Gainclone heaven ?

[CoenP]

I concur with Peter,

LF below the cutoff frequency of the loudspeaker is probably worse than DC. even with 1Volt DC across the terminals, with an 8 ohm (nominal, 7 Ohms DC resistance)  this translates to 140mWatt of power. This is not going to blow your amp on transients.

LF below the c.o. frequency is not well damped by the speaker (if at all). This means high excursions wich is bad for linearity when playing loud, limiting the peaks and not to forget gives rise to doppler effects, modulating the signal.

regards, Coen

[Nick]

Hi,

My system has too much gain and I wanted to ask for comments and advice about losing gain with the least impact on sound quality.

To connect my NOS to my gainclone amps, as I am using phase alignment, I have 5.6 uf capacitors on the inputs to my gainclone modules. The input capacitors work with the 22k ohm input load resistor of the gainclones to give a high pass filter with a -3db power point of 1.3 hz (optimal point for response down to 13hz). All good so far.

The problem I have it that in my system as a whole the 2v output level of the NOS combined with the 28db gain of the gainclones and 105db/m/w response of my speaker means that I have about 15db too much gain in the replay chain.

This means to set a reasonable playback level using only the XXHE software  volume control to generate a "loud" listening level I have to play at a -30db volume setting in XXHE. I think that the 32 bit based volume calculations and 24 bid DAC have the headroom to play at this level but the issue is that -30 db turns the 2 volt full scale NOS output into only 6mv full scale output from the NOS ! I can hear that at this signal amplitude by the time the signal has traversed interconnects and gainclone input caps, sound quality is being impacted. I have been using transformer volume controls (TVC's) to "soak up" some surplus gain but I am very keen to take these out of the system.

My question is dose anyone have any ideas about how I can lose about 15 db of signal level. Ideally I would do this after the signal has been passed through the 5.6uf input caps if the gainclones. The obvious way is to pass the signal through a potential divider with resisters chosen to lose the gain. I have been trying this but so far have not had great results as far as sound quality is concerned (this might just be the resistor values I have tried so far). Does anyone have any ideas on a good way to lose 15 db of gain whilst maintaining sound quality ?

Thanks for any suggestions.

Nick.

[Leo]

Hello Nick,

I do not have the values and numbers of the resistors for you at this moment but it is fairly easy to change the gain of the clone to lower values. If I remember correctly you can't go too low as the amplifier would become unstable but a fair range of adjustment is possible. Probably this was discussed on DIYaudio.
Just my 2 cts
Leo

[PeterSt]

Leo, I guess that doesn't help much because it seems (to me) that Nick is missing some point in the "Gainclone basics".

So Nick, assumed I'm correct on this (which feels almost impossible - just saying) the Gainclones operate with gain resistors. See the datasheet. So, adjust those - done.
And *then* what Leo told applies.

Does this help, or was it only stupid of me that I thought you overlooked the obvious ?

Regards,
Peter

[Scroobius]

Hey Nick - you can set your gain to whatever you need without affecting quality. The circuit diagram of your amps is below (I think!) and the gain you currently have will be:

Gain is  =1+Rf/R3, so 1+(22.1K/680)=33.5dB

You can change Rf and R3 to give a lower gain without affecting quality.

I would have thought best to change Rf and leave R3 at 680ohm. If Rf is the same spec i.e. Caddock as in the original kit then quality should not be affected.

Cheers

Paul

[CoenP]

Nick,

You really can't lower the gain of the amps to any extend. Iirc the datasheet tells that below 20X (26dB) the chip becomes unstable. You win 7dB at best.

A additional resistor devider at the input is easily calculated but introduces some unwanted series resitance, noise and minute distortions. I would go for a low as possible impedance that the opt of the NOS allows (1K?).

Another option is the transformer route. My opinion is that is can be done at very high cost, but some may argue that it will never work.

Both are a bit like introducing a fixed pre-amp in the chain.

If loud is -30 dB you might consider an unity gain amp (a buffer). I don't know if these a commercially available.

Otoh I am shure lowering the gain somewhat will work. Peter demoes on very sensitive speakers (his own and Berts latest) and i've had read no reports on having too much gain.

Fyi I have similar gain issues with my tube amp and have not found another solution that to redesign the amp with lower gain.

Regards, Coen

[PeterSt]

Things are not completely comparable. IIRC my Gainclones are 30W and utilize a gain of 21dB. Can be 3dB less at least (I once had it like that).

Next I will play easily with -30dBFS (but -34.5 would be minimum by nature).

Lastly, nobody is really bothered by my playback level. That the room is some 290m3 does matter somewhat but not all that much (because of the directivity of the horns).

But where it all starts is my gain of 21 which is siginificantly less as the 38 I think I read from Nick's.

Peter

[Nick]

Leo, Paul, Coen, Peter,

Thanks all for your help with my gain question.

It's interesting that pointers are towards increasing feedback to reduce active gain and using passive attenuation, looking at it I think I may need to use both to drop 15db.

The gain factor of my gainclones 33.3 or 30.4db (thanks Paul). As Coen pointed out the LM3775 according the data sheet should not really go below 26db, so changing the feedback resistors could yield a 4db reduction. I'm worried that I may end up listening to a lot of negative feedback, but its easy to try so this could be option 1.

