How did Dynaco get into the discussion? Are you guys taking some funny thing this time of the day?Actually Soong, you have now provided enough information for us to put your measurements in their proper context.
The extra info I've asked for is in case you might want to address some of your measurement issues. eg the earthing on your Dynaco is obviously different from your MyRef and also how you've connected your measuring gear. Those versed in the art can deduce that from your measurements alone.
Your sound card and whether it has balanced in/outs determine how easy it is to sort out these problems but if you feel there are no problems, there is nothing else to be said.
All connections are single ended as so are the MyRef and XL280. Load is driven by the same amplifier in all cases.
Yes I have purchased the boards and I have explained exactly why I did as well. The difference is I am not trying to redesign it on a basis of golden eared feedback and thinking I know better, combined with thinking I can infer performance off a single graph.
Thinking that you can infer anything about the performance from your graphs is either genius or lunacy.
I started the build thread... I also stated up front that I was not going to be giving lots of waffly carp about how it sounded, or invite self proclaimed 'reviewers' to help me decide which brand of capacitors to use. My installation is almost unique and the other person running the same setup uses class D so any listening notes would only be of interest to my ego and that other person.
If you're expecting an effects box that will color the sound, the Modulus-86 is not for you.
If you are expecting a straight wire with gain and high output current drive capability, the Modulus-86 and Parallel-86 would be good choices. Neither the MOD86 nor the PAR86 color the sound. They're sonically transparent by design.
~Tom
You know, the Hafler XL280 was advertised to be straight wire with gain. Apparently it had not accomplished the goal. The MyRef is a step closer after some optimisation. Perhaps you want to do the Hafler straight wire test? I actually got a kit for that as well. But the fact is, different straight wires really behave differently.
What I would recommend is find a lounge that has live, unamplified performance, sit there and really feel what the performers are doing. The listen to your system and compare the feel. I spent lots of time in Dallas doing that.
Last edited:
...and I have the measurements to back up my claim that the Modulus-86 is as close to a straight wire with gain as you can get today.
~Tom
David Hafler was a founder of Dynaco. I forgot he'd changed the name of his second company. Mea maxima culpa
BTW, an important reason why he could claim "straight wire with gain" is that he tested this proposition with DBLTs. You may wish to try this with you own Golden Pinnae too
Soong, I believe your position & authority on the topic is now very clear .. as is ours. Any further discussion would be useless as you don't believe you have a problem with your measurements.
____________________________
The one item not settled is why are you interested in Zo < 0R1 when you are using current drive.
My main interest in following this thread is I want to make a current source that has Zo > 1k to 20kHz and also 1ppm THD20k. Probably (the 2nd criteria is) an impossible target. I'll settle for the first criteria.
An important requirement would be a very clear idea about grounding & earthing .. as Tom would have agonised and spent a lot of time optimising for Mod-86
As a real beach bum, I have problems doing stuff with solder but the existing SPICE models of LM3886 don't seem to model some of the stuff I need. eg I can't have a Zobel for Zo > 1k at 20kHz and I know from real life that a Zobel is required for unconditional stability into real loads with 3886.
Tom has loadsa good stuff on the topic in http://www.diyaudio.com/forums/chip-amps/217790-lm3886-effect-compensation-network-cc-rf2-cf.html ... but alas, nothing that helps someone after a current drive amp.
Last edited:
The TINA-TI model for the LM3886 is quite decent as far as stability is concerned. It's a bit optimistic on the phase margin. Except for the peak output voltage, anything related to the supply pins is not modeled. The mute behavior is not modeled. So it's far from perfect. However, for general stability and transient sims, I find the model to be more than adequate.
1 kohm @ 20 kHz would mean a 8 nF Zobel cap. I don't know if that's enough. You can run a sim to figure that out. Include the parasitics on the resistors (1-2 pF end-to-end and 10 nH in series - so (R+L)||C) in the feedback network. You should be able to see the effects of the Zobel that way.
~Tom
Trouble is, a lot of people appear to have an idea in their head of what a "perfect" audio system should sound like, and that's what they are always comparing with. An anecdote of mine is that a madly keen audiophile came to listen to what I was doing, ohhh, about 10 years ago. For the first hour or so he made some disparaging remarks, and had a bit of a frown - but then it started to sink in ... what he was listening to wasn't a "spectacular hifi" but just natural, unembellished sound, no funny artifacts or exaggeration sticking out all the time - the sound wasn't attacking you, you could just soak it up without effort ...
As he was leaving he explained to me this realisation ...
Tom
What might the differences be when contemplating a Parallel 86 vs a pair of Mod 86 based monoblocks?
(my apologies for pulling the thread away fom the Saga of Soon)
(I do have to admit, the statement "I was quite obvious to me when they explained it" said it all to me and was lovingly cut and pasted into my quote book ....)
What might the differences be when contemplating a Parallel 86 vs a pair of Mod 86 based monoblocks?
