Midifying SSL EQ with Digital Potentiometers

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Gertius

Well-known member
Joined
Feb 22, 2013
Messages
105
Location
Germany
Hi all,
I´m currently looking for a way to replace the pots on an SSL EQ (currently I have a 82e132 board here) with digital pots.
I have already tinkered around with an Arduino and have the digital Midi control side down.
For a digital pot I recently found and ordered this part to try out:
https://www.microchip.com/wwwproducts/en/MCP45HV51
http://ww1.microchip.com/downloads/en/DeviceDoc/20005304A.pdf
It allows for an high analog voltage of +-18dB.

My question at the moment is if the current rating of the part is high enough:
- 12.5 mA (for 10 kOhm)
-  6.5 mA (for 50 kOhm and 100 kOhm)

Does anybody have knowledge of how much current flows through the pots of an SSL EQ?
Or a way of how I could find it out?

Thanks,
Christian
 
Gertius said:
Hi all,
I´m currently looking for a way to replace the pots on an SSL EQ (currently I have a 82e132 board here) with digital pots.
I have already tinkered around with an Arduino and have the digital Midi control side down.
For a digital pot I recently found and ordered this part to try out:
https://www.microchip.com/wwwproducts/en/MCP45HV51
http://ww1.microchip.com/downloads/en/DeviceDoc/20005304A.pdf
It allows for an high analog voltage of +-18dB.

My question at the moment is if the current rating of the part is high enough:
- 12.5 mA (for 10 kOhm)
-  6.5 mA (for 50 kOhm and 100 kOhm)
Unlikely to be a problem. Current will be limited by circuit as used.
Does anybody have knowledge of how much current flows through the pots of an SSL EQ?
Or a way of how I could find it out?
read a schematic, and factor for typical pot values. Again unlikely to be a problem.
Thanks,
Christian
Nice parts... last time I looked at designing with DPOTs they were not running on 36v process technology. (I think I used 15V parts and scaled voltage down 6dB, then back up again for output.  )

I scanned through 50+ pages of performance curves and did not see a single analog linearity curve (distortion), but otherwise curves for wiper resistance changes seem pretty typical for transfer gate based technology.

JR
 
I don't think this would work very well. I don't know if the current limits would be a problem. But digipots are just not designed to be used in a conventional circuit like an SSL EQ. If you design a circuit around the digipot I think performance would be quite good. But if you scan over the datasheet you will see there are a lot of error tables and stay capacitance and things that will ultimately equate to distortion. These things are designed to be driven from low source impedance and loaded with high source impedance. If you don't, the variations in resistance and capacitance will result in distortion and high impedance in general means you're going to pickup noise.

Then you have other issues like the fact that you would need like 10+ digipots. Trying to cram 10 mixed signal ICs into one spot is a challenge. That's a lot of control lines. Yeah, those digipots have address pins but I think you would still probably need I/O expanders. Maybe if you stack them with bus made of header pins where one plugs into the next would help. But no matter what you did I think that you would ultimately get some noise.

I like digitally controlled analog stuff. If done correctly it can work very well. But digipots in a high impedance circuit is a bridge too far. I would stick to relays. It don't recommend it for an SSL EQ because you would need literally 30-something relays to make a decent number of frequency / level steps and that is with compounding steps (eg 32 steps from 5 relays using 2^5=32). But for routing or switching or even stepped volume or gain controls they can be nice.
 
squarewave said:
I don't think this would work very well. I don't know if the current limits would be a problem. But digipots are just not designed to be used in a conventional circuit like an SSL EQ. If you design a circuit around the digipot I think performance would be quite good. But if you scan over the datasheet you will see there are a lot of error tables and stay capacitance and things that will ultimately equate to distortion. These things are designed to be driven from low source impedance and loaded with high source impedance. If you don't, the variations in resistance and capacitance will result in distortion and high impedance in general means you're going to pickup noise.

