There are plenty of good parts readily available. Good transistors like BC550, BC560 for IPS, KSC3503, 1381 for predriver and MJL1302/3281. Those have very good characteristics. I am not talking about getting the any transistors. These have good beta characteristics and low capacitance. Do you think you can make distinct improvement using "the" transistors?
Right, exactly. Those are "the" transistors. There are also plenty of transistors available that can withstand the voltages and currents, but will result in a poorer-performing amplifier due to the double-whammy of higher open-loop distortion (primarily due to beta vs Ic characteristics) and lower loop gain (due to higher capacitance and lower fT). So, you do need to be careful with transistor selection and not just use any old thing.
The topics of conversation is about not able to find "the" transistor and has to include SM parts. Those that I mentioned are so common and readily available from everywhere. Too many posts I've seen people looking for one obsolete transistor. There got to be ones that are readily available that is just as good.
There seems to be a lot of snake oil effect to this. I was told BC550 is better than KSA992. I looked at the data sheet, I sincerely doubt people can tell the difference using in the IPS circuit. Only reason I went with the BC5xx is because I like the pin out better, that the base and collector is not reversed that make layout slightly more difficult. So if one day the BC5xx are no longer available, I won't think twice to use the KSA and KSC. I sure would not get the SM version of BC5xx even if they were available.
It looks like I did not put it clearly. Off course OPS should be simulated separately to get the best of it, but what I tried to explain that most important part is how much of open loop bandwidth you get in the IPS+VAS and that will correct the OPS distortion though the GNFB. To achieve this most important parts are transistors used, type of compensation, and circuitry configuration. For the VFA symmetry is not important and I agree with D. Self in that, but for the CFA symmetry is so natural and easy to achieve. Surprisingly even compensation is easy to implement to get more than 80 dB of the loop gain(negative feedback) at 20 kHz. It is not easy to find good source of high voltage small transistors, and specially those for VAS/TIS. If the power supply voltage is higher than +-40 V than there is not to much choice and good source for not fake ones.
You can see an amp or more of base currents into the output devices. The driver has to support that, etc backwards. These base currents are often non-linear.
Even if the final current draw from the Vas is uamps, if this is non-linear and the Vas has k's of Zout, you end up with non-linear Vas output.
Just suggested to clear this possibility by inserting an ideal buffer between Vas and OPS in the sim and see what happens to the THD.
Edit: another check would be where the distortion is 10 x higher - at the output only, or also at the Vas?
Jan
Last edited:
That's where design come into play. Say for +/-80V rail, you can use cascode to split the voltage so each transistor doesn't see the full voltage. Only the output of the VAS can see the full swing, the rest of the IPS don't swing much.
Worry about capacitance change of the transistor, again cascode fix a lot as it keeps the collector of the main transistor constant and the Ccb stay constant. It's also explained in Mr. Cordell's book using cascode at the output of the VAS really lowed the distortion as I explained above. So you can use a low voltage transistor as output of VAS, then use a high voltage cascode transistor that does not have as good characteristics for the cascode.
So it's the design, design to cover any short coming of the components.
2sa2222
Some specs are superb:
10A, 25W, 119pF Cob, 200MHz fT
But the knee in the SOA starts at 7V ! and 50Vce0,
The fT only exceeds 100MHz for Ic >60mA
The 2sc is even better in some respects, fT and Cob.
A high bias ClassAB, or a ClassA, headphone amplifier.
Last edited:
I agree, that is good advice. The support one has in a day job makes all the difference. I think going to SMT, the biggest advantage is real estate and reduced cost of mass production. It also has big advantages for RF designs, at least in the Gigahertz range.
Cheers,
Bob
Hi Harry,
Thanks for this feedback. For IPS, one of the inevitable tradeoffs is voltage rating vs. suitability for IPS. This is usually especially the case for JFETs and dual monolithic BJTs. The answer then is often cascoding of the IPS, which, although an addition to complexity, can improve overall performance not only by the usual benefits of cascoding, but also by allowing the use of more suitable, lower-voltage IPS devices. If one wants to get slightly fancier with the cascode, one can use a "driven" cascode, where the cascode bases are driven by a replica of the feedback signal to further improve common-mode issues. Of course, the cascode bases can also be driven by a buffered and level-shifted version of the LTP tail.
