tvr, have you tried an ideal current source to see its effect on your overall circuit performance? Your results with LEDs and standard diodes are suspicious, particularly the absence of effects from shunt C (and btw you can make that as big as you like --- take advantage of Simland for the purposes of education!).
There can be, of course, compensatory effects in circuits from the departure from the ideal in one place complementing something nonideal elsewhere.
Hi,
Use the one I posted above and subtract 0.6V from the needed voltage overhead. Returning the gate to the base has pretty much the same effect as to the emitter, but needs 0.6V less voltage less from the FET (or adds 0.6V across the BJT's CE).
Whichever way you look at it, it helps except possibly at very high frequencies.
Ciao T
Use the one I posted above and subtract 0.6V from the needed voltage overhead. Returning the gate to the base has pretty much the same effect as to the emitter, but needs 0.6V less voltage less from the FET (or adds 0.6V across the BJT's CE).
Whichever way you look at it, it helps except possibly at very high frequencies.
Ciao T
I forgot to mention Bob Ludwig, as another name well represented by great LP's. You must have folks like this on your side of the ocean that we never hear of.
I think Bob signs (stamps) many of his masters in the leadout grooves.
Ideal source plotted as .. ta-da: ideal. So that was not very informative.
As expected T's circuit was about 4 times better. Very good to capture. Of course, none of these really apply to the Hafler because if it's funky diode string bias. I did model ripping that all out and will be plugging in the "winners" to see what happens with the rest of the amp there. This gives me much to think and visualize, as the goal here is to really catch how and why this works, not just get a better design.
In shunt caps, the sim did not show much effect at all, and none above 1uF. I have no idea about the quality of the LED model. I just switched to Cordell's 1N4148 model and had to adjust the emitter resistor. It actually made them much worse.
Another question. In a real amp, should the ccs span from the rail to ground, or to the other rail? It would seem to me that in a real amp, it might provide better psrr or if nothing else, not be disturbing signal ground.
I'll go visit Walt's site. He has never steered me wrong.
As expected T's circuit was about 4 times better. Very good to capture. Of course, none of these really apply to the Hafler because if it's funky diode string bias. I did model ripping that all out and will be plugging in the "winners" to see what happens with the rest of the amp there. This gives me much to think and visualize, as the goal here is to really catch how and why this works, not just get a better design.
In shunt caps, the sim did not show much effect at all, and none above 1uF. I have no idea about the quality of the LED model. I just switched to Cordell's 1N4148 model and had to adjust the emitter resistor. It actually made them much worse.
Another question. In a real amp, should the ccs span from the rail to ground, or to the other rail? It would seem to me that in a real amp, it might provide better psrr or if nothing else, not be disturbing signal ground.
I'll go visit Walt's site. He has never steered me wrong.
Actually, if the emitter resistor is >> 1/gm (gm of the bipolar), the increase in output impedance is quite dramatic at the JFET drain for the gate return to emitter configuration--- the JFET output C essentially goes away except at very high frequencies. Strays will dominate, and they have the happy property of not varying with drain voltage, thus generating no nonlinear distortions. I'll publish some data soon (it probably exists elsewhere on the site, perhaps within the thread jcx referenced, so apologies in advance for redundancy).
One can also concatenate FETs for even higher output R and tiny C, but this further adds to the voltage burden. Certainly overkill for a tail current generator.
Brad
PS: Whew! The ad refresh bug from hell has gone away!
In your CIRCUIT, not as an isolated I source! Sheesh! What is the effect of an ideal tail current generator in your overall circuit? If an ideal one gives you POORER overall results, it may be VERY informative.
I came to a similar conclusion regarding a single transistor CCS. So I tried a few instead of my trusty, beloved Motorola MPSA 56/06 and evetually discovered that I was getting better results with Siemens BC 639/640.
They are similar to their MPSA counterparts, but, according to their Data Sheets, don't stop at 60 MHz but go on to 100 MHz. Also cheap and readily available.
If you want to filter the twin diodes at the trannie's base, use a Wima type 100 nF/63V polyester capacior. Bigger is not always better.
So he's probably not ready with the ~comprehensive study I was privy to a bit of. I haven't heard from him for a while.
Touche!
The magic words.
Come on, no CCS is an end unto itself, it's merely a part of a bigger whole. When investigating results, surely you need to do so for the whole, not just one small part of it, no matter how important it may seem, because in reality, each and every part is important. You do everything just right, and one lone part reduces excellent to the mundane.
Oooooohhh, you nasty man ...
You just added some schinken to my already rich table. Back to the simulator, THIS I have to run through its paces.
One question, though - will it be AC or DC coupled? You did say a lot was yet to come for as complete project, so I may be jumping the gun.
