D1 IV Mosfet

[MJPAudio]

Hi,

Currently having a PCM63 dac with D1 iv stage, after many years i would like to build another one and see if i can improve. (better power supplies, caps, wiring, pcb layouts etc)

Well i'm at it i was wondering if there would be a improvement possible with a different mosfet. (lower impedance for the dac)
ZVN2210g? Instead of irf610? Let me know your thoughts or suggestions.

Thanks Martin

 

[Zen Mod]

you can go only up - IRFP240

 

[metalman]

I built a D1 I/V stage using ZVN3310's. I dropped the supply rails to +/- 22V on mine. It sounds pretty good to my ears, but I've never heard an actual D1 to compare.

 

[Sabrosa]

I actually posted yesterday asking for suggestions for high transconductance surface mount MOSFETs to use in an ES9038PRO-capable D1 I/V stage and just came across this thread this morning.

I've put together a list of devices to test for transconductance under the operating conditions of the circuit and would love any recommendations for additional MOSFETs to add!

The supply rails are +/-6V, operating current is 1.09A, and max gate threshold is ~4V, so I've mostly focused on MOSFETs at or around logic level.

Here's the list so far:

IRL40S212
IRL40SC228
IRL60S216
IRFS7430PbF
IRFS7534PbF
IRF2204S
PSMN1R5-30BLE
PSMN3R4-30BLE
PSMNR90-30BL
CSD19535KTT
CSD19536KTT
CSD18510KTT
CSD18542KCS

 

[Nelson Pass]

Transconductance seems to be a vary important criterion for you.

Since this is a Common Gate Mode circuit, transconductance is the
reciprocal (mathematically and literally) of the input impedance, we
ask the question - what input impedance are you trying to achieve
and why?

 

[Sabrosa]

Thanks Nelson, appreciate you taking a look.

In order for the ESS DACs to stay in current-output mode they need to be loaded into as close to a virtual ground as possible. Ideally in the mOhm range, given the low output impedance of the DAC (~200R). Higher input impedance pushes the chips into voltage-output mode and prevents you from reaching the data sheet DNR and THD specs.

I suppose a better way to frame my question and search might be, given that I need a high transconductance MOSFET with a low-ish gate threshold voltage, are there any other audio-beneficial characteristics I should be considering? Or specific lines of MOSFETs known to be audio-friendly when in the signal path under these conditions?

 

[2 picoDumbs]

Maybe 2sc4793 instead of mosfet

 

[Nelson Pass]

I would be interested in seeing any data that shows the performance
degradation into loads greater than 0 ohms.

 

[2 picoDumbs]

I suppose a better way to frame my question and search might be, given that I need a high transconductance MOSFET with a low-ish gate threshold voltage, are there any other audio-beneficial characteristics I should be considering? Or specific lines of MOSFETs known to be audio-friendly when in the signal path under these conditions?

TO264 Lateral Mosfet.
Probably want to bias around half an amp.

Maybe also worth checking out SiC Jfets in to247 package. United SiC

 

[abraxalito]

SSM3J355R has typical gm in excess of 10S @ 1A however it would need cascoding as its limited to 1W dissipation being a thermally enhanced SOT-23. Oh and its also a P-channel device. Threshold voltage <-1V.

 

[Sabrosa]

I would be interested in seeing any data that shows the performance
degradation into loads greater than 0 ohms.

I was mistaken above when I said that both DNR and THD are affected by the output stage input impedance. I believe I picked that up from another thread on the forum. According to an ESS white paper and the ES9018 data sheet only THD increases between current and voltage modes, from -120dB to -108dB at 0dBFS.

The reason given for this in the white paper is that current mode "cancels a slight on-chip resistor voltage coefficient." Dustin Forman, one of the ES9018 designers, also posted about this awhile back:

For the output stage:

I dont think we ever claimed a current source, but rather a "current mode". The current mode is simply when the current going in and out the pin of the chip is sensed. This mode has the benefit of cancelling 2nd and 3rd harmonics of some of the internal ananlog circuitry. The "voltage mode" is when the pin of the chip has a voltage on it that is being sensed. While this has the 2nd and 3rd hormics (at the -100dB level or so), some people have even claimed this mode is more "tube-like" Its all personal prefference.

I haven't been able to find any data that indicates the maximum load for keeping the chip in current mode. It all seems anecdotal at best -- anywhere from, "must be <1 ohm", to "less than a few ohms", or "you'll definitely be in voltage mode at 10 ohms".

