Hi Mr. Cordell
That's the difference between a real job and DIY. In the real job, you have all the support and resources. I think nothing to switch over to SM. Particular when I was the manager, I have tech and engineers to do the dirty work. Now I have a temp contracting job from the old company and I am all by myself. I designed and layout the board already, I am seriously thinking about chicken out and quit the testing and rework when the board comes back!!!
But it's is not good to quit mid stream, I had to go through with it to completion. But I really don't want to deal with it in my old age. Particular I have 0402, SOT23-5 and a few components that have solder pads under the components, not BGAs but bad enough.
My advice is people really need to sit down, bread board with really SM parts to see. I am not talking about using 0603, just use 0805, SOIC8 type that's the largest components of SM. Experience it themselves first. I had to do breadboarding with 0603, it's painful. Try remove a few parts and replace parts on an old computer or cell phone pcb before they dive in.
Ha ha, try removing a SOIC8 and replace with a new one. Then come back and talk!!!
The question is still whether you really need "that" part for the design? Your book really showcase step by step how distortion be lowered on each step of improvement. It's all about the design. That's why I like your book so much. You arm people with theory and result so people know which way they want. To me, individual transistor is not even secondary importance, more like on the third or forth importance.
Last edited:
Wow Bob that’s so exciting! Congratulations!
I hope that I have a while before becoming a grandad (daughter is 5 1/2 and son is 3 3/4).
Glad to hear that the book is still somewhere there on the list of “things to do”. I’m really looking forward to edition II.
In terms of availability, it’s a matter of moments to check to see if major distributors that have low MOQs such as Farnell, Digikey or Mouser stock a part. If none of these do, the part may not be “unobtanium” but will be “verydifficulttoobtanium” for many readers.
For VAS/TIS and output stage pre-drivers the usual-suspect Fairchild parts KSA1220A, KSA1381, KSC2690A, KSC3503 are still available. At lower voltages, the 2SA2222SG/2SC6144SG pair from On Semi (formerly Sanyo parts) could make for a very fast BJT output stage.
In terms of IPS, I’d be interested in thoughts on monolithic bipolar duals such as the SSM2212 and SSM2220. A discussion on how to compare noise specifications wouldn’t go amiss as many transistors have a specified “Noise Figure” in dB whist parts such as the SSM2212 are specified in nV/sqrt(Hz)
SMT isn't as hard to work with as everyone makes it sound. My home etch stuff works great in a toaster oven. Most reworking short of BGA and devices with the large ground pad underneath can be done with a couple good standard irons. More miserable parts can be removed after a short bake cycle in a toaster over after heat sensitive parts are removed.
It's actually not hard if you don't have the silly notion to reuse the part. Just slide a small very hot iron under the side of the first leg and push. The legs will bend up one at a time until they are all disconnected.
I used to be afraid of anything smaller than 0805 but now I can do 0402 and fine-pitch TSSOP without the need for a microscope. I have recently soldered 0201 parts (ceramic capacitors) with the aid of a microscope.
For rework, a hot-air station is really a must. It’s also a good idea to get a 2 oz copper PCB if you think you’re going to be reworking a lot.
For rework, a hot-air station is really a must. It’s also a good idea to get a 2 oz copper PCB if you think you’re going to be reworking a lot.
Since I found the FZT657/757 , I might go all SMD for my VAS/LTP's.
Output stages , I already use outputs as drivers .... just the pre-drivers (FZT's).
Can't really avoid thru hole large decoupling , flyback diodes. 2W 2-10R's -cheap
thick film ..... the emitter resistors would have to be many in parallel . PITA.
Best just to keep the OPS thru hole.
I'm just a bit worried about my favorite 3503/1381's , E grade is already
“verydifficulttoobtanium”.
FZT's are available in all three of their grades.
OS
Output stages , I already use outputs as drivers .... just the pre-drivers (FZT's).
Can't really avoid thru hole large decoupling , flyback diodes. 2W 2-10R's -cheap
thick film ..... the emitter resistors would have to be many in parallel . PITA.
Best just to keep the OPS thru hole.
I'm just a bit worried about my favorite 3503/1381's , E grade is already
“verydifficulttoobtanium”.
FZT's are available in all three of their grades.OS
some really nice op amps of the recent decade or so simply aren't offered in DIP at all
50 mil lead soic spacing isn't a big problem by hand - I do use 0805 but haven't tried smaller
I've hand swapped a few parts even in tssop/ssoic - harder to solder individual pins but the flood and wipe method seems to be fine
can do hack job with aluminum foil mask and a hot air gun to remove fine pitch parts without cutting - but I often do just xacto the body free of the board and mop up the legs with solder wick
do need optics now that I've had LASIC and getting too old for much focus range
50 mil lead soic spacing isn't a big problem by hand - I do use 0805 but haven't tried smaller
I've hand swapped a few parts even in tssop/ssoic - harder to solder individual pins but the flood and wipe method seems to be fine
can do hack job with aluminum foil mask and a hot air gun to remove fine pitch parts without cutting - but I often do just xacto the body free of the board and mop up the legs with solder wick
do need optics now that I've had LASIC and getting too old for much focus range
But I think the ones for this forum should be available in DIP. The high speed ones should not even be in DIP. I just designed using Ti THS3201 1.8GHz CF amp. I don't even want to touch it even if it comes with DIP.
We use a different class of opamp here. We worry about distortion and maybe a little bit of noise.......really, only a little bit, signal is so large that noise is never even on my radar. We never need low bias current, low offset type. LM4562 pretty is the answer here.