Then you have other issues like the fact that you would need like 10+ digipots. Trying to cram 10 mixed signal ICs into one spot is a challenge. That's a lot of control lines. Yeah, those digipots have address pins but I think you would still probably need I/O expanders. Maybe if you stack them with bus made of header pins where one plugs into the next would help. But no matter what you did I think that you would ultimately get some noise.

I like digitally controlled analog stuff. If done correctly it can work very well. But digipots in a high impedance circuit is a bridge too far. I would stick to relays. It don't recommend it for an SSL EQ because you would need literally 30-something relays to make a decent number of frequency / level steps and that is with compounding steps (eg 32 steps from 5 relays using 2^5=32). But for routing or switching or even stepped volume or gain controls they can be nice.
I designed an automatic mixer (self adjust mix levels) using DPOTs and was unsure of DPOT vs VCA so put both in my first prototype. The DPOTs worked so well I never populated the VCA versions.

I used dual DPOTs (2 per package) but do not how many versions microchip supports. 

With serial com protocol you can send update data to all of the DPOTs, then selectively latch them one at a time or on groups... I updated both sides of stereo channels at the same time, with same data.

I worked with SPI instead of I2S but similar.  I2S supports two address pins so you can control 1 of 4 devices at the same time with one data push, so 12 devices might be controlled with 3 different latch lines in batches of 4.

JR 
 
JohnRoberts said:
I worked with SPI instead of I2S but similar.  I2S supports two address pins so you can control 1 of 4 devices at the same time with one data push, so 12 devices might be controlled with 3 different latch lines in batches of 4.
So that's 3 latch, 2 address, data, clock, VDD and DGND which is a 9 pin bus for 12 devices, 10 for 24, 11 for 48, etc. Not horrible. But presumably there would be other stuff like regular potentiometers connected to analog inputs, tactile buttons, an LED driver, etc so I think you would still run out of pins on the Arduino pretty quick.

Yeah, there's no doubt a digipot can be made to work great IF the circuit is designed around the digipot. If the source impedance is vaguely high or if the load impedance is too low, distortion could easily be measurable in a circuit that is supposed to be high performance analog. And I think it would be really hard to keep digital noise out also in large part because of the lack of control over impedances. Meaning even if the layout was perfect, you might still get digital bleed. You would have to shutdown the clock / mute during adjustments.
 
squarewave said:
So that's 3 latch, 2 address, data, clock, VDD and DGND which is a 9 pin bus for 12 devices, 10 for 24, 11 for 48, etc. Not horrible. But presumably there would be other stuff like regular potentiometers connected to analog inputs, tactile buttons, an LED driver, etc so I think you would still run out of pins on the Arduino pretty quick.
No you are thinking parallel not serial....  and the two address bits are hard wire programming pins on the DPOTs that you set high or low in the PCB layout to establish a 1 of 4 unique device address. Then 2 bits inside the serial data stream, tell each DPOT which data to ignore or load.

To pass I2S serial data involves 1 clock, 1 data, and 1 latch enable line. I proposed using 3 separate latch lines to address 12 DPOTs in 3 groups of 4.  So clock and data in parallel to all and 3 unique latch lines, one latch per group of 4 so 5 pins total.

The latch selects which group of 4 DPOTs and the 2 bit address embedded in the data stream header selects which one of four within the selected group gets latched.

Of course you could make it more complicated if you want.  I never worked with an arduino so have no idea how flexible it is.
Yeah, there's no doubt a digipot can be made to work great IF the circuit is designed around the digipot. If the source impedance is vaguely high or if the load impedance is too low, distortion could easily be measurable in a circuit that is supposed to be high performance analog.
and I repeat I was pleasantly surprised by how good the technology was maybe 10 years ago, it is probably better now (but I have no personal experience with the new microchip parts... I have used many of their micros.)

