CraigBuckingham

Why not run an earth line dedicated just for your hi-fi system? Out the window or through the floor or over your shoulder even.

Just read the Yahoo review. It's no surprise that an iPhone may sound better than a Pono. Most analogue design by big corporations, if you can call using a semiconductor manufacturer's application notes "design", these days usually only sounds good by accident. I have a box full of earphones, some $99 Philips ones in there also with all the bragging rights from the marketing gurus, with a set of earphones that came with a $60 Sagem FM radio with no bragging rights that blows them all into the weeds. It is with great trepidation that I buy headphones, earphones, speakers based on price. With speaker it's usually better to build a set.

Looks like a shit load compression to me. If the 20 second lead in is compared, very slight increase in amplitude, almost nothing. The main content after 20 seconds is compressed to high heaven. The modern compression sickness that's like the bubonic plague. Gotta keep the main demographic (and profit group) happy, kids walking around in noisy environments with portables. Pono player, well it's going to maybe sound better than other portables, but it's never going to be pro or hi-end. It's full of op-amps, feedback and no class A and no class full stop.

nopants, sorry for the late reply, not always reading all the threads here frequently. I would prefer you don't post anything as "my mods", instead maybe you can try one thing at a time and if you like the result, post it as something you tried without any reference to me. It's a clone and I don't want to actively promote diversion away from the intent of the thread and project.

Before Self touted optimally biased class AB stages to reduce crossover distortion it was common knowledge in the industry for a much longer time that higher bias in a class AB output stage sounded better. I don't believe that Self's publishing of these articles changed the way AB output stages sounded with respect to bias current. They were nice looking graphs however. Really nice woodwork and great story about it also.

That makes things easier.

I do prefer isolated tabs. Does make mounting easier. It goes without saying that isolated heatsinks are a solution. But if it's a clone amplifier or a repair then that may not be how the original was designed. The data sheet device parameter curves are interesting to me to their differences given that both are stated as being Triple Diffused Planar types.

I believe under the JIS system that Sanyo had exclusive use of the 2SC3675 numbering. The right side transistor is obviously not Sanyo. The one on the left looks like a perfect copy or genuine. If the numbering is not etched below the surface and printed instead or the tab and leads are not all copper, completely non-magnetic it's a fake.

Since it is off topic and is probably not what the majority of members would agree with I'll refrain further from any suggestions keeping with the spirit of the topic being a clone. It would be more appropriate to create another topic or probably keep it to PMs.

I thought spritzer would be first to call me out on this. Anyway, 1.5V offset is way too much, there is a fault.

UFN, there is obviously a fault. For a 2N5564 with a data sheet specification for maximum offset of 5mV between the pair, this circuit should see a maximum of 25.6mV output offset for a worst case 2N5564 with the servo disconnected. This thing doesn't need a servo if it's using a matched input pair like the 2N5564. The servo can be removed and a trimmer potentiometer used to trim R3 and R4 to shift the drain currents of the input pair to null the input offset and therefore the output offset. R45 could be removed and R46 replaced with a 750Ω resistor to keep the gain the same. With the added benefit of removing the LF pole created by C8, C9 not the least the extra distortion caused by those electrolytics. So a true DC low frequency response.

That's the same Douglas Self that had a whole lot of articles published in Electronics World regarding optimally biased class AB output stages? Let's just say that there are many famous amplifier designers that don't follow that mantra. EDIT: Changed Wireless World to Electronics World. (Showing my age)

This is a mocking thread right? €2999 (discounted from €3999 ) for an op-amp based headphones amplifier? Forget how good the 64 step relay based attenuator may possibly be. Once it's gone through those op-amps you can kiss your quality reproduction good-bye. EDIT: Stick with KG and others discrete or tube designs and you will be light years ahead.

I would just like to add, I thought I mentioned it in my previous post. In my experience good quality multi-turn trimpots like the one used in the KSA5 are usually very reliable on the trimmer contact. Much more so than other types. However, I would always prefer a design that didn't take unnecessary risks.

Ok Purk, no need to prototype, it's a just a matter of simple design. See below. Grawk, thank you, I never said I wanted to make a KSA5 derived amp. I actually said I wouldn't use that circuit. I do know what the word "clone" means, however thank you for clarifying what the thread was about. xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx In the original design, if the pot wiper goes open circuit for any reason, such as doing adjustment to the OP (output stage) bias, the OP will go to the maximum bias the Vbe multiplier circuit will allow. As you may know this is not at bad as a 3 terminal pot with wiper connected to the base. For this arrangement a wiper contact failure would result in the OP stage shorting the +ve and -ve rails to each other. I have seen this arrangement in some commercial equipment. At a minimum It generally results in destroyed OP stages. For the KSA5 an open wiper would cause about 500mA current through each OP transistor, however this may not be realised due to the power supply reserve to supply it. If the power supply can source the current it would result in almost 10 watts power dissipation in each OP device. The heatsinks used for the OP devices appear to be Aavid Thermalloy part number 513002B02500G with a thermal resitance of 13.40°C/W. Assuming 500mA bias and 20 volt rails the MJ15030 OP devices would be running junction temperatures 159°C above ambient. This level of heat exceed SOA and would destroy the OP transistors very quickly. Even if the power supply could only supply half the current, 250mA the OP devices would have junction temperatures 80°C above ambient. Whilst inside SOA at 25°C if the amp was in a hot environment it would be running near the limits of the OP device limits of SOA. Here is a modified Vbe multiplier that avoids the flaw in the original design for anyone that is interested. The modified circuit would result in an under biased condition for the OP stage for a wiper failure scenario. It only requires resistor value changes and possible a reverse mounted transistor depending on the type used. EDIT: Correction - mistake made on diagram text. I mistakenly though the base was centre pin. Looks as though the collector and emitter pins would need to be crossed with the MJE15031 PNP and it mounted in the convential way (not reversed). The Toshiba would need to be mounted in reverse and the collector and emitter leads crossed. EDIT2: Updated image text to correct error mentioned in previous edit. (GMT: 01:14 18-09-2014)

That's great and a lovely job as well. Credit where credit is due. However you could have had the same physical layout and with some finesse with component values and the transistor used for the Vbe multiplier fixed up the blunder in the original design that has potential to compromise the output stage reliability. Just because Krell designed a potential hand grenade would it be fitting of a clone to leave it that way? I don't want a clone or a clone of a clone. I wouldn't use the circuit Krell used.

