You will probably be surprised, but definitively, and on all the tries i did in my life, SMPS were better than any oversized linear PSUs.
If you are obliged to go with linear, try a cap multiplier stabilized with a simple Zener (and a fast power transistor).
Avoid regulated PSUs.
Andrej was driven to the same conclusion.
One thing that was not tried was a regulated SMPS isolated by a cap multiplier.
If it works, it could be more economical.
In all the situations, a soft start is a requisite to give a chance to the expensive big caps to wear as long as possible.
I always remember: An amp will never be better than its power supply ;-)
Last edited:
I don't understand what you are saying, a cap multiplier is not a cap multiplier if a zener was added (at least where I think you will put it). That it becomes a voltage regulator with fixed voltage. Cap multiplier is a voltage regulator too, so you suggest to avoid it then suggest to use it??
dadod, it was just to explain the schematic that was too lazy to upload. A cap multiplier with an optional zener or TL431 in // with the cap at the base. Just there to avoid any voltage in excess.
I call this stabilized, not regulated.
What i suggest to never use is any regulation with global feedback. See what i mean ?
I don't know exacly why they don't sound as good to my ears. Conflicts with the feedback of the amps VS the PSU ones during transients?
Last edited:
1)VAS current trim.R28/R29 "steal" VAS
current from the Cascodes , lowering those values (trimmer) will
lower the cascode current with more (I) for the LED's.
You WON'T be able to see this on Q13/Q16 emitter resistors. To
calculate the cascode current , subtract the LED forward current
from the current you measure on R24/R31.
Another way I told still4given is to place a 10R between IPS/OPS PD+
and measure VAS current directly.
The full range on R28 should be 10K=3.8ma to 100k 5.8ma. Your CCS's
are running "hotter" 1.8ma .... you should have 4.5 - 6.5ma range.
Edit - the RED LED Vf must be 1.8v (newer red's).
R17 - "rbridge" .... you can reduce this to 5 to 10R and get 300khz
square-waves that look like your 50khz ones !!
If you want to "max" this one out - 10-12pf for C7-8 and 5R for R17.
This will give 4.2mhz CLG and 1000V/uS slew with a little overshoot.
PS - this will "flat out" beat most forum CFA's and most definitely the VFA's.
You can even degenerate the diamond (and bridge) Re's - 27-33R and get
>>1Kv/us (1.2Kv/us is the record).
OS
Please look at this schematic.
2) I must mention that from starting amplifier up to the warm the VAS current isn't stable but vary from 400mA up to 850mA AND IT DEPENDED from warm conditions of Q5,Q6,Q7,Q8.
3)Lower value for R17(5R) INCREACE VAS CURRENT to 12mA (Q5-Q8)runs very hot.
Making this i can't see any big difference in square wave .
4)Using 12pf for C7,C8 the amplifier is rock stable,suppose you can use even lower capacitance.
5)Today i have my friend's opinion......"This is something special ,i can't believe that is an amplifier with common parts".
Attachments
Last edited:
dadod, it was just to explain the schematic that was too lazy to upload. A cap multiplier with an optional zener or TL431 in // with the cap at the base. Just there to avoid any voltage in excess.
I call this stabilized, not regulated.
What i suggest to never use is any regulation with global feedback. See what i mean ?
I don't know exacly why they don't sound as good to my ears. Conflicts with the feedback speed of the amps VS the PSU ones during transients?
I don't buy it, I use a PS regulator(not stabilized) with negative feedback and the amp sounds better with it.
Why do you call it stabilized not regulated. Regulated does not automatically means feedback, it could be with without. Your schematic is a regulator too. If you want to avoid voltage excess at the output, than your zeners should not go in to action for normal input voltage, but in this case I don't see any need of it. A big cap at the transistor bases will prevent a voltage peaks to pass to the output. If the zeners are always in action then this is voltage regulator with fixed output voltage or how we call it here in Croatia, voltage stabilizer.
Best wishes, Damir
new set of measurements.
Attachments
Can You repeat one square wave measurement on 8 r with small cap in parallel please ( someting like 0,47uf or similar)
Thanks
Ok - to make all of us happy , I DON'T like that temperature thing.
To compensate , notice Q17/18 in the schema (below). Add these , change
R5-8 to 390R ,R4 to 6.8k.
The "2Q" CCS has a VERY negative temperature co-efficient.
(below 2) you see it nearly negates VAS current change from 0C -50C.
OS
Attachments
Apex, you torture the amplifier, this is one on the worst loads you can put on it, and I must say not a very common IRL-load, most speakers rise in impedance as frequency increases, only OLD quads, ML and other electrostatic's may look in that direction. Glad to see the amplifier hold up, but then I was quite confident it would as probing with a TIAN had more than 80 degrees phase-margin. This is a good design. 
A-fusion
A-fusion
MiiB
Amp must survive the difficult load test, good practice is to measure the current/voltage drop across emiter resistor, the peaks can be very high and can kill the o/p tranies easly. Driving a test load 8R resistor with 0.5m of cable is easy, some of the speakers aren't easy to drive so the test must be done anyway.
I do not know what is the acceptable level of overshot/oscillations with capactive load test but IMHO it should look a bit better (not sure).
In this case amp is extremly fast so it would be hard to get better results.
Amp must survive the difficult load test, good practice is to measure the current/voltage drop across emiter resistor, the peaks can be very high and can kill the o/p tranies easly. Driving a test load 8R resistor with 0.5m of cable is easy, some of the speakers aren't easy to drive so the test must be done anyway.
I do not know what is the acceptable level of overshot/oscillations with capactive load test but IMHO it should look a bit better (not sure).
In this case amp is extremly fast so it would be hard to get better results.
Attachments
A feedback amplifier has an inductive output impedance. The better the amplifier, the higher the Q of this inductance. A good amplifier will ring for longer with a capacitor at the output.
To prevent this requires putting some lossy component between the amplifier and capacitor. The L//R network qualifies, but making it large enough to permanently eliminate resonance with capacitors would add too much resistance between the amplifier and load.
Bottom line. if you amp resonates or has overshoot with a capacitor on the output, it's a GOOD thing. It means the feedback is working well. It's not the amp that's the problem, it's your load.
Please all these will be more clear if it possible to draw a simple schematic? (English isn't my first language so i can't understand well).
Thanks.
Thimios , I want to wrap this up.
Try the different CCS to make it thermally stable.
Half this amp is the NAD/NX CFA. That one would stabilize because in/NFB
cancelled. This is a full bridge , no such cancellation is native to it.
But , the "2Q" seems to be just the right amount of thermal cancellation.
OS
Try the different CCS to make it thermally stable.
Half this amp is the NAD/NX CFA. That one would stabilize because in/NFB
cancelled. This is a full bridge , no such cancellation is native to it.
But , the "2Q" seems to be just the right amount of thermal cancellation.
OS