Transfer gates are very mature technology and even DPOTs have been around for decades. This is the same technology used inside digital controlled mic preamps and they don't seem to get complaints. 
And I think it would be really hard to keep digital noise out also in large part because of the lack of control over impedances. Meaning even if the layout was perfect, you might still get digital bleed. You would have to shutdown the clock / mute during adjustments.
I am not encouraging this as an easy DIY project to modify an existing EQ, but i would not making sweeping assumptions that it couldn't work...  8)

Probably a good idea to try a scratch free standing design, perhaps nicked from a SSL EQ schematic. Might be easier than cutting DPOTs into an existing EQ. Get one working in the wild before trying to cut it into existing hardware.  8)

JR
 
JohnRoberts said:
No you are thinking parallel not serial....  and the two address bits are hard wire programming pins on the DPOTs that you set high or low in the PCB layout to establish a 1 of 4 unique device address. Then 2 bits inside the serial data stream, tell each DPOT which data to ignore or load.
No, I'm talking about serial. But I didn't understand what you meant about the address pins. I have hardwired address pins so I know what you mean. It suffices to say that an atmega is going to provide about 20 I/Os so ...
JohnRoberts said:
Transfer gates are very mature technology and even DPOTs have been around for decades. This is the same technology used inside digital controlled mic preamps and they don't seem to get complaints.  I am not encouraging this as an easy DIY project to modify an existing EQ, but i would not making sweeping assumptions that it couldn't work...  8)
I didn't say it absolutely would not work. But that's just a hedge. I don't really think it would.

The analog gates have actually not improved that much over the years. They are just regular mosfets arranged in clever ways. The on resistance has dropped but at the expense of gate capacitance. At least that looks like the trend. This suggests that they're just paralleling lots of gates into one. So the issue boils down to on resistance flatness. The fact that the datasheet didn't even have on resistance flatness plots is a bad sign. So a high impedance source or low impedance load is going to cause distortion. If on resistance is super low (some of the analog gates like ADG1414 are less than 10 ohms) then capacitance will be high and again the source / load resistance is going to cause loss.

I know I don't need to explain this to you of course. I'm just stating why I don't think it will work. I don't want folks to go through unnecessary pain only to realize it won't work when they could be working on some that would (relays).
 
squarewave said:
No, I'm talking about serial. But I didn't understand what you meant about the address pins. I have hardwired address pins so I know what you mean. It suffices to say that an atmega is going to provide about 20 I/Os so ...I didn't say it absolutely would not work. But that's just a hedge. I don't really think it would.

The analog gates have actually not improved that much over the years. They are just regular mosfets arranged in clever ways. The on resistance has dropped but at the expense of gate capacitance. At least that looks like the trend. This suggests that they're just paralleling lots of gates into one. So the issue boils down to on resistance flatness. The fact that the datasheet didn't even have on resistance flatness plots is a bad sign.
seriously,,, there was over 50 pages of data curves. They show the common TG curvature for on resistance. But what is common for 36V CMOS?  :eek: :eek:    http://ww1.microchip.com/downloads/en/DeviceDoc/20005307A.pdf
So a high impedance source or low impedance load is going to cause distortion. If on resistance is super low (some of the analog gates like ADG1414 are less than 10 ohms) then capacitance will be high and again the source / load resistance is going to cause loss.
I repeat I know nothing about this specific part but Microchip has made some interesting CMOS stuff... I used a LDO 3.3V cmos regulator in my battery powered tuner...  They also make some CMOS opamps that might not suck (for battery low current-low  voltage apps) but I haven't checked them out either.
I know I don't need to explain this to you of course. I'm just stating why I don't think it will work. I don't want folks to go through unnecessary pain only to realize it won't work when they could be working on some that would (relays).
Pain helps people learn for themselves. 

I learned that DPOTs don't suck as bad as I feared, years ago.  I was so worried I put VCAs in the prototype design JIC.

