There s none that can be removed.
At best, you could remove the four mosfets used as cascodes,
but then, the fets of the differential just standing a mere 25V,
you would have to stick to a lower voltage supply not exceding +-25V..
I have studied this design; it is artful, and while it seems complex, I don't believe it could be done any simpler. Input stage loads are all resistive; VAS is Lender connected, simple; output stage is EF, conventional. It has low loop gain, courtesy no CCS loads and FET VAS, very low high order distortion artefacts, and Bode plot reveals second pole at 1.5MHz with phase shift around -120 degrees. These are outstanding figures for a low feedback design with no visible compensation, and soft clip is a bonus I did not expect.
Furthermore, this is a scaleable design. Nigel, I think your comments are not justified. If I didn't have a SS zero global fb design of my own, I'd be tempted to speak with PMA for rights.
Cheers,
Hugh
Furthermore, this is a scaleable design. Nigel, I think your comments are not justified. If I didn't have a SS zero global fb design of my own, I'd be tempted to speak with PMA for rights.
Cheers,
Hugh
I was thinking you could get rid of the P side of the input stage, but that wouldn't get you much by the time you stuck all the current sources back in. Maybe Nigel would use resistors.
I wonder about the DC servo output taking 10% of the gain set divider current. C11 and the quality of the opamp are the only things that keep the output still under load in the audio band, so I guess you'd have to at least somewhat like the sound of the opamp too.
I wonder about the DC servo output taking 10% of the gain set divider current. C11 and the quality of the opamp are the only things that keep the output still under load in the audio band, so I guess you'd have to at least somewhat like the sound of the opamp too.
My simulator says that it has an open loop gain of about 94db,
with a -3db point at 3.7Khz..
I did use the 2SA1943/SC5200 as output pairs, but those
can only influence the high frequency roll off.
Although i find that simpler is not the better , i m not a fan
of exagerately complex designs.
VAS buffer would indeed be useless if not preceeded by
a current mirror , since the gain is limted by the differential
resistors that goes to the supply line, not by the Vas input
impedance.
Of course, as you point it , such radical changes would
undoubtly affect stability in a large extent with surely only
marginal gains in respect of the added complexity..
Wahab,
I was talking of LOOP GAIN and the HF pole, the one so important to HF stability (<-180 degrees), not the dominant pole at 3.7KHz, which is about right for -45 degrees phase shift. In the early pages, PMA published the loop gain figures; I used them. Please reread what I have said; my meaning should be very clear.
One could criticise this design for excessive phase shift at 100KHz, but it's probably academic....
I was talking of LOOP GAIN and the HF pole, the one so important to HF stability (<-180 degrees), not the dominant pole at 3.7KHz, which is about right for -45 degrees phase shift. In the early pages, PMA published the loop gain figures; I used them. Please reread what I have said; my meaning should be very clear.
One could criticise this design for excessive phase shift at 100KHz, but it's probably academic....
Ok ,Hugh
Some misunderstanding with no consequences..
I didn t go too far on the sims, since i suspect that
there s some innacurracy in the models i have for
the IRFs , particularly the Cgs and Cgd in dynamic
conditions...
Dear philmagnotta,
the most expensive components are toroidal transformers, case with heatsinks, double-sided PCBs, PSU capacitors, and output transistors.
In case we speak about the PA4, i.e. dual mono version with 4 output pairs per channel, prices approx. are:
- 80 USD per 1pc toroidal transformer (when produced here in the Czech Republic),
- 250 USD case with heatsinks,
- 220 USD all the PCBs.
I would roughly estimate the material costs at some 900 USD.
Best regards,
Hi Bonsai - yes, but the difference in SOA between MJLs 21193/4 and 1302/3281 is quite small (please have a look at datasheets of both). There is quite a big difference in SOA when the 1302/3281 is a Toshiba 2SA/2SC. I have chosen MJL1302/3281 for the higher power version with 4 output pairs.
Thanks, wahab. The CCIF IMD 19+20kHz is about 3-4 times lower with the faster devices.
pavel:
kudos on a very nice design; i really appreciate all the performance pictures (as opposed to simulation plots
).
i agree with an earlier poster that clipping looks pretty good.
i am curious why use the 2SK413/2SJ118 mosfets? i'm looking for insight into the rationale behind post 43:
http://www.diyaudio.com/forums/solid-state/155628-solid-solid-state-power-amplifier-5.html#post1994138
sorry if you stated why earlier - i must have missed it. it's been a while since i looked at the data sheets for these - i think i remember compared to international rectifier hexfet parts, the 2SK413/2SJ118 have much lower capacitances & compared to hitachi/renasas laterals, the 2SK413/2SJ118 might have a little more transconductance at the operating currents you are using and don't have the zero tempco at ~100mA? i likely can't get 2SK413/2SJ118, but do have some new old stock supertex mosfets that may work well in this position.
mlloyd1
kudos on a very nice design; i really appreciate all the performance pictures (as opposed to simulation plots
). i agree with an earlier poster that clipping looks pretty good.
i am curious why use the 2SK413/2SJ118 mosfets? i'm looking for insight into the rationale behind post 43:
http://www.diyaudio.com/forums/solid-state/155628-solid-solid-state-power-amplifier-5.html#post1994138
sorry if you stated why earlier - i must have missed it. it's been a while since i looked at the data sheets for these - i think i remember compared to international rectifier hexfet parts, the 2SK413/2SJ118 have much lower capacitances & compared to hitachi/renasas laterals, the 2SK413/2SJ118 might have a little more transconductance at the operating currents you are using and don't have the zero tempco at ~100mA? i likely can't get 2SK413/2SJ118, but do have some new old stock supertex mosfets that may work well in this position.
mlloyd1
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.