• Disclaimer: This Vendor's Forum is a paid-for commercial area. Unlike the rest of diyAudio, the Vendor has complete control of what may or may not be posted in this forum. If you wish to discuss technical matters outside the bounds of what is permitted by the Vendor, please use the non-commercial areas of diyAudio to do so.

Swordfishy/ASPEN FETZILLA power amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Jump to Latest
AKSA

AKSA

Member
Joined 2001
aksaonline.com
Hmm, Ron,

This is perplexing and I'm sorry you are having this trouble. Most sources connect their shield for both channels, signal earth and power earth on the modules is the same.

I will admit that the amp was designed to either use two separate transformers, or one with four galvanically isolated secondaries. But I think we have an earth loop at signal/power level here.

Ensure that power ground of each of the two modules is connected, and disconnect ONE of the shields on one, since it will be joined inside the source anyway and then joined to both module grounds via star earth.

Make sense?

Hugh
 
R

RonR

Member
Joined 2005
Hi Andrew,

I measured output noise tonight with the recommended 2 transformers, speakers connected and inputs shorted: 2.5mV AC rising to 2.7mV when fully warmed up.
Then I got carried away listening to music, so didn't get the diagrams done.

Hi Hugh,

Your explanation makes perfect sense, but I need to swap back to single transformer configuration for this test. I'll have time on Thursday.

Diagrams to follow.

Cheers,
Ron
 
Last edited:
M

masood

Member
Joined 2010
AKSA said:
Folks,

I have collaborated with Greg Peters, Lineup and Mikelm to design a new power amp of 50W which uses a jfet front end, a mosfet voltage amp, and lateral fet output devices. This amp runs on 36V rails, has integrated power supply rails, and is configured to lie flat on a flat backed heatsink, screwed directly onto the output devices.

The board is 230mm x 69mm, borrows heavily from the NAKSA layout, and while it does not have anything like the power is a good amp set up specifically for FETs, which many like. It was conceived largely as an intellectual exercise, but has survived 115 pages of intense scrutiny on the SS forum, with varied input from lots of clever people.

I will offer these boards (which aren't cheap as they are premium quality) for sale through my website in due course. In the meantime, would those interested please register their interest, along with the numbers required, so that I can get a feel for how many to order.

This board has taken many tens of hours but because humans are fallible may well have mistake(s). If you see something glaring, please let me know in this forum. If you have strong design issues you wish to bring up, then do so on the SS forum, and be aware that Greg and I will not from this point forward, for obvious reasons, change the design. Greg has done lots of listening and CRO tests and feels this is as good as it gets. For myself, I will be most interested to compare it to my own designs; the bar is high!!

16th July 2010: Strategic design update for the benefit of new visitors

Here are the design approaches of Greg Peters, Mikelm and myself so far. I would point out this has all been the original brainchild of Lineout, the redoutable northern Swede who while he lives in one of the coldest places on the planet has an almost feverish, active mind.

#1 1st stage: singleton input device. After much discussion and a few trials, Greg has opted for a jfet, as this gives very low transconductance, around 20mS, and this keeps loop gain low which seems important to the good sound of this amplifier despite higher THD figures.

#2 IIRC Lineout's original design called for a single ended input and VAS with complementary push pull output stage using DC coupling throughout. Essentially this meant eliminating the feedback shunt cap, a 1000uF beast that Mikelm in particular disliked. The problem is that offset control without a servo - which adds complexity and detracts from the circuit simplicity - is very poor with DC coupling as DC gain is no longer unity, but the reciprocal of the gain. In amp design the DC offset control is not a trivial design problem, as it has the potential (literally!) to destroy speakers if poor. Accordingly, Greg has chosen to continue with AC coupling, so back in goes the cap, and DC offset is now very good. I endorse this choice strongly.

#3 Lineout originally chose the IRFP9610 p mosfet as VAS. Briefly the advantages of a bipolar device such as the 2SA1360 were mooted, but Greg noticed, and Andrew confirmed, that a mosfet VAS required no Miller capacitance for rock solid stability. It also gave a more 'lively' sound with palpable vitality. I went looking for a better mosfet VAS, one with a lower current rating to improve its linearity at very low drain currents, typically around 12mA in this application. Greg had suggested the Zetex ZVP3110A, but it's only a TO92 and limited to 740mW. In this application, and with the original 20mA VAS current selected, this gave very high dissipation, too high, so I felt the 2W smd Zetex ZVP2110G would be superior as it was essentially the same chip in a SOT223 high power package. Eventually we decided that since the Ciss of each output device gate was bootstrapped, 20mA of VAS current was not needed, and ultimately we've settled on just less than 12mA, a good compromise.

#4 Originally Lineout had selected a bootstrap CCS for the VAS, but this seemed to detract from the bass response of the amp, though it does sound very good on vocals. Paul Bysouth, a clever Melbourne engineer and good friend, suggested using the Supertex DN2530 as a CCS/bootstrap hybrid in an attempt to get the best of both worlds - good bass, and engaging vocals. The jury is still out on this clever modification, but it's incorporated into the pcb as T3 and on Monday Greg will be able to tell us how it sounds.

#5 I decided to offer a pcb as a means of getting Aspen's name out there, and thought it might be convenient to put independent power supplies for each rail directly onto the amp pcb. These power supplies use ultra fast soft recovery diodes and RC decoupling between the two filter caps on each rail; this adds some refinement to the sound quality. Since Greg had also found that it sounded better at 300mA than 100mA quiescent, and better again at 1A, I felt that appreciable distance between the output devices was required so that we could install the board directly onto a longish 0.42C/W heatsink. The Fetzilla board is therefore 230mm long and only 69mm deep, with 120mm between the outputs, and this gives useful options to set quiescent current nominally at 500mA with 36V rails, or 1A with 24V rails in cooler climates than Oz, all with a simple convection heatsink and no fans.

#6 Every precaution is taken in the board layout to ensure unconditional stability. This includes loading the amplifier with electrostatic speakers, known to be very difficult loads. Phase margin is high, and unity gain loop frequency is just over 700KHz. There is considerable ground plane, and some separation of input/VAS stages and output. An output Zobel, inductor and parallel damping resistor is fitted to the pcb, and the bias generator is range limited by a 4.7V zener diode to prevent overcurrent blowups. There is also provision for experimentation with lag compensation (Miller cap) and phase lead (the JLH feature), making this design highly suited to a novice learning the ropes of audio amp design.

#7 I have suggested polarising both signal coupling caps, at the input and the fb shunt positions. This approach improves sound quality hugely, particularly if the polarising voltage is greater than the peak AC voltage passing through the cap. Accordingly, the input gate is set at around 2.5V positive to achieve the necessary zero output offset.

#8 This circuit is deliberately low loop gain (around 30dB), and very simple, using only one type of semiconductor, ALL of them FETS. This should draw considerable interest, as it appeals to many to own an amp which is designed for a different style of semiconductor. Most are bipolar.

The pcb is 70um copper, two sides, 2mm FR4, overlay both sides and soldermasked. I will attach the smd VAS prior to despatch.

I have attached the parts list for this amplifier as a Word file, and the very latest pcb as a gif.

Cheers,

Hugh


PS For the very latest, as of 26th March 2012, see here: Latest Schemat March 12

__________________
Aspen Amplifiers P/L (Australia)
Aspen Amplifiers[/QU

dear sir ,
can u send me this pcb in pdf form by this i etch pcb. buying pcb by paypal is not possible by me because i do not have my account in bank.my english is poor so plz pardon me.your ever masood.thanks
Click to expand...
 
A

AndrewT

R.I.P.
Joined 2004
RonR said:
............I measured output noise tonight with the recommended 2 transformers, speakers connected and inputs shorted: 2.5mV AC rising to 2.7mV when fully warmed up.
Click to expand...
Using my DMMs those noise voltages are very high, the opposite of dead quiet.

I expect <=0.1mVac using my DMMs.
If I get 0.7mVac after inserting the amp in to a chassis, then I know something is wrong. In that case I will strive to get noise down to <=0.3mVac

Your values are ~20dB worse than anything I would accept, and ~30dB worse than most of my amplifiers.
 
W

Woody1911a1

Member
Joined 2010
ok , my comments were based on audible observation . i can turn my amp , pre amp and a source such as my new cd player (nothing playing) and crank the volume up all the way and there is absolutely NOTHING , nada , zero anything audible coming out of my speakers .

now measuring with my BK precision dmm set too 200mv scale , amp on .1mv ac . pre amp turned on same . cd player turned on - same . speakers were still connected if that matters .

also my questioning whether a transformer with four galvanically isolated secondaries was really the same as two separate transformers was not meant to question your or Hugh's expertise , but more so the general idea .

here we have several people building one way or the other and results seem to say that two separate transformers out perform .

cheers Woody
 
Last edited:
AKSA

AKSA

Member
Joined 2001
aksaonline.com
This was a group project, with multiple designers. Consequently there are numerous methods chosen to kill the cat. Mikelm, for example, uses a CCS on the front end and direct coupling throughout, the original topology Lineout proposed. He swears by it and loves the avoidance of a large shunt cap on the feedback network. There are some other, less formal mods as well out there. The pcb I layed out uses cap coupling throughout, and is the 'cooking pot', non-esoteric version. This is done for reasons of conservative engineering and was the approach both Swordfishy and I preferred. It is a preference, it is not set in stone.

The pcb, however, reflects my usual expensive preference for two separate power transformers, giving two separate power supplies, one for the positive rail, another for the negative rail. Star earth is not connected to the center tap of a transformer, a feature of this approach, and CRC supplies are used. There is no attempt made on the pcb to separate signal from power earth; they are co-located at the central point on the pcb. I have floated a takeoff (back to back diodes and a 10R resistor, all in parallel) from power earth to chassis; this is not compulsory and may be omitted since most transformers are pot (insulation) tested to 4000V. Earthing is a black art, and most angles were covered, but some experimentation may be required. If some here have amps that are dead silent, then we can safely assume it is possible.

That said, Ron, how are you getting along? If with one channel connected all is well, but things fall over with two, then clearly there is an earth loop at the amp level and I'd be connecting both star earths and disconnecting ONE of the signal input shields, OR, separate the two star earths completely, and connect signal input shields on both channels.

Hope this helps,

Hugh


PS: Masood: I don't have gerbers, but there are plenty of gifs of the pcbs on the thread. You can work from these to lay out your own pcb, or, if you prefer, email me and I will send you the native pcb file.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Top Bottom