Spritzer, I don't understand how you knew the caps in my SRM-313 were good and I replaced them with bad ones. I don't recall you being in my workshop when I removed them. Or how you gained access to the measuring equipment in my workshop when I measured those capacitors. Do you know what a choke is and how it can be of benefit in a power supply? I have added to the community, you just don't understand what I am adding. If you speak to anyone with knowledge on the subject matter they will confirm that you are incorrect about your assumptions.

So you don't want to fix the mistake? Whatever makes you happy. KG should know what's wrong with it. I wouldn't be running it that way.

Spritzer I read the introductory letter. I didn't start on you, it was the other way around, you went out of your way to start on me. You read enough to jump in and make a joke about my post. Then you started beating your chest about your engineering prowess and with the put downs. But when it was time to cough up on your experience, you took off like a scared dog with his tail between his legs. Grawk, you don't see the hypocrisy? It's a Boston thing is it?

Did anyone fix up that poorly designed Vbe multiplier in the original schematic that KG published on the 1st page of this thread?

Spritzer, I was waiting for your engineering experience on the topics that were raised. You were after all lecturing me on how to do it and criticising my endeavours. I mean you have a truck load of posts here, I can't find it though. You only contribution was a disparaging vague initial post towards me. You came out all guns blasting and then lost interest real quickly and went posting elsewhere. Seems like the introductory letter doesn't work both ways.

"Head Case" forums . I completely missed the significance of the website and forum name until I read your and luvdunhill's posts. Thanks for the laugh guys.

I apologise, I can read and type fast so it doesn't seem so long to me. I suppose it's a blessing to be able to read, some people can't read at all, so you could take some solace in that. If this is par for the course I'll try and slow the pace down if that helps.

Yes, it was the K340s of 79 vintage. The 313/404 combo I had were brand new in the box from a distributor in Japan. They probably measured well, but lacked colour and sounded grainy. The K340 weren't perfect but were a lot more musical and smooth. Since the source was the same I put it down to the 313. The caps and chokes were enough to make it jump ahead of the K340s, but not enough in my opinion to justify the price to performance ratio. The feeling I had was that amplification for the the electrostatics was possibly going to be challenging using semiconductors and that a vacuum tube (or hybrid) design was maybe the better way to go. I didn't end up using them for R&D work and instead used a speaker setup. The choke mod was a pi CLC network. So, rectifier -> parallel electo -> series choke (inductor) -> parallel electro. The chokes made a big improvement over the standard filter capacitor arrangement. I A/Bed that, only took a few seconds to win me over. Still have those chokes in my spares somewhere. Local transformer manufacturer made them to my specifications.

Now onto some engineering which is of more interest to me. spritzer, I don't have a schematic of the 323. I know the layout of the 313 however so I will provide a response to your assertions on how to improve it based on that. No issue on increased class A biasing. Well known to improve performance generally. However safe operating areas and ripple current increases on existing designs need to be recalculated to ensure reliability. I am sure KG wouldn't miss that. Finding good capacitors is necessary if you want to obtain quality audio reproduction. If you are not going replace sub-standard ones in existing designs then there may always be a limitation. As to the Jensens being garbage, maybe they have changed since 2006. But the gold Jensens were much better than the standard IIRC Nichicon in the 313, and Elna cerafines. It was an easy pick out of the bunch. I had/have some cheap Taiwanese brand that beat all the best audio grade caps I have put against them. But they were not available in high voltage versions and I didn't have time to muck around with a relatively expensive headphone set that were being challenged by some garage sale $10 AKG-340s. Sure the Stax just pulled ahead after the mods, which included some customs chokes I designed and had made. I apologise if I missed your posts on your choke modified power supplies spritzer. I am interested, how did you find choke based unregulated supplies versus regulated ones when you did A/B testing with the same unit? In my experience regulating the supplies can be an improvement, but it's 2 steps forward and one step back. And that one step back can't be fixed unless the regulators are removed. This may appear cryptic, let me explain another way. There are better ways to improve the power supply rejection of the amplifier, series pass regulators using an active device from my experience are about the 4th best way to do it. There are at least another 3 ways to improve on that. The advantage of the differential amplifier design is that power supply noise induced into the amplifier circuit can be made to be virtually all common mode. If the power supply noise is analysed in the differential and common mode domains a lot of optimisation can be made by thinking of it that way. Single ended circuits can also benefit by some of these methods, even though there is no common mode domain. Servos are another can of worms. If at all possible they should be avoided which I am sure most would agree is common-sense. I know KG believes in not putting the correction back into the node used for the feedback loop. But even the terrible artefacts that IC op-amps display when used as amplifiers can be heard when used as a servo in high resolution systems. By putting them on a DC node used for bias stabilisation is better but there are ways to decouple them so the output is not directly used. Spritzer, sorry if I missed your optimised servo designs. Can show me where to find the ones that you optimised over the ones KG uses on the DC nodes of the current sinks/sources. Too big a post again? I'll end it here.