A New Take on the Classic Pass Labs D1 with an ESS Dac

[qusp]

looking at using the semisouth jfets (R550 or R085) and PPS films on mine instead of wimas and IRF610. will try the IRF540 as well.

hows your SMD design going owen if you are still monitoring this? maybe too late for my own build, but i'm making another ackodac for a fellow compatriot and thought if you werent too far off would use your PCB for the lineout and another design for headout

 

[PierreQuiRoule]

looking at using the semisouth jfets (R550 or R085) and PPS films on mine instead of wimas and IRF610. will try the IRF540 as well.

It would be interesting to compare the voltage at the output of the DAC (input of the I/V) for all these devices - under the same bias condition.

From what I gather from the posts, the current-output ESS DAC produces less distortion when driving a virtual ground. So the higher the transconductance of the device, the lower the impedance seen by the DAC, and the lower the distortion. But I am sure there are factors at play

If I may ask, is there a particular reason for not using the IRF610? If it is availability, I can share some.

 

[qusp]

It would be interesting to compare the voltage at the output of the DAC (input of the I/V) for all these devices - under the same bias condition.

indeed it would, I may just do that at some point

From what I gather from the posts, the current-output ESS DAC produces less distortion when driving a virtual ground. So the higher the transconductance of the device, the lower the impedance seen by the DAC, and the lower the distortion. But I am sure there are factors at play

this is true, the lower the impedance of the input of the I/V stage, the more like a current source the sabre looks. and we are talking really low here, so yes these experiments are interesting indeed.

If I may ask, is there a particular reason for not using the IRF610? If it is availability, I can share some.

for the reasons above partially, but also just to put my mark on the build, the 610 isnt really capable of what I desire as far as high resolution output either. I will be trying the semisouth devices, the IRF540 and also the toshiba 2SK170 (a few in parallel) as recommended by Papa.

thanks for the offer though, nah they are easily available, but I think I will start with a different device. I plan on trying several devices in there to see which works best for my purposes. just need to think on how I might rig up some mini T092 heatsinks and pump Papa for more info on the specific tweaks needed to drop the R550 in will keep you in the loop. most other parts for the D1 will be here on friday, so will post some pics sometime next week

 

[lauret]

The only advantage the Mosfet has over the JFET in this
application is a lower impedance at the Source pin, being that
the Mosfet has a much higher transconductance. This results
in lower voltage seen at the output of the DAC. If the DAC is
not sensitive to this, then there is no advantage.

This is from another thread, but why would you use JFETs in the light of this quote?

 

[opc]

Hi Guys,

I'm going to side with lauret on the jfet issue, and the more I muck around with the circuit, the more it appears that performance is closely linked to transconductance. One of the best measuring devices was still the original RJK mosfet I used, and it's also the highest transconductance fet I've tried.

On that note, I have a few IRF1324's here that I think are going to do well, or at least they'll help me verify my theory on high transconductance. I'll post results tomorrow when I get them soldered in and tested.

The main board layout is done, and it's designed to drop the new buffalo board right on top, and use some 0.1" jumpers to connect the two. I'm just sending the Gerbers out for quotes this evening.

Cheers,
Owen

 

[qusp]

well actually I was looking at JFETs due to nelsons advice,but I guess it was late and I missed the relevance. but yes indeed if a mosfet will present a lower impedance and higher transconductance to the sabre, this will help to maintain current mode, so thats the way to go. I knew the relevance of the low impedance already, but i'm fairly new to all manner of mosfet circuits, as I havent needed the power so much in the past (headphones) so stuck to low current JFETs, but its becoming clear the sabre is a strange beasty and new tactics must be used to attain that magical 140db.

arrggh, I think i'll grab both I actually already have some IRF540 on the way, so its a matter of suck it and see I guess. I guess some of you havent noticed my bowerbird tendencies yet. I never just choose one direction, but rather several and use what works best for me, regardless of the numbers. plus i've been looking for an excuse to use the sic parts, guess I can just use a vishay sic mosfet. one problem with some mosfets for me though , is the fact I use a lot of high resolution source, I do not want an upper limit presented by the I/V stage when the rest of my rig is capable of much more. thus playing with the D1 cap RC values too.

