the frugal 1 have gotten it down to
<15ma cold to party hot
. heatsink comp will work but
that resistor will have to be "fiddled with" to get just the right
coefficient.
(not on main sink)
BTW... I won't have it done till late next week parts ordered,
just cleaned out my resistor box so far.
OS
2...
1.
Have you ever seen how many jumpers are on an HT receiver ?? As far as their function, I would rather
do a jumper than reroute 3 or 4 traces to go around something. I only
have 4 in my LPT/ VAS, but It enabled me to allow for real
short paths between stages.
An externally hosted image should be here but it was not working when we last tested it.
All the big jumpers in front are the rails (poor mans double sided board) which could be eliminated with a larger board (4X6).
As far as Vbe , you misunderstood me..I take the comp right off a
output device,not a driver like CFP.. this is as close as you can get to a T-trak without one (AKSA and many precision amps
do this, even D. self recommends this) A much tighter "spread'
of comp can be obtained as on the frugal 1 have gotten it down to
<15ma cold to party hot
. heatsink comp will work butthat resistor will have to be "fiddled with" to get just the right
coefficient.
I hope you meant EF, because CpF likes separate drivers+Vbe.
(not on main sink)
BTW... I won't have it done till late next week parts ordered,
just cleaned out my resistor box so far.
OS
lineup said:
Yes but they are expensive. I used the 696B before and of the 25 I ordered, the first 2 I pulled out had nearly identical gain.
ostripper said:
ostripper said:
The distances saved in most cases are insignificant. To place a jumper to save 2 cm of trace is not the way I like to do it. I want to eliminate all jumpers if possible.
I see a board layout like a puzzle that needs to be solved. Trying to avoid jumpers will ultimately result in a better layout because you need to place the parts more efficiently.
The board I posted above is nearly identical to Patchwork in size - 137mm (5.4") x 75mm (3") - very space efficient for a 4 output amp. Still, given a bit more time to resolve the "puzzle" it could be made even smaller and more efficient, but the results would not be worth the time invested.
ostripper said:
As far as Vbe , you misunderstood me..I take the comp right off a
output device,not a driver like CFP.. this is as close as you can get to a T-trak without one (AKSA and many precision amps
do this, even D. self recommends this) A much tighter "spread'
of comp can be obtained as on the frugal 1 have gotten it down to
<15ma cold to party hot
that resistor will have to be "fiddled with" to get just the right
coefficient.
I hope you meant EF, because CpF likes separate drivers+Vbe.
(not on main sink)
BTW... I won't have it done till late next week parts ordered,
just cleaned out my resistor box so far.
OS
Ok, I thought you had the Vbe off on it's own (picture ESP 3A). Sorry.
Of course I will adjust the values in the Vbe multiplier to what works best - most likely just copy what I have already done in Patchwork. As for tracking accuracy, I can't say I have any problem with the arrangement above. It gives rock steady idle current.
As for CFP, having the Vbe on the main heatsink will work fine as long as the drivers are there too. They don't need to be on a separate sink.
I'm busy over the weekend with work, so I'll not be getting much done. I may spend some time puzzling the puzzle though to make it prettier.
MJL21193 said:
ostripper said:
Are you sure about 0.0001% THD improvement?
Did you read D Self book?
Did you read what MikeB has done to improve his symasym?
John, taking the ground for low current stages in the way you did it is a BIG mistake.
Anyway, it's your design .... do what you want.
Quite right. What is the text, Mihai?
Another technique using the same principle and saving on decoupling caps is to simply run a 10R resistor and back to back diodes from A to B, and use Star Earth so formed to reference input and fb return.
You must keep well away from the heavy charge pulses, which are typically around 50A for very short periods of each reservoir cap charge cycle. As it is, these high currents interfere with speaker earth return current, causing intermodulation.
Hugh
Another technique using the same principle and saving on decoupling caps is to simply run a 10R resistor and back to back diodes from A to B, and use Star Earth so formed to reference input and fb return.
You must keep well away from the heavy charge pulses, which are typically around 50A for very short periods of each reservoir cap charge cycle. As it is, these high currents interfere with speaker earth return current, causing intermodulation.
Hugh
input ground,power grounds
I don't think separating gounds is good idea(decoupling the decoupling),you lose your reference point.reverse the thinking in the text,maybe the input grounds interfere with each other,higher impedance after all,ie the diagram shown is not "starry" enough.
regards
max
I don't think separating gounds is good idea(decoupling the decoupling),you lose your reference point.reverse the thinking in the text,maybe the input grounds interfere with each other,higher impedance after all,ie the diagram shown is not "starry" enough.
regards
max
ostripper said:
"Figure 6.5 shows the effect on an otherwise Blameless amplifier handling 60 W/8, with 220 μF rail decoupling capacitors; at 1 kHz distortion has increased by more than ten times, which is quite bad enough. However, at 20 Hz the THD has increased at least 100-fold, turning a very good amplifier into a profoundly mediocre one with one misconceived connection.
When the waveform on the supply rails is examined, the 100 Hz ripple amplitude will usually be found to exceed the pulses due to Class-B signal current, and so some of the distortion on the upper curve of the plot is actually due to ripple injection. This is hinted at by the phase-crevasse at 100 Hz, where the ripple happened to partly cancel the signal at the instant of measurement. Below 100 Hz the curve rises as greater demands are made on the reservoirs, the signal voltage on the rails increases, and more distorted current is forced into the ground system."
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.