There's a good argument to be made that even if you're operating in voltage mode the performance degradation is small enough not to be audible. Run the DAC into an amplifier with 0.01% THD and you're above both of those thresholds anyway. Although I'm sure there are plenty who would take the other side.

My original thinking in compiling the list of potential MOSFETs was that it's a good excuse to build a test circuit and measure them, since it's not something I've done before. And if I end up with higher transconductance as a result, so much the better. Even a MOSFET that measures 0.25 S will result in an an input impedance of 4 ohms, so I assume you can't go too wrong no matter what.

I re-read your Practical Mosfet Testing for Audio guide last night and remembered this paragraph:

We will want to test in both Common Source and Common Drain modes, which comprise the bulk of usage where we care very much about quality and matching. We can perhaps ignore measurements for the usual Common Gate Mode of operation, inferring their values from the others, and noting that as Cascode devices, they contribute little of their own character to the circuit performance.

If most any modern MOSFET will provide sufficient transconductance, and since this is a Common Gate Mode circuit none of the other MOSFET characteristics affect the performance, then will pretty much any device, assuming a Vgs that allows for proper biasing, perform identically?

I wanted to make sure I wasn't missing something obvious, like "you should definitely use a planar/lateral MOSFET", or "never use a HEXFET in this situation", or some other important characteristic that those with more experience are aware of.

 

[Nelson Pass]

Thanks for the reference, it is largely what I assumed. If we are dealing
with 2nd and 3rd harmonics at .001% or so, I wouldn't be that concerned
about the quality of sound, although the chip maker would see it as a
hit on their spec sheet.

A push pull jfet I/V like the Zen I/V is about 20 ohms with the Toshiba/Linear
Systems parts, and an ordinary Mosfet in a T0-220 package in Common Gate
mode is maybe 5 to 10 ohms, and something like an IRFP240 will deliver less
than an ohm if substantially biased.

The larger transconductances carry proportional capacitances, so probably
you will look for the sweet spot between the two.

One nice thing about the larger sorts of Fets is that the capacitance soaks
up the switching transients, acting as an output filter, apparently without
degradation to to the sound quality.

 

[nigelwright7557]

Improvements depend on how bad the original design was.
If it was good you could be wasting your time and money.

If it was bad then there is room for improvement like upgraded power supply.

Different mosfets will help uprate the power along with a better power supply with possibly higher voltages.

I don't think pcb layout will help much unless there are ground loops or thinner than recommended tracks.

 

[PRR]

> find any data that indicates the maximum load for keeping the chip in current mode

ES9012 / ES9018 Reference 32-bit Audio DAC Datasheet
"Differential current output is equivalent to a differential voltage source in series with a 781.25 Ohm resistor. The differential voltage source has a peak-to-peak output range of 3.05V (0.924*AVCC) and an output offset of 1.65V (AVCC/2)."

By a 10:1 thought, anything under 78 Ohms is "practically current mode", and will tend toward the 120 number rather than the 108 number.

<78 Ohms could be a 2N2222 at 1mA, easy. (BJTs will always beat FETs at the same non-minuscule current.) IRF610 (and similar) can be 1.2 Ohms but at 2 Amps. At 0.1A, maybe 3 Ohms.

 

[Nelson Pass]

Cheers for that.

 

[lonevoyager]

By a 10:1 thought, anything under 78 Ohms is "practically current mode", and will tend toward the 120 number rather than the 108 number.

<78 Ohms could be a 2N2222 at 1mA, easy. (BJTs will always beat FETs at the same non-minuscule current.) IRF610 (and similar) can be 1.2 Ohms but at 2 Amps. At 0.1A, maybe 3 Ohms.

Just a small note, the 781.25 Ohm figure is per channel and for the ES9018.

The ES9038PRO (mentioned in OPs other thread) has a quarter of that (~195 Ohm per channel), giving an output impedance in stereo (4 channels in parallel) of <50 Ohm and in mono (all 8 channels) of <25 Ohm.

 

[lonevoyager]

And another small note, I just came across the following in another thread discussing I/V converters:

There is an article in the magazine Linear Audio where Erno Borbely and Sigurd Ruschkowski have measured THD and harmonics for different loads on the ESS9018 current output DAC.

