An Olson audition amplifier inspired build (completed)

[Gabe001]

Disclaimer: I'm not a professional at this

A few weeks ago I came across a circuit for an Olson audition amplifier, and I decided to build one using a similar design. It struck me as quite interesting, no feedback, excellent measurements, easy to build, all octal and highly sensitive, meaning I could drive it from my phone directly. Previous experience indicates that this requires an input sensitivity of around 1v p-p (350mv RMS). I suspect the voltage is limited so that they aren't sued for causing hearing loss.

The original Olson circuit is attached. Of course, since he worked at RCA, he no doubt had a few constraints with regards to components and valves which likely explains his choice of valves (6f6s) for the output stage. Luckily, I have no brand loyalty, and decided to use what I had in stock i.e a pair of 6v6s. Olson runs a quad of 6f6s in triode mode. I thought this would be a good opportunity to run a pair of 6v6s in ultralinear mode and managed to locate a 10k OPT with 20% taps from china (mableaudio) at a very reasonable price

Olson set out to build a rather low power but high quality amp. He states that one does not need more than 5w to recreate an orchestra in the living room! My ultralinear transformer would sacrifice some output power for more hi-fi sound. The project was build on a chassis purchased from a forum member. I had all the valves in stock.

The original Olson circuit is shown in the attached pdf.
The circuit with my modified output stage is shown in the attachments. The circuit was constructed in LT spice and simulated. Further modifications were necessary as my HT was much lower than Olson's, so I used a solid state rectifier, upped the smoothing capacitors, reduced the resistors and modifies the cathodyne resistors as otherwise 6sn7 at lower HT wouldn't have drawn enough current. I implemented a bit of global NFB to reduce the input sensitivity and used a 100k pot instead of a 47k pot + 47k resistor.

The underchassis layout is attached (I'll be the first to admit that I'm not an artist!) and a few pictures of the final product including the wooden cradle I made for it. Speaker sockets allow for 2x8 ohm speakers connected in parallel, and the left and right inputs are summed using 2x220ohm resistors.

Continued...

[Gabe001]

Continued....

These are the THD figures I've got using free software available online. The amp was fed a 1khz signal from my phone and this was analysed using the computer's sound card. A 10:1 voltage divider was used to limit the voltage into the computer.

THD 1w 0.11%
THD 4w 0.35%

THD remains below 1% until 5.8W power output, then rises rapidly. The 6v6s (russian equivalents) are running at around 275v on anode and plate.

Frequency response shown in attachmemts
Input sensitivity around 1V p-p with around 5db feedback
There is a slight background hum only audible with ear to speaker at full volume

Regarding sound quality, my wife says that it sounds better than my other non ultralinear 6v6 p-p amp. I must admit I can't hear much of a difference myself

Any comments or constructive criticism welcome. If there's anything I've done wrong kindly let me know.

Ps: I know that the volume control is in within the NFB loop,but I'm using so little feedback that it hardly matters.

Thanks
Gabriel

[Diabolical Artificer]

Regarding sound quality, my wife says that it sounds better than my other non ultralinear 6v6 p-p amp

Your wife sounds amazing. Re how amps sound, I've found the differences between valve & transistor amps and between one valve amp and another are subtle. It can take a bit of time before my ears get accustomed to an amp and I can spot differences. Your wife has a good ear by the sound of it.

Good work Gabriel. I can only see one "fault", looks like the ground is connected to the chassis in two places but I might be seeing things. Your resistors don't all lie the same way, EG percentage line not all facing the same way but that's a very very minor quibble.Your ground bus isn't 100% straight, another minor thing and some components are "flapping around in the breeze" a Davism and another not very important observation. Noticed your OP is connected to two speakers, have you taken into account your load will be halved?

All the above apart from the potential chassis ground loop are not important, what matters is that you built it, it works and the figures look very good too. I'll have a look at that Snowflake software, I usually use Soundcard Scope or Arta like you have for the FR. The sidebands on the FFT are odd though, maybe a artefact of the software or AI.

