Transistor amp from scratch

[swpaskett]

[Just Drill some holes and the rest can be excavated using a file or rasp] -- I just let a sheet-metal shop punch it out for me, along with all of the other holes and cutouts and inserting the necessary PEM-hardware. Works excellently EVERYTIME!!!

/

I try to buy IECs and other rectangular stuff that can be mounted from the front, thus covering any sloppiness in making the hole. When I need pretty, I start by drilling the corner holes to match the radius of the connector. Then cut aluminum with a small slitting saw (machinest supply houses sell them) mounted on a mandrel in a Dremel or similar tool, staying well inside the lines, then clean up with a file. For steel the slitting saw is replaced with a fiber composite cutting wheel. I can cut an IEC in under five minutes. The nibblers work great on thin aluminum but won't cut more than a few holes in 16 ga mild steel, and your hand will hate you for it.

 

[swpaskett]

PCB

Back when I started at Motorola in the early '80s we were making controlled impedance traces on PCB material with Xacto knives and peeling up unwanted copper with pliers.
Then they turned me loose on my first multi-layer board; 11 layers laid out by hand with 4X tape and "dolls", pre-printed sticky pads for ICs and other multi-pinned things. The layout tech I worked with took about 12 weeks to complete the full tapeout on that board, and we spent another 4 just checking it. Our order of 12 boards came back and we had shorted track all over the place. We went back and double checked the layout. It was perfect and all design rules had been followed. We ordered another dozen boards and this time they came back with 8 good boards. While we waited the reliability guys did their work, ripping off layers to find the first set of boards contained a layer that hadn't been etched!
It still boggles my mind that a man could lay out 11 layers. He was a real craftsman and it was shear luck that I got to work with him -- the most patient man in the world.

 

[MidnightArrakis]

Back when I started at Motorola in the early '80s we were making controlled impedance traces on PCB material with Xacto knives and peeling up unwanted copper with pliers.
Then they turned me loose on my first multi-layer board; 11 layers laid out by hand with 4X tape and "dolls", pre-printed sticky pads for ICs and other multi-pinned things. The layout tech I worked with took about 12 weeks to complete the full tapeout on that board, and we spent another 4 just checking it. Our order of 12 boards came back and we had shorted track all over the place. We went back and double checked the layout. It was perfect and all design rules had been followed. We ordered another dozen boards and this time they came back with 8 good boards. While we waited the reliability guys did their work, ripping off layers to find the first set of boards contained a layer that hadn't been etched!
It still boggles my mind that a man could lay out 11 layers. He was a real craftsman and it was shear luck that I got to work with him -- the most patient man in the world.

[we were making controlled impedance traces on PCB material with Xacto knives and peeling up unwanted copper with pliers] -- Been there.....done that!!!

[11 layers laid out] -- 11-layers??? Really??? And, you got away with that with nothing happening during the fabrication? I thought that the "standard" was -- an EVEN number of layers --!!!

[by hand with 4X tape] -- My layouts were either 2X or 4X depending upon either the size of the PCB (large boards were 2X, while small boards were 4X) or due to its features (as in needing to use 10-mil tracks). I had to hand-tape a 16-layer board for a defense contractor that were going to be used within U.S. fighter jets for the early -- IFF -- (Interrogated Friend or Foe) system.

[and "dolls", pre-printed sticky pads for ICs and other multi-pinned things] -- "Bishop Graphics", right??? I even still have my nearly 3-inch thick "Bishop Graphics" PCB Footprint Book lost in a closet around here somewhere.

[The layout tech I worked with took about 12 weeks to complete the full tapeout on that board] -- I worked at AMPEX in Redwood City, CA back in 1980 and I had met the guy who hand-designed the motherboard for the AMPEX M1200 tape deck. He told me that it took him -- 1-YEAR -- to layout that board and get it into production!!!