They measured 0,006%THD with 205R load, 0,0012% THD with 22R load, and 0,00066% THD at 1R1 load. With 205R loads all harmonics where present but with the other two loads only 2nd 2nd 3rd harmonics. This is using balanced mode and 1 kHz and 96 kHz sampling-

Russ White, designer of the Buffalo DACs has also stated that load should be less than one Ohms for lowest THD. 0 dBFS.

I have seen other documentations for this on this site but cannot remember them all.

For mono mode, ie 32 mAp-p swing on top of 16mADC, I think that around 50-65 mA should pass through the semiconductors in the IV converter.

BTW, this DAC load should preferably be <1Ihms up to 100 kHz or higher.

 

[Nelson Pass]

I would say the data is encouraging for the 22 ohm or less figure.

 

[Sabrosa]

Thanks for the reference, it is largely what I assumed. If we are dealing
with 2nd and 3rd harmonics at .001% or so, I wouldn't be that concerned
about the quality of sound, although the chip maker would see it as a
hit on their spec sheet.

A push pull jfet I/V like the Zen I/V is about 20 ohms with the Toshiba/Linear
Systems parts, and an ordinary Mosfet in a T0-220 package in Common Gate
mode is maybe 5 to 10 ohms, and something like an IRFP240 will deliver less
than an ohm if substantially biased.

The larger transconductances carry proportional capacitances, so probably
you will look for the sweet spot between the two.

One nice thing about the larger sorts of Fets is that the capacitance soaks
up the switching transients, acting as an output filter, apparently without
degradation to to the sound quality.

Thank you Nelson for the advice! Are there any guidelines to stay within as far as MOSFET capacitance goes? I've been searching around for information on how capacitance affects the circuit when in common gate mode.

Interestingly, the IRL40S212, which came out on top for THD in my simulations, also has lower capacitance (Ciss=8320pF, Coss=1050pF, Crss=790pF) than most other devices I looked at. I don't give too much credence to the simulations since model quality tends to vary, but I thought that was interesting and was curious if it was related.

D2PAK looks to be among the largest of the surface mount MOSFETs, other than the few devices available in D3PAK.

> find any data that indicates the maximum load for keeping the chip in current mode

ES9012 / ES9018 Reference 32-bit Audio DAC Datasheet
"Differential current output is equivalent to a differential voltage source in series with a 781.25 Ohm resistor. The differential voltage source has a peak-to-peak output range of 3.05V (0.924*AVCC) and an output offset of 1.65V (AVCC/2)."

By a 10:1 thought, anything under 78 Ohms is "practically current mode", and will tend toward the 120 number rather than the 108 number.

<78 Ohms could be a 2N2222 at 1mA, easy. (BJTs will always beat FETs at the same non-minuscule current.) IRF610 (and similar) can be 1.2 Ohms but at 2 Amps. At 0.1A, maybe 3 Ohms.

Thanks for that! I hadn't put it together on my own from the data sheet.

Just a small note, the 781.25 Ohm figure is per channel and for the ES9018.

The ES9038PRO (mentioned in OPs other thread) has a quarter of that (~195 Ohm per channel), giving an output impedance in stereo (4 channels in parallel) of <50 Ohm and in mono (all 8 channels) of <25 Ohm.

And another small note, I just came across the following in another thread discussing I/V converters:

Great find on the Linear Audio article. Although the circuit I'm looking to implement is for use with a ES9038PRO, I used the ES9018 characteristics in the examples above since its data sheet is more freely available (silly NDAs).

 

[Sabrosa]

I have a potentially dumb question about HEXFETs. Are they always vertical devices but with either a planar or trench gate structure?

While looking through the Infineon power MOSFETs I noticed their product chart designates either “Planar” or “Trench” for each HEXFET under the Technology column. It’s usually one or the other, but some devices have both a planar and trench version, such as IRL3705ZS (trench) and IRL3705NS (planar).

When people say that HEXFETs are suitable for linear audio use, are they always referring to planar gate HEXFETs, or are there instances where trench gate HEXFETs are used?

I’m assuming the main concerns with trench gates are the high capacitance and limited safe operating area, but are there others?

I ask because the newish Infineon StrongIRFET HEXFETs, such as IRL40S212, have trench gates but with lower capacitance than usual and “rugged silicon”. The Infineon product guide says that the StrongIRFETs are aimed at replacing planar power MOSFETs. Is there any reason the StrongIRFET line would be unsuitable for use in the D1 circuit?

In a common gate circuit, I’d guess we are most concerned with keeping source to drain capacitance reasonably low? In general, is there a threshold we should aim to stay within if some capacitance is considered beneficial?