on that note Owen, where are you planning on setting the LPF? I need to maintain at least 192 and preferably 384khz due to my studio use and there is a balancing act there that must be respected.

and lauret, i'm using transformers on the output of my buff II at the moment and enjoy it very much, as far as the above its far from ideal, but its very pleasing all the same. good numbers please me a great deal, but they do not rule me

 

[qusp]

I know why I didnt see it; because it wasnt there I started my own thread, for those who didnt see it and Nelson said

I spent some time with the R550 last weekend. As an output
device it has more bandwidth than the other power parts such
as the R100's or IRFP240's, and this is due to lower capacitance.
With that comes greater distortion at lower frequencies. For
this it would be natural if your source impedance was high or
you were building a tweeter amp.

The IRF610 still has lower capacitance, but also much lower
transconductance, about 1/5 as much.

I don't know what a D1B1 is, but if it's the IV for a DAC, then
you are better off with JFETs like the 2SK170's, maybe a
few in parallel.

so yeah that was enough for me to want to grab some; along with some mosfets already ordered, TO TRY

 

[lauret]

I know why I didnt see it; because it wasnt there I started my own thread, for those who didnt see it and Nelson said



so yeah that was enough for me to want to grab some; along with some mosfets already ordered, TO TRY

Please let us know what works best for you so we all can learn!

 

[lauret]

on that note Owen, where are you planning on setting the LPF? I need to maintain at least 192 and preferably 384khz due to my studio use and there is a balancing act there that must be respected.

This interests me as well. Pierre changed a couple of capacitors to stay at the same cutoff frequencies. Should there also be changes to the circuit by Owen?

To keep the same cutoff frequencies (RC) for the low pass filters, the following capacitors were adjusted accordingly:
- C40, C45 increased from .01 to .033 uF
- C15, C16 increased from .0027 to .0086 uF

 

[opc]

Hi Guys,

I certainly didn't mean to discourage you qusp, you should, by all means, try every possible device you can and see what works best. I was just generally trying to convey the trends I've seen in the 8 devices I've tried it with.

As I mentioned above, the lateral BUZ fets were some of the best I've tried, and there was no rhyme or reason to it so far as I could see. Keep in mind that things like gate charge and the associated capacitances don't seem to carry much importance here since the gate and source don't actually swing by any appreciable amount. For that same reason, I don't think that "faster" devices (lower gate charges and capacitances) would provide any significant benefit in bandwidth. I will check this though.

As for the capacitor values, I came upon mine by simulation, but I'll confirm with a few measurements tomorrow. As qusp mentioned, it's a tradeoff between how much HF garbage you're willing to let through, and how high you want your response to extend. I'll work out the values for -1dB at 50k, 100k, 200k, and 400k providing the circuit makes it that high.

Cheers,
Owen

 

[lauret]

As I mentioned above, the lateral BUZ fets were some of the best I've tried, and there was no rhyme or reason to it so far as I could see. Keep in mind that things like gate charge and the associated capacitances don't seem to carry much importance here since the gate and source don't actually swing by any appreciable amount. For that same reason, I don't think that "faster" devices (lower gate charges and capacitances) would provide any significant benefit in bandwidth. I will check this though.

Check the distortion at lower frequecies as hinted by Nelson Pass:

I spent some time with the R550 last weekend. As an output
device it has more bandwidth than the other power parts such
as the R100's or IRFP240's, and this is due to lower capacitance.
With that comes greater distortion at lower frequencies. For
this it would be natural if your source impedance was high or
you were building a tweeter amp.

The IRF610 still has lower capacitance, but also much lower
transconductance, about 1/5 as much.

I don't know what a D1B1 is, but if it's the IV for a DAC, then
you are better off with JFETs like the 2SK170's, maybe a
few in parallel.