[contained a layer that hadn't been etched!] -- Back in the early 1990's as PCB-design was entering the phase of being done on computers, not only did you have to supply a set of GERBER files, but before the days of the now "standard" -- RS-274X -- file format, you had to also provide the PCB-fabricator with a text-code file containing all of your aperture designations and settings. That GERBER file format was called "RS-274D". At that point-in-time, I was working with an "RF" communications company that designed and built specialized "RF" communications gear for all sorts of "unknown and covert" 3-letter U.S. Government agencies. One time, a PCB-fabricator technician entered in a typo when transcribing a GERBER-file "Aperture Code" into their photoplotting equipment. That typo ended up rotating the pads of the IC chips 90-degrees, so instead of the pads looking like a "ladder", they looked like "skid-marks" and were basically shorting all of the IC pads together!!! The PCB-fabricator had to "eat" the cost of that failed batch of PCB's.

When the industry finally got around to using the -- RS-274X -- file format that contained "embedded" aperture data, that was such a relief!!!

[boggles my mind that a man could lay out 11 layers] -- As mentioned earlier, try 16-layers and with "Split Power & Ground" layers as well. At the end of each day when hand-taping a PCB back then, I would always have pieces of the crepe tape used for the tracks stuck all over my arms and shirt!!!

[the most patient man in the world] -- Many times during those early years of -- hand-taping -- PCB's, I heard people make comments something like, "To be a PCB Designer, you need to have the patience of Job"!!!

Sorry for the "thread clutter noise".

/

 

[swpaskett]

[11 layers laid out] -- 11-layers??? Really??? And, you got away with that with nothing happening during the fabrication? I thought that the "standard" was -- an EVEN number of layers --!!!



[and "dolls", pre-printed sticky pads for ICs and other multi-pinned things] -- "Bishop Graphics", right??? I even still have my nearly 3-inch thick "Bishop Graphics" PCB Footprint Book lost in a closet around here somewhere.

Yikes, I have been away from hardware too long and am probably using the term "layers" incorrectly. The board had a ground plane, a power plane, and nine signal planes, if I remember right, but all I am really certain of is that there were two digits in the count of planes or layers or whatever is the right terminology. The old brain is a mite foggy these days.

I remember Bishop Graphics. I used a lot of that stuff in school, but the stuff Moto used was made in-house by the graphic artists and printed on clear stickback material they got somewhere. Moto was really into vertical integration up through the '70s -- they wanted to own the entire supply chain and build vs. buy decisions almost always went to build, no matter the cost. That is how they got into the semiconductor business. Remember those days? Now they have divested themselves down to nothing. Same with HP. I hate it when I see the greats die like that, one piece at a time. Tektronix seems to be the only heavyweight that has kept itself intact from when I started this adventure 45 years ago.

 

[KIDVOX]

i think we know what you mean by a layer. no matter what its called

 

[Mr. Moscode]

Me, too. I did tape up with Bishop peel and stick in the 80s at New York Audio Labs but only 2 sided. I later started doing them in McDraw which did layers so I could do hole map, thru hole layer, top and bottom, print aret, etc.. The board makers made masks from the hole layer artwork by shooting it out of focus. Output was postscript which they printed to film.

In the late 90s J Chavez came out with Osmond PC for the Mac and that served my needs. Mr. Chavez even rescued a crashed file for me. I don't think he's continued to keep it current. It does multi-layer and outputs Gerber files. I never inputed net lists since my boards were fairly simple audio boards and I wanted to control trace placement.

Even though I'm trying to be retired I still use it on occasion. I just have to keep it on my current Macs.

 

[Mr. Moscode]

IECs - I would think it's not too hard to make FR4 board with the cutout for the IEC that could mount in a round hole chassis punch knock out. Probably an hour or 2 in PCB software. Doesn't even need to be round, just big enough to cover the hole with wings for mounting screws.

 

[uplinkal]

[we were making controlled impedance traces on PCB material with Xacto knives and peeling up unwanted copper with pliers] -- Been there.....done that!!!

[11 layers laid out] -- 11-layers??? Really??? And, you got away with that with nothing happening during the fabrication? I thought that the "standard" was -- an EVEN number of layers --!!!

[by hand with 4X tape] -- My layouts were either 2X or 4X depending upon either the size of the PCB (large boards were 2X, while small boards were 4X) or due to its features (as in needing to use 10-mil tracks). I had to hand-tape a 16-layer board for a defense contractor that were going to be used within U.S. fighter jets for the early -- IFF -- (Interrogated Friend or Foe) system.