One last observation, the Olson amp has V1 & V2 running at low gain maybe because there's a paralleled OP stage and two gain stages, you could have increased the gain here and got more than 5w out,but I guess you know that and are happy with it as is.

Again, well done and look forward to your next build, Andy.

[Gabe001]

Hi Andy

Thanks for your comments. I can confirm that the ground is connected to the chassis at a single spot. Also I did compensate for the 2 speakers in parallel with regards to the load

The software you called snowflake (which is just the file name I used to save the project) is REW, roomeqwizard. It's available for free online and is pretty good but a bit more complex than Arta. I end up using them both. They give similar results.

Regarding the readings I seem to get best results by inputting the 1khz signal through my phone instead of the computer soundcard, which tends to increase distortion figures by about 0.15%. the phone 1khz sine wave signal seems cleaner unless I crank the phone volume up in excess of 80%. The best results seem to be the phone at minimum volume and the amp at close to max.

The lower than expected output I'm sure is down to the ultralinear OPT quality - the grid to grid swing is in excess of 30v which should be enough to drive the Russian 6v6s. I could also have probably done with more HT on the 6v6 anode and grid (300v instead of 275) but that would compromise the smoothing, it's plenty loud for domestic use and I want to stay married.

According to my wife, she can tell the difference from the vocals of Kurt Cobain (nirvana). She says that this one sounds more like she is used to. I think it may be due to the difference in harmonics between the two amps, the other one has a more prominent second harmonic which I suppose can alter slightly the tonality of the vocals. This is my theory anyway.

This was a fun project, although it ended up taking longer than expected. I have a very run down Mullard 5-10 chassis in the garage which I may do next if the opt and mains transformers are good. I'm actually undecided between the mullard or building a stereo amp in 'Armstrong' style. I'm a bit apprehensive trying to build something from scratch with masses of gain and masses of negative feedback as it seems incredibly easy to cock them up, which is why I did this more forgiving one first.

[joebog1]

HHMM why not build a 5-10 as is ?. THEN change it. A decent built 5-10 is a pretty fine amp that suited many many grandfathers of that time. ( They are grandfathers now, I mean ).

REmember wheels are round, and your design will most likely be out of round. At least for the time being.
My computer is for computing, not audio!!.
Digital audio sounds woeful!! including many " analysis " programs.
I use a CRO and my good old fashioned VTVM to analyse performance.

So instead of curry powder, or pepper sauce, use your growing knowledge.
My pun at "spice" modelling!!.

NICE first attempt at bending something from an "article " .

Joe

[Gabe001]

Hi Joe. The 5-10 will be built stock. No modifications. I don't see the point of modifying something that's as good as it gets (although some may argue that the gain may be too high for modern sources)

I agree with you re digital audio. I can hear the MP3 compression artefacts on certain songs and it can be painful.

[Valvepower]

Hi,

Nice job, well done…

I’ve amassed about half dozen homemade 5-10’s and they all work OK. I rescued them before the transformers were harvested

If you follow the later version of the 5-10 and get it going, you can tweak it to meet your needs. As a guide I have the circuit for the STD amplifier, which basically uses the 5-10 circuit with a triode connected EF86 to decrease the input sensitivity.

Regards
Terry

[bikerhifinut]

Originally Posted by Valvepower

Hi,

Nice job, well done…

I’ve amassed about half dozen homemade 5-10’s and they all work OK. I rescued them before the transformers were harvested

If you follow the later version of the 5-10 and get it going, you can tweak it to meet your needs. As a guide I have the circuit for the STD amplifier, which basically uses the 5-10 circuit with a triode connected EF86 to decrease the input sensitivity.

Regards
Terry

Just a thought Terry,
In much the same way as I decided to follow the HJ Leak path on my EL34 build rather than the usual 5-20, as I personally think the leak designs seem to need less "tweaking" if built pretty much to the original circuit. Would it be an idea to follow that route.
I only ask as you are very aware of the growing pains on my build, although it does work a treat now. And isnt really a 5-10 nor a Leak due to a couple of tweaks to lower the overall sensitivity.
Andy

[Robert Gribnau]

Nice build, Gabriel!

Maybe this article is usefull for comparing your measured results (if you didn't already find this article yourself).

