There is no need for drain or source resistors when using LATFETs.
In 1982 I designed LATFET-amps with 18 paralleled output pairs and no source resistors at al.
In 1982 I designed LATFET-amps with 18 paralleled output pairs and no source resistors at al.
Just as an interest, this amp power is fully regulated by series pass transistors, there is also a microprocessor board for both amplifier and speaker protection. The Transformer is an 800VA toroidal transformer that I designed and my company built it for me as I wanted it. This was a complete project that I worked on for about 1 year, everyday, besides the little MD duties being nice to customers.
There is something more about emitter degeneration.
Besides local feedback that linearizes forward transconduction by reducing it,
this reduced forward transconduction is beneficial for increased stability as the bandwidth of this stage is increased
shifting the corresponding pole to higher frequencies.
I'd really appreciate this knowledge would be more widespread.
The pre=amp is a symmetrical dual differential topology developed in the late eighties. The problem however is that the first stage would determine to actual THD and the power amp will amplify the pre-amp distortion and then add its own. It is a math series. Same goes for noise. The reason I could possible not hear any difference is that the preamp distortion is far worse. When I used the power amp directly from my DAC the difference really became very noticeable both DAC and amp having very similar THD and IMD. I made some discreet op amps for the DAC later years.
laterals are somewhat self-limiting and if the N to N and P to P match are "close enough", you can get away with no source resistor. Haflers did it for years.
i like your design; kind of a Hitachi databook circuit with extras. also nice that you use "junk box" parts; nothing really exotic.
but I'll ask a few questions anyway (not meaning to cause any grief):
- current mirror in the diff amp input stage + current mirror in the VAS? what helps define the voltage and currents? many folks trade off some of that very high open loop gain for quiescent DC stability but i don't see anything like a common mode feedback loop here. maybe you're using the mismatches to help with that?
- i know you said your data was taken prior to clipping, but how is the clipping recovery with 2-pole compensation approach?
- are you really comfortable with 55V rails with MPSA06 and MPSA56 rated 80V?
- you are running the diff amp currents unusually high; any reason you didn't use a cascoded, high GM jfet and get some extra RF immunity?
- just in case, it might be good to put some small resistance in series with Q5 base to eliminate possibility of oscillation
- i'm curious why you chose the current mirrors you used vs, for example, the 3 transister "helpered" mirror that cordell often uses?
I thank you guys for contributions so far, it lets my brain do a little work to keep up. Don't worry I red very slow. Nice points M Lloyd there will be a continuation in the development theoretically since I have now completed the old amp updates. I may have to make new PCBs later, but I will leave it as is.
I would like to test one suggestion at a time since a bundle of changes all at once can get you completely lost. If you look closely you will see the original Hitachi amp in it. That is exactly where it started late 80s
I would like to test one suggestion at a time since a bundle of changes all at once can get you completely lost. If you look closely you will see the original Hitachi amp in it. That is exactly where it started late 80s
It should have very good THD, providing it is stable into 1 ohm. This is because any phase shift caused in the speaker may look like a very low load for an instant, but it is all it takes to cause instability if the compensation caps are too small.
H
HAYK
Distorsion is very low on your sims because due to C2 there s no actual AC degeneration in the IPS whose standing current,
and hence transconductance, is huge, and also because of R61/62 as well as C24/26 and C7/C25 being exageratly low .
You could as well put two capacitors in parralel with R63/64 and the result would be the same, that is, degeneration for DC
but not for for AC.
Unity gain and unity loop gain frequencies are shifted at a much higher frequency by the non damped C23 huge value,
this is clearly visible in the frequency response, so rather than improving stability this capacitor decrease the gain
and phase margins.
As a consequence this amp is eventually very marginaly stable because of the output inductance that is not even damped
with a resistance in parralel, but there s no doubt that the slightest capacitive loading would morph it in some RF
AM transmitter.
