Goldmund Mods, Improvements, Stability

Jump to Latest
#443
Nico Ras said:
Here are my final comments regarding the amplifier, and no I don't think we are cloning ELD, I think we are developing an amplifier that uses the best of the two designs in order to achieve a performance that will be better than each on their own.

Front End.

Although you have made the right decision on the front end I would like to throw in another two pennies worth.

I have a feeling that using the DC coupling and FETs stands Goldmund in good stead regarding to sound. If one considers that there is no phase change in the bass region (unlike capacitor coupling of input and NFB) may just be why their amp sounds good. At low frequencies our phase perception is much more acute than at higher frequencies.

VAS

The ELD driver using medium power transistors and double Wilson performs better than small signal transistors followed by a driver. It is more linear, easier to compensate.

Output

I would insist that those wanting the most out of the amp uses four pairs of devices with source resistors. Those wanting 100-120 watt can omit two pairs.

Power supply Front end

I have done some listening on my class A headphone amp and HD800 cans, using both a shunt and a series regulator. The shunt regulator sounds healthier, don't ask what that means, it just does.

The front end should be run at 10 -15 V higher than the output stage. For this we need a small transformer of 0-9, 0-9 V (two secondaries) and about 10 VA or so feeding a full wave bridge and followed by at least 1000 uF series resistor (filtering) and another 1000uF cap rated at the operating voltage you choose, followed by the shunt regulator.

The secondaries of the small transformer would be connected in series with the secondaries of the main transformer, thus would always be be 9VAC higher, no matter what main transformer size you choose.

Main power supply

We know that lateral mosfets will only swing around 10V from rails, unlike BJTs therefor the need to run the driver stage at 10 or so voltage higher so that we can get close to rail swing on the outputs.

This has the advantage that we do not have to use an 80V supply to the output devices and it can be dropped to a more manageable and available 65V rails (or lower if you like) The transformer will be more cost effective and if you choose correctly 63V capacitors will do.

Interestingly enough and you can go try it practically, the reservoir caps used for the output stage does not have to be mega farad, you need enough energy to cater for the loudest and lowest signal, hence a 4700 uF capacitor sounded nicer and more fluid than the 40 800 uF that I was using.

I hate hacking equipment that I am using, but I did it for the sake of offering advice. The front end is regulated and that is where it matters most. The amp also sounds more dynamic. Much better than using a regulator on the whole amp as I have been doing.

I hate adjectives but unfortunately thats how we communicate. Therefore dump my use of a regulated amp, it is a waste of money.

I would further use two toroidal transformers with a single secondary for each amp. and mount them on top of each other. Feed the primaries anti phase so that any stray magnetic fields cancel. I do this always else you will have problems with these large transformers near the circuitry.

I have never done this, but it may be possible to fit the transformers into a cut-off coffee tin and pot them which will reduce mechanical vibration and the mount them onto the chassis using a rubber insulator or something similar to decouple them from the rest of the amp, maybe even another box.
Click to expand...

I was hoping to try out your output regulator, always wanted to but never found a pcb for one.
 
#444
One more thing I like about the reverse buffer: positive and negative slewing is more equal, as opposed to the Darlington where all the slewing power is focused in one direction. Less phase distortion, theoretically. Downside is that the driver must have Iq enough to feed the slave - which in this case is not a problem.

BTW, the 2N5769/5771 pair are very fast. The Ft isn't specified in the datasheets, except old ones, but for the NPN it is around 300-500MHz and the PNP is at an incredible 600-800MHz. I developed my own models for these to get DC characteristics right:

.model 2N5769_k NPN(Is=1.2f Xti=3 Eg=1.16 Vaf=100 Bf=88 Rb=45 Ikf=200m Ise=33f Ne=1.7 Xtb=1.7 Br=1.365 Nc=2 Isc=0 Ikr=0 Rc=.6 Cjc=2.83p Mjc=86.19m Vjc=.75 Fc=.5 Cje=4.5p Mje=.2418 Vje=.75 Tr=1.073u Tf=227.6p Itf=.3 Vtf=4 Xtf=4 Vceo=15 Icrating=200m mfg=Fairchild)
.model 2N5771_k PNP(Is=1.2f Xti=3 Eg=1.16 Vaf=100 Bf=125 Rb=45 Ikf=100m Ise=62f Ne=1.7 Xtb=1.7 Br=1.365 Nc=2 Isc=0 Ikr=0 Rc=3.75 Cjc=2.77p Mjc=.1416 Vjc=.75 Fc=.5 Cje=2.65p Mje=.3083 Vje=.75 Tr=4.033n Tf=118.5p Itf=.5 Vtf=3 Xtf=6 Rb=10 Vceo=15 Icrating=200m mfg=Fairchild)
Click to expand...

For both, optimal Hfe is below 2mA.

- keantoken
 
#447
Pardon my quadruplepost.

Actually, using a Jfet CCS decreases slew rate, because it keeps the VAS balanced, meaning one side can't draw current through the tail to compensate for the other being slow. Symmetrical slewing isn't improved either, contrary to my thoughts.