Option 2 is to apply passive attenuation on the inputs. The choice here would be to apply the attenuation before the input caps or after the caps and before the Gainclone modules. I guess after the caps would be best so the signal level is higher as it traverses the caps. The problem here is that I have just bought these rather nice (and expensive) mundorf silver gold oil caps which are the correct size for the standard gain clone module 22k input load. I like Coens suggestion about keeping the "input load" of a resistor potential divider as low as the NOS output can handle to keep series resistance of the input as low as possible. To lose 9 db on the input I could go with a potential divider of 820 and 270 ohms. The problem is at these values I would need input caps of 100uf a bit bigger than the new 5.6uf  mundorfs. Hummmm a bit to think about here.

Thanks guys for the help so far.

Nick.

[CoenP]

Option 2 is to apply passive attenuation on the inputs. The choice here would be to apply the attenuation before the input caps or after the caps and before the Gainclone modules. I guess after the caps would be best so the signal level is higher as it traverses the caps.

I like Coens suggestion about keeping the "input load" of a resistor potential divider as low as the NOS output can handle to keep series resistance of the input as low as possible. To lose 9 db on the input I could go with a potential divider of 820 and 270 ohms. The problem is at these values I would need input caps of 100uf a bit bigger than the new 5.6uf  mundorfs. Hummmm a bit to think about here.

You're welcome!

I would first try to increase feedback on the amp and assess then if you are really in need of more attennuation.
The way the LM works is that I think a little extra feedback will not hurt sq (au contraire).

I see no reason to be afraid of low signal levels through the caps. If you need the divider please put it before the cap, you can keep you precious gold cap for minimum distortion. AFAIK larger voltages across caps can cause audible minute deforming of the dielectricum, so I'd rather keep them small.

regards, Coen

[Scroobius]

Nick - this is a quote from Peter Daniel on the Audiosector forum about gain maybe As discussed above R1 will affect SQ but its easy enough to try it. Also I suppose that R1 at 22k may affect the L/C equation for the input caps but again easy enough to work out.

Quote
It's best to change the gain with R3, but I wouldn't go over 1k6 or so. If you need additional gain trimming, increase R1 (22k will give you -6dB gain reduction)
Unquote

Maybe 22k input R will not be ideal given the low output impedance of NOS1 - any comments Peter?

Am looking forward to hearing your system. After all the mods I have made to my system recently I have some spare amp boards (populated and burned in) I measured the quiescent offset voltages so if you measure the offset voltages on your amps we may (but we will need some luck) be able to find a combination where the Left Right offsets are reasonably well matched (at least for your bass units) which I assume (?) do not have capacitors in series with them in the crossover. I have matched my amps to within 5 mv offset for the bass units and it was worth doing (previously I had about 60mv offset difference) removing "confusion" in the stereo image and just sounded more coherent overall.

Paul

[Nick]

Coen,

Thanks again, I'll start with the feedback loop. I’ve been messing about with zero feedback valve amps for a few years before the gainlones came along so the idea of adding more feedback didn’t sit too well. It’s good to know that more feedback in a gainclone could be good for sound quality  .

Also its interesting what you are saying about putting the passive potential divider before the caps.

Time to get some parts to give it a go.

Regards,

Nick.

[PeterSt]

Maybe 22k input R will not be ideal given the low output impedance of NOS1 - any comments Peter?

Well, why not ? It almost looks like you want it less ? (knowing that 47K and up will be normal for main amps)
Anyway, less is OK just the same. But I'd stop at 1K to be on the safe side (NOS1 is ~33R). And otherwise you'd hear it on the too tinny sound.

Nick, by all means try to avoid any (more) resistor in the signal path.
Also, the application of a feedback resistor is a most normal thing for these amplification means. It would be true though that there's an optimum for THD (and R value). Well, that's what I think. And when it is, it should be in the datasheet.

I would first try to increase feedback on the amp and assess then if you are really in need of more attennuation.

I agree with this. So, change RF as long as you can, before touching anything else.

Peter

[Scroobius]

For anyone interested this is my latest Gainclone layout.

As you will see there are four amps for bi amping. Recently I rebuilt the amp modules selecting LM's to give small offset between channels (all in the interest of better phase alignment). I took the opportunity to simplify the layout of the amps by removing the aluminium heat sinks and bolting the LM's direct to the case. There is a little more internal wiring now but not much. The aluminium sinks were only there originally because I was lazy having used the Al sinks to breadboard test the amps first I just transferred them straight into the case. Much better without them.

One important thing I removed the plug and socket I used to have in the power cord that runs from the power supply to the amp. The reason is that if one of the power lines is lost to the amp 35V can appear across a speaker drive unit. That was a risk not worth taking so everything in the power delivered to the amps is hard wired.

The SQ nowadays still surprises me and I have to say when I look at the amps and think how much they cost to build I do smile from ear to ear as they "wipe the floor" with much much more expensive amps.

[Nick]

Hey Paul,

Thanks for posting the internal shots, we have ended up with very similar layout of the modules but in the next version I build am going to put the transformers in a separate case like yours. Good point about hard wiring the psu connections I think I will do that too.

Now i have to work out how to get a good layout with 8 modules and 4 transformers 

Cheers,

Nick.

[PeterSt]

Paul,

Stupid question : are those transformers loose in there ? And if not, how do they stick ?
But if yes, try the sound when putting them on top of eachother (but have some pad in between them).

Regards,
Peter