===
The last time I used a relay in a design of my own (for power off bypass) was back in the 80s..

There was one in an inherited design I had to support this century, same application.

JR 
 
Gertius said:
Does anybody have knowledge of how much current flows through the pots of an SSL EQ?
Or a way of how I could find it out?
Max voltage is about 30Vp-to-p, or 11V rms, which results in 1.1mA into a 10k pot.
Worst-case could be a parametric EQ where the frequency is set at maximum. The impedance would be about 5k, for a current of 2.2mA.
 
JohnRoberts said:
seriously,,, there was over 50 pages of data curves. They show the common TG curvature for on resistance. But what is common for 36V CMOS?  :eek: :eek:
Ok now. You had to go digging for that datasheet so take it easy on me.

But from looking at that datasheet I actually don't even see what I would really be interested in. There are lots of plots of resistance vs. wiper code but what I would really want to know resistance vs. voltage across the gate (or on the case of a pot the voltage across the wiper and whatever end terminal). This is never perfectly flat which means as the signal swings up and down, the resistance changes! Now usually it's not that much. If it's only a few ohms and the load is relatively high impedance (or source impedance is low) then it's not a problem. But what if the load is like 1K? That is perfectly plausible in a conventional audio circuit. Is there going to be no additional distortion down at 80 or 90dB down? That's not so clear. To know for sure, I would have to create a simple digipot circuit and carefully measure distortion and response with various source and load impedances.
 
ruffrecords said:
Anyone thought about using a Digipot in a compressor?

Cheers

Ian
I designed a 6 channel automatic mixer using DPOTs for gain control. While not a limiter it involved tens of dB gain manipulation to multiple channels.  I anticipated adding zero crossing coordination to reduce zipper noise but did not hear any so didn't have to turn that on. (Caveat i was processing mostly speech so better than music for concealing gain step/perturbations).

The DPOT was very respectable sounding (IMO) but in 2019 why not just do it completely inside the digital domain?

The company I designed the AM for was convinced their customers would pay more for analog technology... bzzt, they were wrong as digital offers huge feature set bang for the buck. I tried to drag them into digital technology and digital controlled analog was the first step on that path. The AM never saw the light of day as their analog business faded before this reached the market.

JR
 
JohnRoberts said:
why not just do it completely inside the digital domain?
Phase shift / delay?

An ADC/DAC round trip is usually somewhere around 5ms which is a phase shift of 180 degrees at 100 Hz, 1800 degrees at 1KHz, etc. So mixing that back with the original signal will cause comb filtering at low frequencies which could suck the tone out of a channel (or make it sound better but you should have a choice in the matter).

Personally I think that all channels should make the transition between analog to digital (or between digital to analog) together  so that they are all delayed the same. So that would mean you can't use something that's DSP and then go back to analog and just do things however without regard for phase shift / delay.

Another reason for not using DSP would be IF it does not sound good. Personally speaking I think there are many things that do not sound as good implemented in DSP. I think a DSP SSL EQ would sound good. It's pretty simple filtering which DSP is known to do well. But anything that exhibits non-linear behavior is potentially difficult to model. So a guitar amp with a little edge to it is a great example where I think DSP fails to deliver. Or even maybe something like an 1176 compressor because it's basically a wave-shaper at faster settings. I think an 1176 could be done well in DSP but I think it would also be very easy to screw up and make the model too simple.
 
Hi guys,
thanks for all your input!
I basically have all the parts available already, I have a spare 132 EQ card, the card edge connector, power supply etc.
I´ve ordered two of those chips with 10k, so that would be enough to get the low pass filter of the EQ going.
When the pots have arrived it´s my plan to just wire them to the lpf and see how that works.
Even if there would be a little digital noise while adjusting, maybe it wouldn´t even be a show stopper if it´s quiet when not changing values.

I just wanted to make sure that it´s unlikely to go up in flames because of the current limitation.
From there I´m gonna take it step by step, to keep investment low in case I run into a wall anyway.