SOIC is easier, good for those that find it not hard to deal with SM. I worked a lot with SM and I consider myself having good hand eye coordination and have very steady hands. I found it hard. It took so long to work on SMD parts.
Last edited:
Don't get me wrong, if designed and layout correctly, SMD always out performs TH. there is no if and buts about it. You eliminates all the lead length, parasitics inductance and capacitance, less area to be susceptible to noise pick up and radiate out. But these advantage is questionable at best when comes to audio that works barely to 1MHz ( and that's crazy already).
For work, I would not even consider TH, But I am at home, with a Weller station, a magnifying light only...........More importantly, I don't have technicians and engineers to kick around!!! At home, I only have my big boss............So I am the kickee..........Not the kicker.
For work, I would not even consider TH, But I am at home, with a Weller station, a magnifying light only...........More importantly, I don't have technicians and engineers to kick around!!! At home, I only have my big boss............So I am the kickee..........Not the kicker.
Last edited:
I think Audio DAC I/V is demanding enough that some of the smt only chips should offer real technical performance improvements - particularly the badly advertised linearized front end ADA489x series
I don't recall Bob showing any of the monolithic op amp front end audio power amp architectures with floating supplies, common source/emitter outputs so popular in touring sound reinforcement PA
I don't recall Bob showing any of the monolithic op amp front end audio power amp architectures with floating supplies, common source/emitter outputs so popular in touring sound reinforcement PA
Last edited:
Another strategy I used successfully is to carefully cut the pins off the body in situ with an Xacto knife. You can then pick off the body and remove each pin in turn from the board. This way you don't need to use a lot of heat and you the PCB tracking.
As to SMD, I find myself using selected SMD parts, for instance decoupling caps on opamps. It is very nice to use an SMD cap on the PCB backside directly on the pins, impossible to do with TH parts. Also things like gate stoppers; SMD directly on the pins on the back!
Jan
Last edited:
You can't look at an OPS in isolation if the amp use GNFB. More of GNFB you have, less THD you get, but all needed gain is coming from IPS, so it's quite important to have good parts there specially those used for VAS with low Cob. It is quite easy to get 80 dB of the Loop Gain at 20 kHz with TPC/TMC and OIC, and with this THD20k drops below 10 ppm even with cheap MOSFET like IRFP240/9240 in the OPS. No need for Error Correction inside the OPS. I came to conclusion that this goal is easier achieved with CFA.
Damir
I think I can. I did extensive simulations. I first did different configurations of IPS and VAS and GNFB from the output of the VAS back to the IPS with 20K resistor and look at the distortion. Then I use the exact IPS/VAS, add the 3EF OPS that has unity gain and drive a 4ohm resistor load and did the simulation. The distortion is at least 10 times higher. Since the OPS is 3EF, loading on the VAS is very low. The distortion has to be dominated by the OPS.
This could be easily verified by inserting an (ideal) buffer between Vas and OPS.
An effect often overlooked is that while the 3EF output voltage may be nice distortion free, the internal base currents may be quite nonlinear, and these nonlinear currents on the Vas Zout may cause distortion. Unless the (pre) drivers run heavily in class A and/or the Vas output is buffered.
Jan
I simulated with VAS idle at 10mA and it's a push-pull VAS. That's very high current for a 3EF OPS. The predriver I used were KSC3503 and 1381 which has low input capacitance and being EF, it's bootstrapped.
I thought both the pre-drivers and drivers on a typical EF3 were fully
class A ? What is "heavily" (most of the time ??).
I only see uA base currents on my EF3 ? How would this load the VAS ?
OS
EF3 usually has @ .05% THD. What dadod meant is with enough GNFB
(blameless w/TMC) you can get close to the IPS's loopback (standalone)
THD.
I've simulated (and ran) some of my IPS's with a class A OPS. 1ppm.
The TMC blameless can get close with an EF3 @ 5ppm.
Without some sort of EC (like syn08's PGP , or PAX) , you never eliminate
all of it. That last few PPM is the Xover distortion.
But why are we "chasing" this 1 ppm. 30ppm amps sound just fine.
Some are even preferable to the blameless (to me).That's what counts !
OS
(blameless w/TMC) you can get close to the IPS's loopback (standalone)
THD.
I've simulated (and ran) some of my IPS's with a class A OPS. 1ppm.
The TMC blameless can get close with an EF3 @ 5ppm.
Without some sort of EC (like syn08's PGP , or PAX) , you never eliminate
all of it. That last few PPM is the Xover distortion.
But why are we "chasing" this 1 ppm. 30ppm amps sound just fine.
Some are even preferable to the blameless (to me).That's what counts !
OS
The characteristics of the individual transistors can be very important. Especially for the output stage, but the performance of all three stages of your typical three-stage “IPS + VAS/TIS + Output stage” amplifier can be influenced by the transistor being used. Assuming bipolar devices throughout:
In the OPS, selecting devices that minimise “beta droop” will significantly improve distortion.
In the OPS, selecting devices with low parasitic capacitance and high Ft will allow for an increase in unity loop-gain frequency (ULGF), increasing audio-band loop gain and reducing distortion.
In the VAS/TIS, devices with low Cob help to reduce distortion as their non-linear Cob will be swamped by the compensation capacitance and therefore have less of an affect on amplifier performance. Low Cob devices in the VAS/TIS also help to make the stage faster allowing for higher ULGF.
For the IPS, devices with “flat” beta vs Ic will avoid creating distortion that can be caused by drawing non-linear currents through the impedances seen at the amplifier inputs.