The original Hitachi was 20 V/uS or better. The BCV 61 can have emitter balancing resistors , BCV 61 is better than hand selected mirrors.The second current mirror can be 1N4007.
The 10 K+ 5n6 is selected at 50 kHz for least distortion.
9 mA is more than enough. The amp will dip above 20 kHz full power . At 5 watts it won't .
It will drive 2 R. The 4700 uF can be what you want.
The under driving the FET's is deliberate.
Exicon 10 N/P 20.
No tail current source is a choice.
High gain is a choice.
The power fets, either lateral or vertical have a really nasty nonlinear drive requirement as they start conducting. There is a point where the cap of the junction is increasing and amplified by the miller effect or some such and you need a bunch of chage to push to get it moving. On the Spectral DMA 100 and DMA 50 I had to add FET drivers to have enough drive to isolate that transistion from the VAS. Hitachi ignored it and I think that was a problem. Goldmund did the drive using bipolar devices. The Mimesis is very similar to a DMA50 only with more bipolars where I used FET devices.
A quick check for the first 10,000uF 50V that comes up at Farnell reveals 5nH ESL and 25 milliohms ESR.
A 6 inch 2mm dia wire has about 100nH L.
5nH at 20kHz is about 0.65 milliohms, the 6 inch wire inductance at 20kHz is about 13 milliohms.
Perspective, guys!
Jan
Jan:
You are right about perspective and the need to look at the system. The self inductance of the wire in free space is different from that of two conductors relatively close to each other. My real point was to figure out how much current is needed for the load and if the bigger cap would make a difference. I think a lot of the effort on big caps is not producing the desired results since its focused on the wrong problem. Stick a current probe on a speaker cable and measure the peak currents. Anything else is conjecture based on imperfect and possibly incorrect models.
Demian
True, but anything that gets 'returned' to the xformer HAS to go through one or two diodes. And these are cut off 80% or more of the time, when the secondary AC is below the cap value. Only when the ac peaks above the cap voltage, do the diodes open up to top up the cap. Only then, and it's only a small fraction of the mains cycle, can any current 'return' to the xformer.
Surely this is basic stuff?
jan
Can't disagree with that Demian. There is a way that the cap size might impact bass reproduction. If the cap is so small that it discharges appreciably during an extended bass note, the ripple valley may become so low that the bass note may be modulated with ripple. I don't know if that happens in reality, never measured it. Should be easy to sim though.
Jan
That's one of the reasons that large MOSFETs are so successful in class A. They are always on and above the Cgs step.
jan
Actually, my PSU decoupling on the board, and after the big value calps, is 100uF // 3.3 uF // 100 nF // 1R+220nF next to the output section.
1R+220 nF to GND gets rid of stray inductance quite well, however, the values given apply to two Fisher & Tausche 10,000 uF/63V caps in parallel. For other caps and other cap arrangements, this will vary.
What will not vary is the effect once you get it right. The high range "opens up", becomes as fluid as the electronics can manage, no harshness, no spitting. Works well as an afterthought in standard commercial products as well, I've done it often enough to know (though installing it can sometimes be a real pain). The C value will vary between 680 and 220 nF for best effect.
Another little something I picked up from the Otala/Lohstroh amp.
Yes, a four ohm resistive load with 10,000uF gives about 4 Hz. Not so great.
It is a bit more difficult than many here suggest: It is NOT just the low frequency response, it is ALSO the change in damping factor with lower frequency, and as well as the peak current available for a given voltage drop in order to not lose too much continuous or peak current. Peak current can be much more than the continuous current with a fair number of loudspeakers, and so must be provided for, IF you want performance under most loads.
Of course, if you have an 'easy' load, then the power supply requirements are not so high, and you can get away with a 200W transformer and 10,000 uf of capacitance in the main supply. My cheap amps provide this and more. My better amps require a large supply, all the way up to the limits of the AC service available.
Of course, if you have an 'easy' load, then the power supply requirements are not so high, and you can get away with a 200W transformer and 10,000 uf of capacitance in the main supply. My cheap amps provide this and more. My better amps require a large supply, all the way up to the limits of the AC service available.
< .05 Ohms
It may help some to start with some basics for high performance amp's PS design. Walt Jung with help from his friends did the Listen, Test, Change, Re-listen route on amps. I also used this info in phono preamp and MC pre-pre ---> Suggesting that all using reg PS place them AT the circuit and not remote as the wiring Z alone can kill this requirement plus added resonances etc.
PS Zo at the circuit/amplifying stage:
Must be under .05 Ohms (50milliOhms) over the frequency range of interest; AT The Circuit being powered.
THx-RNMarsh
It may help some to start with some basics for high performance amp's PS design. Walt Jung with help from his friends did the Listen, Test, Change, Re-listen route on amps. I also used this info in phono preamp and MC pre-pre ---> Suggesting that all using reg PS place them AT the circuit and not remote as the wiring Z alone can kill this requirement plus added resonances etc.
PS Zo at the circuit/amplifying stage:
Must be under .05 Ohms (50milliOhms) over the frequency range of interest; AT The Circuit being powered.
THx-RNMarsh
Last edited:
I think you are definitely on the spot, Demian. Not only related to your Spectral amp work, but in general.
In my view, one of the key tricks to pull off at any cost is to make sure that each following stage effects the previous stage as little as possible.
For example, a three stage output a la Locanthi does that all on its own. Not perfect, but it does a fairly good job of separating the VAS from the output devices.
I am not against using a buffer stage or device between the input stage and the VAS, depending on the topology. And running the VAS at what might seem to be an unnecsarily high current does make it less susceptible to power stage demands. In short, make it as stable as you possibly can before you even start considering global NFB.
I'll take your word for it.
But bypassing a large cap with a small one is, I think, as old as electronics. It seems to have been done like forever, and I venture to say, everywhere. It's done in the PSU lines, it's done on te input (typically a say 47uF bipolar bypassed by 100 nF polypropylene) and it's done in the NFB loop (main say 220 uF bipolar bypassed by say 470 or 680 nF, etc). They call it "phase correction".
I remember reading literally 40 years ago a text about some Japanese tweakers who claimed that the best cap (if one must have it) is a composite cap, say, usinf 1 uF polypropylene // 1 uF polycarbonate // 1 uF polyethilene or some such to obtain a 3 uF value. Their claim was that enagles the composite cap to be musically neutral, and if one must combine unequal values, then make the polycarbonate the biggest, as it is extremely fast (if memory serves, speed is like 5,000 V/uS, or some such).
Yes, and to do this accurately some series resistance must be used with the bypass cap. That wasn't addressed in the ARC patent.
The right values of R and C for the bypass network can be made to cancel the main cap's inductance. Of course, the bypass C must have low inductance.
(Bold by DVV)
And therein lies the key problem - how big and muscular do you want it? IN theory, as big and as muscular as possible, but that soon starts to cost serious money and eats up a lot of space.
As the designer, I imagine you are on a slow turning spit every time you get down to it for a new amp. I imagine it's not very often you are told that price is no object, just go and make some magic.
You have practically nothing to work with. No idea which speakers, no idea what they are like as an amp load, no idea at what power levels will it have to work most of the time, and so forth.
BTW --- via L-T-C-Re-L.... it appeared the point of diminishing returns in a PA PS is 100K microfarads.... [for the fixed transformer/Amp size used.]
I have no idea about the details but it seems you cant get enough C there. Note, a common value is 10K mfd. 10K seems to take care of the ripple issue (along with the PSRR) but, as JC said, other factors are affected as well.
-RM
Last edited:
- Status
- Not open for further replies.