d3a gyrator loaded parafeed design
Nov 29, 2010 at 12:27 PM Thread Starter Post #1 of 21


500+ Head-Fier
Feb 29, 2008
Hi all.
Thought i would post this here.
There has been some discussion on diyaudio about the sonic merits of a mosfet gyrator Vs a CCS anode load for triodes (or in this case a triode wired pentode - the D3a)
I have already built a CCS loaded d3a amp so now i am building a gyrator version to try. The salas HV shunt is used for B+, leds for bias.
circuit here

Power supply is shown below. the only change is that i am using 30r instead of the 56 shown, this gives 60ma total, 40ma for the load and 20ma for the shunt (it needs a margin)

I have just about finished the perf board prototypes. The HV shunt is done and tested so we are nearly there.
IF anyone spots any issues with the circuit, please shout. Rather than go round in circles, there has already been a discussion about the relative merits of parafeed Vs single end OPT, so lets not go through that topic here. 
Cost is mostly dependant on the output transformers, whilst testing i am using some super cheap edcors ($20) but if this shows promise then electraprints will be purchased.
My guess is all parts, excluding the case can be bought for <$200
Nov 30, 2010 at 2:21 AM Post #3 of 21
it could, or you could use an led, or a bypassed resistor.
the gyrator is actually a constant voltage device at dc, so set the plate voltage and then use the bias as you normally would to set operating conditions. personally i will be using a single red led and bias the tubes at about 20ma. I may have to use an ir led to get a low enough plate voltage. 
interestingly, the gyrator does not have the PSRR of a ccs, so the salas shunt (or any other reg) is essential really for a headamp. if you wanted to do it more cheaply, it would only take another mosfet acting as a vreg to improve the circuits PSRR a lot, and thaat would allow you to have cheap components in the PS, just a CRC would probably be ample.
error in schematic: the cap should be connected between the 2k2 and the 2m2 resistor, not to the gate directly as oscillations are likely.
Nov 30, 2010 at 12:25 PM Post #4 of 21
Interesting. What kind of steel would you be considering for the Electraprints?
Have you ever considered using an active current source/sink(I am referring to an adapted Aleph CS)?
Lots of good reports on parafeed I've been reading lately.
Nov 30, 2010 at 12:40 PM Post #5 of 21
10k : 300r / 32r M6 from jack or the mu metals from sowter. I had a pair from jack that went into an amp for a mate and the sound that thing made was jaw dropping. The edcors do a job, but they are audibly compromised for sure once you have listened to the electraprints (i still prefer them to a big electrolytic
I'll go and have a poke in the aleph thread - thanks for the pointer. 
I have also just finished a C3g(s) amp, its a bit rough looking but i will post some pics later. the c3g is more polite than the d3a (it doesnt sing to itself without the use of stoppers more pins than i care for!) and the finished result is superb.... but I think the d3a just edges it for detail.
Nov 30, 2010 at 10:53 PM Post #6 of 21
Thanks for sharing your schematic. Unfortunately, you're probably sailing  over the heads of most of the people here (me included). A little general information about the use of a gyrator as an anode load would be appreciated. I know that a gyrator is roughly an electronic choke without the ability to store energy.The trade-offs between a plate choke and a CCS are well known. CCS generally wins in all areas except for swing and voltage drop. I assume a gyrator would compare similarly to a CCS against a plate choke, but how do you expect the gyrator to compare to the CCS?  
Like your shunt regulator. Any idea if it could be scaled  up to a 600v range?
Dec 1, 2010 at 2:35 AM Post #7 of 21

Originally Posted by FrankCooter /img/forum/go_quote.gif
I assume a gyrator would compare similarly to a CCS against a plate choke, but how do you expect the gyrator to compare to the CCS?  
Like your shunt regulator. Any idea if it could be scaled  up to a 600v range?