The IPS with CMs will not define precisely the VAS standing current, this can be partially solved by setting
an adequate voltage drop in R13/14 or dides on the other side of the CM, but there s other issues here.
There are other things that are not satisfactory but the ones i mentioned are the most critical,
i would also mention C28 which short R10 on AC mode and render the filtering useless, and the output RC network
R43/C12 wich is connected to the wrong side of the inductance as to be rendered useless a well.
Sorry to tell you, but that s really a lot of missteps, indeed a pleasantly looking schematic has never been synonymous
of a well functioning circuitry.
and hence transconductance, is huge, and also because of R61/62 as well as C24/26 and C7/C25 being exageratly low .
You could as well put two capacitors in parralel with R63/64 and the result would be the same, that is, degeneration for DC
but not for for AC.
Unity gain and unity loop gain frequencies are shifted at a much higher frequency by the non damped C23 huge value,
this is clearly visible in the frequency response, so rather than improving stability this capacitor decrease the gain
and phase margins.
As a consequence this amp is eventually very marginaly stable because of the output inductance that is not even damped
with a resistance in parralel, but there s no doubt that the slightest capacitive loading would morph it in some RF
AM transmitter.
The IPS with CMs will not define precisely the VAS standing current, this can be partially solved by setting
an adequate voltage drop in R13/14 or dides on the other side of the CM, but there s other issues here.
There are other things that are not satisfactory but the ones i mentioned are the most critical,
i would also mention C28 which short R10 on AC mode and render the filtering useless, and the output RC network
R43/C12 wich is connected to the wrong side of the inductance as to be rendered useless a well.
Sorry to tell you, but that s really a lot of missteps, indeed a pleasantly looking schematic has never been synonymous
of a well functioning circuitry.
I simply copied the amps that I made in around 2002 into microcap and the started optimising each of the values in microCAP it took a very long time because optimising on value throws other out and the process was performed many times over a period of a week. I have now replaced the values in the real amps and they are working and I closed up the amp and I have been listen to it since bout 3 a.m. my time. They are running hotter than before because they have been biased higher. I will wait for the room temp to increase to about 30 deg later today and take the temp readings on the heat sinks just for good measure and set the temp switch for the fans to activate.
Thanks HYAK, unfortunately this amp you offered is completely different and I cannot modify an existing amplifier such as this on the PCBs, but I am halfway through capturing the schematic into MicroCAP and once simulated and seeing if there is some improvement, I might lay out a PCB and test listen to it.
Hi WAHAB, I beg to differ with some of your claims. The amp is used on apogee scintillas which has a nominal load of I ohm, and can dip to less than 0.3 ohms. At the time I bought them they were rated the best available at any price. The amp has been driving them for more that 20 years, it shows absolutely no instability nor strain. Your phase margin has to extend unity gain frequency by a fair margin else you will have instability. This phase margin does not dip at all in the simulation, that I have never come across and would check it and what is responsible for this. In reality I have no output inductor in the actual amp, never bothered since it was unconditionally stable when I first made it. I am not arguing the theoretical but the actual. Furthermore, when I was still working and had a full R&D lab available, I never saw any problems with it before putting it in service 20 years ago. I see no problem with your theories either and would try these under simulation but there is nothing really that needs solving. Whether in practice it is two orders of magnitude worse than what MicroCAP suggests. I will post the simulations of the original amp without optimisation in the next post. At the moment I am not impressed or otherwise, it sounds the same as it did before. I have not tried it on the KEF Concerto or the Rodgers LS3/5a because of mobility problems and need someone to move the equipment into my listening environment for me. Unfortunately the scintillas are monsters to move because of their size.
Last edited:
I cannot remember the exact date (year) I did this simulation, but the original design (a little simpler) was commercialised and marketed by a boutique brand still present, maybe with the mods you suggested. It obviously does not look as crude as my creation as they probably employed some industrial artist to detail their product aesthetics.