[Gabe001]

That's a great article Robert. Seems like 6w is the norm with 20% taps.

[cathoderay57]

Another fine result Gabriel. Keep up the good work!
Jerry

[Gabe001]

Well, I found this for a quad 2 and it looks like I have a log way to go. It looks incredible. I'm not sure how it's possible to get that degree of performance! There's almost nothing in the 50 and 100hz bands

[Diabolical Artificer]

Can't get on with REW, the cal process is tortuous, Arta is better. Ignor Joe, he's an old fuddy duddy, PC software can do a frequency response in 30 seconds, it takes hours with a DMM + paper. That said ARTA et al only go up to 20khz, where frequency response gets interesting is at the extremes, IE under 50hz over 20khz, plot that with THD and you see what an amp is really up to. So yes, Joe has a point, but you have to be really experienced to see THD on a scope and have a damn good scope that can trigger and display @ 10hz and under.

Well, I found this for a quad 2 and it looks like I have a log way to go

Ahh, but can you hear the difference? I've mucked about testing my amps, tweaking this component here that component there but with my ears and my speakers I'm b*ggered if I can tell the difference.

The other thing is to be able to hear the subtle differences as shown in an FFT plot or whatever you have to sit down maybe for hours on end training your ears, tweaking room EQ, swopping speakers and what have you. I'm more into rushing about, Grandad dancing whilst I stir the soup, supping beer & puffing a smoke and having a bit of a sing song. Who wants to turn into a robot for listening, that way madness lies. Music is about life not clinical analysis. Think I've contradicted myself there : )

Andy.

[joebog1]

Think I've contradicted myself there : )

AGAIN Andy . Ahh well, without the fiddlers there would never have been a Leak, Quad or Andy's other fine amps that still carry huge weight in listening tests.

Joe

PS. yes an old fuddy duddy, BUT I still like my good sound.

JB.

[Gabe001]

I'm a bit of I tinkerer I guess, and there are some quick wins. I've identified the 100hz spike to inadequate smoothing on the anode of the cathodyne, so I should be able to bring that down quite easily. Also, according to Roberts article there may be some mileage in increasing the bias from -17.5v (currently) to -19/-20v with regards to thd/total power. I'm confident I can make it a bit better. It will never be a quad but at least it won't be embarrassed by one.

[cathoderay57]

Well, if you're playing Nirvana, maybe the distortion is deliberate . For a subjective listening test I find classical piano music quite useful. Some notes seem to emphasise any distortion present. Cheers, Jerry

[Valvepower]

Hi,

Andy is right the Leak TL12+ is also a viable circuit to copy.

Andy did a cracking job with his home brew amplifiers. Yes, there were growing pains, but no more than I’d expect for a ground-up build like Andy did.

Terry

Originally Posted by bikerhifinut

Just a thought Terry,
In much the same way as I decided to follow the HJ Leak path on my EL34 build rather than the usual 5-20, as I personally think the leak designs seem to need less "tweaking" if built pretty much to the original circuit. Would it be an idea to follow that route.
I only ask as you are very aware of the growing pains on my build, although it does work a treat now. And isnt really a 5-10 nor a Leak due to a couple of tweaks to lower the overall sensitivity.
Andy

[Gabe001]

Thank you. I've improved the HT ripple and increased the 6v6 bias to 19v and I'm happy with the finished project. The maximum I can squeeze out If f it is 6.3watts which is close enough to expectations based upon Robert's article. I don't think Olson would have been disappointed - the parameters match the ones he described with a THD of 0.1% at 1w and 0.26% at 4w

My next build will be a stereo ultralinear el84 p-p model. I'm leaning towards the Armstrong 220 but the leak tl12+ looks interesting too.

[Valvepower]

Hello,

I’ve an Armstrong 220 and it’s a nice amplifier and worth looking at. From memory it uses the Floating Paraphase phase inverter.

Terry

[Diabolical Artificer]

From memory it uses the Floating Paraphase phase inverter.

I've always been leery of the FP, sounds like a karate move, looks over complicated but is probably fine.

Andy.