I've not built one, so I am as curious as you to hear what Adam has to say. But, I would expect it to be similar to a CCS in terms of performance.  As I understand it, basically, a gyrator allows you to set plate voltage while the tube finds its own current, while a CCS allows you to set current while the tube finds it's voltage. I'd think the gyrator might be advantageous for DC coupling because of that.
The shunt should scale, providing you can find parts that can handle the voltage. Only change I might think about would be a better CCS feeding it, though all things considered it should be fine.

Dec 1, 2010 at 2:42 AM Post #8 of 21

Have you ever considered using an active current source/sink(I am referring to an adapted Aleph CS)?

Thinking about this, if you effectively CT the source resistor on the lower dn2540 (just half the value and use 2 resistors) then take the output from that point, so that not give us the anti triode / aleph config (or something close).
Frank :
Thanks for the reply.
The idea is basically a version of wavebourn (diyaudio member - a very clever man!) SVCS or stabilised voltage current source.

It sims like a plate choke (very much like) but obviously has the limitations of CCS in that it wont 'store' energy so B+ needs to be high enough for required swing. 
To date i dont know the sonic benefits until its fired up. Others that use them say it sounds quite different to a CCS and imparts more of the triode signature. You'll also find views that a simple ring of two ccs is better technically. Proof will be in  the listening.
Re the salas shunt reg, scaling up to 600v may be a problem. I know people have taken it over 400v with high v silicons ..... but unless you can find fets that can take the voltage i am not sure.
Dec 1, 2010 at 10:56 AM Post #9 of 21
I havnt read up on gyrators too much, so take this a good guess:
Would replacing R5 in the OP schematic with a CCS improve PSR? It looks like the most obvious path for PS noise to enter the circuit compared to a more conventional Depletion mode CCS. 
Dec 1, 2010 at 12:37 PM Post #10 of 21
yes you could, not much current needed so a little lnd150 or maybe even a crd could be used, or to cut down on parts, forget the cascode.
As i am suing the salas HV shunt reg, i am not to concerned about the PSRR.
Dec 1, 2010 at 11:14 PM Post #11 of 21
Keep us posted,  not sure what you mean as to where to place the misdrawn cap.  Also how do you "dial" this gyrator in (trimpot across R1)?   Sounds like the typical hook up to a 9V battery won't work with this,  I guess what is the setup procedure ?
Frank, the MJE350 limits the Salas reg to around 300V,  you can go a bit over 400V with an [size=10pt]MPSA94 ,   the thing also requires a huge heatsink  if you want it to last a tube burning out.   For example the amp I'm working on which is 20 mA per channel I have an 8x3x2 heatsink,  but the payoff is extremly low B+ output impedance.[/size]
Dec 2, 2010 at 2:48 AM Post #12 of 21
regal, the cap as drawn is between the the anode of the d3a and the gate of the dn2540. it should really be to the left of the gate stopper (inductance of the longish cap lead will make the dn2540 oscillate if the wrong side of the stopper.
re 'dialing in'. This is a constant voltage device so i will just go for the plate voltage i want, and as you would normally just use cathode bias to set the operating point (at the fixed voltage). Very approx the plate voltage is that created by the voltage divider on the left minus vgs i think. in the opening post, the resistors R5 and R6 are infact a 500k pot. the 2M2 is two 1M and a 200k in series (the 2m2 i had were measuring all over the place  - maybe a limitation of my dmm?)
revised schematic

Dec 2, 2010 at 6:58 AM Post #13 of 21
Thanks I think the constant voltage may make a lot of sense for parafeed hybrids,  the dc servos usually junk up the hybrid stage but without voltage fluctuations on the parafeed cap might be able to get away with a nice diamond buffer sans dc servo.
Dec 2, 2010 at 1:12 PM Post #14 of 21
here is a c3g version...with what looks like a neat way of biasing with very very ac impedance (although 2.5v is the only value, but thinking about it, find a device with a suitable vgs and stick in there....) 


Users who are viewing this thread