Would you suggest driving the amp with a square wave signal, and look for ringing (under damped response) and so on?
Does anyone have any boards spare for the 100W version? If not, would anybody be interested in buying some from me if I order some?
Thanks valley711, could you ship these to the UK, or would that prove too pricey? I'd want two, of course.
Thanks, I would prefer ENIG, but it's no big deal. It will come down to price I expect, and whether there is anyone in the EU who'd share a PCB order with me. How much do you want for them and what will it cost to post them to the UK?
Hey Suzy, reading about your testing of the 100 W version, I see you swapped R48 for a 5k1 MELF (http://www.suzyj.net/2017/01/measuring-mosfet-power-amplifier.html) but that isn't reflected in the construction notes. Were there other changes you made?
It was part of the diagnostics in tracking down the poor thd that I was seeing. It's explained in the blog post. Reducing the feedback resistance from 15K to 5K1 should have reduced THD by a couple of dB. It was much more than this due to using a thin film resistor in place of the crappy thick film one. There's no need to use a 5K1 resistor in practice. Just use a thin film 15K one.
Thanks for the explanation, sorry, I didn't spot it in the post - I had seen the discussion about thin film vs thick film, but hadn't spotted the return to a 15k resistor. Really looking forward to building this amp!
I have completed a build of two modules that use the circuit layout shown in post 352. Some minor changes to the compensation and servo but otherwise no issues. A 200kHz squarewave test (with input filter removed) show a slight ring but it is well damped and only lasts one cycle. So it is fast but no hint of instability.
It should be noted that a heatsink is needed on the drivers as the second differential pair runs about 350mW per device. This is much less than the drivers in the original AEM6000 which according to LTSpice are over 1 watt each. I believe running the VAS and drivers at much lower current than in the original design is part of the magic that Suzyj brought to this party.
I did set the output bias at 100mA which I believe was used in the AEM6000. This did further reduce the distortion at both 1K and 10K from that observed with 50mA bias. The reletively massive heatsink of the Disipante case is just slightly warm to the touch at the higher bias.
The sound is wonderful. I had my Yamaha B2 in the system for a week while the new modules were being installed in the amp and again I can't say that there is any difference in the sound. The distortion does measure lower than in the earlier modules, but if it is detectable in hearing tests one would need a proper A/B test setup to instantly go between the two designs.
It should be noted that a heatsink is needed on the drivers as the second differential pair runs about 350mW per device. This is much less than the drivers in the original AEM6000 which according to LTSpice are over 1 watt each. I believe running the VAS and drivers at much lower current than in the original design is part of the magic that Suzyj brought to this party.
I did set the output bias at 100mA which I believe was used in the AEM6000. This did further reduce the distortion at both 1K and 10K from that observed with 50mA bias. The reletively massive heatsink of the Disipante case is just slightly warm to the touch at the higher bias.
The sound is wonderful. I had my Yamaha B2 in the system for a week while the new modules were being installed in the amp and again I can't say that there is any difference in the sound. The distortion does measure lower than in the earlier modules, but if it is detectable in hearing tests one would need a proper A/B test setup to instantly go between the two designs.
In the blog, it shows lowest distortion with the lower R value MELFs. Are there stability issues with chasing the lowest distortion?
Are these ok for the JFET at Q1?
https://www.digikey.co.uk/en/produc...eEFa9Ib4b9Y1zdn-pglEpV2p03nv4Ugr_A2kURXdPm8JT
https://www.digikey.co.uk/en/produc...eEFa9Ib4b9Y1zdn-pglEpV2p03nv4Ugr_A2kURXdPm8JT
I've just ordered 5 100 watt boards from JLC pcb in black, enig finish. If anyone in the UK wants a couple at cost, please PM me
Suzyj, I just made a couple of PCBs for your amp as well as the original AEM6000 amp. I have a couple of questions about heatsinking. On your board, is it sufficient to put small TO-220 heat sinks on the transistors? On the AEM6000 board, I thought of having all the TO-126 and TO-220 transistors on a common heat sink bar, as you can see below. Is this a good or bad idea? And thank you for a superb series of amp threads!
This PCB I made to replace the amp modules in a Roksan Kandy chassis I bought. Thus the irrelevant name.
This PCB I made to replace the amp modules in a Roksan Kandy chassis I bought. Thus the irrelevant name.
Based on LTSpice modeling of different variations of the AEM6000, the dissipation of the drivers is quite dependant on the current you choose to run them at. I believe Tilbrook's original design ran at a high enough current that the drivers dissipated over one watt. He mounted them on the main heatsink, if memory serves.