[and "dolls", pre-printed sticky pads for ICs and other multi-pinned things] -- "Bishop Graphics", right??? I even still have my nearly 3-inch thick "Bishop Graphics" PCB Footprint Book lost in a closet around here somewhere.

[The layout tech I worked with took about 12 weeks to complete the full tapeout on that board] -- I worked at AMPEX in Redwood City, CA back in 1980 and I had met the guy who hand-designed the motherboard for the AMPEX M1200 tape deck. He told me that it took him -- 1-YEAR -- to layout that board and get it into production!!!

[contained a layer that hadn't been etched!] -- Back in the early 1990's as PCB-design was entering the phase of being done on computers, not only did you have to supply a set of GERBER files, but before the days of the now "standard" -- RS-274X -- file format, you had to also provide the PCB-fabricator with a text-code file containing all of your aperture designations and settings. That GERBER file format was called "RS-274D". At that point-in-time, I was working with an "RF" communications company that designed and built specialized "RF" communications gear for all sorts of "unknown and covert" 3-letter U.S. Government agencies. One time, a PCB-fabricator technician entered in a typo when transcribing a GERBER-file "Aperture Code" into their photoplotting equipment. That typo ended up rotating the pads of the IC chips 90-degrees, so instead of the pads looking like a "ladder", they looked like "skid-marks" and were basically shorting all of the IC pads together!!! The PCB-fabricator had to "eat" the cost of that failed batch of PCB's.

When the industry finally got around to using the -- RS-274X -- file format that contained "embedded" aperture data, that was such a relief!!!

[boggles my mind that a man could lay out 11 layers] -- As mentioned earlier, try 16-layers and with "Split Power & Ground" layers as well. At the end of each day when hand-taping a PCB back then, I would always have pieces of the crepe tape used for the tracks stuck all over my arms and shirt!!!

[the most patient man in the world] -- Many times during those early years of -- hand-taping -- PCB's, I heard people make comments something like, "To be a PCB Designer, you need to have the patience of Job"!!!

Sorry for the "thread clutter noise".

/

Jerry, if you are there, any news on my Gately SM6 mixer. It's been nearly a year since I heard from you.

 

[DaveP]

While you have been down memory lane, I have at last got this amp working but it has an oscillation problem.

These are the voltages on each channel as it stands and it passes audio with no load connected until I wind up the input, then it breaks into oscillation. It does not like an 8 ohm load either as that causes oscillation.

I have been working on the board out of the chassis so maybe that doesn't help?
Also the heat sinks are not earthed at the moment, should they be?
Any advice appreciated.
best
DaveP

 

[JohnRoberts]

welcome back...

That looks like more than 200mA of class A current, high in my judgement. Once again the class A currents aren't the same suggesting there is some 45mA going somewhere else and unaccounted for.

(19.45 -19.4)/.22 =227 mA in R13, (19.4- 19.36)/.22 =181mA in R12 , 45 mA is leaving the output node through a different path.
===
I am unsure about the values of c5 220nF across base-emitter of TR2 and unmarked 100pf from collector to base of TR2.

Instability or oscillation in an amplifier with negative feedback is when the delay in the forward path results in 180' of phase shift to the negative feedback signal, flipping it to positive feedback, before the open loop gain drops below unity gain so it can't support oscillation.

The 100pF capacitor rolls off open loop gain and more capacitance there should improve stability margin. C5 looks like it is adding lag in the forward path and perhaps working against stability. Confirm that these values are an exact copy of a working design. The unnumbered 100pf may have been added to stabilize the original design if it wasn't stable.

That class A current sounds wrong and I would try to figure out where the missing 45mA is going. You could confirm that c9 is not leaking by measuring for DC voltage across the speaker. Also maybe measure DC voltage at C8 should be 19.4.

good luck.


JR

 

[Audio1Man]

Hi Dave

I have not tracked all posts; however, I would add local PS BYPASS ON THE PCB, .1uf min.
Check if stable, temp increase the 100pf cap on TR2 to 500pf and see if is stable.

Duke

 

[DaveP]

That looks like more than 200mA of class A current, high in my judgement. Once again the class A currents aren't the same suggesting there is some 45mA going somewhere else and unaccounted for.