Next, not using drivers means that at max slew behavior the VAS will be saturating into the FETs. This does not bode well for TIMD. Can someone speak of how the sound changes when using drivers or not?

- keantoken
 
#449
Luke said:
I was hoping to try out your output regulator, always wanted to but never found a pcb for one.
Click to expand...

You are still welcome to do so. I have reconnected everything back in my amp like before, I will do a final assesment during our summer holidays and then make a permanent change.

Now until today, there where nothing better than a regulator. You learn something everyday
 
#450
keantoken said:
Pardon my quadruplepost.

Actually, using a Jfet CCS decreases slew rate, because it keeps the VAS balanced, meaning one side can't draw current through the tail to compensate for the other being slow. Symmetrical slewing isn't improved either, contrary to my thoughts.

Next, not using drivers means that at max slew behavior the VAS will be saturating into the FETs. This does not bode well for TIMD. Can someone speak of how the sound changes when using drivers or not?

- keantoken
Click to expand...

What are you simulating????? Have you read anything?
 
#451
I have read every post up until now, what am I missing?

Here is the schematic I'm simulating. I tried replacing R5 (tail resistor) with a Jfet CCS. The Cdom connection is... Experimental.

- keantoken
 

Attachments

  • Goldmund_mod1.png
    Goldmund_mod1.png
    127.7 KB · Views: 565
#453
I successfully pasted the lateral models but they bias "funny". V mosfets (irf's) need several volts to be forward biased ... these L fets .. I can't get to stop conducting. Here is what I have below. Works good , but the bias is "screwy". 🙁
OS
 

Attachments

  • GM.gif
    GM.gif
    21.5 KB · Views: 527
#454
Why the use of current source in the input stage....to me they degrade sound as opposed to pure resistor based circuits...i do feel they improve PSSR.. but here We operate with a super good shunt regulated power supply for the front end...So i would like to keep it as simple as absolutely and only have the cascode and current mirrors to keep control over the two legs and to shield the input jfets from excess voltage...

I dont have too much knowledge of the jfets used here.. but for the SK170...there seems to be a sweet spot win 5-6 mA and a cascode voltage off 9-12 V...here the current in each leg is a little more than 1 mA...is this deliberate...the sweet spot of these jfets...?
 
Last edited:
#455
Guys, if anyone is interested I will start a thread for building ELD. I am offering this design for anyone who wants a well specified great sounding amplifier.

I would like Andrew and Jacco lurking around just to make sure that we do not stray too much off target. This will be a practical building thread. ELD has existed for some years, it is a low distortion, high slew, quite powerful and very stable amplifier.

I have made and tested some small modifications that has come out of the wash here on my existing system and they have made improvements.

I would like someone to take responsibility of designing the PCB (because of the mods, mine is outdated) as well as the mechanical (I will post the mechanicals, but it is too expensive for DIY there are a lot of machining) as I do not have that much spare time available this late in the year.

We will discuss the new power supply that would improve perceived sound quality.

I have the facilities to make PCBs, to manufacture customized transformers for laser cutting and sheet metal work, but not for machining face plates (I will look if there are water jet cutting facilities near us) If there are more than 10 members interested in making ELD then I will do the above else it is up to the individual.

I am not on a money making spree, I have enough of it but I do not want to load the factory with nonsense work, production is essentially full.

I need to get as many wanting a group buy before the end of this month so it can be loaded onto the factory and mailed off to you before our factory shuts down on 16 December. Before you buy anything I will provide the manufacturing cost and you can check it against what you can get it for (I cannot compete with the Chinese) and decide where the group buys.

I will go as far as to say that I will arrange for the components for the complete amp as if it is a kit. I am taking on a huge commitment and will only do it if there are enough members interested, else it is a non starter.

You can e-mail me from today until next week Friday showing your vote. On Friday if there is enough then the thread starts and the ball is rolling, we have four weeks to complete the project and you can start listening on your new Christmas amplifier. I do not know what the cost of shipping will be, but if it includes transformer chassis and all it is not going to be cheap, maybe look at your own courier service.

nico_ras@digisec.co.za
 
Last edited:
#456
MiiB said:
Why the use of current source in the input stage....to me they degrade sound as opposed to pure resistor based circuits...i do feel they improve PSSR.. but here We operate with a super good shunt regulated power supply for the front end...So i would like to keep it as simple as absolutely and only have the cascode and current mirrors to keep control over the two legs and to shield the input jfets from excess voltage...

I dont have too much knowledge of the jfets used here.. but for the SK170...there seems to be a sweet spot win 5-6 mA and a cascode voltage off 9-12 V...here the current in each leg is a little more than 1 mA...is this deliberate...the sweet spot of these jfets...?
Click to expand...

MiiB,

I quite agree with you. As well as with the FET input stage.
 
#460
I have no practical experience with the bootstrapped resistor CCS....the reason for my statement with the simple resistor type, is from experiments with a moving coil preamplifier..where the pure resistor version was clearly more expansive and dynamic.. than CCS'ed version...And here we are contemplating a Salas..superb shunt...so power supply ripple rejection should not be an issue...
 
Last edited:
You must log in or register to reply here.
Top Bottom