As for protection circuits, perhaps I can contribute? I have two basic schemes worked out in detail (NS comparator based and fully discrete, both for nominal 100 wpc amps), just assemble and go, with options. If you're interested, let me know, I'll be more than pleased to pitch in.
For what it's worth, I use NI Multisim 10.something, and just in case you do too, it's available in that format as well, ready to go.
Hi,
Subjectively, yes.
Actually, two reasons.
First is that each output stage PCB has space for 4pcs of big electrolytics, two per rail.
Second is that I'm a tube guy and if I see two identical capacitors on the same rail I feel someone has forgotten the filter choke that goes inbetween.
So I had some small torroids made from some manganese gosh knows what cores that can be bodged into the empty spaced on the PCB between the cap's and which give me 2mH @ 10A DC with 50mOhm DCR (I did verify these parameters using a very high grade LCR meter)...
Actually, the effect on noise and ripple is non-trivial.
Connecting the cap's in parallel without chokes gives 2V PP 100Hz ripple and that with very buzzy sawtooth shape, the LC one gives 260mV peak-peak with basically a sinewave shape that is not visibly distorted, both at 10A DC.
It is mainly the absence of higher harmonics that I am after, not so much the reduced 100Hz components.
Incidentally the Cap's i procured are Elna "For Audio" (Genuine) 18,000uF/71V, which measured in terms of ESR, leakage, loss and inductance MUCH BETTER than generic 22,000uF/63V, when D. Self measured Elna "For Audio" he found much lower distortion than generics.
PSU will be:
Bridge (using some TO247 superfast supersoft rectifiers) ->
Cap ->
choke in each line (2 * 1mH) ->
2nd Cap
(Star)ground established at the 2nd Cap pairs suitable terminals.
BTW, the frontend uses 3,300uF/22mH/3,300uF before the CCS & Shuntregulator PSU (LM317 CCS plus a CFB Shunt) with around 125mA in the CCS's for both channels frontends (yes, this is with around 15mA per channel total current for the frontend).
In many ways I do take great care with powersupplies and it usually pays off... Great care does not mean to me brainlessly throwing big caps at the PSB, BTW.
I also have a massive NEC-Toking 25A line filter for the Amp plus the parts for an in-line DC trap (68mF Cap and massive bridge).
Ciao T
Subjectively, yes.
Actually, two reasons.
First is that each output stage PCB has space for 4pcs of big electrolytics, two per rail.
Second is that I'm a tube guy and if I see two identical capacitors on the same rail I feel someone has forgotten the filter choke that goes inbetween.
So I had some small torroids made from some manganese gosh knows what cores that can be bodged into the empty spaced on the PCB between the cap's and which give me 2mH @ 10A DC with 50mOhm DCR (I did verify these parameters using a very high grade LCR meter)...
Actually, the effect on noise and ripple is non-trivial.
Connecting the cap's in parallel without chokes gives 2V PP 100Hz ripple and that with very buzzy sawtooth shape, the LC one gives 260mV peak-peak with basically a sinewave shape that is not visibly distorted, both at 10A DC.
It is mainly the absence of higher harmonics that I am after, not so much the reduced 100Hz components.
Incidentally the Cap's i procured are Elna "For Audio" (Genuine) 18,000uF/71V, which measured in terms of ESR, leakage, loss and inductance MUCH BETTER than generic 22,000uF/63V, when D. Self measured Elna "For Audio" he found much lower distortion than generics.
PSU will be:
Bridge (using some TO247 superfast supersoft rectifiers) ->
Cap ->
choke in each line (2 * 1mH) ->
2nd Cap
(Star)ground established at the 2nd Cap pairs suitable terminals.
BTW, the frontend uses 3,300uF/22mH/3,300uF before the CCS & Shuntregulator PSU (LM317 CCS plus a CFB Shunt) with around 125mA in the CCS's for both channels frontends (yes, this is with around 15mA per channel total current for the frontend).
In many ways I do take great care with powersupplies and it usually pays off... Great care does not mean to me brainlessly throwing big caps at the PSB, BTW.
I also have a massive NEC-Toking 25A line filter for the Amp plus the parts for an in-line DC trap (68mF Cap and massive bridge).
Ciao T
This reminds me of a job I got designing an ultralinear triangle wave generator for a pulse width modulator. One of the requests was that it be crystal-controlled, but also extraordinarily stable in amplitude (which is a tall order). The design was successful in meeting the objectives, but I did not openly question the assumptions of the customer: that somehow this would make the class D amplifier in which it would be used, perfect
Hi,
As Bugs Bunny often says, "Ain't I a stinker!" (Discordians worship Bugs Bunny as an incarnation of d*G)...
DC.
The input will be THE INPUT (RCA Jack), output to binding posts via a trick dual relay (a monster PCB mounted one rated to break 10A DC at substantial voltages and a CP-Clare Mercury Wetted Reed relay which will switch on with a small delay).