(my apologies for pulling the thread away fom the Saga of Soon)
(I do have to admit, the statement "I was quite obvious to me when they explained it" said it all to me and was lovingly cut and pasted into my quote book ....)
Yay... Signal. Thank you for increasing the SNR of the thread.
The main difference is power. A PAR86 amp will deliver 60 W in 8 Ω, 120 W in 4 Ω on a ±35 V supply. 120 W into 2 Ω if you lower the power supply to ±28 V. The PAR86 is also the best candidate for a bridged amp.
For the MOD86, I recommend running it at ±28 V. This gives you 38 W into 8 Ω and a bit over 60 W into 4 Ω. Operation with 2 Ω load is possible but not recommended.
Tom
My main interest in following this thread is I want to make a current source that has Zo > 1k to 20kHz and also 1ppm THD20k. Probably (the 2nd criteria is) an impossible target. I'll settle for the first criteria.
An important requirement would be a very clear idea about grounding & earthing .. as Tom would have agonised and spent a lot of time optimising for Mod-86
As a real beach bum, I have problems doing stuff with solder but the existing SPICE models of LM3886 don't seem to model some of the stuff I need. eg I can't have a Zobel for Zo > 1k at 20kHz and I know from real life that a Zobel is required for unconditional stability into real loads with 3886.
Tom has loadsa good stuff on the topic in http://www.diyaudio.com/forums/chip-amps/217790-lm3886-effect-compensation-network-cc-rf2-cf.html ... but alas, nothing that helps someone after a current drive amp.
Hi There,
I have simulated a current drive amp based on 3886 and 49720 in a composite mode, this is based on specific models which I derivated from the data sheet bode plots. That way you can model yourself the zobel and also the additional network needed in the feedback loop to (kina) linearize the current curve. Then you get around 1k ohm zout at 20khz, with good phase margin.
I ve made a PCB based on this, with current measurement with a 0.5 ohm resistor in series with ground. I ve also made possible to measure the current in series with an diff amplifier ina134 as I wanted to bridge.
when it comes to performances and grounding, I ve tried to be very careful in avoiding effect of high current return in caps and connecting the signal gnd to the bottom of the measurement resistor, but at the end of the day I see with my poor measurement chain that the noise floor is not that good. I m missing equipment to tune all of that but I think you can achieve a good thd+n by having the amp close to the driver, minimizing cables as the driver is in the feedback loop.
I give you the link on the french forum.
AMPLI DIY DUAL LM3886 TENSION/COURANT - bi/tri amp asservi sur le forum Amplis et Préamplis du site Homecinema-fr.com - 30056403 - 1056
Page 1 contains schematics, last pages shows some noise diagrams, and somewhere in between you ll find spices model and so one. good luck for translation, unfortunately I cannot afford too much time to develop here on diyaudio.
at the moment, I put this current drive idea on hold as the concern is really about finding the composite architecture and pcb layout which enables a 10ppm thd+n while having the driver in the feedback loop, feel free to share your findings.
thanks to Tom for his hard and inspirational work
I never said I wanted to look at Zo. I just want to look at the Mod response when measured as described to show how it performs with reaching loads.David Hafler was a founder of Dynaco. I forgot he'd changed the name of his second company. Mea maxima culpa
BTW, an important reason why he could claim "straight wire with gain" is that he tested this proposition with DBLTs. You may wish to try this with you own Golden Pinnae too
Soong, I believe your position & authority on the topic is now very clear .. as is ours. Any further discussion would be useless as you don't believe you have a problem with your measurements.
____________________________
The one item not settled is why are you interested in Zo < 0R1 when you are using current drive.
My main interest in following this thread is I want to make a current source that has Zo > 1k to 20kHz and also 1ppm THD20k. Probably (the 2nd criteria is) an impossible target. I'll settle for the first criteria.
An important requirement would be a very clear idea about grounding & earthing .. as Tom would have agonised and spent a lot of time optimising for Mod-86
As a real beach bum, I have problems doing stuff with solder but the existing SPICE models of LM3886 don't seem to model some of the stuff I need. eg I can't have a Zobel for Zo > 1k at 20kHz and I know from real life that a Zobel is required for unconditional stability into real loads with 3886.
Tom has loadsa good stuff on the topic in http://www.diyaudio.com/forums/chip-amps/217790-lm3886-effect-compensation-network-cc-rf2-cf.html ... but alas, nothing that helps someone after a current drive amp.
I think the LM3886 is an interesting chip. I have discussed and shown some interesting aspects while playing around with the MyRef. Also posted some in other LM3886 threads.
Talking about Zobels. I feel that using RCR instead of CR from output to ground would be better. However, I am not going into any discussion on this issue.
I think comparison against a straight wire with gain is quite vague and misleading because different straight wires have different characteristics. I think the test jig for the Hafler is used for the adjustment purpose, but is really not ideal when you use it with driver or headphone loading as described.
Last edited: