Kismet 2A3SE amp

What people are working on at the moment
Morgan Jones
User
Posts: 148
Joined: Fri Mar 25, 2016 10:58 am

#46 Re: Kismet 2A3SE amp

Post by Morgan Jones »

Sh*tting frog, Batman! She is spot-on. That description nails it and brought a smile to my face. I found that 90mph was ideal to take the strain off my wrists. I was pulled over by a police car whilst going up the M1 in 1988 and told, "I've been following you for ten miles and you were doing 90mph all the way," I explained that being an Italian machine, the mph numbers on its speedo were almost impossible to read in the dark and I hadn't yet got used to converting kph mentally into proper speeds. I did not say, "Well, you weren't worried about my riding then" or any other smart-arse comments. He quizzed me on where I was going (not the race meeting at Donnington Park that weekend, but to see fiancee) and let me go. I rode more uncomfortably after that. I doubt that a copper would show that discretion now.
User avatar
Paul Barker
Loony Bin!
Posts: 9110
Joined: Mon May 21, 2007 9:42 pm

#47 Re: Kismet 2A3SE amp

Post by Paul Barker »

Well Vry and I did that from the 212 output transformers primary direct coupled to Vr’s Martin Logan, in the presence of Morgan. It sounded woderful. A Cellist was playing I loved it. As it is Im almost getting that sound as it is badly matched the 57’s are the king of such music types anyway. Sure I can tell the trebble isnt ideal but its all about the midrange someone said.
"Two things are infinite, the universe and human stupidity, and I am not yet completely sure about the universe." – Albert Einstein
User avatar
Paul Barker
Loony Bin!
Posts: 9110
Joined: Mon May 21, 2007 9:42 pm

#48 Re: Kismet 2A3SE amp

Post by Paul Barker »

Morgan Jones wrote: Mon Jul 08, 2024 6:36 pm Sh*tting frog, Batman! She is spot-on. That description nails it and brought a smile to my face. I found that 90mph was ideal to take the strain off my wrists. I was pulled over by a police car whilst going up the M1 in 1988 and told, "I've been following you for ten miles and you were doing 90mph all the way," I explained that being an Italian machine, the mph numbers on its speedo were almost impossible to read in the dark and I hadn't yet got used to converting kph mentally into proper speeds. I did not say, "Well, you weren't worried about my riding then" or any other smart-arse comments. He quizzed me on where I was going (not the race meeting at Donnington Park that weekend, but to see fiancee) and let me go. I rode more uncomfortably after that. I doubt that a copper would show that discretion now.
Friend of mine bought a Ferari Dino insured it and drove it home in the early hours. Pair of traffic cops stopped him to tell him they were jealous, and to complement him on getting the right documents all legal for the trip.
"Two things are infinite, the universe and human stupidity, and I am not yet completely sure about the universe." – Albert Einstein
User avatar
MistyBlue
User
Posts: 155
Joined: Mon Feb 12, 2024 9:38 pm
Location: North Somerset

#49 Re: Kismet 2A3SE amp

Post by MistyBlue »

I got the 2A3SE amp turned over for the first part of surgery. Removed the Arduino relay and replaced the faulty blue “angel eye” power switch. I fitted these to all of my self-built HiFi stuff, and they have all failed (RS 111-5828). The Chinese ones got from Ali Express seem to be cheaper and better – go figure.

The HT was offloading 322V at the 200uF cap and 250V at the 2A3 anodes. Removing the 680R resistor, as suggested, gave 316V at the 200uF and 311V at the 2A3 anodes, a little higher than the target, but I’ve left it there for the moment. I will probably re insert a 100R resistor just to drop the HT to around 300V when it turns up in a few days.

The heatsink didn’t need too much in the way of modification for fitting the new regulators, just needed 4 holes drilling and tapping. I’ve also 3D printed a cap clamp for the 4 22000uF ALS30s for the raw DC supplies.

2A3SE Amp002.jpg
2A3SE Amp002.jpg (115.71 KiB) Viewed 576 times

Couldn’t hear too much in the way of increased headroom after the HT mod, but at least there is less waste heat being generated. The regulators are working well and gave the sound a slightly softer edge than previously, which was welcome. At least the filament supplies work as intended this time!