The voltages I gave are not exact for the emitter resistors, in order to set the quiescent current, I have fed the amp with a series resistor of 100 ohms.
The driver TR2 draws 49mA so I set the output stage via a difference with 6.4V across the 100 ohms, thus 64mA minus 49mA = 15mA.

What I don't understand is why it was not oscillating until I changed the circuit by adding the base stoppers and the circuit around TR3.

The poles have the following values:- R7/C4 =33.8kHz, R6/C5 =1.5kHz, R14/C8 =338kHz, R11/C6 =15.9MHz. This was supposed to be a simple circuit!

I will carefully try a few things to see if I can find the problem without smoking the devices. Any ideas welcome.
best
DaveP

 

[DaveP]

OK, I have made some progress!
Audioman's suggestion of a 0.1uF on the power supply stops the oscillation. Thank you and i would like to know why this works.
best
DaveP

 

[JohnRoberts]

OK, I have made some progress!
Audioman's suggestion of a 0.1uF on the power supply stops the oscillation. Thank you and i would like to know why this works.
best
DaveP

PSRR (power supply rejection ratio).... When the output stage draws current, inductance in the PS feed results in a voltage drop. Poor power supply rejection of the crude amplifier circuit allows that PS node changing voltage to feed back into the input. The 0.1uF ceramic disc makes the PS rail a low impedance at HF preventing the feedback coming through the PS rail.

The voltages I gave are not exact for the emitter resistors, in order to set the quiescent current, I have fed the amp with a series resistor of 100 ohms.
The driver TR2 draws 49mA so I set the output stage via a difference with 6.4V across the 100 ohms, thus 64mA minus 49mA = 15mA.

I am not sure I understand this logic...To make an accurate measurement of class A current the best measure is the voltage drop across the 0.22 emitter degeneration resistors (at rest with 0V AC audio input). That resistor voltage divided by 0.22 will tell you the class A current. Ideally the drop across both emitter resistors should be the same.

JR

 

[DaveP]

I am not sure I understand this logic...To make an accurate measurement of class A current the best measure is the voltage drop across the 0.22 emitter degeneration resistors (at rest with 0V AC audio input). That resistor voltage divided by 0.22 will tell you the class A current. Ideally the drop across both emitter resistors should be the same.

The 0.22 resistors are only 10% components and the voltage across them is very small and it varies quite a lot making it difficult to decide on the actual value. This is why I measured it the way I described above.
best
DaveP

 

[DaveP]

I had this going OK without a load connected, but with an 8 ohm load there is still waveform distortion and oscillation.

I tried to find the source of the problem but it ended up in smoke for the third time and I've had enough.

Thank you for your time and helpful comments but this design is far too fragile for me.

I think the fundamental problem is probably my layout with some obscure capacitance or inductance in the wiring causing the problem. It actually worked as a copy of the original circuit and it drove 9W into 8 ohms, but since adding all the safety features it can't bear a load being attached, so its time to call it a day.
best
DaveP

 

[JohnRoberts]

With a speaker load connected you need to consider the current flow coming back into the ground from the speaker. Any resistance in the ground path can cause voltage drops. If these signal related voltages find their way back into the amplifier inputs it can affect stability.

Have fun

JR

 

[DaveP]

For completeness, This is a successful 90W amp I made a few years back, with no problems.

The layout was very similar.


best
DaveP

 

[ruffrecords]

Topologically that is almost identical to the one you have just built. The only obvious addition is C4, the 100pF from collector to base of T2. Might be worth adding it to the present design to see if it tames the oscillation. By the way, what is the frequency of oscillation?

Cheers

Ian

 

[DaveP]

Topologically that is almost identical to the one you have just built. The only obvious addition is C4, the 100pF from collector to base of T2. Might be worth adding it to the present design to see if it tames the oscillation. By the way, what is the frequency of oscillation?

The waveform is like a sinewave with the centre chopped out and replaced by a mega Hz section, it was too high for me to measure the frequency.

The main difference is that the transistors are all high voltage/power types. Also note that the zobel network is after the output cap, not before.
best
DaveP