 

[qusp]

Please let us know what works best for you so we all can learn!

absolutely, i'm not one to hoard information if I think its meaningful; as long as it doesnt fall outside the bounds of my NDA and owen, I think you know why I think high resolution is very attainable

Hi Guys,

I certainly didn't mean to discourage you qusp, you should, by all means, try every possible device you can and see what works best. I was just generally trying to convey the trends I've seen in the 8 devices I've tried it with.

oh of course Owen, its all good mate, I was simply explaining my motives and thought the fact I was flexible in my direction was a point worth mentioning; particularly to Lauret. happy to take onboard any practical experience I may lack, didnt mean to sound cross

As I mentioned above, the lateral BUZ fets were some of the best I've tried, and there was no rhyme or reason to it so far as I could see. Keep in mind that things like gate charge and the associated capacitances don't seem to carry much importance here since the gate and source don't actually swing by any appreciable amount. For that same reason, I don't think that "faster" devices (lower gate charges and capacitances) would provide any significant benefit in bandwidth. I will check this though.

think we are entering a new area here mate and there may well be a n unexpected hero in all this, due to a new paradigm, previously it was quite appealing to have an almost infinitely high input impedance, but this has changed with the sabre, making buffers of many types unusable

As for the capacitor values, I came upon mine by simulation, but I'll confirm with a few measurements tomorrow. As qusp mentioned, it's a tradeoff between how much HF garbage you're willing to let through, and how high you want your response to extend. I'll work out the values for -1dB at 50k, 100k, 200k, and 400k providing the circuit makes it that high.

great mate! thanks, yeah I ordered some 500R for R27/34 and some 0.0027 and 0.001 for C15/C16. got this from mouser, who have a pretty limited range of films, these are the MKP and FKP I got, but will use the evox/Rifa PPS for the final build, even thought about socketing the positions for these, but worry about the added inductance and capacitance effecting any results I get.

I guess youve made your parts choices for the BOM already owen, but wondered if you are aware of the newish expended range of panasonic SMD PPS film caps available with higher voltage ranges than previously and in sizes as small as 0805. would be perfect for this design I think mate and wouldnt require any changes to the layout as they come in 0805, 0603 even I think (nut not sure on that one), 1210, 1206, right up to 2824. also the sanyo oxicap and poscap are pretty special, though probably too low voltage.

you are still going mostly SMD here hey mate?? with options in some positions for through-hole. hope you stuch to your guns there mate, I love SMD these days. as long as there is those 2 positions for larger more exotic films i'm happy with every other part being SMD and i'm sure I can just jerry-rig the pin headers to fit the ackodac or just use short molex

Cheers,
Owen[/QUOTE]

 

[anbello]

... i'm using transformers on the output of my buff II at the moment and enjoy it very much, as far as the above its far from ideal, but its very pleasing all the same. good numbers please me a great deal, but they do not rule me

Hi qusp,
I wolud like to know which kind of transformers are you using with the buffalo DAC and in which way they are connected (with or without resistor/capacitor on primary and/or secondary)

Thanks
Andrea

 

[qusp]

O-netics from Bud Purvine who is a member here under the name BudP. they are 600:600, with dual primaries and dual secondaries. I have simply tied both primaries and secondaries together in series and connected the output if the buff II directly to the primary and the output directly to the secondary (balanced output to my active monitors) I tried with trimpots from each leg to ground on the primary, but didnt find any improvement I could put my finger on, so removed them. the impedance of the transformer itself is doing the I/V (as well as the dac outputting mostly voltage at this point anyway)

 

[PierreQuiRoule]

Understanding D1 I/V

Just thought I would summarize my understanding of the filters in the D1 I/V circuit for you folks to review and correct me if I am wrong.

C15 and R27 in parallel with the gate impedance of Q2 form a low pass filter. The MOSFET gate impedance is high enough to be negligible. The -3 dB cutoff frequency of the filter is thus approximately 1 / (2 pi R27 C15), which for the stock D1 schematic gives about 40 kHz. Same frequency for the modified I/V circuit values of Post #81 (higher C15 making up for lower R27).

The role and size of C40 are a bit more nebulous. I think C40 and R26 in parallel with the input resistance of Q3 form another low pass filter to reject RF transients from the DAC’s digital circuitry. The input resistance Rs of a MOSFET in common gate configuration is approximately the inverse of its transconductance (cf. “common gate” in Wikipedia). Mr. Pass states here http://www.diyaudio.com/forums/pass-labs/77040-pass-labs-d1-5.html#post1233833 that the transconductance of an IRF610 at 10mA is about .1S, which translates into a resistance Rs of 10 ohms. 10 mA is close to the operating condition of the stock D1. Hence here the contribution of R26 is negligible and the -3 dB cutoff frequency is approximately given by 1 / (2 pi Rs C40), or about 1.6 MHz. For the circuit values of Post #81 the -3 dB frequency is approximately 500 kHz, perhaps a bit less because of the slightly higher transconductance of the MOSFET due to the 30 mA bias current.