I tried using the 12db octave slope on the active crossover to get a feel for how that slope will sound. Sounded a bit warm, muddy and over-blown in the upper bass/lower mid region. In contrast, the upper section sounded better, more detailed and more open. On investigating with VituixCAD simulating the crossover, that is exactly what happens to the response in the LF. Tweaking the pots on the board does little to compensate for it, so I’m going to put it back to 24db octave for now and try some minor mods on the passive crossover ready for trying the coupling/cathode bypass cap mods on the 2A3SE.

I’m also in the process of designing a PCB in Kicad that will fit inside the 2A3 amp (or the active crossover enclosure) incorporating a 12db octave LP filter for outputting to the bass amp along with a JFET preamp stage based on FE2022 to equalise the level to the 2A3 to get back the headroom I’ve lost. This is based on the sims I did on the cap modded 2A3 and the sallen key active filter. I will have to test it will work before I commit to PCB fabrication with the modules I have.
"When we're about to do something stupid, we like to catalogue the full extent of our stupidity, for future reference." - Commander Susan Ivanova, Babylon 5
User avatar
Nick
Site Admin
Posts: 15859
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#50 Re: Kismet 2A3SE amp

Post by Nick »

and they have all failed
Out of interest how did they fail? I use a similar but non latching (was Bulgin, now Chinese (probably Chinese in both cases)) but only switch an input to a controller and I drive the LED from the same controller so I can sequence startups. The ones you have are 5A rated, but I wonder what the actual switching current is for the amp. Did you have a cap across them to remove switching arc's?
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
User avatar
MistyBlue
User
Posts: 155
Joined: Mon Feb 12, 2024 9:38 pm
Location: North Somerset

#51 Re: Kismet 2A3SE amp

Post by MistyBlue »

The led light is the bit that failed. That was the bit I really liked about them. The switch itself remains OK. No cap across it though.
"When we're about to do something stupid, we like to catalogue the full extent of our stupidity, for future reference." - Commander Susan Ivanova, Babylon 5
Morgan Jones
User
Posts: 148
Joined: Fri Mar 25, 2016 10:58 am

#52 Re: Kismet 2A3SE amp

Post by Morgan Jones »

MistyBlue wrote: Wed Jul 17, 2024 12:04 pm I’m also in the process of designing a PCB in Kicad that will fit inside the 2A3 amp (or the active crossover enclosure) incorporating a 12db octave LP filter for outputting to the bass amp along with a JFET preamp stage based on FE2022 to equalise the level to the 2A3 to get back the headroom I’ve lost.
I imagine the LP filter is at quite a low frequency, perhaps 100Hz. That means it will have significant delay. The reason most subwoofers sound awful is that they are out of step with the rest of the music. If you want it to be right, you need dastardly digits to delay the signal to the main loudspeakers to equalise the unavoidable delay to the subwoofer. Professional digital crossovers like the Behringer DCX2496 incorporate adjustable delays, although their DACs aren't brilliant. Another way to deal with the problem is to have two DACs, one for bass and one for everything else and feed the "everything else" DAC via an obscenity delay such as the Sonifex RB-DS2. It all gets a bit fiddly because professional gear expects to see AES/EBU digits on XLRs. Arguably, the subwoofer DAC doesn't need to be as special as the "everything else" DAC. Something to watch out for when going down this path is to minimise the latency in the DACs and certainly avoid DACs that include a time base corrector (and therefore variable latency). But if a job's worth doing, it's worth doing well.

I've just looked up the price of the Sonifex and it's alarmingly expensive. Perhaps one of the forum's digital demons knows how to delay an S/PDIF data stream by up to about 50ms using a Raspberry or somesuch?
User avatar
Nick
Site Admin
Posts: 15859
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#53 Re: Kismet 2A3SE amp

Post by Nick »

I've just looked up the price of the Sonifex and it's alarmingly expensive. Perhaps one of the forum's digital demons knows how to delay an S/PDIF data stream by up to about 50ms using a Raspberry or somesuch?
Well, probably better to convert to i2s and then push that through a shift register. Simple enough job for a FPGA. Though for that matter should be easy for a pi if you have a spdiff interface board and a circular buffer. 50ms of 96k 24 bit 2 chan would fit into a Pi's memory space easy enough.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
User avatar
MistyBlue
User
Posts: 155
Joined: Mon Feb 12, 2024 9:38 pm
Location: North Somerset

#54 Re: Kismet 2A3SE amp

Post by MistyBlue »

Thanks Morgan/Nick.