BTW, one useful exception to the voltage tradeoff for amplifiers of modest power is the LSK489 dual monolithic N-channel JFET, which has a voltage rating of 60V.
Cheers,
Bob
IMO --- unless I have a severe space problem.... which i never will in audio .... I'm sticking to TH parts only.
-RNM
I tend to agree, but parts availability for TH parts is getting a little more difficult - at least for some transistors and ICs.
Cheers,
Bob
Some of us are now considering self powered speakers where the amplification and network are now inside the enclosure and there the smaller footprint of surface mount devices mixed with higher power output devices becomes a very useful goal. Board size becomes an issue to think about. Internal volume is limited and keeping everything as small as is reasonable becomes part of the goal.
Sorry but I do not think this is good advise for the DIY community. It maybe good advise as a professional designer. You are scarring other DIY members from doing smt because you are lacking the necessary skills and or equipment. I just finished building a couple of TPA3116 amps "Weiner". See the group buy section. Gary is a very skilled EE, that I have learnt somethings from. It took me a good part of a day to solder up two weiner's. It used some 0402 parts. I used the big eye mag and my jewelers loupe for inspection when I soldered up the SSOP. Both worked first time, so it is entirely possible from my experience. If your design is 100% perfect, which is what we are to strive for, then you may be lucky and not have to do any rework. Okay i am not talking about a 1152 bga, 4-pi xray, that is for the pro's. Of course you need a hot plate or hot air equipment, for bga but for the majority of smt packages, some simple tools is all that one needs, like two fine tip irons to start with. Sparkfun sells a decent Haikko rework machine clone.
Gary taught us all to put the pcb in a skillet/fry pan, put on the hot stove top and remove a ssop pkg. How easy is that?
The move in the industry is to leadless, DFN/QFN. Why? saves more space, leads take up space. I am able to do DFN by hand, but I have years of experience because I take them on and learn. Not to say they ain't a PITA.
I have yet to need to use a stencil/paste/reflow oven for my smt work. Although they have there merits and I am not going to invest into a pick 'n place for my low volume DIY projects.
Sometimes the device you want to use is only available in one package, so if you want to use it you are forced to learn how to process the part. I does not have to be expensive.
I do not mind helping out others learn to use smt. That is why we are all here. Sounds like Bob C. chose not to learn or hone his smt soldering skills, on the job, rather let someone else do it for him
As a company(manager), that was probably the best use of his time. I remember at the Motorola factory, no techs/EE's were allowed to rework any customer product.I am very lucky in this sense, since I am myopic, so I can pick fly dung out of pepper or read the values on a 0805 by eye. Older people really do need good optical aids.
All for smt!!
Happy soldering
Rick
Last edited:
Gathering from several of the above post, is it safe to say that virtually no improvement can be made to most of the current OPS's that are on the forum? Just use the IPS to correct the short comings of the OPS, including or discarding the cross over distortion. I am hoping out loud that OStripper and Bob Cordell and especially Dadod would pursue with their collective great and vast expertise something that closely approaches the outstanding IPS's in and by the way of architectures and such they have shared. I know it is a tough cookie to attack, there are a lot more complex variables to address. Multiple devices, each with slightly different characteristics trying their best to work in unison to produce a excellent sound. Just like the members of the forum.
Gathering from several of the above post, is it safe to say that virtually no improvement can be made to most of the current OPS's that are on the forum? Just use the IPS to correct the short comings of the OPS, including or discarding the cross over distortion. I am hoping out loud that OStripper and Bob Cordell and especially Dadod would pursue with their collective great and vast expertise something that closely approaches the outstanding IPS's in and by the way of architectures and such they have shared. I know it is a tough cookie to attack, there are a lot more complex variables to address. Multiple devices, each with slightly different characteristics trying their best to work in unison to produce a excellent sound. Just like the members of the forum.
I was going to move on to the EC OPS. I have both E Stuart's and Bob's
EC simulating correctly. E. Stuart's is both EC AND AB2.
But
- They are "picky" about what IPS is used. It seems that a custom
IPS is needed to derive the best performance from the EC.
- What are my goals ? Better SQ and longevity . The EF3 can give
me both. I've listened and listened , my OEM's are already quite outclassed.
What more do I want , to just see 1/2ppm on a readout or bragging
rights ?
_ So I decided to further "harden" the EF3 and devote more time on
a better IPS. I'm already able to have quite the class A region with
my 5 pair. I quite sure I'm actually hearing single digit PPM at a
normal listening level. What else would I want ? - oh , a better
DAC (source).
OS
Sorry but I do not think this is good advise for the DIY community. It maybe good advise as a professional designer. You are scarring other DIY members from doing smt because you are lacking the necessary skills and or equipment. I just finished building a couple of TPA3116 amps "Weiner". See the group buy section. Gary is a very skilled EE, that I have learnt somethings from. It took me a good part of a day to solder up two weiner's. It used some 0402 parts. I used the big eye mag and my jewelers loupe for inspection when I soldered up the SSOP. Both worked first time, so it is entirely possible from my experience. If your design is 100% perfect, which is what we are to strive for, then you may be lucky and not have to do any rework. Okay i am not talking about a 1152 bga, 4-pi xray, that is for the pro's. Of course you need a hot plate or hot air equipment, for bga but for the majority of smt packages, some simple tools is all that one needs, like two fine tip irons to start with. Sparkfun sells a decent Haikko rework machine clone.
Gary taught us all to put the pcb in a skillet/fry pan, put on the hot stove top and remove a ssop pkg. How easy is that?
The move in the industry is to leadless, DFN/QFN. Why? saves more space, leads take up space. I am able to do DFN by hand, but I have years of experience because I take them on and learn. Not to say they ain't a PITA.
I have yet to need to use a stencil/paste/reflow oven for my smt work. Although they have there merits and I am not going to invest into a pick 'n place for my low volume DIY projects.
Sometimes the device you want to use is only available in one package, so if you want to use it you are forced to learn how to process the part. I does not have to be expensive.
I do not mind helping out others learn to use smt. That is why we are all here. Sounds like Bob C. chose not to learn or hone his smt soldering skills, on the job, rather let someone else do it for him
I am very lucky in this sense, since I am myopic, so I can pick fly dung out of pepper or read the values on a 0805 by eye. Older people really do need good optical aids.
All for smt!!
Happy soldering
Rick
Hi Rick,
You are right - my eyesight is terrible, whether I wear glasses or not. Your points are very well-taken, however. I have not done SMT DIY, but I would do it with some of the larger SMT components like 0805. A mixed SMT/TH design is also an option. I agree, with SMT we really want to get it right the first time. Even so with TH PCB. I always recommend doing a prototype on perfboard that has a similar component placement and wiring as the PCB is to have. At that point we can still find a lot of mistakes and do tweaks easily. ICs that are unavailable in TH can be put on perfboards with adapters.
Are you using metal film SMT resistors for highest audio quality?
Cheers,
Bob
The Athena loudspeakers that I did about 8 years ago were self-powered - see my web site. Space was an issue, but getting the heat out was the biggest issue. Each cabinet had four 125-watt MOSFET power amplifiers in it, plus a 3.5-way active crossover, active baffle step compensation and active EQSS LF frequency shaping (essentially a Linkwitz Transform).
Nowadays, most active speakers use class-D amplifiers, so the tradeoffs are different.
Cheers,
Bob
I was going to move on to the EC OPS. I have both E Stuart's and Bob's
EC simulating correctly. E. Stuart's is both EC AND AB2.
But
- They are "picky" about what IPS is used. It seems that a custom
IPS is needed to derive the best performance from the EC.
- What are my goals ? Better SQ and longevity . The EF3 can give
me both. I've listened and listened , my OEM's are already quite outclassed.
What more do I want , to just see 1/2ppm on a readout or bragging
rights ?
_ So I decided to further "harden" the EF3 and devote more time on
a better IPS. I'm already able to have quite the class A region with
my 5 pair. I quite sure I'm actually hearing single digit PPM at a
normal listening level. What else would I want ? - oh , a better
DAC (source).
OS
These are good points. How the IPS/VAS fits the output stage can be important.
When I did my MOSFET power amplifier with error correction, the main reason I used EC was because I wanted to get the benefits of the MOSFETs in the output stage while getting very low distortion by reducing the effects of transconductance droop in the MOSFET output stage. That way I was able to get 0.0006% THD at 20kHz while still using MOSFETs (and at a reasonable bias current of 150mA and one output pair).
Cheers,
Bob