I will include a proper build out filter, this will be mainly a filter to stop RF ingress (so it will look more like a Ham Project - in/out terminated with zobel at CI, tapped choke with cap to ground at centre, I'm looking at a few 100KHz cutoff).
In my case this will be done as a rebuild of an existing Amp, it has protection circuits, basically a classic V/I foldback limiter NOT limiting but triggering speaker-protection/mute, mute will be CP-Clare Reed Relay, these can switch at around 1KHz IIRC.
So protection will be completely non-invasive up to the point where it trips.
As this payware and the company has something else, I have to make do with Tina-TI and LTSpice... Lest you'd have a link to a t0rr3nt or 3mul3 file containing a trial version that works enough to sim this Amp...
Ciao T
Oooooohhh, you nasty man ...
As Bugs Bunny often says, "Ain't I a stinker!" (Discordians worship Bugs Bunny as an incarnation of d*G)...
DC.
The input will be THE INPUT (RCA Jack), output to binding posts via a trick dual relay (a monster PCB mounted one rated to break 10A DC at substantial voltages and a CP-Clare Mercury Wetted Reed relay which will switch on with a small delay).
I will include a proper build out filter, this will be mainly a filter to stop RF ingress (so it will look more like a Ham Project - in/out terminated with zobel at CI, tapped choke with cap to ground at centre, I'm looking at a few 100KHz cutoff).
In my case this will be done as a rebuild of an existing Amp, it has protection circuits, basically a classic V/I foldback limiter NOT limiting but triggering speaker-protection/mute, mute will be CP-Clare Reed Relay, these can switch at around 1KHz IIRC.
So protection will be completely non-invasive up to the point where it trips.
As this payware and the company has something else, I have to make do with Tina-TI and LTSpice... Lest you'd have a link to a t0rr3nt or 3mul3 file containing a trial version that works enough to sim this Amp...
Ciao T
So does the L replace bigger caps or let me ask differently is CLC with small caps better than CRC. With really big caps ?
Still trying to wrap my head around the L value in an CLC , everyone seems to have an arbitrary size when doing theirs ....
Hi,
This is largely an aesthetic issue, I prefer to span the rails, wherever possible.
BTW, note in my Amp concept the two things that have come straight from this thread, the bootstrapped J-Fet CCS (the J-Fet CCS's are already there in the original, I merely added four resistors for the bootstrap - this should give a massive improvement on apparent CCS impedance though, even with 1% resistors) and the parallel transistors in the VAS (after Samuel Groners work I fancy the 2SC2240/2SA970 as VAS, they have a very "tubey" HD spectrum).
Ciao T
This is largely an aesthetic issue, I prefer to span the rails, wherever possible.
BTW, note in my Amp concept the two things that have come straight from this thread, the bootstrapped J-Fet CCS (the J-Fet CCS's are already there in the original, I merely added four resistors for the bootstrap - this should give a massive improvement on apparent CCS impedance though, even with 1% resistors) and the parallel transistors in the VAS (after Samuel Groners work I fancy the 2SC2240/2SA970 as VAS, they have a very "tubey" HD spectrum).
Ciao T
Hi,
Well, that depends on your definition of "better".
My Amp will already have 4pcs 18,000uF per channel (72,000uF per channel in magazine Spec's) so how much bigger can we realistically get?
Chokes store energy, resistors do not.
Chokes give a 2nd order filter, resistors a 1st.
The so-called "really big cap's" have their own problems.
As I am not trying to drive speakers with 75dB/2.83V/1m efficiency and sub 1 ohm bass impedance I do not see the need for stupidly big PSU Cap's.
As a rule, bigger values are better, but for me (existing chassis/PCB's that must accept modification) size was the main constraint, so I asked the factory that made the samples to give me something for 10A DC, 50mOhm DCR per choke and the given size core and the maximum "L" that would go for this. They came out at 1mH appx...
Ciao T
Well, that depends on your definition of "better".
My Amp will already have 4pcs 18,000uF per channel (72,000uF per channel in magazine Spec's) so how much bigger can we realistically get?
Chokes store energy, resistors do not.
Chokes give a 2nd order filter, resistors a 1st.
The so-called "really big cap's" have their own problems.
As I am not trying to drive speakers with 75dB/2.83V/1m efficiency and sub 1 ohm bass impedance I do not see the need for stupidly big PSU Cap's.
As a rule, bigger values are better, but for me (existing chassis/PCB's that must accept modification) size was the main constraint, so I asked the factory that made the samples to give me something for 10A DC, 50mOhm DCR per choke and the given size core and the maximum "L" that would go for this. They came out at 1mH appx...
Ciao T
The ideal model in the circuit acted like an ideal. The current through the tail was perfectly uniform. Not informative. I did not bother putting it in the isolated model.
- Status
- Not open for further replies.