I was in danger of another one of my classic over-look and forget mistakes that lead me up the garden path. I was meant to look at that phase shift going on in LTspice when I modelled it…..

The crossover frequency is currently at around 160Hz. Ideally it should be at 200Hz (as per Troels). At the moment, one of the things that is seriously impressive about the new speakers is the timing of the bass. I’ve never heard anything this good timing wise – not even from my friends Quad ESL989s.

Interesting stuff re digital. I am really trying to avoid that if possible unless it can be done at the “front end” ie at the initial DAC(s). Funnily enough, I have an RTZ build on the cards with PCM2DSD which is FPGA based, but I’m not sure whether it can accommodate delays or it’s just standalone. I remember one of the designers saying that the spartan FPGA on PCM2DSD is full. I did think of getting another PCM2DSD board made for the DSC2 that I currently use, so I can A/B test the 2 DACs easily.
….obscenity delay….
I like that. Sums up what I feel about it! Although it’s all useful tools.

I am not well versed at all with the digital stuff, and I have to rely on copying the work of others. So, unless there is a simple circuit I can get fabricated and solder up that can do the delay, and/or some simple uploading firmware type thing, I’m pretty stuffed.

Having said that, a Raspberry Pi is something I’ve been meaning to play around with for a while. I did consider one for controlling the x and y motion of the surface grinder, but it was simpler to use an Arduino.

One of the reasons for building the speakers was to get away from a software based (IR) filter used for lifting the bass level. Trying to implement that across several software platforms was doable, but a pain. Really, a global Windows over riding audio management system would have been better. Any software-based stuff needs to work across several software platforms as the system is used for movies, games, DAW etc as well as music. Then there is the whole software/OS update trashing settings etc that can make things go wrong very quickly, so I’m not too keen on that either. I remember looking at a rather longwinded article on ASR forum about Windows OS and wrestling control of the audio from it, but again, one overriding update that changes the options and its curtains.

Someone once said: “Hardware eventually fails, software eventually works…”

To which I’ll add, “until it’s updated”.

Makes a second order passive LP filter on the bass look like a simple option!

Or amplification stages placed on each output on the crossover, which I’m now seriously considering.

I do want to have a listen to the FE2022 front end as it can be used as a general-purpose amplification stage, and can run from the single rail 24V supply for the crossover. I only bought 2 boards though with the matched JFETs, with the primary intention of trying them with the MoFo. The cost of postage from the US has become prohibitively high as I discovered recently, so I got myself a JFET matcher along with some extra matched JFETs. I’m not against using op amps, but the system is pretty revealing of anything that’s lacking in the SQ department, so I’m nervous about just bunging some OPA2134s or whatever in there, despite the convenience.

Worked for Linkwitz though……am I missing something? :)
"When we're about to do something stupid, we like to catalogue the full extent of our stupidity, for future reference." - Commander Susan Ivanova, Babylon 5
Morgan Jones
User
Posts: 148
Joined: Fri Mar 25, 2016 10:58 am

#55 Re: Kismet 2A3SE amp

Post by Morgan Jones »

I fully understand your feelings on software. If you're happy with the timing of your bass without dastardly digits, then leave it well alone. For my part, I used the Behringer DCX2496 twenty years ago on my tranmission lines and was very surprised at the audible difference wrought by pulling tweeter and woofer into time alignment. All of a sudden, odd noises suddenly resolved themselves into musicians kicking microphone stands etc. And that was with a crossover frequency of 860Hz, so it was only 1ms or so. I would expect lower crossover frequencies to be more of a problem. There are probably better digital crossovers now than the DCX2496, but it's a terrific tool for finding out what a loudspeaker needs.

Obscenity delay was the name for it in broadcast. Broadcasters used to interpose a 5s or so delay into phone-ins just in case they got a foul-mouthed character on the other end. The delay allowed them to prevent obscenties reaching the transmitter. I expect they still use it.

I'm curious about your matched FETs; I've made a couple of FET VTO testers.