I have not measured the above. So if this is wrong please feel free to jump in. What I am sure of, is that it sounds good .

 

[qusp]

yeah it was mainly the LPF at R27/C15 and co that was of interest to me, because I cant live with a LPF of 40khz, so I set the values there to give either ~110khz or ~220khz I changed the R down by 1/3 from 1.5K to 500R and have bought 0.0027, 0.0015 and for interest 0.001uf as well. these are not the final caps anyway as I will be using PPS for them, but I figured this was a good place to start. I just used a quick RC calc rather than plotting it out, because I was low on time and it wasnt crucial for this round.

thanks for the comprehensive rundown tho, that will come in handy; provided its correct lol

 

[opc]

Hi Guys,

If there isn't already a saying like "The simplest things in life are often the most complicated", then I'm going to claim it, and apply it here.

I spent the better part of the day trying a few different things with the buffered D1B1, and I found out a few really neat things.

1. Everything in this circuit affects everything else!!!!!! There is not a single part that you can change without tossing the balance of the circuit out the window.

2. The transconductance of the mosfets is important, and affects both the filter values, distortion, and voltage swing at the output of the DAC.

I measured four different devices today, and I can say that higher transconductance fets are going to perform better in this circuit. I'll post THD+N graphs tomorrow, but the difference was more than 10dB from the lowest transconductance to highest transconductance devices.

I also measured the voltage swing at the DAC output with both an IRF610 and a IRF1324. It was 42mV P-P with the 610 and 12mV P-P with the 1324. Higher TC devices present a lower impedance to the DAC.

The filters are very messy to deal with. The value of the upper caps running from +18V to the drain, works in concert with the value of the lower caps running from source to ground. Making the lower caps large, causes the response to peak before rolloff, making them small causes a lower order rolloff with no peaking. Getting the ratio just right causes the rolloff to be steep, while maintaining flat response in the passband. I'll post some measurements tomorrow to better illustrate this.

The mosfet chosen will change the FR, so every mosfet will need it's own values of caps to get a desired rolloff. I did have some luck with simulation, but when I put the values in the actual circuit I had to tweak the values a little to get a flat response.

With the IRF1324's, a 400ohm source and 182ohm drain resistor, the cap values were 27nF up top and 220nF at the bottom. This cause a 0.05dB bump upwards at 15kHz, and the response was down 0.1dB at 20kHz falling steeply thereafter. I don't remember the -3 and -6 points, but I'll check tomorrow.

Anyhow, I've got a lot more work to do, and I'm going to spend some time trying to sort out why everything does what it does.

Measurements to follow tomorrow with the un-buffered D1 using the IRF1324's and properly tuned C values for a few different -1db points.

Cheers,
Owen

 

[qusp]

oh joy!! :S maybe i'll be omitting the buffer section afterall. burt lucky I have craploads of the normal metal films just because I do, I bought numbers of different cap values and with more on the way, I suppose the dielectric will also effect the circuit in something like this by the sounds of it. also bought several devices with plans for more. so lucky I dont have my heart set on anything but the best sound I can find for my ears

 

[PierreQuiRoule]

Owen - Thank you for your measurements, analysis and post!

2. The transconductance of the mosfets is important, and affects both the filter values, distortion, and voltage swing at the output of the DAC.

Oh yes! Of course! transconductance varies with frequency... So different FET types have difference transconductance curves and thus require different value of C40. I had missed this completely till now.

 

[PierreQuiRoule]

The role and size of C40 are a bit more nebulous... Hence here the contribution of R26 is negligible and the -3 dB cutoff frequency is approximately given by 1 / (2 pi Rs C40), or about 1.6 MHz.

Had a hunch I was wrong. Indeed I was. Filtering is provided by the FET's transconductance too!