Anyway thanks again, very valuable input here!

Cheers,
Christian
 
squarewave said:
Ok now. You had to go digging for that datasheet so take it easy on me.
No I looked at the data sheet (all 50 pages) before I posted, this is the first 36v DPOT I've ever seen.
But from looking at that datasheet I actually don't even see what I would really be interested in. There are lots of plots of resistance vs. wiper code but what I would really want to know resistance vs. voltage across the gate (or on the case of a pot the voltage across the wiper and whatever end terminal). This is never perfectly flat which means as the signal swings up and down, the resistance changes! Now usually it's not that much. If it's only a few ohms and the load is relatively high impedance (or source impedance is low) then it's not a problem. But what if the load is like 1K? That is perfectly plausible in a conventional audio circuit. Is there going to be no additional distortion down at 80 or 90dB down? That's not so clear. To know for sure, I would have to create a simple digipot circuit and carefully measure distortion and response with various source and load impedances.
I get that you are not a fan of DPOTs ... how would you automate an EQ?

JR
 
The only significant issue I see with these is that their taper is linear. In most cases they would be used in rheostat mode.
Let's say you want to have a midrange frequency control with an extended range of 150Hz to 6kHz.
That's a 1:40 ratio. You would have a 250r resistor in series with the 10k pot. The last stop would be 6kHz, and the one before would be 3kHz! That's a quite poor resolution...
Even if you opt for a limited range, for example 1:10,  for let's say a 20-200Hz shelving. The last stop would be 200Hz, and the one before 160. I know that many are content with switchable inductor-based EQ's that often have less resolution, but the big point about parametrics is that they offer pinpoint resolution.
 
abbey road d enfer said:
The only significant issue I see with these is that their taper is linear. In most cases they would be used in rheostat mode.
Let's say you want to have a midrange frequency control with an extended range of 150Hz to 6kHz.
That's a 1:40 ratio. You would have a 250r resistor in series with the 10k pot. The last stop would be 6kHz, and the one before would be 3kHz! That's a quite poor resolution...
Even if you opt for a limited range, for example 1:10,  for let's say a 20-200Hz shelving. The last stop would be 200Hz, and the one before 160. I know that many are content with switchable inductor-based EQ's that often have less resolution, but the big point about parametrics is that they offer pinpoint resolution.

Thanks for voicing your concern! I don´t know if it would change things in your opinion, but the pots are also available as 50kOhm and 100kOhm versions. Other than that the SSL uses a 25kOhm pot as well, which I´d plan to realize by connecting two 50kOhms in parallel.

Regarding the linear taper, the part (in each version) has 256 linear steps across it´s range, and on page 71 of the data sheet there is even a table how to realize log behaviour (with less points of course).
 
Gertius said:
Thanks for voicing your concern! I don´t know if it would change things in your opinion, but the pots are also available as 50kOhm and 100kOhm versions. Other than that the SSL uses a 25kOhm pot as well, which I´d plan to realize by connecting two 50kOhms in parallel.
The nominal value of teh pots is not a concern; the problem of resolution is just teh fact that with 256 points, the step between each point represents a more and more significant change when approaching the end of travel.

Regarding the linear taper, the part (in each version) has 256 linear steps across it´s range, and on page 71 of the data sheet there is even a table how to realize log behaviour (with less points of course).
Yes, log but with a poor resolution.
 
> with a poor resolution.

You know this but lurkers may miss it:

You run a 10:1 (or 2:1 or 30:1) range on the pot, then the computer switches capacitors and runs the pot again from the top.
 
PRR said:
> with a poor resolution.

You know this but lurkers may miss it:

You run a 10:1 (or 2:1 or 30:1) range on the pot, then the computer switches capacitors and runs the pot again from the top.
That's the obvious answer, but here it implies a big mod of an existing piece of gear.
 
Back
Top