It's best to remember that Linkwitz's active system was a very long time ago. Things have moved on since.
User avatar
MistyBlue
User
Posts: 155
Joined: Mon Feb 12, 2024 9:38 pm
Location: North Somerset

#56 Re: Kismet 2A3SE amp

Post by MistyBlue »

The JFET matcher was purchased as a kit from here:

https://stompville.co.uk/cart/jfet-matcher-kit

Along with some matched pairs of 2SK170BLs. Many of Nelson’s circuits use J113 in matched pairs which are available from Mouser relatively cheaply in packs of 100. I’m sure I will be soulfully enriched by going through them and grading them! :lol:

JFET Matcher001.jpg
JFET Matcher001.jpg (97.97 KiB) Viewed 460 times

I think what I’m going to do is amplify the xover sections with whatever gives the best SQ and transparency and just be happy. I might try designing a passive LP option for being driven by the Sugden/MoFo. It will depend on cost of the coil/caps though.
"When we're about to do something stupid, we like to catalogue the full extent of our stupidity, for future reference." - Commander Susan Ivanova, Babylon 5
Morgan Jones
User
Posts: 148
Joined: Fri Mar 25, 2016 10:58 am

#57 Re: Kismet 2A3SE amp

Post by Morgan Jones »

You've probably already discovered that FET VTO is a bit tricky to measure. My circuit is similar to theirs, but very slightly more sophisticated. I test at IDS = 100nA for leaded FETs. I use a voltage reference to set the current reliably, rather than half battery voltage. Even so, I see self-heating (yes, even at 100nA!) so for testing SMD FETs, I test at 10nA. J112 can oscillate, so I added some bits to prevent that. I test at known (and adjustable) VDS. I also added a connection to go to a (screened) resistance box to allow me to change test current. I have a screening can that goes over the ZIF socket and that helps in protecting against air currents. Hot fingers change things too, so for critical measurements, I wait a minute. My testers can switch between four FETs. With care, I can get reliable measurements to 1mV on leaded FETs. I find the best way to find pairs is to bin the FETs roughly into perhaps ten bins (1oz baccy tins are your friend), then if you need closer matching than that, go through a tin to find pairs. That way, I managed to find three J111 matched to 1mV (VTO of about 8V). I had to go through about 300 to get that, mind. And after all that, the idea didn't work...
User avatar
MistyBlue
User
Posts: 155
Joined: Mon Feb 12, 2024 9:38 pm
Location: North Somerset

#58 Re: Kismet 2A3SE amp

Post by MistyBlue »

Thanks Morgan, some great information here!

I haven’t actually tried it yet, I only assembled it about a week ago. I had read about testing and how the results can vary depending on temperature etc. – I came across some pretty elaborate temperature controlled jigs which I thought was going a little bit too far for my uses. For me, it’s just about getting it right enough to give a good result for whatever circuit I’m playing with. TBH, I was more curious about having a go for future uses.

I was thinking of 3d printing a shroud to go over the board, but I like your screening can for air currents idea which I will factor in. I also like the resistance box idea. Thanks!

“And after all that, the idea didn't work...”

Oh crap!
"When we're about to do something stupid, we like to catalogue the full extent of our stupidity, for future reference." - Commander Susan Ivanova, Babylon 5
User avatar
Nick
Site Admin
Posts: 15859
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#59 Re: Kismet 2A3SE amp

Post by Nick »

Yep, they are very good temperature sensors. The JFET front end I use on a phono stage caused me all sorts of pain when I was first testing, with odd VLF wandering of the output. Only when I realized that the signal was correlated with moving near the PCB and it going away when I put the lid on did I realize it was air current on the jfet.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
Morgan Jones
User
Posts: 148
Joined: Fri Mar 25, 2016 10:58 am

#60 Re: Kismet 2A3SE amp

Post by Morgan Jones »

2mm thick EVA foam is your friend for reducing the problems of air currents. When I've matched a pair of leaded JFETs, I Araldite them together (clamp them with a plastic clothes peg). Then enclose them in foam, held on with thin heatshrink. Voltage references can have a foam house built around them (and foam on the underside of the board too). A sharp scalpel, a steel rule (to use as a cutting guide), a cutting mat, and a little care allow significant improvements.

Resistance boxes are often wood. Looks lovely, but it doesn't electrostatically screen. But you can line them with kitchen foil and connect that to the top plate to give a complete screened box. I replaced the connectors on my resistance box with BNCs. That way, I can make a screened connection all the way from the biscuit tin that encloses my test circuit to the resistance box. I glued bits of chipboard together to make an insert for my boxes, then folded the kitchen foil over it ("hospital corners" - ask your parents) to avoid gaps, then inserted it.
Post Reply